gOlogy: fix debug binds in auto-inc-dec
[official-gcc.git] / gcc / ChangeLog
blob2c5abd39a4ba92daacd6dde9be2a23f9eeba3e08
1 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
3         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
4         (reg_next_debug_use): New.
5         (attempt_change): Propagate adjusted expression into affected
6         debug insns.
7         (merge_in_block): Track uses in debug insns.
8         (pass_inc_dec::execute): Allocate and release
9         reg_next_debug_use.
11 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
13         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
14         (_mm512_fixupimm_round_pd): Update parameters and builtin.
15         (_mm512_maskz_fixupimm_round_pd): Ditto.
16         (_mm512_fixupimm_round_ps): Ditto.
17         (_mm512_maskz_fixupimm_round_ps): Ditto.
18         (_mm_fixupimm_round_sd): Ditto.
19         (_mm_maskz_fixupimm_round_sd): Ditto.
20         (_mm_fixupimm_round_ss): Ditto.
21         (_mm_maskz_fixupimm_round_ss): Ditto.
22         (_mm512_fixupimm_pd): Ditto.
23         (_mm512_maskz_fixupimm_pd): Ditto.
24         (_mm512_fixupimm_ps): Ditto.
25         (_mm512_maskz_fixupimm_ps): Ditto.
26         (_mm_fixupimm_sd): Ditto.
27         (_mm_maskz_fixupimm_sd): Ditto.
28         (_mm_fixupimm_ss): Ditto.
29         (_mm_maskz_fixupimm_ss): Ditto.
30         (_mm512_mask_fixupimm_round_pd): Update builtin.
31         (_mm512_mask_fixupimm_round_ps): Ditto.
32         (_mm_mask_fixupimm_round_sd): Ditto.
33         (_mm_mask_fixupimm_round_ss): Ditto.
34         (_mm512_mask_fixupimm_pd): Ditto.
35         (_mm512_mask_fixupimm_ps): Ditto.
36         (_mm_mask_fixupimm_sd): Ditto.
37         (_mm_mask_fixupimm_ss): Ditto.
38         * config/i386/avx512vlintrin.h:
39         (_mm256_fixupimm_pd): Update parameters and builtin.
40         (_mm256_maskz_fixupimm_pd): Ditto.
41         (_mm256_fixupimm_ps): Ditto.
42         (_mm256_maskz_fixupimm_ps): Ditto.
43         (_mm_fixupimm_pd): Ditto.
44         (_mm_maskz_fixupimm_pd): Ditto.
45         (_mm_fixupimm_ps): Ditto.
46         (_mm_maskz_fixupimm_ps): Ditto.
47         (_mm256_mask_fixupimm_pd): Update builtin.
48         (_mm256_mask_fixupimm_ps): Ditto.
49         (_mm_mask_fixupimm_pd): Ditto.
50         (_mm_mask_fixupimm_ps): Ditto.
51         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
52         * config/i386/i386-builtin.def: Update builtin definitions.
53         * config/i386/i386.c: Handle new builtin types and remove useless ones.
54         * config/i386/sse.md: Update VFIXUPIMM* patterns.
55         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
56         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
57         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
58         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
59         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
60         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
61         * config/i386/subst.md:
62         (round_saeonly_sd_mask_operand4): Add new subst_attr.
63         (round_saeonly_sd_mask_op4): Ditto.
64         (round_saeonly_expand_operand5): Ditto.
65         (round_saeonly_expand): Update.
67 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
69         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
71 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
73         PR rtl-optimization/87871
74         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
76 2018-11-05  Paul Koning  <ni1d@arrl.net>
78         * doc/sourcebuild.texi (target attributes): Document new "inf"
79         effective target keyword.
81 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
83         * config/s390/s390.c (s390_register_move_cost): Increase costs for
84         moves involving the CC reg.
86 2018-11-05  Richard Biener  <rguenther@suse.de>
88         PR tree-optimization/87873
89         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
90         argument.
91         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
92         * tree-vect-loop.c (vect_transform_loop): When splitting the
93         loop exit also create forwarder PHIs for constants.
94         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
95         Handle constant to_arg, add extra checking we match up the correct
96         PHIs.
98 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
100         * config/s390/s390.md: QImode and HImode for load on condition.
102 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
104         * config/s390/predicates.md: Fix typo.
105         * config/s390/s390.md: Allow immediates for load on condition.
107 2018-11-05  Martin Liska  <mliska@suse.cz>
109         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
110         * bitmap.h (struct bitmap_usage): Likewise.
111         * ggc-common.c (SCALE): Remove.
112         (LABEL): Likewise.
113         (struct ggc_usage): Use SIZE_AMOUNT. And update
114         compare method.
115         * ggc-page.c (SCALE): Remove.
116         (STAT_LABEL): Remove.
117         (ggc_print_statistics): Use SIZE_AMOUNT.
118         * gimple.h (SCALE): Remove.
119         (LABEL): Likewise.
120         * input.c (ONE_K): Remove.
121         (ONE_M): Likewise.
122         (SCALE): Likewise.
123         (STAT_LABEL): Likewise.
124         (FORMAT_AMOUNT): Likewise.
125         (dump_line_table_statistics): Use SIZE_AMOUNT.
126         * mem-stats.h (struct mem_usage): Likewise.
127         * rtl.c (dump_rtx_statistics): Likewise.
128         (rtx_alloc_counts): Change type to size_t.
129         (rtx_alloc_sizes): Likewise.
130         (rtx_count_cmp): New.
131         (dump_rtx_statistics): Sort first based on counts.
132         * tree.c (tree_nodes_cmp): New.
133         (tree_codes_cmp): New.
134         (dump_tree_statistics): Sort first based on counts.
135         * system.h (ONE_K): New.
136         (ONE_M): Likewise.
137         (SIZE_SCALE): Likewise.
138         (SIZE_LABEL): Likewise.
139         (SIZE_AMOUNT): Likewise.
140         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
141         * tree-dfa.c (dump_dfa_stats): Likewise.
142         * tree-phinodes.c (phinodes_print_statistics): Likewise.
143         * tree-ssanames.c (ssanames_print_statistics): Likewise.
144         * tree.c (dump_tree_statistics): Likewise.
145         * vec.c (struct vec_usage): Likewise.
146         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
147         trigger a -Werror=format-overflow with
148         --enable-gather-detailed-stats.
150 2018-11-05  Martin Liska  <mliska@suse.cz>
152         * mem-stats.h (mem_alloc_description::release_instance_overhead):
153         Return T *.
154         * vec.c (struct vec_usage): Register m_element_size.
155         (vec_prefix::register_overhead): New arguments: elements and
156         element_size.
157         (vec_prefix::release_overhead): Subtract elements.
158         * vec.h (struct vec_prefix): Change signature.
159         (va_heap::reserve): Pass proper arguments.
160         (va_heap::release): Likewise.
162 2018-11-05  Martin Liska  <mliska@suse.cz>
164         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
165         style.
166         * vec.c: Likewise.
168 2018-11-05  Richard Biener  <rguenther@suse.de>
170         * tree-scalar-evolution.h (final_value_replacement_loop): Update
171         prototype.
172         * tree-scalar-evolution.c (final_value_replacement_loop): Return
173         whether anything was done.
174         (scev_const_prop): Remove constant propagation part, fold
175         remains into ...
176         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
177         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
178         conditionally.
180 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
182         PR tree-optimization/87859
183         * gimple-ssa-store-merging.c (struct merged_store_group): Add
184         only_constants and first_nonmergeable_order members.
185         (merged_store_group::merged_store_group): Initialize them.
186         (merged_store_group::do_merge): Clear only_constants member if
187         adding something other than INTEGER_CST store.
188         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
189         stores with order >= first_nonmergeable_order.  Use
190         merged_store->only_constants instead of always recomputing it.
191         Set merged_store->first_nonmergeable_order if we've skipped any
192         stores.  Attempt to merge overlapping INTEGER_CST stores that
193         we would otherwise skip.
195         PR sanitizer/87837
196         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
198 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
200         PR target/87853
201         * config/i386/emmintrin.h (__v16qs): New to cope with option
202         -funsigned-char.
203         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
204         (_mm_cmplt_epi8): Likewise.
205         (_mm_cmpgt_epi8): Likewise.
207 2018-11-05  Richard Biener  <rguenther@suse.de>
209         PR rtl-optimization/87852
210         * fwprop.c (use_killed_between): Only consider single-defs of the
211         use whose definition statement dominates the use.
213 2018-11-05  Martin Liska  <mliska@suse.cz>
215         PR web/87829
216         * doc/invoke.texi: Remove options that are
217         not disabled with -Os.
219 2018-11-05  Martin Liska  <mliska@suse.cz>
221         PR c/87811
222         * doc/extend.texi: Update constrain about the last argument
223         of __builtin_expect_with_probability.
225 2018-11-05  Martin Liska  <mliska@suse.cz>
227         PR c/87811
228         * predict.c (expr_expected_value_1): Verify
229         that last argument is a real constants and emit
230         error.
232 2018-11-05  Martin Liska  <mliska@suse.cz>
234         PR gcov-profile/77698
235         * ipa-profile.c (ipa_profile): Adjust hotness threshold
236         only in LTO mode.
238 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
240         PR tree-optimization/86572
241         * builtins.c (c_strlen): Handle negative offsets in a safe way.
243 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
245         PR tree-optimization/87672
246         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
247         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
249 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
251         PR middle-end/58372
252         * cfgexpand.c (pass_expand::execute): Move the call to
253         finish_eh_generation in front of the call to expand_stack_alignment.
255 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
257         * common/config/i386/i386-common.c (processor_alias_table): Add
258         znver2 entry.
259         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
260         (case ${target}): Add znver2.
261         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
262         -march=native recognize znver2 processors.
263         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
264         * config/i386/i386.c (m_znver2): New definition.
265         (m_ZNVER): New definition.
266         (m_AMD_MULTIPLE): Includes m_znver2.
267         (processor_cost_table): Add znver2 entry.
268         (processor_target_table): Add znver2 entry.
269         (get_builtin_code_for_version): Set priority for
270         PROCESSOR_ZNVER2.
271         (processor_model): Add M_AMDFAM17H_ZNVER2.
272         (arch_names_table): Ditto.
273         (ix86_reassociation_width): Include znver2.
274         * config/i386/i386.h (TARGET_znver2): New definition.
275         (struct ix86_size_cost): Add TARGET_ZNVER2.
276         (enum processor_type): Add PROCESSOR_ZNVER2.
277         * config/i386/i386.md (define_attr "cpu"): Add znver2.
278         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
279         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
280         (ix86_adjust_cost): Add znver2.
281         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
282         * gcc/doc/extend.texi: Add details about znver2.
283         * gcc/doc/invoke.texi: Add details about znver2.
285 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
287         PR target/87079
289         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
290         pattern.
292 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
294         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
295         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
297 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
299         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
300         shrink-wrapping when -mtrack-speculation.
302 2018-11-02  Richard Biener  <rguenther@suse.de>
304         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
305         (pop_cost_one_pair): Do not free pair.
306         (pop_best_coalesce): Likewise.
307         (create_coalesce_list): Initialize obstack.
308         (delete_coalesce_list): Free obstack.
309         (find_coalesce_pair): Obstack-allocate coalesce pairs.
310         (add_cost_one_coalesce): Likewise.
311         (struct live_track): Remove bitmap pointer indirections.
312         (new_live_track): Adjust.
313         (delete_live_track): Likewise.
314         (live_track_remove_partition): Likewise.
315         (live_track_add_partition): Likewise.
316         (live_track_live_p): Likewise.
317         (live_track_process_def): Likewise.
318         (live_track_clear_base_vars): Likewise.
320 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
322         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
323         * configure: Regenerate.
324         * config.in: Regenerate.
325         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
326         HAVE_AS_REGISTER_PSEUDO_OP guard.
327         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
328         !HAVE_AS_REGISTER_PSEUDO_OP support.
329         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
330         guard.
332 2018-11-02  Richard Biener  <rguenther@suse.de>
334         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
335         with is_type_die.
337 2018-11-02  Richard Biener  <rguenther@suse.de>
339         PR tree-optimization/87776
340         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
341         executable when iterating but running into rpo-vn-max-loop-depth
342         and not eliding the iteration.
344 2018-11-30  Jan Hubicka  <jh@suse.cz>
346         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
347         TYPE_DECL.
349 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
351         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
352         prototype.
353         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
354         Combine with rs6000_address_for_fpconvert.
355         (rs6000_address_for_fpconvert): Combine with
356         rs6000_force_indexed_or_indirect_mem.
357         (rs6000_expand_vector_init): Change function call from
358         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
359         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
360         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
361         (floatsi<mode>2_lfiwax_mem): Ditto.
362         (floatunssi<mode>2_lfiwzx): Ditto.
363         (floatunssi<mode>2_lfiwzx_mem): Ditto.
364         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
365         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
366         (fix_trunc<mode>si2_stfiwx): Ditto.
367         (fixuns_trunc<mode>si2_stfiwx): Ditto.
368         (float_<mode>si2_hw): Ditto.
369         (floatuns_<mode>si2_hw): Ditto.
370         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
371         (vsx_splat_<mode>): Ditto.
373 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
375         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
376         at end of assembler input text.
377         * configure: Regenerate.
379 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
381         PR tree-optimization/87826
382         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
383         negative or larger or equal to type's precision.
385 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
387         * opts.c (default_options_table): Do not enable
388         OPT_fdelayed_branch at -Og.
389         * doc/invoke.texi (-fdelayed-branch): Document it.
391 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
393         * optabs-libfuncs.c (build_libfunc_function_visibility):
394         New, split out from...
395         (build_libfunc_function): ... here.
396         (init_one_libfunc_visibility): New, split out from ...
397         (init_one_libfunc): ... here.
399         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
400         scratch register need not be early-clobber.  Document the reason
401         why we cannot use ST<OP>.
403 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
405         PR bootstrap/82856
406         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
407         line for second argument of AC_DEFINE_UNQUOTED.
408         * doc/install.texi (Tools/packages necessary for modifying GCC):
409         Update to autoconf 2.69 and automake 1.15.1.
410         * aclocal.m4, config.in, configure: Regenerate.
412 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
414         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
415         initialization.
417 2018-10-31  Martin Liska  <mliska@suse.cz>
419         PR driver/83193
420         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
421         Add new argument invalid_extension.
422         (aarch64_get_all_extension_candidates): New function.
423         (aarch64_rewrite_selected_cpu): Add NULL to function call.
424         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
425         new argument.
426         (aarch64_get_all_extension_candidates): New function.
427         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
428         argument invalid_extension.
429         (aarch64_parse_cpu): Likewise.
430         (aarch64_print_hint_for_extensions): New function.
431         (aarch64_validate_mcpu): Provide hint about invalid extension.
432         (aarch64_validate_march): Likewise.
433         (aarch64_handle_attr_arch): Pass new argument.
434         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
435         (aarch64_handle_attr_isa_flags): Likewise.
437 2018-10-31  Richard Biener  <rguenther@suse.de>
439         PR middle-end/70359
440         PR middle-end/86270
441         * tree-outof-ssa.c (insert_backedge_copies): Restrict
442         copy generation to useful cases.  Place the copy before
443         the definition of the backedge value when possible.
445 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
447         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
448         * config/arc/arc.c (arc_active_insn): New function.
449         (check_store_cacheline_hazard): Likewise.
450         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
451         (arc_override_options): Disable delay slot scheduler for older
452         A7.
453         (arc_store_addr_hazard_p): New implementation, old one renamed to
454         ...
455         (arc_store_addr_hazard_internal_p): Renamed.
456         (arc_reorg): Don't combine into brcc instructions which are part
457         of hardware hazard solution.
458         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
459         (tune_arc700): Likewise.
460         * config/arc/arc.opt (arc7xx): New tune value.
461         * config/arc/arc700.md: Improve A7 scheduler.
463 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
465         * config/arc/arc.c (arc_override_options): Remove
466         TARGET_COMPACT_CASESI.
467         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
468         (CASE_VECTOR_MODE): Likewise.
469         (CASE_VECTOR_PC_RELATIVE): Likewise.
470         (CASE_VECTOR_SHORTEN_MODE): Likewise.
471         (CASE_VECTOR_SHORTEN_MODE1): Delete.
472         (ADDR_VEC_ALIGN): Update.
473         (ASM_OUTPUT_CASE_LABEL): Undefine.
474         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
475         (TARGET_BI_BIH): Define.
476         (DEFAULT_BRANCH_INDEX): Likewise.
477         * config/arc/arc.md (casesi): Rework to accept BI/BIH
478         instructions, remove compact_casesi use case.
479         (casesi_compact_jump): Remove.
480         (casesi_dispatch): New pattern.
481         * config/arc/arc.opt: Add mbranch-index option. Deprecate
482         compact_casesi option.
483         * doc/invoke.texi: Document mbranch-index option.
485 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
487         * config/arc/arc.c (arc_get_tp): Remove function.
488         (arc_emit_call_tls_get_addr): Likewise.
489         (arc_call_tls_get_addr): New function.
490         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
491         * config/arc/arc.md (tls_load_tp_soft): Remove.
492         (tls_gd_get_addr): Likewise.
494 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
496         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
497         (mulsi3_600_lib): Remove pattern.
498         (umulsi3_highpart_600_lib_le): Likewise.
499         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
500         (umulsidi3): Remove call to umulsidi3_600_lib.
501         (umulsidi3_600_lib): Remove pattern.
502         (peephole2): Remove peephole using the above deprecated patterns.
504 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
506         PR target/87374
507         * config/arm/arm.c (arm_option_check_internal): Disable the combined
508         use of -mslow-flash-data and -mword-relocations.
509         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
510         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
511         flag_pic.
512         * doc/invoke.texi (-mword-relocations): Mention conflict with
513         -mslow-flash-data.
514         (-mslow-flash-data): Reciprocally.
516 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
518         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
519         16-byte modes held in GP registers to use an even regno.
521         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
522         (aarch64_atomic_ldop_supported_p): Remove.
523         (aarch64_gen_atomic_ldop): Remove.
524         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
525         Fully expand LSE operations here.
526         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
527         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
528         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
529         and use ATOMIC_LDOP instead; use register_operand for the input;
530         drop the split and emit insns directly.
531         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
532         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
533         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
535         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
536         (aarch64_gen_atomic_ldop): Don't call it.
537         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
538         Use aarch64_reg_or_zero.
539         (aarch64_atomic_exchange<ALLI>): Likewise.
540         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
541         operand 0; use aarch64_reg_or_zero for input; merge ...
542         (@aarch64_atomic_swp<ALLI>): ... this and remove.
544         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
545         (aarch64_split_compare_and_swap): Use it.
546         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
547         test oldval against the proper predicate.
548         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
549         Use nonmemory_operand for expected.
550         (cas_short_expected_pred): New.
551         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
552         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
553         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
554         (aarch64_plushi_operand): New.
556         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
557         Force oldval into the rval register for TARGET_LSE; emit the compare
558         during initial expansion so that it may be deleted if unused.
559         (aarch64_gen_atomic_cas): Remove.
560         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
561         Change =&r to +r for operand 0; use match_dup for operand 2;
562         remove is_weak and mod_f operands as unused.  Drop the split
563         and merge with...
564         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
565         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
566         (@aarch64_atomic_cas<GPI>): Similarly.
568 2018-10-31  Richard Biener  <rguenther@suse.de>
570         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
571         using ABSU_EXPR.
573 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
575         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
576         * config/aarch64/aarch64.md: Include saphira.md
577         * config/aarch64/saphira.md: New file for pipeline description.
579 2018-10-30  Martin Sebor  <msebor@redhat.com>
581         PR middle-end/87041
582         * gimple-ssa-sprintf.c (format_directive): Use %G to include
583         inlining context.
584         (sprintf_dom_walker::compute_format_length):
585         Avoid setting POSUNDER4K here.
586         (get_destination_size): Handle null argument values.
587         (get_user_idx_format): New function.
588         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
589         functions, including user-defined with attribute format printf.
590         Use %G to include inlining context.
591         Set POSUNDER4K here.
593 2018-10-30  Jan Hubicka  <jh@suse.cz>
595         * params.def (lto-partitions): Bump from 32 to 128.
597 2018-10-30  Jan Hubicka  <jh@suse.cz>
599         * tree.c
600         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
601         head in file.
602         (free_lang_data_in_type): Forward declare.
603         (fld_type_variant_equal_p): New function.
604         (fld_type_variant): New function
605         (fld_incomplete_types): New hash.
606         (fld_incomplete_type_of): New function
607         (fld_simplfied-type): New function.
608         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
609         (free_lang_data): Allocate and free fld_incomplete_type; update call
610         of free_lang_data_in_decl.
612 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
614         * gcov.c (output_lines): Remove duplicate line.
616 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
618         * config/rs6000/rs6000.md (bswapdi2): Force address into register
619         if not in indexed or indirect form.
620         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
621         (bswapdi2_store): Ditto.
622         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
623         helper function.
624         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
625         Prototype for helper function.
627 2018-10-30  Martin Sebor  <msebor@redhat.com>
629         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
630         (target, pragma GCC optimize, pragma GCC target): Ditto.
632 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
634         * doc/extend.texi: Fix prototype and description of
635         __builtin_expect_with_probability.
637 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
639         * cgraph.h (clone_function_name_1): Replaced by new
640           clone_function_name_numbered that takes name as string; for
641           privatize_symbol_name_1 use only.
642           (clone_function_name): Renamed to
643           clone_function_name_numbered to be explicit about numbering.
644           (clone_function_name): New two-argument function that does
645           not number its output.
646           (clone_function_name): New three-argument function that
647           takes a number to append to its output.
648         * cgraphclones.c (duplicate_thunk_for_node):
649           (clone_function_name_1): Renamed.
650           (clone_function_name_numbered): Two new functions.
651           (clone_function_name): Improved documentation.
652           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
653         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
654         * final.c (final_scan_insn_1): Use the new clone_function_name
655           without numbering.
656         * multiple_target.c (create_dispatcher_calls): Ditto.
657           (create_target_clone): Ditto.
658         * omp-expand.c (grid_expand_target_grid_body): Ditto.
659         * omp-low.c (create_omp_child_function_name): Ditto.
660         * omp-simd-clone.c (simd_clone_create): Ditto.
661         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
662           new clone_function_name without numbering.
664 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
666         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
667         Assert that the allocation size is not zero.
669 2018-10-30  Richard Biener  <rguenther@suse.de>
671         PR tree-optimization/87800
672         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
673         non-induction or reduction PHIs.
675 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
677         * config/aarch64/falkor-tag-collision-avoidance.c
678         (execute_tag_collision_avoidance): Call df_note_add_problem.
680 2018-10-30  Martin Liska  <mliska@suse.cz>
682         * doc/extend.texi: Fix typo in documentation
683         of __builtin_expect_with_probability.
685 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
687         PR c++/87721
688         * input.c (get_substring_ranges_for_loc): Detect if
689         linemap_resolve_location gives us a NULL map, and reject
690         this case.
692 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
694         * config.gcc (xstormy16-*-elf): Set tm_d_file.
696 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
697             Martin Sebor  <msebor@redhat.com>
698             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
700         * doc/ux.texi (Quoting): New subsection, adapted from material at
701         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
702         MartinSebor and ManuelLopezIbanez.
703         (Fix-it hints): Note that fix-it hints shouldn't be marked for
704         translation.
706 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
708         PR middle-end/87469
709         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
710         max value.
712 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
714         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
716 2018-10-29  Paul Koning  <ni1d@arrl.net>
718         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
720 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
722         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
723         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
724         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
725         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
726         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
727         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
728         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
729         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
730         Change 'vector' to '__vector'.
731         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
732         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
733         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
734         _mm_avg_pu16): Likewise.  And, whitespace corrections.
736 2018-10-29  Richard Biener  <rguenther@suse.de>
738         PR tree-optimization/87785
739         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
740         internal defs.
742 2018-10-29  Olivier Hainque  <hainque@adacore.com>
744         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
746 2018-10-29  Olivier Hainque  <hainque@adacore.com>
748         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
749         ports configuration macro, defaults to "gnu".
750         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
751         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
753 2018-10-29  Olivier Hainque  <hainque@adacore.com>
755         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
757 2018-10-29  Richard Biener  <rguenther@suse.de>
759         PR tree-optimization/87790
760         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
761         (vect_make_slp_decision): Adjust.
762         (vect_slp_analyze_bb_1): Likewise.
763         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
764         edges.
766 2018-10-29  Richard Biener  <rguenther@suse.de>
768         PR tree-optimization/87785
769         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
770         and processing.
771         (vect_build_slp_tree): Likewise.
772         (vect_gather_slp_loads): New function.
773         (vect_analyze_slp_instance): Gather loads separately from the
774         SLP tree build.
776 2018-10-29  Martin Liska  <mliska@suse.cz>
778         * Makefile.in: Make dependency to json.o.
779         * doc/gcov.texi: Document new JSON format, remove
780         old intermediate format documentation.
781         * gcov.c (struct function_info): Come up with m_name and
782         m_demangled_name.
783         (function_info::function_info): Initialize it.
784         (function_info::~function_info): Release it.
785         (main): Rename flag_intermediate_format to flag_json_format.
786         (print_usage): Describe --json-format.
787         (process_args): Set flag_json_format.
788         (output_intermediate_line): Remove.
789         (output_intermediate_json_line): Likewise.
790         (get_gcov_intermediate_filename): Return new extension
791         ".gcov.json.gz".
792         (output_intermediate_file): Implement JSON emission.
793         (output_json_intermediate_file): Implement JSON emission.
794         (generate_results): Use ::get_name for function name.
795         Handle JSON output file.
796         (read_graph_file): Use ::get_name instead of cplus_demangle.
797         (read_count_file): Likewise.
798         (solve_flow_graph): Likewise.
799         (add_line_counts): Likewise.
800         (accumulate_line_counts): Use new flag_json_format.
801         (output_function_details): Use ::get_name instead of cplus_demangle.
802         (output_lines): Likewise.
803         * json.cc (test_writing_literals): Add new tests.
804         * json.h (class literal): Add new boolean constructor.
806 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
808         PR rtl-optimization/87701
809         PR rtl-optimization/87780
810         * combine.c (make_more_copies): Rewrite.
812 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
814         * doc/generic.texi (ABSU_EXPR): Document.
815         * match.pd (absu(x)*absu(x) -> x*x): Handle.
816         (absu(absu(X)) -> absu(X)): Likewise.
817         (absu(-X) -> absu(X)): Likewise.
818         (absu(X)  where X is nonnegative -> X): Likewise.
820 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
822         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
823         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
824         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
825         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
826         (s-tm-texi): Also check timestamp on d-target.def.
827         (generated_files): Add TM_D_H and d-target-hooks-def.h.
828         (build/genhooks.o): Also depend on D_TARGET_DEF.
829         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
830         variables.
831         * config/aarch64/aarch64-d.c: New file.
832         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
833         Define.
834         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
835         prototype.
836         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
837         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
838         * config/arm/arm-d.c: New file.
839         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
840         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
841         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
842         * config/arm/t-arm (arm-d.o): New rule.
843         * config/default-d.c: New file.
844         * config/glibc-d.c: New file.
845         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
846         * config/i386/i386-d.c: New file.
847         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
848         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
849         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
850         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
851         * config/i386/t-i386 (i386-d.o): New rule.
852         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
853         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
854         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
855         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
856         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
857         * config/mips/mips-d.c: New file.
858         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
859         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
860         * config/mips/t-mips (mips-d.o): New rule.
861         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
862         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
863         * config/powerpcspe/powerpcspe-d.c: New file.
864         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
865         New prototype.
866         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
867         Support GNU D by using 0 as the language type.
868         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
869         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
870         * config/riscv/riscv-d.c: New file.
871         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
872         prototype.
873         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
874         * config/riscv/t-riscv (riscv-d.o): New rule.
875         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
876         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
877         * config/rs6000/rs6000-d.c: New file.
878         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
879         prototype.
880         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
881         Support GNU D by using 0 as the language type.
882         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
883         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
884         * config/s390/s390-d.c: New file.
885         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
886         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
887         * config/s390/t-s390 (s390-d.o): New rule.
888         * config/sparc/sparc-d.c: New file.
889         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
890         prototype.
891         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
892         * config/sparc/t-sparc (sparc-d.o): New rule.
893         * config/t-glibc (glibc-d.o): New rule.
894         * configure: Regenerated.
895         * configure.ac (tm_d_file): New variable.
896         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
897         * doc/contrib.texi (Contributors): Add self for the D frontend.
898         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
899         * doc/install.texi (Configuration): Mention libphobos as an option for
900         --enable-shared.  Mention d as an option for --enable-languages.
901         (Testing): Mention check-d as a target.
902         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
903         name suffixes.  Mention d as a -x option.
904         * doc/sourcebuild.texi (Top Level): Mention libphobos.
905         * doc/standards.texi (Standards): Add section on D language.
906         * doc/tm.texi: Regenerated.
907         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
908         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
909         * dwarf2out.c (is_dlang): New function.
910         (gen_compile_unit_die): Use DW_LANG_D for D.
911         (declare_in_namespace): Return module die for D, instead of adding
912         extra declarations into the namespace.
913         (gen_namespace_die): Generate DW_TAG_module for D.
914         (gen_decl_die): Handle CONST_DECLSs for D.
915         (dwarf2out_decl): Likewise.
916         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
917         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
918         kinds of aggregates.
919         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
920         * genhooks.c: Include d/d-target.def.
922 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
924         PR target/85669
925         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
926         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
927         (STACK_DYNAMIC_OFFSET): Likewise.
929 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
931         PR target/80024
932         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
933         error message.
935 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
937         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
938         return type and other typos.
940 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
942         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
943         a shorter sequence with fewer branches.
944         (emit_final_str_compare_gpr): Ditto.
946 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
948         * config/rs6000/tmmintrin.h: New file.
949         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
951 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
953         * config/rs6000/mmintrin.h: Enable 32bit compilation.
954         * config/rs6000/xmmintrin.h: Likewise.
956 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
958         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
960 2018-10-26  Richard Biener  <rguenther@suse.de>
962         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
963         and wrapper.
964         (vect_mark_slp_stmts_relevant): Likewise.
965         (vect_detect_hybrid_slp_stmts): Likewise.
966         (vect_bb_slp_scalar_cost): Likewise.
967         (vect_remove_slp_scalar_calls): Likewise.
969 2018-10-26  Jan Hubicka  <jh@suse.cz>
971         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
972         (warn_types_mismatch): Fix walk of DECL_NAME.
973         (odr_types_equivalent_p): Fix overactive assert.
975 2018-10-26  Richard Biener  <rguenther@suse.de>
977         PR tree-optimization/87105
978         * tree-vectorizer.h (_slp_tree::refcnt): New member.
979         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
980         refcnt.
981         (vect_create_new_slp_node): Initialize refcnt to one.
982         (bst_traits): Move.
983         (scalar_stmts_set_t, bst_fail): Remove.
984         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
985         (vect_build_slp_tree): Add bst_map argument and lookup
986         already created SLP nodes.
987         (vect_print_slp_tree): Handle a SLP graph, print SLP node
988         addresses.
989         (vect_slp_rearrange_stmts): Handle a SLP graph.
990         (vect_analyze_slp_instance): Adjust and free SLP nodes from
991         the CSE map.  Fix indenting.
992         (vect_schedule_slp_instance): Add short-cut.
994 2018-10-26  Martin Liska  <mliska@suse.cz>
996         PR testsuite/86158
997         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
998         addr_expr and not with pointers.
1000 2018-10-26  Jan Hubicka  <jh@suse.cz>
1002         * tree.c (free_lang_data_in_type): Only check main variants.
1003         * ipa-devirt.c (warn_odr): Make static.
1004         (types_same_for_odr): Drop strict variant.
1005         (types_odr_comparable): Likewise.
1006         (odr_or_derived_type_p): Look for main variants.
1007         (odr_name_hasher::equal): Cleanup comment.
1008         (odr_subtypes_equivalent): Add warn and warned arguments; check main
1009         variants.
1010         (type_variants_equivalent_p): break out from ...
1011         (odr_types_equivalent): ... here; go for main variants where needed.
1012         (warn_odr): ... here; turn static.
1013         (warn_types_mismatch): Compare mangled names of main variants.
1014         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
1015         (type_with_linkage_p): Sanity check that we look at main variant.
1016         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
1017         * tree.h (types_same_for_odr): Drop strict argument.
1019 2018-10-26  Richard Biener  <rguenther@suse.de>
1021         PR tree-optimization/87746
1022         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1023         Simplify and fix WRT strided store groups with size not
1024         equal to step in element count.
1025         (vect_analyze_group_access_1): Dump the whole group.
1027 2018-10-25  Carl Love  <cel@us.ibm.com>
1029         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
1030         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
1031         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
1032         precicion entry for each overloaded builtin.
1033         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
1034         VSCEDPUO): Rename overloaded name.
1035         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
1036         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
1037         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
1038         define_expand for xscmpexqp instruction.
1039         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
1041 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1042             Jinsong Ji <jji@us.ibm.com>
1044         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
1045         function with vec_sl.
1046         (_mm_slli_epi32): Likewise.
1047         (_mm_slli_epi64): Likewise.
1048         (_mm_srai_epi16): Replace deprecated function with vec_sra.
1049         (_mm_srai_epi32): Likewise.
1050         (_mm_srli_epi16): Replace deprecated function with vec_sr.
1051         (_mm_srli_epi32): Likewise.
1052         (_mm_srli_epi64): Likewise.
1053         (_mm_sll_epi16): Replace deprecated function with vec_sl.
1054         (_mm_sll_epi32): Likewise.
1055         (_mm_sll_epi64): Likewise.
1056         (_mm_sra_epi16): Replace deprecated function with vec_sra.
1057         (_mm_sra_epi32): Likewise.
1058         (_mm_srl_epi16): Replace deprecated function with vec_sr.
1059         (_mm_srl_epi32): Likewise.
1060         (_mm_srl_epi64): Likewise.
1062 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1063             Jinsong Ji <jji@us.ibm.com>
1065         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
1066         comparison operators with vec_cmp* for compatibility due to
1067         unfortunate history; clean up formatting and use types more
1068         appropriately.
1069         (_mm_sll_epi32): Likewise.
1070         (_mm_sll_epi64): Likewise.
1071         (_mm_srl_epi16): Likewise.
1072         (_mm_srl_epi32): Likewise.
1073         (_mm_srl_epi64): Likewise.
1075 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1076             Jinsong Ji <jji@us.ibm.com>
1078         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
1079         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
1080         __vector __bool int.  Use vec_cmpgt in preference to deprecated
1081         function vec_vcmpgtfp.
1082         (_mm_max_ps): Likewise.
1084 2018-10-25  Jeff Law  <law@redhat.com>
1086         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
1087         if returning false.
1089 2018-10-25  Martin Sebor  <msebor@redhat.com>
1091         * doc/extend.texi (aligned): Expand attribute description.
1092         (Alignment): Rename section.  Discuss function arguments.
1094 2018-10-25  Jan Hubicka  <jh@suse.cz>
1096         * ipa-devirt.c (main_odr_variant): Remove.
1097         (hash_odr_name, types_same_for_odr, types_odr_comparable,
1098         odr_name_hasher::equal, odr_subtypes_equivalent_p):
1099         Drop use of main_odr_variant.
1100         (add_type_duplicate): Silence confused warnings on integer types.
1101         (get_odr_type): Always look for main variant.
1102         (register_odr_type): Simplify.
1104 2018-10-25  Richard Biener  <rguenther@suse.de>
1106         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
1107         Initialize ng to silence error with release checking bootstrap.
1109 2018-10-25  Richard Biener  <rguenther@suse.de>
1111         * tree-if-conv.c: Include tree-ssa-sccvn.h.
1112         (tree_if_conversion): Run CSE on the if-converted loop body.
1114 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1116         * config/s390/constraints.md (ZL): New constraint.
1117         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
1118         operands.
1119         * config/s390/s390.md (movdi_larl): Remove.
1120         (movdi_64): Add the LARL alternative.
1122 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1124         PR bootstrap/87747
1125         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
1126         (RTX_CODE_HWINT_P): New macro.
1127         (rtx_code_size): Use RTX_CODE_HWINT_P ().
1129 2018-10-25  Jan Hubicka  <jh@suse.cz>
1131         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
1132         is anonymous.
1134 2018-10-25  Richard Biener  <rguenther@suse.de>
1136         PR tree-optimization/87665
1137         PR tree-optimization/87745
1138         * tree-vectorizer.h (get_earlier_stmt): Remove.
1139         (get_later_stmt): Pick up UID from the original non-pattern stmt.
1141 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
1143         * options.texi (Deprecated): Move list to Var section.
1145 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
1146             Jinsong Ji <jji@us.ibm.com>
1148         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
1149         __vector long to __vector long long.
1150         (_mm_cvtpd_ps): Likewise.
1151         (_mm_cvttpd_epi32): Likewise.
1152         (_mm_cvtpi32_pd): Likewise.
1153         (_mm_unpackhi_epi64): Likewise.
1154         (_mm_unpacklo_epi64): Likewise.
1156 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
1158         PR rtl-optimization/87720
1159         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
1161 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
1163         * gimple-ssa-isolate-paths.c
1164         (find_implicit_erroneous_behavior): Do not change code if the
1165         pass is running for warnings only.
1166         (find_explicit_erroneous_behavior): Likewise.
1168 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
1170         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
1171         Define as rs6000_mangle_decl_assembler_name.
1172         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
1173         long double to IEEE long double, switch the names of the long
1174         double built-in functions to be <func>f128 instead of <func>l.
1176 2018-10-24  Martin Sebor  <msebor@redhat.com>
1178         * doc/extend.texi (nonnull): List no-argument form.  Reference
1179         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
1181 2018-10-24  Richard Biener  <rguenther@suse.de>
1183         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
1185 2018-10-24  Martin Liska  <mliska@suse.cz>
1187         PR tree-optimization/84436
1188         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
1189         Remove.
1190         (switch_conversion::contains_linear_function_p): New.
1191         (switch_conversion::build_one_array): Support linear
1192         transformation on input.
1193         * tree-switch-conversion.h (struct switch_conversion): Add
1194         contains_linear_function_p declaration.
1196 2018-10-24  Richard Biener  <rguenther@suse.de>
1198         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
1199         if its argument is CONSTANT_CLASS_P.
1201 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1203         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
1204         it is wrong for forward declarations.
1206 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1208         * config/s390/s390.c (s390_check_qrst_address): Add the missing
1209         SYMBOL_REF_P () check.
1211 2018-10-24  Richard Biener  <rguenther@suse.de>
1213         PR tree-optimization/87105
1214         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
1215         dump classification.
1216         (vect_analyze_data_ref_accesses): Handle duplicate loads and
1217         stores by splitting the affected group after the fact.
1218         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
1219         fail the SLP build because of size constraints.
1221 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1223         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
1224         * configure: Regenerate.
1225         * config.in: Regenerate.
1226         * varasm.c (mergeable_string_section): Use readonly_data_section
1227         if linker doesn't support SHF_MERGE with alignment > 8.
1228         (mergeable_constant_section): Likewise.
1230 2018-10-24  Richard Biener  <rguenther@suse.de>
1232         PR tree-optimization/84013
1233         * tree-ssa-structalias.c (struct msdi_data): New struct for
1234         marshalling data to walk_stmt_load_store_ops.
1235         (maybe_set_dependence_info): Refactor as callback for
1236         walk_stmt_load_store_ops.
1237         (compute_dependence_clique): Set restrict info on all stmt kinds.
1239 2018-10-24  Martin Liska  <mliska@suse.cz>
1241         * cgraph.c (cgraph_node::dump):
1242         Remove reduntant dumps and make tp_first_run dump more compact.
1244 2018-10-24  Richard Biener  <rguenther@suse.de>
1246         PR tree-optimization/87665
1247         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
1248         to reflect reality.
1250 2018-10-12  Jeff Law  <law@redhat.com>
1252         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
1253         for H8/S.
1255 2018-10-23  Richard Biener  <rguenther@suse.de>
1257         * tree-vrp.c (add_assert_info): Guard dump_printf with
1258         dump_enabled_p.
1259         * gimple-ssa-evrp-analyze.c
1260         (evrp_range_analyzer::record_ranges_from_incoming_edge):
1261         Use value_range::ignore_equivs_equal_p.
1263 2018-10-23  Richard Biener  <rguenther@suse.de>
1265         PR tree-optimization/87105
1266         PR tree-optimization/87608
1267         * passes.def (pass_all_early_optimizations): Add early phi-opt
1268         after dce.
1269         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
1270         addition to debug stmts.
1271         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
1272         and abs replacement early.
1273         * tree-cfg.c (gimple_empty_block_p): Likewise.
1275 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
1277         PR target/86383
1278         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
1279         specified to configure.
1280         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
1282 2018-10-23  Richard Biener  <rguenther@suse.de>
1284         PR tree-optimization/87700
1285         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
1287 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
1289         PR target/87674
1290         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
1291         second argument from __mmask16 to __mmask8.
1292         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
1293         _mm_mask_packs_epi32): Likewise.
1294         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
1295         Likewise.
1296         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
1298 2018-10-23  Richard Biener  <rguenther@suse.de>
1300         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
1302 2018-10-23  Richard Biener  <rguenther@suse.de>
1304         PR tree-optimization/86144
1305         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
1306         over simd attribute.
1308 2018-10-23  Richard Biener  <rguenther@suse.de>
1310         PR tree-optimization/87693
1311         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
1312         the case we do not find the taken edge.
1314 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
1315             Jinsong Ji  <jji@us.ibm.com>
1317         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
1318         (_mm_store_pd): Use unaligned vector type for pointer cast.
1319         (_mm_maskmoveu_si128): Likewise.
1320         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
1321         (_mm_store_ps): Use unaligned vector type for pointer cast.
1323 2018-10-22  Paul Koning  <ni1d@arrl.net>
1325         * symtab.c (symtab_node::increase_alignment): Correct max
1326         alignment check.
1328 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
1330         PR tree-optimization/87633
1331         * match.pd: Do not generate unordered integer comparisons.
1333 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1335         PR rtl-optimization/87600
1336         * combine.c: Add include of expr.h.
1337         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
1338         register to a pseudo.
1339         (make_more_copies): New function, add a copy to a new pseudo after
1340         the moves from hard registers into pseudos.
1341         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
1342         later.  Call make_more_copies.
1344 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
1346         * lra-constraints.c (process_alt_operands): New local array,
1347         matching_early_clobber.  Check matching_early_clobber before
1348         decrementing reject, and set matching_early_clobber after.
1350 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1352         PR target/87598
1353         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
1354         call output_addr_const and hope for the best.
1356 2018-10-22  Richard Biener  <rguenther@suse.de>
1358         * gimple-ssa-evrp-analyze.c
1359         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
1360         smarter about what ranges to use.
1361         * tree-vrp.c (add_assert_info): Dump here.
1362         (register_edge_assert_for_2): Instead of here at multiple but
1363         not all places.
1365         * gcc.dg/tree-ssa/evrp12.c: New testcase.
1366         * gcc.dg/predict-6.c: Adjust.
1367         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
1368         * gcc.dg/tree-ssa/vrp02.c: Likewise.
1369         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
1371 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
1372         Richard Biener  <rguenther@suse.de>
1374         * bitmap.h: Update data structure documentation, including a
1375         description of bitmap views as either linked-lists or splay trees.
1376         (struct bitmap_element_def): Update comments for splay tree bitmaps.
1377         (struct bitmap_head_def): Likewise.
1378         (bitmap_list_view, bitmap_tree_view): New prototypes.
1379         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
1380         tree_form fields.
1381         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
1382         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
1383         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
1384         released bitmap element here.
1385         (bitmap_element_free): Remove.
1386         (bitmap_elt_clear_from): Work on splay tree bitmaps.
1387         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
1388         this function similar ones such that linked-list bitmap implementation
1389         functions are grouped.
1390         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
1391         and moved for grouping.
1392         (bitmap_list_insert_element_after): Renamed from
1393         bitmap_elt_insert_after, and moved for grouping.
1394         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
1395         (bitmap_tree_link_left, bitmap_tree_link_right,
1396         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
1397         bitmap_tree_link_element, bitmap_tree_unlink_element,
1398         bitmap_tree_find_element): New functions for splay-tree bitmap
1399         implementation.
1400         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
1401         Renamed and moved, see above entries.
1402         (bitmap_tree_listify_from): New function to convert part of a splay
1403         tree bitmap to a linked-list bitmap.
1404         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
1405         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
1406         (bitmap_find_bit): Remove.
1407         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
1408         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
1409         Handle splay tree bitmaps.
1410         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
1411         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
1412         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
1413         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
1414         bitmap_intersect_compl_p, bitmap_ior_and_compl,
1415         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
1416         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
1417         corresponding changes to use linked-list specific bitmap_element
1418         manipulation functions as applicable for efficiency.
1419         (bitmap_tree_to_vec): New function.
1420         (debug_bitmap_elt_file): New function split out from ...
1421         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
1422         (bitmap_print): Likewise.
1424         PR tree-optimization/63155
1425         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
1426         SSA edge worklists.
1427         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
1428         in tree-view.
1430 2018-10-22  Martin Liska  <mliska@suse.cz>
1432         PR tree-optimization/87686
1433         Revert
1434         2018-08-29  Martin Liska  <mliska@suse.cz>
1436         * tree-switch-conversion.c (switch_conversion::expand):
1437         Strenghten assumption about gswitch statements.
1439 2018-10-22  Martin Liska  <mliska@suse.cz>
1441         * ipa-icf.c (sem_item::compare_attributes): Remove.
1442         (sem_item::compare_referenced_symbol_properties): Use
1443         attribute_list_equal instead.
1444         (sem_function::equals_wpa): Likewise.
1445         * ipa-icf.h: Remove compare_attributes.
1447 2018-10-22  Richard Biener  <rguenther@suse.de>
1449         PR middle-end/87682
1450         * mem-stats.h (mem_usage::operator==): Fix pasto.
1452 2018-10-22  Richard Biener  <rguenther@suse.de>
1454         PR tree-optimization/87640
1455         * tree-vrp.c (set_value_range_with_overflow): Decompose
1456         incomplete result.
1457         (extract_range_from_binary_expr_1): Adjust.
1459 2018-10-22  Martin Jambor  <mjambor@suse.cz>
1461         * tree-eh.h (stmt_could_throw_p): Add function parameter.
1462         (stmt_can_throw_external): Likewise.
1463         (stmt_can_throw_internal): Likewise.
1464         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
1465         (lower_eh_constructs_2): Likewise.
1466         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
1467         (stmt_can_throw_external): Likewise.
1468         (stmt_can_throw_internal): Likewise.
1469         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
1470         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
1471         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
1472         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
1473         (pass_lower_eh_dispatch::execute): Pass cfun to
1474         stmt_can_throw_external.
1475         (cleanup_empty_eh): Likewise.
1476         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
1477         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
1478         stmt_can_throw_external instead of pushing it to cfun.
1479         (symbol_table::create_edge): Likewise.
1480         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
1481         stmt_can_throw_internal.
1482         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
1483         to stmt_could_throw_p.
1484         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
1485         stmt_can_throw_internal.
1486         (pass_store_merging::execute): Likewise.
1487         * gimple-ssa-strength-reduction.c
1488         (find_candidates_dom_walker::before_dom_children): Pass cfun to
1489         stmt_could_throw_p.
1490         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
1491         stmt_can_throw_internal.
1492         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1493         to stmt_can_throw_external.
1494         (check_stmt): Pass cfun to stmt_could_throw_p.
1495         (check_stmt): Pass cfun to stmt_can_throw_external.
1496         (pass_nothrow::execute): Likewise.
1497         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1498         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1499         stmt_can_throw_internal.
1500         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1501         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1502         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1503         * tree-complex.c (expand_complex_libcall): Pass cfun to
1504         stmt_could_throw_p and to stmt_can_throw_internal.
1505         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1506         * tree-inline.c (copy_edges_for_bb): Likewise.
1507         (maybe_move_debug_stmts_to_successors): Likewise.
1508         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1509         stmt_could_throw_p.
1510         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1511         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1512         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1513         stmt_can_throw_internal.
1514         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1515         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1516         stmt_could_throw_p.
1517         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1518         stmt_can_throw_internal.
1519         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1520         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1521         stmt_could_throw_p.
1522         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1523         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1524         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1525         (convert_mult_to_fma_1): Likewise.
1526         (convert_to_divmod): Likewise.
1527         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1528         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1529         * tree-ssa-propagate.c
1530         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1531         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1532         (maybe_optimize_range_tests): Likewise.
1533         (linearize_expr_tree): Likewise.
1534         (reassociate_bb): Likewise.
1535         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1536         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1537         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1538         (handle_char_store): Likewise.
1539         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1540         stmt_can_throw_internal.
1541         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1542         stmt_could_throw_p.
1543         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1544         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1545         (vectorizable_simd_clone_call): Likewise.
1546         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1547         (gimple_stringop_fixed_value): Likewise.
1549 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
1551         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1552         literal pool references.
1553         (s390_check_qrst_address): Adapt to the new behavior of
1554         s390_loadrelative_operand_p ().
1556 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1558         PR target/72782
1559         * config/i386/sse.md (*andnot<mode>3_bcst): New.
1561 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1563         PR target/72782
1564         * config/i386/sse.md (*<code><mode>3_bcst): New.
1566 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1568         PR target/72782
1569         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1570         V4DI, V16SI and V8DI.
1571         (*sub<mode>3<mask_name>_bcst): New.
1572         (*add<mode>3<mask_name>_bcst): Likewise.
1574 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
1575             Jinsong Ji  <jji@us.ibm.com>
1577         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1578         __m64 with __vector unsigned long long for compatibility.
1579         (_mm_movemask_epi8): Likewise.
1580         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1581         (_mm_cvttps_pi32): Likewise.
1582         (_mm_cvtpi32_ps): Likewise.
1583         (_mm_cvtps_pi16): Likewise.
1584         (_mm_loadh_pi): Likewise.
1585         (_mm_storeh_pi): Likewise.
1586         (_mm_movehl_ps): Likewise.
1587         (_mm_movelh_ps): Likewise.
1588         (_mm_loadl_pi): Likewise.
1589         (_mm_storel_pi): Likewise.
1590         (_mm_movemask_ps): Likewise.
1591         (_mm_shuffle_pi16): Likewise.
1593 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1595         PR target/72782
1596         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1597         __builtin_ia32_vfnmsubpd512_mask.
1598         (_mm512_mask_fnmsub_round_pd): Likewise.
1599         (_mm512_fnmsub_pd): Likewise.
1600         (_mm512_mask_fnmsub_pd): Likewise.
1601         (_mm512_maskz_fnmsub_round_pd): Use
1602         __builtin_ia32_vfnmsubpd512_maskz.
1603         (_mm512_maskz_fnmsub_pd): Likewise.
1604         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1605         (_mm512_mask_fnmsub_round_ps): Likewise.
1606         (_mm512_fnmsub_ps): Likewise.
1607         (_mm512_mask_fnmsub_ps): Likewise.
1608         (_mm512_maskz_fnmsub_round_ps): Use
1609         __builtin_ia32_vfnmsubps512_maskz.
1610         (_mm512_maskz_fnmsub_ps): Likewise.
1611         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1612         __builtin_ia32_vfnmsubpd256_mask.
1613         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1614         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1615         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1616         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1617         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1618         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1619         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1620         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1621         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1622         __builtin_ia32_vfnmsubpd.
1623         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1624         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1625         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1626         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1627         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1628         * config/i386/i386-builtin.def: Add
1629         __builtin_ia32_vfnmsubpd256_mask,
1630         __builtin_ia32_vfnmsubpd256_maskz,
1631         __builtin_ia32_vfnmsubpd128_mask,
1632         __builtin_ia32_vfnmsubpd128_maskz,
1633         __builtin_ia32_vfnmsubps256_mask,
1634         __builtin_ia32_vfnmsubps256_maskz,
1635         __builtin_ia32_vfnmsubps128_mask,
1636         __builtin_ia32_vfnmsubps128_maskz,
1637         __builtin_ia32_vfnmsubpd512_mask,
1638         __builtin_ia32_vfnmsubpd512_maskz,
1639         __builtin_ia32_vfnmsubps512_mask,
1640         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1641         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1642         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1643         __builtin_ia32_vfnmsubpd256.
1644         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1645         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1646         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1647         Likewise.
1648         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1649         Likewise.
1650         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1651         Likewise.
1652         (fmai_vmfnmsub_<mode><round_name>): Likewise.
1654 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1656         PR target/72782
1657         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1658         __builtin_ia32_vfnmaddpd512_mask.
1659         (_mm512_mask_fnmadd_round_pd): Likewise.
1660         (_mm512_fnmadd_pd): Likewise.
1661         (_mm512_mask_fnmadd_pd): Likewise.
1662         (_mm512_maskz_fnmadd_round_pd): Use
1663         __builtin_ia32_vfnmaddpd512_maskz.
1664         (_mm512_maskz_fnmadd_pd): Likewise.
1665         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1666         (_mm512_mask_fnmadd_round_ps): Likewise.
1667         (_mm512_fnmadd_ps): Likewise.
1668         (_mm512_mask_fnmadd_ps): Likewise.
1669         (_mm512_maskz_fnmadd_round_ps): Use
1670         __builtin_ia32_vfnmaddps512_maskz.
1671         (_mm512_maskz_fnmadd_ps): Likewise.
1672         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1673         __builtin_ia32_vfnmaddpd256_mask.
1674         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1675         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1676         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1677         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1678         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1679         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1680         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1681         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1682         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1683         __builtin_ia32_vfnmaddpd.
1684         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1685         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1686         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1687         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1688         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1689         * config/i386/i386-builtin.def: Add
1690         __builtin_ia32_vfnmaddpd256_mask,
1691         __builtin_ia32_vfnmaddpd256_maskz,
1692         __builtin_ia32_vfnmaddpd128_mask,
1693         __builtin_ia32_vfnmaddpd128_maskz,
1694         __builtin_ia32_vfnmaddps256_mask,
1695         __builtin_ia32_vfnmaddps256_maskz,
1696         __builtin_ia32_vfnmaddps128_mask,
1697         __builtin_ia32_vfnmaddps128_maskz,
1698         __builtin_ia32_vfnmaddpd512_mask,
1699         __builtin_ia32_vfnmaddpd512_maskz,
1700         __builtin_ia32_vfnmaddps512_mask,
1701         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1702         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1703         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1704         __builtin_ia32_vfnmaddpd256.
1705         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1706         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1707         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1708         Likewise.
1709         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1710         Likewise.
1711         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1712         Likewise.
1713         (fmai_vmfnmadd_<mode><round_name>): Likewise.
1715 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1717         PR target/72782
1718         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1719         __builtin_ia32_vfmsubpd512_mask.
1720         (_mm512_mask_fmsub_round_pd): Likewise.
1721         (_mm512_fmsub_pd): Likewise.
1722         (_mm512_mask_fmsub_pd): Likewise.
1723         (_mm512_maskz_fmsub_round_pd): Use
1724         __builtin_ia32_vfmsubpd512_maskz.
1725         (_mm512_maskz_fmsub_pd): Likewise.
1726         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1727         (_mm512_mask_fmsub_round_ps): Likewise.
1728         (_mm512_fmsub_ps): Likewise.
1729         (_mm512_mask_fmsub_ps): Likewise.
1730         (_mm512_maskz_fmsub_round_ps): Use
1731         __builtin_ia32_vfmsubps512_maskz.
1732         (_mm512_maskz_fmsub_ps): Likewise.
1733         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1734         __builtin_ia32_vfmsubpd256_mask.
1735         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1736         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1737         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1738         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1739         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1740         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1741         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1742         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1743         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1744         __builtin_ia32_vfmsubpd.
1745         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1746         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1747         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1748         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1749         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1750         * config/i386/i386-builtin.def: Add
1751         __builtin_ia32_vfmsubpd256_mask,
1752         __builtin_ia32_vfmsubpd256_maskz,
1753         __builtin_ia32_vfmsubpd128_mask,
1754         __builtin_ia32_vfmsubpd128_maskz,
1755         __builtin_ia32_vfmsubps256_mask,
1756         __builtin_ia32_vfmsubps256_maskz,
1757         __builtin_ia32_vfmsubps128_mask,
1758         __builtin_ia32_vfmsubps128_maskz,
1759         __builtin_ia32_vfmsubpd512_mask,
1760         __builtin_ia32_vfmsubpd512_maskz,
1761         __builtin_ia32_vfmsubps512_mask,
1762         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1763         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1764         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1765         __builtin_ia32_vfmsubpd256.
1766         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1767         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1768         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1769         Likewise.
1770         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1771         Likewise.
1772         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1773         Likewise.
1774         (fmai_vmfmsub_<mode><round_name>): Likewise.
1776 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1778         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1779         Remove plus.  Renamed to ...
1780         (*sub<mode>3<mask_name>_bcst): This.
1781         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1782         (*add<mode>3<mask_name>_bcst): This.
1784 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1786         PR target/72782
1787         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1789 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1791         PR target/87662
1792         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1793         (_mm_or_epi32): Likewise.
1794         (_mm256_xor_epi32): Likewise.
1795         (_mm_xor_epi32): Likewise.
1796         (_mm256_or_epi64): Likewise.
1797         (_mm_or_epi64): Likewise.
1798         (_mm256_xor_epi64): Likewise.
1799         (_mm_xor_epi64): Likewise.
1801 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
1803         PR target/72782
1804         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1806 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
1808         PR middle-end/87647
1809         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1811 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
1813         * doc/ux.texi: Move @section directly after @node.
1815 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
1817         PR middle-end/85488
1818         PR middle-end/87649
1819         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1820         depend closely nested inside of loop with ordered clause with
1821         a parameter.
1823 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
1825         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1826         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1827         * doc/ux.texi: New file.
1829 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
1831         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1832         be the first CR field allocated.
1834 2018-10-19  Richard Biener  <rguenther@suse.de>
1836         PR target/87657
1837         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1838         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1840 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1842         PR target/72782
1843         * config/i386/sse.md
1844         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1845         (*add<mode>3<mask_name>_bcst_2): Likewise.
1847 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1849         * config/i386/sse.md
1850         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1851         Replace nonimmediate_operand with register_operand.
1852         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1853         Likewise.
1854         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1855         Likewise.
1857 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
1859         PR rtl-optimization/87596
1860         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1861         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1862         for instructions in FROM..TO range.
1864 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
1866         * cfgexpand.c (expand_one_var): Use specific wording in error message
1867         for non-local frame variables.
1868         * stor-layout.c (layout_decl): Do not issue a warning for them.
1870 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
1872         * haifa-sched.c (priority): Add force_recompute parameter.
1873         (apply_replacement): Call priority () with force_recompute = true.
1874         (restore_pattern): Likewise.
1876 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1878         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
1879         HOST_BITS_PER_WIDE_INT.
1880         (test_vector_ops_duplicate): Likewise.
1882 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1884         PR target/72782
1885         * config/i386/sse.md (VF_AVX512): New.
1886         (avx512bcst): Likewise.
1887         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1888         Likewise.
1889         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1890         Likewise.
1891         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1892         Likewise.
1894 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
1896         * doc/invoke.texi (-dumpversion): Improve grammar.
1897         (-dumpfullversion): Make more consistent with -dumpversion.
1899 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
1901         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
1902         Set JUMP_LABEL to the jump insn.
1903         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
1904         Predict emitted jump and add label to jump insn.
1906 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
1908         PR tree-optimization/87562
1909         * input.c (get_substring_ranges_for_loc): Use
1910         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
1911         getting the linemap for the endpoint.  Verify that it's either
1912         in the same linemap as the start point's spelling location, or
1913         at least in the same file.
1915 2018-10-18  Richard Biener  <rguenther@suse.de>
1917         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
1918         feed width-specific load/store costs through ix86_vec_cost.
1919         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
1920         (k8_cost): Likewise.
1921         (bdver_cost): Likewise.
1922         (znver1_cost): Likewise.
1923         (btver1_cost): Likewise.
1924         (btver2_cost): Likewise.
1926 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1928         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
1929         to simplify subreg of vec_merge.
1931 2018-10-18  Richard Biener  <rguenther@suse.de>
1933         * config/i386/i386.c: Fix costing of vector FMA.
1935 2018-10-18  Richard Biener  <rguenther@suse.de>
1937         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
1938         and argument.
1939         (ix86_builtin_vectorization_cost): For vec_construct properly
1940         cost insertion into SSE regs.
1941         (...): Adjust calls to ix86_vec_cost.
1943 2018-10-18  Richard Biener  <rguenther@suse.de>
1945         PR middle-end/87087
1946         Revert
1947         2018-02-07  Richard Biener  <rguenther@suse.de>
1949         PR tree-optimization/84204
1950         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1951         this place.
1953 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1955         PR target/87537
1956         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
1957         of vec_duplicate.
1958         (test_vector_ops_duplicate): Add test for a scalar subreg of a
1959         VEC_MERGE of a VEC_DUPLICATE.
1961 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1963         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
1964         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
1965         * doc/standards.texi (C Language): Document C2X.
1966         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
1967         (rl78_option_override): Handle "GNU C2X" language name.
1969 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1971         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
1972         Document C17 as published in 2018.
1974 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
1976         PR middle-end/87623
1977         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
1978         bail out if both sides do not have the same storage order.
1980 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1982         * bitmap.c (bitmap_head::dump): New.
1983         * bitmap.h (bitmap_head): Add dump().
1984         * gimple-ssa-evrp-analyze.c
1985         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
1986         (evrp_range_analyzer::set_ssa_range_info): Same.
1987         (evrp_range_analyzer::record_ranges_from_phis): Same.
1988         (evrp_range_analyzer::record_ranges_from_stmt): Same.
1989         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
1990         * gimple-ssa-sprintf.c (get_int_range): Same.
1991         (format_integer): Same.
1992         (sprintf_dom_walker::handle_gimple_call): Same.
1993         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
1994         (ipcp_vr_lattice::top_p): Same.
1995         (ipcp_vr_lattice::bottom_p): Same.
1996         (ipcp_vr_lattice::set_to_bottom): Same.
1997         (ipa_vr_operation_and_type_effects): Same.
1998         (propagate_vr_across_jump_function): Same.
1999         (ipcp_store_vr_results): Same.
2000         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
2001         (ipa_print_node_jump_functions_for_edge): Same.
2002         (ipa_get_value_range): Same.
2003         (ipa_compute_jump_functions_for_edge): Same.
2004         (ipa_write_jump_function): Same.
2005         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
2006         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2007         Same.
2008         * vr-values.c (set_value_range_to_nonnegative): Same.
2009         (set_value_range_to_truthvalue): Same.
2010         (vr_values::get_value_range): Same.
2011         (vr_values::set_defs_to_varying): Same.
2012         (vr_values::update_value_range): Same.
2013         (symbolic_range_based_on_p): Same.
2014         (vr_values::op_with_boolean_value_range_p): Same.
2015         (vr_values::extract_range_for_var_from_comparison_expr): Same.
2016         (vr_values::extract_range_from_ssa_name): Same.
2017         (vr_values::extract_range_from_binary_expr): Same.
2018         (vr_values::extract_range_from_unary_expr): Same.
2019         (vr_values::extract_range_from_cond_expr): Same.
2020         (vr_values::extract_range_from_comparison): Same.
2021         (vr_values::check_for_binary_op_overflow): Same.
2022         (vr_values::extract_range_basic): Same.
2023         (vr_values::extract_range_from_assignment): Same.
2024         (compare_ranges): Same.
2025         (compare_range_with_value): Same.
2026         (vr_values::adjust_range_with_scev): Same.
2027         (vrp_valueize): Same.
2028         (vrp_valueize_1): Same.
2029         (vr_values::get_vr_for_comparison): Same.
2030         (vr_values::compare_name_with_value): Same.
2031         (vr_values::compare_names): Same.
2032         (vr_values::vrp_evaluate_conditional): Same.
2033         (find_case_label_ranges): Same.
2034         (vr_values::vrp_visit_switch_stmt): Same.
2035         (vr_values::extract_range_from_phi_node): Same.
2036         (vr_values::simplify_div_or_mod_using_ranges): Same.
2037         (vr_values::simplify_bit_ops_using_ranges): Same.
2038         (test_for_singularity): Same.
2039         (range_fits_type_p): Same.
2040         (vr_values::simplify_cond_using_ranges_1): Same.
2041         (vr_values::simplify_switch_using_ranges): Same.
2042         (vr_values::simplify_float_conversion_using_ranges): Same.
2043         (vr_values::two_valued_val_range_p): Same.
2044         (vr_values::add_equivalence): Move to value_range::equiv_add.
2045         * vr-values.h (vr_values::add_equivalence): Remove.
2046         (VR_INITIALIZER): Remove.
2047         * tree-vrp.c (value_range::set): New.
2048         (value_range::equiv_add): New.
2049         (value_range::value_range): New.
2050         (value_range::deep_copy): New.
2051         (value_range::check): New.
2052         (value_range::equal_p): New.
2053         (value_range::ignore_equivs_equal_p): New.
2054         (value_range::operator==): New.
2055         (value_range::operator!=): New.
2056         (value_range::symbolic_p): New.
2057         (value_range::numeric_p): New.
2058         (value_range::set_undefined): New.
2059         (value_range::set_varying): New.
2060         (value_range::may_contain_p): New.
2061         (value_range::equiv_clear): New.
2062         (value_range::singleton_p): New.
2063         (value_range::intersect): New.
2064         (value_range::dump): New.
2065         (value_range::set_and_canonicalize): New.
2066         (set_value_range): Adjust for value_range API.
2067         (set_value_range_to_undefined): Same.
2068         (set_value_range_to_varying): Same.
2069         (set_and_canonicalize_value_range): Same.
2070         (set_value_range_to_nonnull): Same.
2071         (set_value_range_to_null): Same.
2072         (range_is_null): Same.
2073         (range_is_nonnull): Same.
2074         (range_int_cst_p): Same.
2075         (range_int_cst_singleton_p): Same.
2076         (symbolic_range_p): Same.
2077         (range_includes_zero_p): Same.
2078         (value_range_constant_singleton): Same.
2079         (vrp_set_zero_nonzero_bits): Same.
2080         (ranges_from_anti_range): Same.
2081         (extract_range_into_wide_ints): Same.
2082         (extract_range_from_multiplicative_op): Same.
2083         (set_value_range_with_overflow): Same.
2084         (extract_range_from_binary_expr_1): Same.
2085         (extract_range_from_unary_expr): Same.
2086         (dump_value_range): Same.
2087         (debug_value_range): Same.
2088         (vrp_prop::check_array_ref): Same.
2089         (vrp_prop::check_mem_ref): Same.
2090         (vrp_prop::vrp_initialize): Same.
2091         (vrp_prop::visit_stmt): Same.
2092         (intersect_ranges): Same.
2093         (vrp_prop::visit_phi): Same.
2094         (vrp_prop::vrp_finalize): Same.
2095         (determine_value_range_1): Same.
2096         (determine_value_range): Same.
2097         (vrp_intersect_ranges_1): Rename to...
2098         (vrp_intersect_1): this.
2099         (vrp_intersect_ranges): Rename to...
2100         (value_range::intersect_helper): ...this.
2101         (vrp_meet_1): Rename to...
2102         (value_range::union_helper): ...this.
2103         (vrp_meet): Rename to...
2104         (value_range::union_): ...this.
2105         (copy_value_range): Remove.
2106         * tree-vrp.h (struct value_range): Rewrite into a proper class.
2107         (value_range::vrtype): New.
2108         (value_range::type): New.
2109         (value_range::equiv): New.
2110         (value_range::min): New.
2111         (value_range::max): New.
2112         (value_range::varying_p): New.
2113         (value_range::undefined_p): New.
2114         (value_range::null_p): New.
2115         (value_range::equiv_add): New.
2116         (copy_value_range): Remove.
2118 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
2120         * Makefile.in (SELFTEST_TARGETS): New.
2121         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
2122         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
2123         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
2124         c/Make-lang.in.
2125         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
2126         (selftest-c++-gdb, selftest-c++-valgrind): Move to
2127         cp/Make-lang.in.
2128         * configure: Regenerate.
2129         * configure.ac (selftest_languages): New.
2131 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2133         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
2134         overflow wraps argument.
2135         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
2136         wide_int_range_multiplicative_op.
2137         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
2138         overflow wraps argument.
2139         (wide_int_range_multiplicative_op): Same.
2140         (wide_int_range_lshift): Same.
2141         (wide_int_range_div): Same.
2142         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
2143         (wide_int_range_lshift): Same.
2144         (wide_int_range_div): Same.
2146 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2148         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
2149         use sign as argument.
2150         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
2151         wide_int_range_shift_undefined_p.
2153 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2155         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
2156         Rename to...
2157         (@despeculate_copy<ALLI_TI:mode>): ... This.
2158         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
2159         switch statement.
2161 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2163         * config.gcc: Obsolete *-*-solaris2.10*.
2164         * doc/install.texi (Specific, *-*-solaris2*): Document it.
2166 2018-10-12  Jeff Law  <law@redhat.com>
2168         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
2169         reg + sym +- const_int addressing modes.
2171 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2173         * common.opt (fdiagnostics-minimum-margin-width=): New option.
2174         * diagnostic-show-locus.c (layout::layout): Apply the minimum
2175         margin width.
2176         (layout::start_annotation_line): Only print up to 3 of the
2177         margin character, to avoid touching the left-hand side.
2178         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
2179         minimum margin width, as set by test_diagnostic_context's ctor.
2180         (selftest::test_fixit_insert_containing_newline): Likewise.
2181         (selftest::test_fixit_insert_containing_newline_2): Likewise.
2182         (selftest::test_line_numbers_multiline_range): Clear
2183         dc.min_margin_width.
2184         * diagnostic.c (diagnostic_initialize): Initialize
2185         min_margin_width.
2186         * diagnostic.h (struct diagnostic_context): Add field
2187         "min_margin_width".
2188         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
2189         * opts.c (common_handle_option): Handle
2190         OPT_fdiagnostics_minimum_margin_width_.
2191         * selftest-diagnostic.c
2192         (selftest::test_diagnostic_context::test_diagnostic_context):
2193         Initialize min_margin_width to 6.
2194         * toplev.c (general_init): Initialize global_dc->min_margin_width.
2196 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2198         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2199         Fix usage of "error_at_rich_loc" in the comment.
2201 2018-10-15  Renlin Li  <renlin.li@arm.com>
2203         PR target/87563
2204         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
2205         if-conversioned loop when it contains ifn with types not
2206         supported by backend.
2207         * internal-fn.c (expand_direct_optab_fn): Add an assert.
2208         (direct_internal_fn_supported_p): New helper function.
2209         * internal-fn.h (direct_internal_fn_supported_p): Declare.
2211 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
2213         PR target/87572
2214         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2215         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2216         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2218 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
2220         PR tree-optimization/87022
2221         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
2222         bits in dist vector rather than the first one.
2224 2018-10-15  Richard Biener  <rguenther@suse.de>
2226         PR middle-end/87610
2227         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2228         (visit_loadstore): When a used restrict tag escaped verify that
2229         the points-to solution of "other" pointers do not include
2230         escaped.
2231         (compute_dependence_clique): If a used restrict tag escaped
2232         communicated that down to visit_loadstore.
2234 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
2236         * config/s390/s390.c (s390_expand_vec_init): Force vector element
2237         into reg if it isn't a general operand.
2239 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2241         PR target/87599
2242         * config/i386/sse.md (*vec_dupv2di): Add register source to
2243         movddup.
2245 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2247         PR target/87572
2248         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2249         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
2250         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2251         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2253 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
2255         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
2256         (notice_args_size): Set it in the current trace if no insn that can
2257         throw internally has been seen yet.
2258         (connect_traces): When connecting args_size between traces, allow the
2259         incoming values not to match if there is an insn setting it before the
2260         first insn that can throw internally; in that case, force the creation
2261         of a CFI note on this latter insn.
2263 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
2265         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
2266         the base class, because using the injected-class-name was not clearly
2267         specified until DR 176.
2269 2018-10-12  Paul Koning  <ni1d@arrl.net>
2271         * config/pdp11/pdp11.md (doloop_end): New expander.
2272         (doloop_end_insn): renamed from "doloop_end".
2273         (addqi3): New pattern.
2274         (subqi3): New pattern.
2275         * config/pdp11/predicates.md (incdec_operand): New predicate.
2277 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
2279         PR middle-end/81376
2280         * real.c (format_helper::can_represent_integral_type_p): New function
2281         * real.h (format_helper::can_represent_integral_type_p): Ditto.
2282         * match.pd: New pattern.
2284 2018-10-12  Alexandre Oliva <oliva@adacore.com>
2286         * configure.ac: Introduce --enable-large-address-aware
2287         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
2288         * doc/install.texi: Document it.
2289         * configure, config.in: Rebuilt.
2290         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
2291         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
2292         (LINK_SPEC): Insert it.
2293         * config/i386/mingw-w64.h: Likewise.
2295         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
2297 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
2299         PR rtl-optimization/87600
2300         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
2302 2018-10-12  Paul Koning  <ni1d@arrl.net>
2304         * doc/md.texi (doloop_end): Document that the pattern code may
2305         need to check operand mode.
2307 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
2309         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
2310         to zero-extend between int and floating-point registers.
2311         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
2312         ldp into floating-point registers.  Add type and arch attributes.
2313         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
2314         Use f_loads for type attribute.
2316 2018-10-11  Martin Sebor  <msebor@redhat.com>
2318         * doc/extend.texi (attribute packed): Correct typos.
2320 2018-10-11  Martin Sebor  <msebor@redhat.com>
2322         * doc/extend.texi (attribute flatten): Mention interaction with
2323         noinline.
2325 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
2327         PR target/87156
2328         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2329         Set new_decl virtual flag to zero.
2331 2018-10-11  Martin Sebor  <msebor@redhat.com>
2333         PR middle-end/87593
2334         * doc/extend.texi (attribute format_arg): Discuss using multiple
2335         attributes on a single function.
2337 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
2339         PR tree-optimization/86829
2340         * match.pd (sin (atan (x))): New simplification rules.
2341         (cos (atan (x))): Likewise.
2342         * real.c (build_sinatan_real): New function.
2343         * real.h (build_sinatan_real): Prototype.
2345 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
2347         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
2348         function.
2349         (fold_mergeeo_helper): New helper function.
2350         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
2351         intrinsics.  Correct some whitespace indentation issues.
2353 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
2355         PR target/87511
2356         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2357         Use HOST_WIDE_INT_1U for shift.
2359 2018-10-11  Doug Rupp  <rupp@adacore.com>
2360             Olivier Hainque  <hainque@adacore.com>
2362         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
2363         Pass --relax to the linker for RTPs.
2364         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
2366 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
2367             Jan Hubicka  <jh@suse.cz>
2368             Martin Jambor  <mjambor@suse.cz>
2370         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
2371         the same elements are repeated rather than printing all of them.
2372         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
2373         "repeated" elements.
2374         * read-rtl-function.c (test_loading_repeat): New function.
2375         (read_rtl_function_c_tests): Call test_loading_repeat.
2376         * rtl-tests.c (test_dumping_repeat): New function.
2377         (rtl_tests_c_tests): Call test_dumping_repeat.
2379 2018-10-11  Richard Biener  <rguenther@suse.de>
2381         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
2382         bdver?_cost): Unify to ...
2383         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
2384         * config/i386/i386.c (processor_cost_table): Adjust.
2386 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
2388         PR middle-end/87574
2389         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
2390         the thunk when expanding to GIMPLE.
2392 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2394         * varasm.c (mergeable_string_section): Don't try to move zero-length
2395         strings to the merge section.
2397 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
2399         PR target/87573
2400         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
2402 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
2404         PR target/87550
2405         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2406         to special_args set.
2408 2018-10-10  Richard Biener  <rguenther@suse.de>
2410         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
2411         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
2412         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
2413         and recursing and pattern terminating the recursion on SSE
2414         vector width using ix86_expand_reduc.
2415         (reduc_sminmax_scal_<mode>): Split into part reducing to half
2416         width and recursing and SSE2 vector variant doing the final
2417         reduction with ix86_expand_reduc.
2418         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
2419         with terminating the recursion at AVX level, splitting that
2420         to SSE there.
2422 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
2424         * genmatch.c (error_cb): Rename to...
2425         (diagnostic_cb): ...this, converting int params to enums.
2426         (fatal_at): Update for renaming.
2427         (warning_at): Likewise.
2428         (main): Likewise.
2429         * input.c (selftest::ebcdic_execution_charset::apply):
2430         Update for renaming of...
2431         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
2432         to...
2433         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
2434         converting level and reason to enums.
2435         (class selftest::lexer_error_sink): Rename to...
2436         (class selftest::lexer_test_options): ...this, renaming field
2437         "m_errors" to "m_diagnostics".
2438         (selftest::lexer_test_options::apply): Update for renaming of...
2439         (selftest::lexer_test_options::on_error): ...this, renaming to...
2440         (selftest::lexer_test_options::on_diagnostic): ...this
2441         converting level and reason to enums.
2442         (selftest::test_lexer_string_locations_raw_string_unterminated):
2443         Update for renamings.
2444         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
2445         "reason".
2447 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
2449         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
2450         * config/rs6000/pmmintrin.h: New file.
2452 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
2454         PR tree-optimization/86659
2455         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
2457 2018-10-09  Richard Biener  <rguenther@suse.de>
2459         PR tree-optimization/63155
2460         * tree-ssa-structalias.c: Include tree-ssa.h.
2461         (get_constraint_for_ssa_var): For undefs return nothing_id.
2462         (find_func_aliases): Cleanup PHI handling.
2464 2018-10-09  Richard Biener  <rguenther@suse.de>
2466         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
2467         replacements.
2469 2018-10-09  Martin Liska  <mliska@suse.cz>
2471         * asan.c (asan_emit_stack_protection): If a stack variable
2472         is located in a same file as current function, then emit
2473         line info into variable definition string.
2475 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2477         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
2478         information.
2480 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2482         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
2483         on the thunk.
2485 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2487         PR tree-optimization/86659
2488         * gimple-match.h (struct gimple_match_op): Add reverse field.
2489         (gimple_match_op::set_op): New overloaded method.
2490         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
2491         the REF_REVERSE_STORAGE_ORDER flag on the value.
2492         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2493         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2494         
2495 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
2497         PR middle-end/63155
2498         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2499         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2501 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
2503         PR target/87517
2504         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2505         Defined with __builtin_ia32_vfmaddsubpd512_mask.
2507 2018-10-08  Richard Biener  <rguenther@suse.de>
2509         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2510         cost the same as AVX128 ones.
2512 2018-10-08  Paul Koning  <ni1d@arrl.net>
2514         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2515         (expand_block_move): New function.
2516         * config/pdp11/pdp11.c (output_block_move): Remove.
2517         (expand_block_move): New function.
2518         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2519         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2520         (*movmemhi1): Remove.
2522 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
2524         * config/s390/2827.md: Increase latencies for some FP instructions.
2526 2018-10-08  Richard Biener  <rguenther@suse.de>
2528         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2529         Open a dump scope.
2530         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2531         * tree-vectorizer.h (dump_stmt_cost): Adjust.
2532         (add_stmt_cost): Dump return value of the hook.
2534 2018-10-08  Richard Biener  <rguenther@suse.de>
2536         PR tree-optimization/63155
2537         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2538         (ssa_propagation_engine::ssa_propagate): Remove redundant
2539         bitmap bit clearing.
2541 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
2543         PR rtl-optimization/86939
2544         PR rtl-optimization/87479
2545         * ira.h (non_conflicting_reg_copy_p): New prototype.
2546         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2547         (make_hard_regno_dead): Don't add conflicts for register
2548         ignore_reg_for_conflicts.
2549         (make_object_dead): Likewise.
2550         (non_conflicting_reg_copy_p): New function.
2551         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2552         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2553         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2554         (make_hard_regno_dead): Don't add conflicts for register
2555         ignore_reg_for_conflicts.  Remove special conflict handling of
2556         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
2557         check_pic_pseudo_p and update callers.
2558         (mark_pseudo_dead): Don't add conflicts for register
2559         ignore_reg_for_conflicts.
2560         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2562 2018-10-05  Andrew Waterman  <andrew@sifive.com>
2563             Jim Wilson  <jimw@sifive.com>
2565         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2566         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
2567         new pattern using HONOR_SNANS that emits one extra instruction.
2569 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2571         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2572         patterns): Merge SI and DI patterns to a GPR pattern.
2573         (unnamed define_insn and define_split for record form of that): Merge
2574         to a single define_insn_and_split pattern.
2576 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
2578         PR c++/56856
2579         * input.c (expand_location_to_spelling_point): Add param "aspect"
2580         and use rather than hardcoding LOCATION_ASPECT_CARET.
2581         (get_substring_ranges_for_loc): Handle the case of a single token
2582         within a macro expansion.
2583         * input.h (expand_location_to_spelling_point): Add "aspect" param,
2584         defaulting to LOCATION_ASPECT_CARET.
2586 2018-10-05  Paul Koning  <ni1d@arrl.net>
2588         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2589         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2590         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2591         (pdp11_guard_type): New function.
2593 2018-10-05  Paul Koning  <ni1d@arrl.net>
2595         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2596         * config/pdp11/pdp11.opt (mfloat32): Remove.
2597         (mfloat64): Remove.
2598         * doc/invoke.texi (pdp11 -mfloat32): Remove:
2599         (pdp11 -mfloat64): Remove.
2601 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2603         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2604         (*cmp<mode>_cc_i387): Ditto.
2605         (*cmpu<mode>_cc_i387): Ditto.
2606         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2607         * config/i386/i386.c (ix86_expand_fp_compare): Remove
2608         "scratch" argument.
2609         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2610         Emit x86_sahf_1 pattern.
2611         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2612         (ix86_expand_carry_flag_compare): Ditto.
2614 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2616         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2617         to reg_or_0_operand.  Add "C" constraint.
2618         (*cmpxf_cc_i387): Ditto.
2619         (*cmp<mode>_i387): Change operand 2 predicate
2620         to nonimm_or_0_operand.  Add "C" constraint.
2621         (*cmp<mode>_cc_i387): Ditto.
2622         (*cmp<mode>_0_i387): Remove insn pattern.
2623         (*cmp<mode>_0_cc_i387): Ditto.
2625 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2627         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2628         * config/i386/predicates.md (nonimm_or_0_operand): Rename
2629         from vector_move_operand.  Update all uses.
2631 2018-10-05  Martin Sebor  <msebor@redhat.com>
2633         PR tree-optimization/87490
2634         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2635         consistently.
2637 2018-10-05  Richard Biener  <rguenther@suse.de>
2639         PR tree-optimization/63155
2640         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2641         vertical space in dumpfiles.
2642         * tree-ssa-propagate.h
2643         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2644         * tree-ssa-propagate.c (cfg_blocks_back): New global.
2645         (ssa_edge_worklist_back): Likewise.
2646         (curr_order): Likewise.
2647         (cfg_blocks_get): Remove abstraction.
2648         (cfg_blocks_add): Likewise.
2649         (cfg_blocks_empty_p): Likewise.
2650         (add_ssa_edge): Add to current or next worklist based on
2651         RPO index.
2652         (add_control_edge): Likewise.
2653         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2654         into ...
2655         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
2656         iteration from CFG and SSA edge worklist so we process
2657         everything in RPO order, prioritizing forward progress
2658         over iteration.
2659         (ssa_prop_init): Allocate new worklists, do not dump
2660         immediate uses.
2661         (ssa_prop_fini): Free new worklists.
2663 2018-10-05  Richard Biener  <rguenther@suse.de>
2665         * tree-core.h (tree_block::abstract_flag): Remove.
2666         (tree_block::block_num): Make full 32bits.
2667         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2668         * tree.h (BLOCK_ABSTRACT): Remove.
2669         * dwarf2out.c (gen_lexical_block_die): Remove dead code
2670         resulting from BLOCK_ABSTRACT being always false.
2671         (gen_inlined_subroutine_die): Likewise.
2672         (gen_block_die): Likewise.
2673         * tree.c (block_ultimate_origin): Likewise.
2674         * tree-pretty-print.c (dump_block_node): Remove code dealing
2675         with BLOCK_ABSTRACT.
2676         * tree-ssa-live.c (dump_scope_block): Likewise.
2677         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2678         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2680 2018-10-05   Richard Biener  <rguenther@suse.de>
2682         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2683         is asked for initialize mode to the component mode of the
2684         vector type.
2686 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
2688         PR target/87522
2689         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2690         assembler for -mavx.
2691         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2693 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2695         PR target/87509
2696         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2697         RS6000_BTM_DFP.
2698         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2699         to be DImode.  When using mffscrn, force the operand to a register.
2701 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
2703         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2704         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2705         X87MODEF mode iterator.
2706         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2707         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2708         X87MODEF mode iterator.
2710 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
2712         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2713         -Wno-prio-ctor-dtor.
2715 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2717         * Makefile.in (OBJS): Add opt-problem.o.
2718         * dump-context.h: Include "selftest.h.
2719         (selftest::temp_dump_context): New forward decl.
2720         (class dump_context): Make friend of class
2721         selftest::temp_dump_context.
2722         (dump_context::dump_loc_immediate): New decl.
2723         (class dump_pretty_printer): Move here from dumpfile.c.
2724         (class temp_dump_context): Move to namespace selftest.
2725         (temp_dump_context::temp_dump_context): Add param
2726         "forcibly_enable_dumping".
2727         (selftest::verify_dumped_text):
2728         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2729         (selftest::verify_item):
2730         (ASSERT_IS_TEXT): Move here from dumpfile.c.
2731         (ASSERT_IS_TREE): Likewise.
2732         (ASSERT_IS_GIMPLE): Likewise.
2733         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2734         to...
2735         (dump_context::dump_loc_immediate): ...this new function.
2736         (class dump_pretty_printer): Move to dump-context.h.
2737         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2738         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2739         (temp_dump_context::temp_dump_context): Move to "selftest"
2740         namespace.  Add param "forcibly_enable_dumping", and use it to
2741         conditionalize the use of m_pp;
2742         (selftest::verify_dumped_text): Make non-static.
2743         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2744         (selftest::verify_item): Make non-static.
2745         (ASSERT_IS_TEXT): Move to dump-context.h.
2746         (ASSERT_IS_TREE): Likewise.
2747         (ASSERT_IS_GIMPLE): Likewise.
2748         (selftest::test_capture_of_dump_calls): Pass "true" for new
2749         param of temp_dump_context.
2750         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2751         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
2752         TDF_COMPARE_DEBUG.
2753         * opt-problem.cc: New file.
2754         * opt-problem.h: New file.
2755         * optinfo-emit-json.cc
2756         (selftest::test_building_json_from_dump_calls): Pass "true" for
2757         new param of temp_dump_context.
2758         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2759         (optinfo::emit_for_opt_problem): New function.
2760         (optinfo::emit): Clarity which emit_item is used.
2761         * optinfo.h (optinfo::get_dump_location): New accessor.
2762         (optinfo::emit_for_opt_problem): New decl.
2763         (optinfo::emit): Make const.
2764         * selftest-run-tests.c (selftest::run_tests): Call
2765         selftest::opt_problem_cc_tests.
2766         * selftest.h (selftest::opt_problem_cc_tests): New decl.
2767         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2768         bool to opt_result, converting fprintf messages to
2769         opt_result::failure_at calls.  Add "stmt" param for use by the
2770         failure_at calls.
2771         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2772         (runtime_alias_check_p): Convert return type from bool to
2773         opt_result, converting dump_printf calls to
2774         opt_result::failure_at, using the statement DDR_A for their
2775         location.
2776         (find_data_references_in_stmt): Convert return type from bool to
2777         opt_result, converting "return false" to opt_result::failure_at
2778         with a new message.
2779         * tree-data-ref.h: Include "opt-problem.h".
2780         (dr_analyze_innermost): Convert return type from bool to opt_result,
2781         and add a const gimple * param.
2782         (find_data_references_in_stmt): Convert return type from bool to
2783         opt_result.
2784         (runtime_alias_check_p): Likewise.
2785         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2786         dr_analyze_innermost.
2787         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2788         Convert return type from bool to opt_result, adding a message for
2789         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2790         (vect_analyze_data_ref_dependence): Convert return type from bool
2791         to opt_result.  Change sense of return type from "false"
2792         effectively meaning "no problems" to "false" meaning a problem,
2793         so that "return false" becomes "return opt_result::success".
2794         Convert "return true" calls to opt_result::failure_at, using
2795         the location of statement A rather than vect_location.
2796         (vect_analyze_data_ref_dependences): Convert return type from bool
2797         to opt_result.
2798         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2799         calls to opt_result::failure_at, using the stmt location rather
2800         than vect_location.
2801         (vect_verify_datarefs_alignment): Convert return type from bool
2802         to opt_result.
2803         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
2804         into multiple more-tightly-scoped copies.
2805         (vect_analyze_data_refs_alignment): Convert return type from bool
2806         to opt_result.
2807         (vect_analyze_data_ref_accesses): Likewise, converting a
2808         "return false" to a "return opt_result::failure_at", adding a
2809         new message.
2810         (vect_prune_runtime_alias_test_list): Convert return type from
2811         bool to opt_result, converting dump_printf_loc to
2812         opt_result::failure_at.  Add a %G to show the pertinent statement,
2813         and use the stmt's location rather than vect_location.
2814         (vect_find_stmt_data_reference): Convert return type from
2815         bool to opt_result, converting dump_printf_loc to
2816         opt_result::failure_at, using stmt's location.
2817         (vect_analyze_data_refs):  Convert return type from bool to
2818         opt_result.  Convert "return false" to "return
2819         opt_result::failure_at", adding messages as needed.
2820         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2821         type from bool to opt_result.
2822         (vect_determine_vf_for_stmt): Likewise.
2823         (vect_determine_vectorization_factor): Likewise, converting
2824         dump_printf_loc to opt_result::failure_at, using location of phi
2825         rather than vect_location.
2826         (vect_analyze_loop_form_1): Convert return type from bool to
2827         opt_result, converting dump_printf_loc calls, retaining the use of
2828         vect_location.
2829         (vect_analyze_loop_form): Convert return type from loop_vec_info
2830         to opt_loop_vec_info.
2831         (vect_analyze_loop_operations): Convert return type from bool to
2832         opt_result, converting dump_printf_loc calls, using the location
2833         of phi/stmt rather than vect_location where available.  Convert
2834         various "return false" to "return opt_result::failure_at" with
2835         "unsupported phi" messages.
2836         (vect_get_datarefs_in_loop): Convert return type from bool to
2837         opt_result.  Add a message for the
2838         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2839         (vect_analyze_loop_2): Convert return type from bool to
2840         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
2841         each "goto again;", adding new messages where needed.
2842         Add "unsupported grouped {store|load}" messages.
2843         (vect_analyze_loop): Convert return type from loop_vec_info to
2844         opt_loop_vec_info.
2845         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2846         bool to opt_result.
2847         * tree-vect-stmts.c (process_use): Likewise, converting
2848         dump_printf_loc call and using stmt location, rather than
2849         vect_location.
2850         (vect_mark_stmts_to_be_vectorized): Likeise.
2851         (vect_analyze_stmt): Likewise, adding a %G.
2852         (vect_get_vector_types_for_stmt): Convert return type from bool to
2853         opt_result, converting dump_printf_loc calls and using stmt
2854         location, rather than vect_location.
2855         (vect_get_mask_type_for_stmt): Convert return type from tree to
2856         opt_tree, converting dump_printf_loc calls and using stmt location.
2857         * tree-vectorizer.c: Include "opt-problem.h.
2858         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2859         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
2860         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
2861         enabled, use it to report at the top level "couldn't vectorize
2862         loop" followed by the problem.
2863         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2864         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2865         to opt_result.
2866         (vect_analyze_stmt): Likewise.
2867         (vect_get_vector_types_for_stmt): Likewise.
2868         (tree vect_get_mask_type_for_stmt): Likewise.
2869         (vect_analyze_data_ref_dependences): Likewise.
2870         (vect_enhance_data_refs_alignment): Likewise.
2871         (vect_analyze_data_refs_alignment): Likewise.
2872         (vect_verify_datarefs_alignment): Likewise.
2873         (vect_analyze_data_ref_accesses): Likewise.
2874         (vect_prune_runtime_alias_test_list): Likewise.
2875         (vect_find_stmt_data_reference): Likewise.
2876         (vect_analyze_data_refs): Likewise.
2877         (vect_analyze_loop): Convert return type from loop_vec_info to
2878         opt_loop_vec_info.
2879         (vect_analyze_loop_form): Likewise.
2880         (vect_analyze_slp): Convert return type from bool to opt_result.
2882 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2884         * doc/invoke.texi (-fopt-info): Document new "internals"
2885         sub-option.
2886         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
2887         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
2888         MSG_ALL_KINDS.
2889         (optinfo_verbosity_options): Add "internals".
2890         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
2891         (dump_context::apply_dump_filter_p): New member function.
2892         (dump_context::dump_loc): Use apply_dump_filter_p rather than
2893         explicitly masking the dump_kind.
2894         (dump_context::begin_scope): Increment the scope depth first.  Use
2895         apply_dump_filter_p rather than explicitly masking the dump_kind.
2896         (dump_context::emit_item): Use apply_dump_filter_p rather than
2897         explicitly masking the dump_kind.
2898         (dump_dec): Likewise.
2899         (dump_hex): Likewise.
2900         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
2901         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
2902         (opt_info_switch_p): Update handling of default
2903         MSG_OPTIMIZED_LOCATIONS to cope with default of
2904         MSG_PRIORITY_USER_FACING.
2905         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
2906         masking the dump_kind.
2907         (selftest::test_capture_of_dump_calls): Update test_dump_context
2908         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
2909         than MSG_ALL.  Generalize scope test to be run at all four
2910         combinations of with/without MSG_PRIORITY_USER_FACING and
2911         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
2912         for each of the two values.
2913         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
2914         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
2915         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
2916         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
2917         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
2918         with MSG_PRIORITY_*.
2919         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
2920         dump messages as MSG_PRIORITY_USER_FACING.
2921         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
2922         about the interaction with MSG_PRIORITY_*.
2924 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2926         * varasm.c (output_constant): Add new parameter merge_strings.
2927         Make strings properly zero terminated in merge string sections.
2928         (mergeable_string_section): Don't fail if the last char is non-zero.
2929         (assemble_variable_contents): Handle merge string sections.
2930         (assemble_variable): Likewise.
2931         (assemble_constant_contents): Likewise.
2932         (output_constant_def_contents): Likewise.
2933         (output_constructor_array_range,
2934         output_constructor_regular_field): Adjust call to output_constant.
2935         (output_object_block): Adjust call to assemble_constant_contents
2936         and assemble_variable_contents.
2938 2018-10-04  Martin Liska  <mliska@suse.cz>
2940         PR c/87483
2941         * cgraphunit.c (process_function_and_variable_attributes):
2942         Warn about a function with alias attribute and a body.
2944 2018-10-04  Martin Liska  <mliska@suse.cz>
2946         PR ipa/82625
2947         * multiple_target.c (redirect_to_specific_clone): New function.
2948         (ipa_target_clone): Use it.
2949         * tree-inline.c: Fix comment.
2951 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2953         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
2954         fields.
2955         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
2956         (gcc::dump_manager::register_pass): New member function, adapted
2957         from loop body in gcc::pass_manager::register_pass, adding a
2958         call to update_dfi_for_opt_info.
2959         (gcc::dump_manager::opt_info_enable_passes): Store the
2960         -fopt-info options into the new fields.  Move the loop
2961         bodies into...
2962         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
2963         function.
2964         * dumpfile.h (struct opt_pass): New forward decl.
2965         (gcc::dump_manager::register_pass): New decl.
2966         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
2967         (class gcc::dump_manager): Add fields "m_optgroup_flags",
2968         "m_optinfo_flags", and "m_optinfo_filename".
2969         * passes.c (gcc::pass_manager::register_pass): Move all of the
2970         dump-handling code to gcc::dump_manager::register_pass.
2972 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
2974         PR rtl-optimization/87466
2975         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
2976         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
2977         * doc/tm.texi: Regenerate.
2978         * ira-lives.c (process_bb_node_lives): Use the new target hook.
2979         * lra-lives.c (process_bb_lives): Likewise.
2980         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
2981         Define.
2983 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
2985         * params.c (add_params): Fix initialization.
2987 2018-10-04  Martin Liska  <mliska@suse.cz>
2989         PR gcov-profile/84107
2990         * tree-profile.c (init_ic_make_global_vars):
2991         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
2992         Come up with new ic_tuple* variables.  Emit
2993         __gcov_indirect_call{,_topn} variables.
2994         (gimple_gen_ic_profiler): Access the variable
2995         and emit gimple.
2996         (gimple_gen_ic_func_profiler): Access
2997         __gcov_indirect_call.callee field.
2998         (gimple_init_gcov_profiler): Use ptr_type_node.
2999         * value-prof.c (gimple_ic): Use ptr_type_node.
3001 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3003         PR tree-optimization/85787
3004         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
3005         into this function and add support for detecting multiple phis.
3006         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
3008 2018-10-04  Martin Liska  <mliska@suse.cz>
3010         PR ipa/87491
3011         * ipa-inline.c (inline_to_all_callers_1):
3012         Call ultimate_alias_target for node being inlined.
3014 2018-10-03  Jeff Law  <law@redhat.com>
3016         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
3017         target's wchar_t.
3018         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
3019         * tree.h (get_typenode_from_name): Prototype.
3021 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
3023         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
3024         Change operand 2 predicate to nonimmediate_operand.
3025         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
3027 2018-10-03  Martin Sebor  <msebor@redhat.com>
3028             Jeff Law  <law@redhat.com>
3030         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
3031         initialize it.
3032         (get_string_length): Detect unterminated arrays.
3033         (format_string): Same.
3034         (format_directive): Warn about unterminated arrays.
3035         (handle_gimple_call): Mark statements with no_warning as needed.
3037 2018-10-03  Jim Wilson  <jimw@sifive.com>
3039         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
3040         also define __riscv_abi_rve.  Delete trailing white space.
3042 2018-10-03  Paul Koning  <ni1d@arrl.net>
3044         Enable LRA register allocator for PDP11.
3045         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
3046         (R): Likewise.
3047         (D): Likewise.
3048         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
3049         * config/pdp11/pdp11.opt (-mlra): New option.
3050         * doc/invoke.texi (PDP-11 Options): Document -mlra.
3052 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
3054         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
3055         (*<absneg:code>extend<mode>xf2): Ditto.
3057 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
3059         PR tree-optimization/87415
3060         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
3061         precision fields.
3063 2018-10-02  Jeff Law  <law@redhat.com>
3065         * gimple-fold.c (get_range_strlen): Only set *nonstr when
3066         an unterminated string is discovered.  Bubble up range
3067         even for unterminated strings.
3068         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
3069         indicates the string was not terminated via NONSTR.
3071 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
3073         * tree-vrp.c (extract_range_from_unary_expr): Special case all
3074         pointer conversions.
3075         Do not do anything special for anti-ranges.
3077 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
3079         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
3080         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
3082 2018-10-03  Martin Liska  <mliska@suse.cz>
3084         PR gcov-profile/86109
3085         * coverage.c (coverage_begin_function): Do not
3086         mark lambdas as artificial.
3087         * tree-core.h (struct GTY): Remove tm_clone_flag
3088         and introduce new lambda_function.
3089         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
3091 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3093         PR target/87474
3094         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
3095         P8_VECTOR and VSX are enabled.
3097 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3099         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
3100         0x3907 as CPU model number.
3102 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3104         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
3105         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
3106         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
3107         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
3108         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
3109         * config/s390/s390.md: Likewise. Rename also the cpu attribute
3110         value from arch12 to z14.
3112 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3114         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
3115         (isinfxf2): Ditto.
3116         (isinf<mode>2): Ditto.
3118 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3120         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
3121         before emitting fxam.  Perform calculations in XFmode.
3123 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3125         * match.pd (((X /[ex] A) +- B) * A): New transformation.
3127 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3129         PR middle-end/87319
3130         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
3131         * tree.c (signed_or_unsigned_type_for): Handle complex.
3133 2018-10-02  Jeff Law  <law@redhat.com>
3135         * gimple-fold.c (get_range_strlen): Remove dead code.
3137 2018-10-02  Martin Sebor  <msebor@redhat.com>
3138             Jeff Law  <law@redhat.com>
3140         * builtins.c (unterminated_array): Add new arguments.
3141         If argument is not terminated, bubble up size and exact
3142         state to callers.
3143         (expand_builtin_strnlen): Detect, avoid expanding
3144         and diagnose unterminated arrays.
3145         (c_strlen): Fill in offset of start of unterminated strings.
3146         * builtins.h (unterminated_array): Update prototype.
3148 2018-10-02  Richard Biener  <rguenther@suse.de>
3150         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
3151         of haddv4df, first reduce to SSE width and exploit the fact
3152         that we only need element zero with the reduction result.
3153         (reduc_plus_scal_v2df): Likewise.
3155 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3157         * dojump.h (do_jump): Delete.
3158         (do_jump_1): Likewise.
3159         (split_comparison): Move around.
3160         * dojump.c (do_jump): Make static.
3161         (do_jump_1): Likewise.
3162         (jumpifnot): Move around.
3163         (jumpifnot_1): Likewise.
3164         (jumpif): Likewise.
3165         (jumpif_1): Likewise.
3166         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
3168 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3170         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
3171         insns in the delay slot and add_insn_after for the jump insn.
3173 2018-10-02  Richard Biener  <rguenther@suse.de>
3175         * tree-inline.c (expand_call_inline): Use the location of
3176         the callee declaration for the inline-entry marker.
3177         * final.c (notice_source_line): Remove special-casing of
3178         NOTE_INSN_INLINE_ENTRY.
3180 2018-10-01  Carl Love  <cel@us.ibm.com>
3182         PR 69431
3183         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
3184         (__builtin_mtfsb0): New.
3185         (__builtin_mtfsb1): New.
3186         ( __builtin_set_fpscr_rn): New.
3187         (__builtin_set_fpscr_drn): New.
3188         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
3189         (rs6000_expand_set_fpscr_rn_builtin): Add.
3190         (rs6000_expand_set_fpscr_drn_builtin): Add.
3191         (rs6000_expand_builtin): Add case statement entries for
3192         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
3193         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
3194         RS6000_BUILTIN_MFFSL.
3195         (rs6000_init_builtins): Add ftype initialization and def_builtin
3196         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
3197         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
3198         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
3199         rs6000_mffscdrn): Add define_insn.
3200         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
3201         * doc/extend.texi: Add documentation for the builtins.
3203 2018-10-01  Richard Biener  <rguenther@suse.de>
3205         PR tree-optimization/87465
3206         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
3207         causing branch miscounts.
3209 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3211         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
3212         aarch64_option_default_param):  New.
3213         (params.h): Include.
3214         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
3215         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
3216         stack-clash protection validation code.
3218 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3220         * params.c (validate_param): New.
3221         (add_params): Use it.
3222         (set_param_value): Refactor param validation into validate_param.
3223         (diagnostic.h): Include.
3224         * diagnostic.h (diagnostic_ready_p): New.
3226 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3228         * params.c (set_param_value):
3229         Add index of parameter being validated.
3230         * common/common-target.def (option_validate_param): New.
3231         * common/common-targhooks.h (default_option_validate_param): New.
3232         * common/common-targhooks.c (default_option_validate_param): New.
3233         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
3234         * doc/tm.texi: Regenerate.
3236 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3238         PR target/86486
3239         * config/aarch64/aarch64.c (aarch64_override_options_internal):
3240         Add validation for stack-clash parameters and set defaults.
3242 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3244         PR target/86486
3245         * configure.ac: Add stack-clash-protection-guard-size.
3246         * doc/install.texi: Document it.
3247         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
3248         * params.def: Update comment for guard-size.
3249         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
3250         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
3251         * configure: Regenerate.
3253 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3255         PR target/86486
3256         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
3257         STACK_DYNAMIC_OFFSET): New.
3258         * config/aarch64/aarch64.c (aarch64_layout_frame):
3259         Update outgoing args size.
3260         (aarch64_stack_clash_protection_alloca_probe_range,
3261         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3263 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3265         PR target/86486
3266         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
3267         probe ranges.
3268         * target.def (stack_clash_protection_alloca_probe_range): New.
3269         (stack_clash_protection_final_dynamic_probe): Remove.
3270         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
3271         (default_stack_clash_protection_final_dynamic_probe): Remove.
3272         * targhooks.c: Likewise.
3273         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3274         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
3275         * doc/tm.texi: Regenerate.
3277 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3279         PR target/86486
3280         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
3281         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
3282         aarch64_clamp_to_uimm12_shift): New.
3283         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
3284         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
3286 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3288         PR target/86486
3289         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
3291 2018-10-01  Jeff Law  <law@redhat.com>
3292             Richard Sandiford <richard.sandiford@linaro.org>
3293             Tamar Christina  <tamar.christina@arm.com>
3295         PR target/86486
3296         * config/aarch64/aarch64.md
3297         (probe_stack_range): Add k (SP) constraint.
3298         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
3299         STACK_CLASH_MAX_UNROLL_PAGES): New.
3300         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
3301         stack probes for stack clash.
3302         (aarch64_allocate_and_probe_stack_space): New.
3303         (aarch64_expand_prologue): Use it.
3304         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
3305         (aarch64_sub_sp): Add emit_move_imm optional param.
3307 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
3309         PR tree-optimization/87261
3310         * match.pd: Remove trailing whitespace.
3311         Add (x & y) | ~(x | y) -> ~(x ^ y),
3312         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
3314 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3316         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
3317         constraints.
3318         (ashlsi3_insn): Update instruction constraints.
3319         (ashrsi3_insn): Likewise.
3320         (rotrsi3): Likewise.
3321         (add_shift): Likewise.
3322         * config/arc/constraints.md (Csz): New 32 bit constraint. It
3323         avoids placing in the limm field small constants which, otherwise,
3324         could end into a small instruction.
3326 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3328         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
3329         destination register is not odd-even.
3330         (umaddsidi4_split): Likewise.
3332 2018-10-01  Richard Biener  <rguenther@suse.de>
3334         * tree-inline.c (expand_call_inline): Store origin of fn
3335         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
3336         * tree.c (block_ultimate_origin): Simplify and do some
3337         checking.
3339 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
3341         * config/i386/mmx.md (EMMS): New int iterator.
3342         (emms): New int attribute.
3343         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
3344         EMMS int iterator.  Explicitly declare clobbers.
3345         (mmx_emms): Remove expander.
3346         (mmx_femms): Ditto.
3347         * config/i386/predicates.md (emms_operation): Remove predicate.
3348         (vzeroall_pattern): New predicate.
3349         (vzeroupper_pattern): Rename from vzeroupper_operation.
3350         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
3351         vzeroupper_pattern and vzeroall_pattern predicates.
3353 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
3355         PR rtl-optimization/86939
3356         * ira-lives.c (make_hard_regno_born): Rename from this...
3357         (make_hard_regno_live): ... to this.  Remove update to conflict
3358         information.  Update function comment.
3359         (make_hard_regno_dead): Add conflict information update.  Update
3360         function comment.
3361         (make_object_born): Rename from this...
3362         (make_object_live): ... to this.  Remove update to conflict information.
3363         Update function comment.
3364         (make_object_dead):  Add conflict information update.  Update function
3365         comment.
3366         (mark_pseudo_regno_live): Call make_object_live.
3367         (mark_pseudo_regno_subword_live): Likewise.
3368         (mark_hard_reg_dead): Update function comment.
3369         (mark_hard_reg_live): Call make_hard_regno_live.
3370         (process_bb_node_lives): Likewise.
3371         * lra-lives.c (make_hard_regno_born): Rename from this...
3372         (make_hard_regno_live): ... to this.  Remove update to conflict
3373         information.  Remove now uneeded check_pic_pseudo_p argument.
3374         Update function comment.
3375         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
3376         to conflict information.  Update function comment.
3377         (mark_pseudo_live): Remove update to conflict information.  Update
3378         function comment.
3379         (mark_pseudo_dead): Add conflict information update.
3380         (mark_regno_live): Call make_hard_regno_live.
3381         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
3382         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
3384 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
3386         PR target/87370
3387         * config/i386/i386.c (construct_container): Use TImode for
3388         BLKmode values in 2 integer registers.
3390 2018-09-29  Jeff Law  <law@redhat.com>
3392         * builtins.c (unterminated_array): Pass in c_strlen_data * to
3393         c_strlen rather than just a tree *.
3394         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
3395         Update recursive calls appropriately.  If caller did not provide a
3396         suitable data pointer, create a local one.  When a non-terminated
3397         string is discovered, bubble up information about the string via the
3398         c_strlen_data object.
3399         * builtins.h (c_strlen): Update prototype.
3400         (c_strlen_data): New structure.
3401         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
3402         For a type 2 call, if c_strlen indicates a non-terminated string
3403         use the length of the non-terminated string.
3404         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
3406 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
3408         PR target/87467
3409         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
3410         __m512d type for __A argument rather than __m512.
3412 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
3414         * match.pd (simple_comparison): Don't optimize if either operand is
3415         a function pointer when target needs function pointer canonicalization.
3417 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3419         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
3420         power5 .. power9 to remove indirection.
3421         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
3422         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
3423         ASM_CPU_476_SPEC): Delete.
3424         (ASM_CPU_SPEC): Adjust.
3425         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
3426         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
3428 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3430         * config.in: Delete HAVE_AS_DCI.
3431         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
3432         * config/rs6000/rs6000.h: Ditto.
3433         * configure.ac: Delete HAVE_AS_DCI.
3434         * configure: Regenerate.
3436 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3438         * config.in (HAVE_AS_LWSYNC): Delete.
3439         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3440         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
3441         do it as a .long .
3442         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3443         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
3444         as a .long .
3445         * configure.ac: Delete HAVE_AS_LWSYNC.
3446         * configure: Regenerate.
3448 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3449             Pierre-Marie de Rodat  <derodat@adacore.com>
3451         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
3452         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
3453         (cgraph_node::create_thunk): Add indirect_offset parameter.
3454         (thunk_adjust): Likewise.
3455         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
3456         and initialize the corresponding field with it.
3457         (cgraph_node::dump): Dump indirect_offset field.
3458         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
3459         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
3460         (thunk_adjust): Add indirect_offset parameter and deal with it.
3461         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
3462         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
3463         or if the thunk is external or local.  Fix formatting.  Do not chain
3464         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
3465         if any, in the GIMPLE representation.
3466         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
3467         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
3468         (input_node): Read indirect_offset field.
3469         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
3470         call to thunk_adjust.
3471         * tree-nested.c (struct nesting_info): Add thunk_p field.
3472         (create_nesting_tree): Set it.
3473         (convert_all_function_calls): Copy static chain from targets to thunks.
3474         (finalize_nesting_tree_1): Return early for thunks.
3475         (unnest_nesting_tree_1): Do not finalize thunks.
3476         (gimplify_all_functions): Do not gimplify thunks.
3478 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3480         * opt-suggestions.c (option_proposer::build_option_suggestions):
3481         Release "option_values".
3483 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3485         * coverage.c (get_coverage_counts): Convert problem-reporting dump
3486         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
3487         * dumpfile.c (kind_as_string): New function.
3488         (dump_loc): Rather than a hardcoded prefix of "note: ", use
3489         kind_as_string to vary the prefix based on dump_kind.
3490         (selftest::test_capture_of_dump_calls): Update for above.
3492 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3494         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3495         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
3497 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3499         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3500         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3501         INVALID_REGNUM instead of FPSR_REG.
3502         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3503         * config/i386/i386.md: Update comment of FP compares.
3504         (fldenv): Do not clobber FPSR_REG.
3506 2018-09-28  Richard Biener  <rguenther@suse.de>
3508         * tree.h (BLOCK_ORIGIN): New.
3509         * omp-expand.c (grid_expand_target_grid_body): Assign
3510         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3511         * tree-inline.c (remap_block): Likewise.
3512         * auto-profile.c (get_function_decl_from_block): Simplify
3513         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3514         * langhooks.c (lhd_print_error_function): Likewise.
3515         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3516         Likewise.
3517         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3518         * tree.c (block_nonartificial_location): Likewise.
3519         (block_ultimate_origin): Likewise.
3520         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
3521         no longer needed LTO case.
3523 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
3524             Jan Hubicka  <jh@suse.cz>
3525             Martin Jambor  <mjambor@suse.cz>
3527         * simplify-rtx.c (simplify_merge_mask): New function.
3528         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3529         same masks are used in op1 or op2.
3530         (test_vec_merge): New function.
3531         (test_vector_ops): Call test_vec_merge.
3533 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3535         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3536         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3537         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3538         (cypress_costs): Set it.
3539         (supersparc_costs): Likewise.
3540         (hypersparc_costs): Likewise.
3541         (leon_cost): Likewise.
3542         (leon3_costs): Likewise.
3543         (sparclet_costs): Likewise.
3544         (ultrasparc_costs): Likewise.
3545         (ultrasparc_costs): Likewise.
3546         (niagara_costs): Likewise.
3547         (niagara2_costs): Likewise.
3548         (niagara3_costs): Likewise.
3549         (niagara4_costs): Likewise.
3550         (niagara7_costs): Likewise.
3551         (m8_costs): Likewise.
3552         (TARGET_CAN_FOLLOW_JUMP): Define.
3553         (pass_work_around_errata::gate): Minor tweak.
3554         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3555         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3556         Automaitcally clear MASK_FSMULD mask for V7 processors.
3557         (sparc_can_follow_jump): New static function.
3558         (output_ubranch): Deal with CROSSING_JUMP_P.
3559         (sparc_use_sched_lookahead): Rewrite using switch statement.
3560         (sparc_issue_rate): Reorder.
3561         (sparc_branch_cost): New function.
3563 2018-09-27  Martin Sebor  <msebor@redhat.com>
3565         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3566         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3567         (int_fits_type_p): Same.
3569 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
3571         * config/i386/i386.md (FPCR_REG): Remove.
3572         (UNSPEC_FLDCW): Remove.
3573         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3574         (x86_fldcw_1): Remove insn pattern.
3575         (fnstenv): Do not clobber FPCR_REG.
3576         (fldenv): Ditto.
3577         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3578         (CALL_USED_REGISTERS): Ditto.
3579         (REG_ALLOC_ORDER): Ditto.
3580         (REG_CLASS_CONTENTS): Ditto.
3581         (HI_REGISTER_NAMES): Ditto.
3582         (ADDITIONAL_REGISTER_NAMES): Use defines instead
3583         of numerical constants.
3584         * config/i386/i386.c (regclass_map): Remove fpsr register.
3585         (dbx_register_map): Ditto.
3586         (dbx64_register_map): Ditto.
3587         (svr4_dbx_register_map): Ditto.
3588         (print_reg): Do not handle FPCR_REG.
3590 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
3592         PR target/87149
3593         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3594         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3595         Delete, always treat as true.
3596         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3597         Ditto.  Simplify remaining code.
3598         * config/powerpcspe/powerpcspe.h: Ditto.
3599         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3600         Simplify remaining code.
3601         (rs6000_expand_builtin): Ditto.
3602         * config/rs6000/rs6000.h: Ditto.
3603         * configure.ac: Ditto.
3604         * configure: Regenerate.
3606 2018-09-27  Martin Liska  <mliska@suse.cz>
3608         * coverage.c (get_coverage_counts): Revert the formatting
3609         of missing profile opt info.
3611 2018-09-27  Richard Biener  <rguenther@suse.de>
3613         PR debug/37801
3614         PR debug/87440
3615         * dwarf2out.c (set_block_origin_self): Do not mark outermost
3616         block as we do not output that.
3617         (gen_inlined_subroutine_die): Elide the originally outermost
3618         block, matching what we do for concrete instances.
3619         (decls_for_scope): Add parameter specifying whether to recurse
3620         to subblocks.
3622 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
3623             Tom de Vries  <tom@codesourcery.com>
3625         PR 82089
3627         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3628         STORE_FLAG_VALUE == 1.
3630 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
3632         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3633         constant definitions.
3634         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3635         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3636         ("speculation_barrier"): New expander definition.
3638 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
3640         PR gcov-profile/86957
3641         * common.opt: New warning option -Wmissing-profile.
3642         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3643         * doc/invoke.texi: Document -Wmissing-profile.
3645 2018-09-26  Jim Wilson  <jimw@sifive.com>
3647         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3648         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3650 2018-09-26  Martin Sebor  <msebor@redhat.com>
3652         * tree.c (zerop): Change return type to bool.
3653         (integer_zerop, integer_onep, integer_each_onep): Same.
3654         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3655         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3656         (real_onep, real_minus_onep, chain_index): Same.
3657         (print_type_hash_statistics, type_list_equal): Same.
3658         * tree.h (zerop): Same.
3659         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
3660         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3661         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3662         (real_onep, real_minus_onep, chain_index): Same.
3663         (print_type_hash_statistics, type_list_equal): Same.
3665 2018-09-26  Jim Wilson  <jimw@sifive.com>
3667         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3669 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
3671         PR target/87414
3672         * config/i386/i386.c: Include debug.h and dwarf2out.h.
3673         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3674         call.
3676 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
3678         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3680 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3682         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3683         and FP_SECOND_SSE_REGS.
3684         (REG_CLASS_NAMES): Ditto.
3685         (REG_CLASS_CONTENTS): Ditto.
3686         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3687         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3688         (ix86_preferred_output_reload_class): Ditto.
3689         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3690         clobber constraint to "=&f".
3691         (fix_truncdi_i387): Ditto.
3692         (lrintxfdi2): Ditto.
3693         (fistdi2_<rounding>): Ditto.
3694         (fpremxf4_i387): Change "=u" constraint to "=f".
3695         (fprem1xf4_i387): Ditto.
3696         (sincosxf3): Ditto.
3697         (fptanxf4_i387): Ditto.
3698         (fxtractxf3_i387): Ditto.
3699         (fscalexf4_i387): Ditto.
3700         (atan2xf3): Change "u" constraint to "f".
3701         (fyl2xxf3_i387): Ditto.
3702         (fyl2xp1xf3_i387): Ditto.
3704 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3706         PR target/87439
3707         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3708         for removed I387_MASK_PM entity.
3711 2018-09-26  Jeff Law  <law@redhat.com>
3712         Revert
3713         2018-09-26  Alexey Neyman  <stilor@att.net>
3715         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3716         headers are no longer pulled in by <isl/val.h>.
3718 2018-09-26  Richard Biener  <rguenther@suse.de>
3720         PR debug/87443
3721         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3722         or concrete instance DIE to the tree.  Create abstract origin
3723         attributes also for concrete instances.
3725 2018-09-26  Alexey Neyman  <stilor@att.net>
3727         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3728         headers are no longer pulled in by <isl/val.h>.
3730 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
3732         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3733         Use new helper functions.
3734         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3735         Use new helper functions.
3736         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3737         aarch_mm_needs_release): New declarations.
3738         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3739         aarch_mm_needs_release): New.
3741 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
3743         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3744         (arm32_output_mi_thunk): Deal with long calls.
3746 2018-09-26  Richard Biener  <rguenther@suse.de>
3748         PR debug/87428
3749         PR debug/87362
3750         * tree-inline.c (expand_call_inline): When the location
3751         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3752         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3753         the inserted BLOCK to make inlined_function_outer_scope_p
3754         recognize it.
3755         * dwarf2out.c (add_call_src_coords_attributes): Do not add
3756         coords for reserved locations.
3758 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3760         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3761         (*call_indirect_nonlocal_sysv<mode>): Ditto.
3762         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3763         (*sibcall_nonlocal_sysv<mode>): Ditto.
3764         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3765         (<bd>_<mode>): Ditto.
3766         (<bd>tf_<mode>): Ditto.
3768 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3770         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3771         control string as a list of templates instead of as C code.
3772         (*altivec_movti): Ditto.
3773         * config/rs6000/darwin.md (movdf_low_di): Ditto.
3775 2018-09-25  Jim Wilson  <jimw@sifive.com>
3777         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3778         when target symbol is weak.
3780 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3782         PR c/87387
3783         * builtins.c (unterminated_array): Simplify.
3784         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
3785         where pointer arithmetic is safe.
3787 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3789         PR target/86987
3790         * config/rs6000/altivec.md (altivec_vspltb): Use
3791         const_0_to_15_operand instead of u5bit_cint_operand.
3792         (*altivec_vspltb_internal): Ditto.
3793         (altivec_vspltb_direct): Ditto.
3794         (altivec_vsplth): Use const_0_to_7_operand instead of
3795         u5bit_cint_operand.
3796         (*altivec_vsplth_internal): Ditto.
3797         (altivec_vsplth_direct): Ditto.
3798         (altivec_vspltw): Use const_0_to_3_operand instead of
3799         u5bit_cint_operand.
3800         (*altivec_vspltw_internal): Ditto.
3801         (altivec_vspltw_direct): Ditto.
3802         (altivec_vspltsf): Ditto.
3803         (*altivec_vspltsf_internal): Ditto.
3804         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3805         various splats with the proper size immediate.  Reorder the various
3806         cases by ascending size of immediate, and put all such together.
3808 2018-09-25  Richard Biener  <rguenther@suse.de>
3810         PR debug/83941
3811         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3812         GC-ification.
3813         (maybe_create_die_with_external_ref): Do not create
3814         DW_TAG_imported_unit here.
3815         (add_abstract_origin_attribute): Handle external BLOCK refs.
3816         (dwarf2out_abstract_function): Simplify LTO case.
3817         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3818         rather than using maybe_create_die_with_external_ref.
3820 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
3822         PR target/71278
3823         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3824         (frndintxf2_mask_pm_i387): Ditto.
3825         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3826         Enable for !flag_trapping_math.
3827         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3828         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3829         Change operand 1 predicate to nonimmediate_operand.
3830         (attr "i387_cw"): Remove mask_pm.
3831         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3832         (enum ix86_entity): Remove I387_MASK_PM.
3833         * config/i386/i386.c (ix86_i387_mode_needed): Do not
3834         handle I387_MASK_PM.
3835         (ix86_mode_needed): Ditto.
3836         (ix86_mode_after): Ditto.
3837         (ix86_mode_entry): Ditto.
3838         (ix86_mode_exit): Ditto.
3839         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3841 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
3843         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3844         to_update_switch_stmts to vNULL instead of calling create on them
3845         immediately.
3847 2018-09-25  Richard Biener  <rguenther@suse.de>
3849         PR tree-optimization/87402
3850         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3851         (visit_phi): Re-instantiate handling of supposed to be VARYING
3852         but non-VARYING backedge value.
3854 2018-09-25  Richard Biener  <rguenther@suse.de>
3856         PR debug/83941
3857         * dwarf2out.c (struct sym_off_pair): New.
3858         (external_die_map): New global.
3859         (lookup_decl_die): When in LTO create DIEs lazily from the
3860         external_die_map.
3861         (lookup_block_die): New function, create DIEs lazily in LTO.
3862         (equate_block_to_die): New function.
3863         (dwarf2out_die_ref_for_decl): During WPA get the association
3864         from the external DIE map.
3865         (dwarf2out_register_external_die): Record mapping into the
3866         external DIE map.
3867         (maybe_create_die_with_external_ref): New function split out from
3868         DIE generation part of old dwarf2out_register_external_die.
3869         (add_abstract_origin_attribute): Do not return the DIE.  When
3870         in LTO reference externals directly.
3871         (dwarf2out_abstract_function): When in LTO ignore calls for
3872         decls with external DIEs (already present abstract instances).
3873         (gen_call_site_die): Adjust.
3874         (add_high_low_attributes): Likewise.
3875         (gen_lexical_block_die): Likewise.
3876         (gen_inlined_subroutine_die): Likewie.
3877         (gen_block_die): Likewise.
3878         (dwarf2out_inline_entry): Likewise.
3879         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
3880         DIEs.
3882 2018-09-25  Martin Liska  <mliska@suse.cz>
3884         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
3885         integers and not by a float value.
3887 2018-09-25  Martin Liska  <mliska@suse.cz>
3889         PR fortran/87394
3890         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
3891         instead of NULL.
3892         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
3893         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
3894         * godump.c (go_format_type): Remove extra parenthesis.
3896 2018-09-25  Martin Liska  <mliska@suse.cz>
3898         * alias.c (set_dest_equal_p): Remove unused function.
3899         * config/i386/i386.c (def_builtin_pure2): Likewise.
3900         * diagnostic-show-locus.c (class layout): Remove
3901         unused field.
3902         (layout::layout): Likewise here.
3903         * dump-context.h (class temp_dump_context): Likewise.
3904         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
3905         (add_AT_loclistsptr): Likewise.
3906         (add_AT_offset): Likewise.
3907         (get_AT_hi_pc): Likewise.
3908         (is_comdat_die): Likewise.
3909         (type_is_enum): Likewise.
3910         (ceiling): Likewise.
3911         (add_AT_vms_delta): Likewise.
3912         (is_class_die): Likewise.
3913         * edit-context.c (class line_event): Remove unused field.
3914         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
3915         unused function.
3916         * ipa-cp.c (ipa_get_vr_lat): Likewise.
3917         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
3918         (ok_for_base_p_nonstrict): Likewise.
3919         * tree-chrec.c (is_not_constant_evolution): Likewise.
3920         (chrec_fold_poly_cst): Likewise.
3921         * tree-if-conv.c (has_pred_critical_p): Likewise.
3922         * tree-ssa-coalesce.c (print_exprs): Likewise.
3923         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
3924         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
3925         * tree-vrp.c (value_ranges_intersect_p): Likewise.
3926         (value_range_nonnegative_p): Likewise.
3928 2018-09-25  Martin Liska  <mliska@suse.cz>
3930         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3931         Do not handle "GNU Pascal".
3932         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3933         Likewise.
3934         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
3935         from documentation. Likewise.
3936         * dbxout.c (dbxout_range_type): Likewise.
3937         * doc/cpp.texi: Likewise.
3938         * doc/extend.texi: Likewise.
3939         * doc/frontends.texi: Likewise.
3940         * doc/invoke.texi: Remove Pascal entry.
3941         * tree.def (CLEANUP_POINT_EXPR): Likewise.
3942         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
3944 2018-09-25  Martin Liska  <mliska@suse.cz>
3946         PR middle-end/86078
3947         * doc/invoke.texi: Document all parameters and remove default
3948         of the parameters.
3950 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3952         PR bootstrap/87417
3953         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
3954         contains HOST_WIDE_INTs when computing its size.
3956 2018-09-24  Jim Wilson  <jimw@sifive.com>
3958         PR target/87391
3959         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
3960         not TARGET_RVE.
3961         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
3963 2018-09-24  Andrew Pinski  <apinski@marvell.com>
3965         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3966         access prev before checking it for NULLness in the
3967         AARCH64_FUSE_CMP_BRANCH case.
3969 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
3971         PR target/82699
3972         * config/i386/i386.c (rest_of_insert_endbranch): Set
3973         endbr_queued_at_entrance to true and don't insert ENDBR if
3974         x86_function_profiler will be called.
3975         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3976         is true.
3977         * config/i386/i386.h (machine_function): Add
3978         endbr_queued_at_entrance.
3980 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3982         * genattrtab.c (mk_attr_alt): Use alternative_mask.
3983         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
3984         types.
3985         (check_attr_test): Use alternative_mask.
3986         (get_attr_value): Likewise.
3987         (compute_alternative_mask): Use alternative_mask and XWINT.
3988         (make_alternative_compare): Use alternative_mask.
3989         (attr_alt_subset_p): Use XWINT.
3990         (attr_alt_subset_of_compl_p): Likewise.
3991         (attr_alt_intersection): Use alternative_mask and XWINT.
3992         (attr_alt_union): Likewise.
3993         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
3994         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
3995         (simplify_test_exp): Use alternative_mask and XWINT.
3996         (write_test_expr): Use alternative_mask and XWINT, adjust bit
3997         number calculation to support 64 bits.  Generate code that
3998         checks 64-bit masks.
3999         (main): Use alternative_mask.
4000         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
4002 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
4004         PR target/80080
4005         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
4006         RETURN+USE when returning via %r14.
4008 2018-09-24  Martin Liska  <mliska@suse.cz>
4010         * gcov.c (output_lines): Print colorization legend
4011         for both flag_use_colors and flag_use_hotness_colors.
4012         Reword the help.
4014 2018-09-24  Martin Liska  <mliska@suse.cz>
4016         * coverage.c (get_coverage_counts): Use warning_at
4017         with current_function_decl location. Use %qD in warning
4018         message.
4020 2018-09-24  Martin Liska  <mliska@suse.cz>
4022         * memory-block.h (memory_block_pool::release): Annotate with
4023         valgrind that the memory is not accessible.
4025 2018-09-24  Martin Liska  <mliska@suse.cz>
4027         PR sanitizer/85774
4028         * asan.c: Make asan_handled_variables extern.
4029         * asan.h: Likewise.
4030         * cfgexpand.c (expand_stack_vars): Make sure
4031         a representative is unpoison if another
4032         variable in the partition is handled by
4033         use-after-scope sanitization.
4035 2018-09-24  Richard Biener  <rguenther@suse.de>
4037         PR tree-optimization/63155
4038         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
4039         the worklist when the edge of the respective argument isn't
4040         executable.
4042 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4044         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
4045         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
4046         (MASK_CLASS_P): Update for rename.
4047         (MAYBE_MASK_CLASS_P): Ditto.
4048         (REG_CLASS_NAMES): Update.
4049         (REG_CLASS_CONTENT): Update.
4050         * config/i386/i386.c (regclass_map): Update for MASK_REG
4051         and ALL_MASK_REGS rename.
4052         * config/i386/constraints.md (Yk): Update for rename.
4053         (k): Ditto.
4055 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4057         * config/i386/i386.h (enum reg_class): Remove
4058         EVEX_SSE_REGS and MOD4_SSE_REGS.
4059         (REG_CLASS_NAMES): Update.
4060         (REG_CLASS_CONTENT): Update.
4061         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
4062         registers as ALL_SSE_REGS.
4063         (ix86_additional_allocno_class_p): Remove.
4064         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
4065         (ix86_register_priority): Lower priority of EVEX SSE registers.
4066         Use IN_RANGE macro where appropriate.
4067         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
4068         AVX-5124VNNIW checks.
4069         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
4070         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
4071         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
4072         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
4073         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
4074         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
4075         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
4076         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
4077         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
4078         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
4079         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
4080         * config/i386/constraints.md (Yh): Remove.
4082 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4084         * config/i386/i386.c (regclass_map): Declare integer REX registers
4085         as GENERAL_REGS.
4087 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
4089         * doc/service.texi (Service): Switch the fsf.org link to https.
4091 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4093         PR target/86798
4094         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4095         Define to speculation_safe_value_not_needed.
4097 2018-09-21  Florian Weimer  <fweimer@redhat.com>
4099         PR middle-end/81035
4100         * doc/extend.texi (Common Function Attributes): Mention that
4101         noreturn suppresses tail call optimization.
4103 2018-09-21  Jeff Law  <law@redhat.com>
4105         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
4106         vr_values::cleanup_edges_and_switches.
4107         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
4108         vr_values class.
4109         (identify_jump_threads): Remove EDGE_IGNORE handling.
4110         (execute_vrp): Move handling of to_remove_edges and
4111         to_update_switch_stmts into vr_values class member functions.
4112         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
4113         (to_update_switch_stmts): Likewise.
4114         * vr-values.c: Include cfghooks.h.
4115         (vr_values::vr_values): Initialize to_remove_edges and
4116         to_update_switch_stmts.
4117         (vr_values::~vr_values): Verify to_remove_edges and
4118         to_update_switch_stmts are empty.
4119         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
4120         (vr_values::cleanup_edges_and_switches): New member function.
4121         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
4122         function.  Add new data members.
4124 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
4126         PR tree-optimization/87309
4127         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
4128         calls with pflags and alt_flags.
4129         (selftest::test_capture_of_dump_calls): Add test of interaction of
4130         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
4132 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4134         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
4136 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4138         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
4140 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4142         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
4143         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
4144         (VXWORKS_LIBS_RTP): Minor reordering.
4146 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4148         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
4149         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
4151 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4153         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
4154         (PTRDIFF_TYPE): Likewise.
4156 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4158         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
4159         triplet, similar to support for VxWorks7.
4160         * config/vxworks-dummy.h: Provide a default definition
4161         of TARGET_VXWORKS64 to 0.
4163 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4165         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
4166         * config/vxworks-dummy.h: here.
4168 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4170         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
4172 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
4173             Bo Zhou  <zbo.zhou@hisilicon.com>
4175         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
4176         * config/aarch64/aarch64-tune.md: Regenerated.
4177         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
4178         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
4179         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
4181 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
4182             Julian Brown  <julian@codesourcery.com>
4184         * builtins.c (get_builtin_sync_mem): Handle address spaces.
4186 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4188         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
4189         if the call takes a static chain.
4191 2018-09-21  Martin Liska  <mliska@suse.cz>
4193         * auto-profile.c (autofdo_source_profile::read): Do not
4194         set sum_all.
4195         (read_profile): Do not add working sets.
4196         (read_autofdo_file): Remove sum_all.
4197         (afdo_callsite_hot_enough_for_early_inline): Remove const
4198         qualifier.
4199         * coverage.c (struct counts_entry): Remove gcov_summary.
4200         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
4201         do not support GCOV_TAG_PROGRAM_SUMMARY.
4202         (get_coverage_counts): Remove summary and expected
4203         arguments.
4204         * coverage.h (get_coverage_counts): Likewise.
4205         * doc/gcov-dump.texi: Remove -w option.
4206         * gcov-dump.c (dump_working_sets): Remove.
4207         (main): Do not support '-w' option.
4208         (print_usage): Likewise.
4209         (tag_summary): Likewise.
4210         * gcov-io.c (gcov_write_summary): Do not dump
4211         histogram.
4212         (gcov_read_summary): Likewise.
4213         (gcov_histo_index): Remove.
4214         (gcov_histogram_merge): Likewise.
4215         (compute_working_sets): Likewise.
4216         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
4217         it not obsolete.
4218         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
4219         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
4220         (GCOV_HISTOGRAM_SIZE): Remove.
4221         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
4222         (struct gcov_summary): Simplify rapidly just
4223         to runs and sum_max fields.
4224         (gcov_histo_index): Remove.
4225         (NUM_GCOV_WORKING_SETS): Likewise.
4226         (compute_working_sets): Likewise.
4227         * gcov-tool.c (print_overlap_usage_message): Remove
4228         trailing empty line.
4229         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
4230         (output_lines): Remove program related line.
4231         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
4232         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
4233         histogram.
4234         (input_profile_summary): Do not read it.
4235         (merge_profile_summaries): And do not merge it.
4236         (input_symtab): Do not call removed function.
4237         * modulo-sched.c (sms_schedule): Do not print sum_max.
4238         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
4239         removed when histogram method was invented.
4240         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
4241         mode.
4242         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
4243         GCOV coding style.
4244         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
4245         and dump selected value.
4246         * profile.c (add_working_set): Remove.
4247         (get_working_sets): Likewise.
4248         (find_working_set): Likewise.
4249         (get_exec_counts): Do not work with working sets.
4250         (read_profile_edge_counts): Do not inform as sum_max is removed.
4251         (compute_branch_probabilities): Likewise.
4252         (compute_value_histograms): Remove argument for call of
4253         get_coverage_counts.
4254         * profile.h: Do not make gcov_summary const.
4256 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
4258         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
4260 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4262         PR tree-optimization/86990
4263         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
4264         Check that the entire merged store group is made of constants only for
4265         overlapping stores.
4267 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
4269         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
4270         (VTABLE_VERIFICATION_SPEC): Likewise.
4271         (SANITIZER_EARLY_SPEC): Likewise.
4272         (SANITIZER_SPEC): Likewise.
4273         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
4274         * doc/invoke.texi (Link Options): Document -r.
4276 2018-09-20  Richard Biener <rguenther@suse.de>
4278         PR middle-end/87054
4279         * gimplify.c (gimplify_expr): Retain alignment of
4280         addressable lvalue in dereference.
4282 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
4284         PR bootstrap/87013
4285         * configure.ac: Check for .loc is_stmt support.
4286         * configure, config.in: Rebuilt.
4287         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
4288         if not supported.
4290 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4292         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
4293         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
4294         -misel=no.
4296 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4298         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4299         VECTOR_OTHER.
4300         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
4301         case VECTOR_OTHER.
4303 2018-09-20  Marek Polacek  <polacek@redhat.com>
4305         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
4307 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4309         PR tree-optimization/87288
4310         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
4311         into account when determining PEELING_FOR_NITERS.
4313 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4315         PR tree-optimization/86877
4316         * tree-vect-loop.c (vect_analyze_loop_2): Call
4317         vect_verify_datarefs_alignment.
4319 2018-09-19  Marek Polacek  <polacek@redhat.com>
4321         * doc/invoke.texi: Document -Wclass-conversion.
4323 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
4325         * config/pa/pa.c (pa_adjust_priority): Delete.
4326         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
4328         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
4329         (atomic_storehi): Likewise.
4330         (atomic_storesi): Likewise.
4331         (atomic_loaddi): Restore compare and swap exchange loop code.
4333 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
4335         PR rtl-optimization/86902
4336         * combine.c (try_combine): When changing the CC mode used, don't change
4337         an unrelated mode in other_insn to that new CC mode.
4339 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
4341         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
4342         with %T in place of calls to dump_generic_expr.
4343         (prune_runtime_alias_test_list): Likewise.
4344         (create_runtime_alias_checks): Likewise.
4345         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
4346         (vect_analyze_data_ref_dependence): Likewise.
4347         (vect_slp_analyze_data_ref_dependence): Likewise.
4348         (vect_record_base_alignment): Likewise.  Use %G in place of call
4349         to dump_gimple_stmt.
4350         (vect_compute_data_ref_alignment): Likewise.
4351         (verify_data_ref_alignment): Likewise.
4352         (vect_find_same_alignment_drs): Likewise.
4353         (vect_analyze_group_access_1): Likewise.
4354         (vect_analyze_data_ref_accesses): Likewise.
4355         (dependence_distance_ge_vf): Likewise.
4356         (dump_lower_bound): Likewise.
4357         (vect_prune_runtime_alias_test_list): Likewise.
4358         (vect_find_stmt_data_reference): Likewise.
4359         (vect_analyze_data_refs): Likewise.
4360         (vect_create_addr_base_for_vector_ref): Likewise.
4361         (vect_create_data_ref_ptr): Likewise.
4362         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
4363         (vect_can_advance_ivs_p): Likewise.
4364         (vect_update_ivs_after_vectorizer): Likewise.
4365         (vect_gen_prolog_loop_niters): Likewise.
4366         (vect_prepare_for_masked_peels): Likewise.
4367         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4368         (vect_determine_vectorization_factor): Likewise.
4369         (vect_is_simple_iv_evolution): Likewise.
4370         (vect_analyze_scalar_cycles_1): Likewise.
4371         (vect_analyze_loop_operations): Likewise.
4372         (report_vect_op): Likewise.
4373         (vect_is_slp_reduction): Likewise.
4374         (check_reduction_path): Likewise.
4375         (vect_is_simple_reduction): Likewise.
4376         (vect_create_epilog_for_reduction): Likewise.
4377         (vect_finalize_reduction:): Likewise.
4378         (vectorizable_induction): Likewise.
4379         (vect_transform_loop_stmt): Likewise.
4380         (vect_transform_loop): Likewise.
4381         (optimize_mask_stores): Likewise.
4382         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
4383         (vect_split_statement): Likewise.
4384         (vect_recog_over_widening_pattern): Likewise.
4385         (vect_recog_average_pattern): Likewise.
4386         (vect_determine_min_output_precision_1): Likewise.
4387         (vect_determine_precisions_from_range): Likewise.
4388         (vect_determine_precisions_from_users): Likewise.
4389         (vect_mark_pattern_stmts): Likewise.
4390         (vect_pattern_recog_1): Likewise.
4391         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4392         (vect_record_max_nunits): Likewise.
4393         (vect_build_slp_tree_1): Likewise.
4394         (vect_build_slp_tree_2): Likewise.
4395         (vect_print_slp_tree): Likewise.
4396         (vect_analyze_slp_instance): Likewise.
4397         (vect_detect_hybrid_slp_stmts): Likewise.
4398         (vect_detect_hybrid_slp_1): Likewise.
4399         (vect_slp_analyze_operations): Likewise.
4400         (vect_slp_analyze_bb_1): Likewise.
4401         (vect_transform_slp_perm_load): Likewise.
4402         (vect_schedule_slp_instance): Likewise.
4403         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4404         (vect_mark_stmts_to_be_vectorized): Likewise.
4405         (vect_init_vector_1): Likewise.
4406         (vect_get_vec_def_for_operand): Likewise.
4407         (vect_finish_stmt_generation_1): Likewise.
4408         (vect_check_load_store_mask): Likewise.
4409         (vectorizable_call): Likewise.
4410         (vectorizable_conversion): Likewise.
4411         (vectorizable_operation): Likewise.
4412         (vectorizable_load): Likewise.
4413         (vect_analyze_stmt): Likewise.
4414         (vect_is_simple_use): Likewise.
4415         (vect_get_vector_types_for_stmt): Likewise.
4416         (vect_get_mask_type_for_stmt): Likewise.
4417         * tree-vectorizer.c (increase_alignment): Likewise.
4419 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
4421         * doc/rtl.texi: Adjust vec_select description.
4422         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
4423         non-constant selectors.
4425 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
4427         * config/aarch64/aarch64-protos.h
4428         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
4429         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
4430         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
4431         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
4432         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
4433         (AARCH64_FL_PROFILE): Move index so flags are ordered.
4434         (AARCH64_ISA_RCPC8_4): New flag.
4435         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
4436         to aarch64_offset_9bit_signed_unscaled_p.
4437         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
4438         and use stlur.
4439         * config/aarch64/constraints.md (Ust): New constraint.
4440         * config/aarch64/predicates.md.
4441         (aarch64_9bit_offset_memory_operand): New predicate.
4442         (aarch64_rcpc_memory_operand): New predicate.
4444 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
4446         PR rtl-optimization/87361
4447         * rtlanal.c (nonzero_bits1): Revert accidental change.
4449 2018-09-19  Richard Biener  <rguenther@suse.de>
4451         PR tree-optimization/87349
4452         PR tree-optimization/87342
4453         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
4455 2018-09-18  Marek Polacek  <polacek@redhat.com>
4457         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
4458         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
4460 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4462         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
4464 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4466         PR rtl-optimization/86882
4467         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
4469 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
4471         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
4472         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
4474 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
4476         PR other/87353
4477         * doc/invoke.texi (Link Options): Fix formatting and grammar.
4479 2018-09-18  Richard Biener  <rguenther@suse.de>
4481         PR middle-end/63155
4482         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
4483         (compute_samebase_partition_bases): Likewise.
4484         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
4485         (gimple_can_coalesce_p): Simplify.
4487 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
4489         Handle a library implementation of ffs calling __builtin_ffs.
4490         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
4491         (mmix_init_libfuncs): New function: make __builtin_ffs expand
4492         to __ffsdi2.
4494 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4496         * diagnostic-show-locus.c (class layout_range): Add field
4497         "m_original_idx".
4498         (layout_range::layout_range): Add "original_idx" param and use it
4499         to initialize new field.
4500         (make_range): Use 0 for original_idx.
4501         (layout::layout): Pass in index to calls to
4502         maybe_add_location_range.
4503         (layout::maybe_add_location_range): Add param "original_idx" and
4504         pass it on to layout_range.
4505         (layout::print_any_labels): Pass on range->m_original_idx to
4506         get_text call.
4507         (gcc_rich_location::add_location_if_nearby): Use 0 for
4508         original_idx.
4509         * gcc-rich-location.h (text_range_label::get_text): Update for new
4510         param.
4511         (range_label_for_type_mismatch::get_text): Likewise.
4513 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4515         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4517 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4519         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4520         format_string_diagnostic_t.
4521         (fmtwarn_n): Likewise.
4522         * substring-locations.c
4523         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4524         (format_warning_n_va): Convert to...
4525         (format_string_diagnostic_t::emit_warning_n_va): ...this.
4526         (format_warning_va): Convert to...
4527         (format_string_diagnostic_t::emit_warning_va): ...this.
4528         (format_warning_at_substring): Convert to...
4529         (format_string_diagnostic_t::emit_warning): ...this.
4530         (format_warning_at_substring_n): Convert to...
4531         (format_string_diagnostic_t::emit_warning_n): ...this.
4532         * substring-locations.h (class format_string_diagnostic_t): New
4533         class.
4534         (format_warning_va): Convert to
4535         format_string_diagnostic_t::emit_warning_va.
4536         (format_warning_n_va): Convert to
4537         format_string_diagnostic_t::emit_warning_n_va.
4538         (format_warning_at_substring): Convert to
4539         format_string_diagnostic_t::emit_warning.
4540         (format_warning_at_substring_n): Convert to
4541         format_string_diagnostic_t::emit_warning_n.
4543 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
4544             Bernd Schmidt <bernds_cb1@t-online.de>
4546         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4547         SImode args.
4549 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4551         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4552         operand 0 predicate to nonimmediate operand.
4553         (rint<mode>2_frndint): Remove insn pattern.
4554         (rint<mode>2): Change operand 1 predicate to general_operand.
4555         Extend operand 1 to XFmode and generate rintxf2 insn.
4556         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4557         Do not use X87MODEF mode macro.
4558         (frndintxf2_<rounding>_i387): Rename from
4559         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
4560         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4561         to XFmode and generate significandxf3 insn.
4563 2018-09-17  Richard Biener  <rguenther@suse.de>
4565         PR tree-optimization/87328
4566         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4567         visiting unexecutable backedges when not iterating.
4568         (do_rpo_vn): Mark all edges not executable even when not
4569         iterating.
4571 2018-09-17  Martin Jambor  <mjambor@suse.cz>
4573         PR c/63886
4574         * doc/invoke.texi (Warning Options): Likewise.
4576 2018-09-17  Richard Biener  <rguenther@suse.de>
4578         PR tree-optimization/87301
4579         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4580         clean EH info from leftover copy assignments.
4582 2018-09-17  Martin Liska  <mliska@suse.cz>
4584         PR gcov-profile/85871
4585         * gcov.c (output_intermediate_file): Fix out of bounds
4586         access.
4588 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
4590         * config/arc/arc.c: Object attributes for core4 not reflected
4591         correctly.
4592         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4593         core3).
4595 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
4597         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4599 2018-09-17  Martin Liska  <mliska@suse.cz>
4601         * doc/gcov.texi: Document new option --use-hotness-colors.
4602         * gcov.c (struct source_info): Declare new field.
4603         (source_info::source_info): Set default for maximum_count.
4604         (print_usage): Add new -q option.
4605         (process_args): Process it.
4606         (accumulate_line_info): Save src->maximum_count.
4607         (output_line_beginning): Make color line number if
4608         flag_use_hotness_colors is set.
4609         (output_line_details): Pass default argument value.
4610         (output_lines): Pass src->maximum_count.
4612 2018-09-17  Martin Liska  <mliska@suse.cz>
4614         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4615         Use processor_names table.
4616         * config/i386/i386.c (ix86_default_align): Use
4617         processor_cost_table for alignment values.
4618         (ix86_option_override_internal): Use processor_names.
4619         (ix86_function_specific_print): Likewise.
4620         * config/i386/i386.h (struct processor_costs):
4621         Add alignment values.
4622         (struct ptt): Remove and replace with const char *.
4623         * config/i386/x86-tune-costs.h (struct processor_costs):
4624         Declare default alignments for all costs.
4626 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
4628         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4629         symbolics or VR_VARYING ranges for ABS_EXPR.
4630         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4631         when range will wrap.
4633 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
4635         PR middle-end/86864
4636         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4637         before and after a JUMP_TABLE_DATA.
4639 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
4641         PR middle-end/87188
4642         * dojump.c (do_compare_and_jump): Canonicalize function pointers
4643         when one operand is a function pointer.  Use POINTER_TYPE_P and
4644         FUNC_OR_METHOD_TYPE_P.
4645         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4646         * fold-const.c (build_range_check): Likewise.
4647         * match.pd (simple_comparison): Likewise.
4649 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
4651         PR c/82967
4652         * spellcheck.c (get_edit_distance_cutoff): New function.
4653         (selftest::test_edit_distance_unit_test_oneway): Rename to...
4654         (selftest::test_get_edit_distance_one_way): ...this.
4655         (selftest::test_get_edit_distance_unit): Rename to...
4656         (selftest::test_get_edit_distance_both_ways): ...this.
4657         (selftest::test_edit_distances): Move tests to this new function,
4658         and test some more pairs of strings.  Update for above renaming.
4659         (selftest::get_old_cutoff): New function.
4660         (selftest::test_get_edit_distance_cutoff): New function.
4661         (selftest::assert_suggested_for): New function.
4662         (ASSERT_SUGGESTED_FOR): New macro.
4663         (selftest::assert_not_suggested_for): New function.
4664         (ASSERT_NOT_SUGGESTED_FOR): New macro.
4665         (selftest::test_suggestions): New function.
4666         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4667         tests to selftest::test_edit_distances and call it.  Add calls to
4668         selftest::test_get_edit_distance_cutoff and
4669         selftest::test_suggestions.
4670         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4671         (best_match::consider): Replace hard-coded cutoff calculation with
4672         a call to...
4673         (best_match::get_cutoff): New declaration.
4674         (best_match::get_best_meaningful_candidate): Likewise.
4676 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4678         * builtins.c (fold_builtin_strlen): Remove TODO comment.
4680 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4682         revert:
4683         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4685         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4686         terminated string literal.
4688 2018-09-14  Martin Sebor  <msebor@redhat.com>
4690         * builtins.c (unterminated_array): Handle ARRAY_REF.
4691         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4692         * builtins.h (unterminated_array): Declare extern.
4693         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4694         arrays.
4695         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4696         calls.
4698 2018-09-14  Martin Sebor  <msebor@redhat.com>
4699             Jeff Law  <law@redhat.com>
4701         * builtins.c (unterminated_array): New.
4702         (expand_builtin_strcpy): Adjust.
4703         (expand_builtin_strcpy_args): Detect unterminated arrays.
4704         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
4705         unterminated arrays.
4706         * gimple-fold.h (get_maxval_strlen): Add argument.
4707         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4709         * gimple-fold.c (get_range_strlen): Add argument.
4710         (get_maxval_strlen): Adjust.
4711         * gimple-fold.h (get_range_strlen): Add argument.
4713 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
4715         * config/i386/movdirintrin.h: Fix copyright year.
4717 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
4719         * reg-stack.c: Include regs.h.
4720         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4721         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4722         FIRST_STACK_REG, not DFmode.
4723         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4724         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
4725         (change stack): Default register mode to the reg_raw_mode of
4726         FIRST_STACK_REG, not DFmode.
4727         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4728         (*swapxf): Rename from swapxf.
4730 2018-09-14  Carl Love  <cel@us.ibm.com>
4732         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4733         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4735 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
4737         PR target/87224
4738         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4739         alternatives.
4741 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
4743         PR target/85628
4744         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4746 2018-09-14  Jason Merrill  <jason@redhat.com>
4748         Fix --enable-gather-detailed-mem-stats.
4749         * hash-table.c (hash_table_usage): Change from variable to function.
4750         * hash-table.h: Adjust.
4751         * Makefile.in: Add missing dependencies on hash-table.h.
4753 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4755         PR tree-optimization/87259
4756         PR lto/87283
4757         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4758         execute_cse_reciprocals_1 has tried transforming.
4760 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
4762         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4763         VR_VARYING for PLUS/MINUS_EXPR.
4765 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4767         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4768         formatting.
4770 2018-09-14  Richard Biener  <rguenther@suse.de>
4772         PR middle-end/63155
4773         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4774         bits for the merged partition.
4776 2018-09-13  Martin Sebor  <msebor@redhat.com>
4777             Bernd Edlinger  <bernd.edlinger@hotmail.de>
4779         * builtins.h (c_srlen): Add argument.
4780         * builtins.c (warn_string_no_nul): New function.
4781         (c_strlen): Add argument and use it.  Update recursive calls.
4782         Pass DECL argument to string_constant to get info on non
4783         terminated strings.  Update *NONSTR as needed.
4784         (fold_builtin_strlen): Add argument to calls to c_strlen.
4785         Warn for unterminated arrays.
4786         (warn_string_no_null): Add prototype.
4787         * expr.c (string_constant): Update arguments.  Update recursive
4788         calls appropriately.  Detect missing NUL terminator and outermost
4789         declaration its missing in.
4790         Improve checks for arrays with nonzero lower bound or elements
4791         that are not a single byte.  Simplify offset computation.
4792         Simplify checks for non-NUL terminated strings.
4793         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4794         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4796 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4798         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4799         correctly.
4800         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
4801         argument.  Fix range checks.
4802         * fold-const.h (c_getstr): Adjust protoype.
4803         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4804         string is constant but contains no NUL byte.
4806         * expr.c (string_constant): Adjust function comment.
4807         Remove bogus check for zero termination.
4809         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4811         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4812         (get_constant_size): Don't make STRING_CSTs larger than they are.
4813         (check_string_literal): New check function for STRING_CSTs.
4814         (output_constant): Use it.
4816 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
4818         PR target/86812
4819         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4821 2018-09-13  Richard Biener  <rguenther@suse.de>
4823         PR tree-optimization/87263
4824         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4825         (struct unwind_state): Add max_rpo field.
4826         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4827         Compute max_rpo, the max RPO number a block can be backwards reached
4828         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
4829         separating it from the iterating mode.
4831 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4833         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4834         (rfs_decision): New scheduling decision.
4836 2018-09-13  Richard Biener  <rguenther@suse.de>
4838         PR bootstrap/87134
4839         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4840         (vn_nary_op_insert_pieces_predicated): Do not write useless
4841         valid_dominated_by_p entry outside of the allocated storage.
4843 2018-09-13  Omar Sandoval  <osandov@osandov.com>
4844             Tom de Vries  <tdevries@suse.de>
4846         PR debug/86985
4847         * dwarf2out.c (is_c): New function.
4848         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4850 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
4852         PR target/85628
4853         * config/aarch64/aarch64.md (*aarch64_bfxil):
4854         Define.
4855         * config/aarch64/constraints.md (Ulc): Define.
4856         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4857         Define.
4858         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4859         New function.
4861 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4863         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4864         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4865         aarch64_layout_frame call.
4866         (aarch64_expand_epilogue): Likewise.
4867         (aarch64_initial_elimination_offset): Likewise.
4868         (aarch64_get_separate_components): Likewise.
4869         (aarch64_use_return_insn_p): Likewise.
4870         (aarch64_layout_frame): Remove unneeded check.
4872 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4874         * configure.ac: Only append
4875         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
4876         gcc_config_arguments if it was never reconfigured or last reconfigure
4877         was with different arguments.
4878         * configure: Regenerated.
4880 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4881             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4883         PR middle-end/87290
4884         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
4885         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
4887 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4889         PR tree-optimization/87287
4890         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
4891         X % C == 0 to X % (unsigned) C == 0 optimization to ...
4892         * match.pd (X % C == 0): ... here.  New optimization.
4894 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4896         PR middle-end/82853
4897         * expr.h (maybe_optimize_mod_cmp): Declare.
4898         * expr.c (mod_inv): New function.
4899         (maybe_optimize_mod_cmp): New function.
4900         (do_store_flag): Use it.
4901         * cfgexpand.c (expand_gimple_cond): Likewise.
4903 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
4904             Julian Brown  <julian@codesourcery.com>
4906         PR middle-end/86336
4907         * gimplify.c (gimplify_scan_omp_clauses): Set
4908         target_firstprivatize_array_bases in OpenACC parallel and kernels
4909         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
4910         OpenACC data regions.
4912 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
4914         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
4915         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
4917 2018-09-12  Richard Biener  <rguenther@suse.de>
4919         PR tree-optimization/87280
4920         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
4921         edge but unreachable target.
4922         (do_rpo_vn): For conservatively handling a PHI only mark
4923         the backedge executable but not the block reachable.
4925 2018-09-12  Richard Biener  <rguenther@suse.de>
4927         PR tree-optimization/87266
4928         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
4929         visited blocks.
4931 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
4933         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
4934         constants.
4935         ("trunc<BFP:mode><DFP_ALL:mode>2")
4936         ("trunc<DFP_ALL:mode><BFP:mode>2")
4937         ("extend<BFP:mode><DFP_ALL:mode>2")
4938         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
4939         according to the target operand type.
4941 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4942             Andreas Krebbel  <krebbel@linux.ibm.com>
4944         PR tree-optimization/86844
4945         * gimple-ssa-store-merging.c
4946         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
4947         there are any overlapping stores in between them, make sure they are
4948         also coalesced or we give up completely.
4950 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4952         PR middle-end/87248
4953         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
4954         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
4956 2018-09-12  Tom de Vries  <tdevries@suse.de>
4958         * common.opt (gdescribe-dies): Add option.
4959         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
4960         attribute for artifical and nameless decls.
4961         (dwarf2out_register_external_die): Add description attribute to
4962         external reference die.
4963         (add_desc_attribute): New functions.
4964         (gen_subprogram_die): Add description attribute to
4965         DW_TAG_call_site_parameter.
4966         * tree-pretty-print.c (print_generic_expr_to_str): New function.
4967         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
4968         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
4969         -gno-describe-dies.
4970         (@item -gdescribe-dies): Add.
4972 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4974         * tree-vrp.c (vrp_shift_undefined_p): Remove.
4975         (extract_range_from_binary_expr_1: Call
4976         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
4977         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
4978         depend on sign.
4980 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4982         * gimple-ssa-warn-alloca.c
4983         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
4984         field for ALLOCA_BOUND_*_LARGE.
4986 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4988         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
4990 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4992         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
4993         for clobbers.  Remove obsolete comment.
4995 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4997         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
4998         mpxchk, mpxld and mpxst types.
4999         (define_attr length_immediate): Remove all processing of mpx types.
5000         (define_attr prefix_0f): Ditto.
5001         (define_attr memory): Ditto.
5003 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
5005         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
5006         (log<mode>2): Change operand 1 predicate to general_operand.
5007         Extend operand 1 to XFmode and generate logxf3 insn.
5008         (log10<mode>2): Change operand 1 predicate to general_operand.
5009         Extend operand 1 to XFmode and generate log10xf3 insn.
5010         (log2<mode>2): Change operand 1 predicate to general_operand.
5011         Extend operand 1 to XFmode and generate log2xf3 insn.
5012         (fyl2xp1_extend<mode>xf3_i387): Remove.
5013         (log1p<mode>2): Change operand 1 predicate to general_operand.
5014         Extend operand 1 to XFmode and generate log1pxf3 insn.
5015         (fxtract_extend<mode>xf3_i387): Remove.
5016         (logb<mode>2): Change operand 1 predicate to general_operand.
5017         Extend operand 1 to XFmode and generate logbxf3 insn.
5018         (ilogb<mode>2): Change operand 1 predicate to general_operand.
5019         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
5020         (significand<mode>2): Change operand 1 predicate to general_operand.
5021         Extend operand 1 to XFmode and generate significandxf3 insn.
5023 2018-09-11  Nathan Sidwell  <nathan@acm.org>
5025         * gcc.c (perror_with_name, pfatal_with_name): Delete.
5026         (load_specs): Use fatal_error.
5027         (DELETE_IF_ORDINARY, process_command): Use error.
5028         (execute, run_attempt): Use fatal_error.
5030 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
5032         * diagnostic-core.h (sorry_at): New prototype.
5033         * diagnostic.c (sorry_at): New function.
5035 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
5037         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
5038         by zero as VR_UNDEFINED.
5040 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
5042         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
5043         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
5044         (<sincos>mode2): New expander.
5045         (sincos_extend<mode>xf3_i387): Remove insn pattern.
5046         (sincos -> sin, cos splitters): Remove splitter patterns.
5047         (sincos<mode>3): Change operand 2 predicate to general_operand.
5048         Extend operand 2 to XFmode and generate sincosxf3 insn.
5049         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
5050         Change operand 3 predicate to const1_operand.
5051         (fptan_extend<mode>xf4_i387): Remove insn pattern.
5052         (tanxf2): Update operands in the call to fptanxf4_i387.
5053         (tan<mode>2): Change operand 1 predicate to general_operand.
5054         Extend operand 1 to XFmode and generate tanxf3 insn.
5055         (atan2xf3): Rename from *fpatanxf3_i387.
5056         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
5057         (atan2xf3): Remove expander.
5058         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
5059         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
5060         (atan<mode>2): Change operand 1 predicate to general_operand.
5061         Extend operand 1 to XFmode and generate atanxf3 insn.
5063 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
5065         * config/i386/i386.md (x87/SSE constant load splitter): Use
5066         memory_operand instead of nonimmediate_operand for input operand
5067         predicate.
5069 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
5071         * config/i386/i386.md (float partial SSE register stall splitter): Move
5072         splitter near its instruction pattern.
5073         (float_extend partial SSE register stall splitter): Ditto.
5074         (float_truncate partial SSE register stall splitter): Ditto.
5076 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
5078         PR target/86794
5079         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5080         to speculation_safe_value_not_needed.
5082         PR target/85666
5083         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
5084         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
5085         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
5086         leaf_function_p, instead use has_hard_reg_initial_val.
5088 2018-09-09  Nathan Sidwell  <nathan@acm.org>
5090         * gcc.h (pfatal_with_name): Don't declare here.
5091         * gcc.c (pfatal_with_name): Make static.
5093 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
5095         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
5096         earlyclobber.
5098 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
5100         PR rtl-optimization/85458
5101         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
5102         priority hook to reduce the priority of EXPR.
5104 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
5106         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
5107         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
5108         with X87_ENABLE_FLOAT.  Remove preparation code.
5109         (*float<SWI48:mode><MODEF:mode>2): Rename from
5110         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
5111         math using "enabled" attribute.
5112         (*floatdi<MODEF:mode>2_i387): Rename from
5113         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
5114         enable for 32bit targets only.
5115         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
5116         splitter.
5117         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
5118         as operand 1 predicate.  Rewrite as define_insn_and_split.
5119         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
5121 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5123         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
5124         to fallthru to FLOAT case.
5126 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5128         PR target/86731
5129         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
5130         around folding of vec_sl to handle out of range shift values.
5132 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5134         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
5135         Update callers to gen_fix_trunc<mode>_i387_fisttp
5136         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
5137         nonimmediate_operand.
5138         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
5139         and corresponding splitters.
5140         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
5141         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
5142         (fix_truncdi_i387_with_temp): Remove insn pattern
5143         and corresponding splitters.
5144         (fix_trunc<mode>_i387): Change operand 0 predicate to
5145         nonimmediate_operand.
5146         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
5147         and corresponding splitters.
5148         (*fistdi2_1): Remove.
5149         (fistdi2): Ditto.
5150         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
5151         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
5152         (*fist<mode>2_1): Remove.
5153         (fist<mode>2): Ditto.
5154         (fist<mode>2_with_temp): Remove insn pattern and corresponding
5155         splitters.
5156         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
5157         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
5158         (fistdi2_<rounding>): Change operand 0 predicate to
5159         nonimmediate_operand.
5160         (fistdi2_<rounding>_with_temp): Remove insn pattern
5161         and corresponding splitters.
5162         (fist<mode>2_<rounding>): Change operand 0 predicate to
5163         nonimmediate_operand.
5164         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
5165         and corresponding splitters.
5167         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
5169 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5171         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
5172         the init value.
5174 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5176         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
5177         early gimple folding of vec_splat().
5178         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
5179         * gimple-fold.h: Add an extern define for tree_vec_extract().
5181 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5183         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
5184         wrappers around TREE_TYPE comparisons.
5186 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5188         PR target/80080
5189         * config/s390/predicates.md: Add nonsym_memory_operand.
5190         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
5191         contains a SYMBOL_REF, load it into an intermediate pseudo.
5192         (s390_emit_compare_and_swap): Legitimize operand.
5193         * config/s390/s390.md: Use the new nonsym_memory_operand
5194         with UNSPECV_CAS patterns.
5196 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5198         PR target/80080
5199         * config/s390/s390-passes.def: New file.
5200         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
5201         declaration.
5202         (make_pass_s390_early_mach): Add declaration.
5203         * config/s390/s390.c (make_pass_s390_early_mach):
5204         (s390_option_override): Remove dynamic registration.
5205         * config/s390/t-s390: Add s390-passes.def.
5207 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5209         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5210         Remove UNSPEC_LTREL_BASE check.
5211         (annotate_constant_pool_refs): Likewise.
5212         (find_constant_pool_ref): Likewise.
5213         (find_ltrel_base): Removed.
5214         (replace_ltrel_base): Removed.
5215         (s390_mainpool_finish): Remove replace_ltrel_base call.
5216         (s390_chunkify_start): Remove pending LTREL_BASE logic.
5217         (s390_chunkify_finish): Remove replace_ltrel_base call.
5218         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
5220 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
5222         PR target/86779
5223         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5224         to speculation_safe_value_not_needed.
5226 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
5227             Bernd Schmidt  <bernds_cb1@t-online.de>
5229         * config/nvptx/nvptx-opts.h: New file.
5230         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
5231         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
5232         (ASM_SPEC): Define.
5233         (TARGET_SM35): New macro.
5234         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
5235         correct predicate.
5236         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
5237         values.
5238         (misa=): New option.
5239         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
5241 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
5243         * config/i386/i386.md (truncdfsf2): Remove expander.
5244         (truncdfsf2_with_temp): Ditto.
5245         (truncxf<mode>2): Ditto.
5246         (*truncdfsf_fast_mixed): Remove insn pattern.
5247         (*truncdfsf_fast_i387): Ditto.
5248         (*truncdfsf_mixed): Ditto.
5249         (*truncdfsf_i387): Ditto.
5250         (*truncdfsf2_i387_1): Ditto.
5251         (*truncxfsf2_mixed): Ditto.
5252         (*truncxfdf2_mixed): Ditto.
5253         (*truncxf<mode>2_i387_noop): Ditto. Update callers
5254         to call gen_truncxf<mode>2 instead.
5255         (*truncxf<mode>2_i387): Remove.
5256         (reg->reg splitters): Remove splitter pattern.
5257         (reg->mem splitters): Ditto.
5259         (truncdfsf2): New insn pattern.
5260         (truncxf<mode>2): Ditto.
5262 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5264         * tree-ssa-math-opts.c (is_mult_by): New function.
5265         (is_square_of): Use the above.
5266         (optimize_recip_sqrt): New function.
5267         (pass_cse_reciprocals::execute): Use the above.
5269 2018-09-05  Richard Biener  <rguenther@suse.de>
5271         PR bootstrap/87134
5272         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
5273         to zero-init the emplaced vec.
5275 2018-09-05  Martin Liska  <mliska@suse.cz>
5277         PR tree-optimization/87205
5278         * tree-switch-conversion.c (pass_lower_switch::execute):
5279         Group cases for switch statements.
5281 2018-09-05  Richard Biener  <rguenther@suse.de>
5283         PR tree-optimization/87217
5284         * tree-ssa-sccvn.c (vuse_valueize): New.
5285         (vn_reference_lookup_pieces): Use it.
5286         (vn_reference_lookup): Likewise.
5288 2018-09-05  Nathan Sidwell  <nathan@acm.org>
5290         PR c++/87137
5291         * stor-layout.c (place_field): Scan forwards to check last
5292         bitfield when ms_bitfield_placement is in effect.
5294 2018-09-05  Richard Biener  <rguenther@suse.de>
5296         PR bootstrap/87225
5297         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
5298         return.
5300 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5301             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
5303         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
5304         * config.gcc (extra_objs): Build it.
5305         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
5306         Likewise.
5307         * config/aarch64/aarch64-passes.def
5308         (pass_tag_collision_avoidance): New pass.
5309         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
5310         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
5311         (aarch64_classify_address): Remove static qualifier.
5312         (aarch64_address_info, aarch64_address_type): Move to...
5313         * config/aarch64/aarch64-protos.h: ... here.
5314         (make_pass_tag_collision_avoidance): New function.
5315         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
5316         New tuning flag.
5318 2018-09-05  Martin Liska  <mliska@suse.cz>
5320         * doc/gcov.texi: Update documentation of humar
5321         readable mode.
5322         * gcov.c (format_count): Print one decimal place, it provides
5323         more fine number of situations like '1G' vs. '1.4G'.
5325 2018-09-05  Martin Liska  <mliska@suse.cz>
5327         PR target/87164
5328         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
5329         * optc-gen.awk: Allow 'Var' for Deprecated options in order
5330         to generate a MASK value.
5332 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
5334         PR debug/86593
5335         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
5336         if frame pointer isn't used.
5337         (compute_frame_pointer_to_fb_displacement): Likewise.
5339 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
5341         PR target/87198
5342         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
5343         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
5344         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
5345         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
5346         and OPTION_MASK_ISA_XSAVEC_UNSET.
5348 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
5350         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
5351         XOR operations in NAND case.
5353 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5355         * wide-int-range.cc (wide_int_range_convert): New.
5356         * wide-int-range.h (wide_int_range_convert): New.
5357         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
5358         code into wide_int_range_convert.
5359         (extract_range_into_wide_ints): Do not munge anti range constants
5360         into the entire domain.  Just return the range back.
5362 2018-09-04  Martin Liska  <mliska@suse.cz>
5364         * genmatch.c (output_line_directive): Add new argument
5365         fnargs.
5366         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
5368 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5370         * doc/invoke.texi (Option Summary): Add whitespace.
5372         * doc/invoke.texi (Option Summary): Add -Waligned-new.
5374 2018-09-04  Richard Biener  <rguenther@suse.de>
5376         PR tree-optimization/87211
5377         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
5378         backedge value we're supposed to treat as VARYING also number
5379         the PHI to VARYING in case it got a different value-number already.
5381 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5383         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
5384         (extract_range_from_binary_expr_1): Do not call
5385         vrp_can_optimize_bit_op.
5386         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
5387         static.
5388         (wide_int_range_get_mask_and_bounds): New.
5389         (wide_int_range_optimize_bit_op): New.
5390         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
5391         (wide_int_range_bit_and): Same.
5392         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
5393         (wide_int_range_optimize_bit_op): New.
5394         (wide_int_range_get_mask_and_bounds): New.
5396 2018-09-04  Richard Biener  <rguenther@suse.de>
5398         PR tree-optimization/87176
5399         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
5400         variable.  When value-numbering a virtual PHI node make sure
5401         to not value-number to the backedge value.
5403 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5405         * doc/extend.texi (Long Long, Hex Floats): Document support for
5406         long long and hex floats in more recent versions of ISO C++.
5408 2018-09-03  Richard Biener  <rguenther@suse.de>
5410         PR tree-optimization/87177
5411         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
5412         cleanup.
5414 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5416         * bb-reorder.c (edge_order): Convert to C-qsort-style
5417         tri-state comparator.
5418         (reorder_basic_blocks_simple): Change std::stable_sort to
5419         gcc_stablesort.
5421 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5423         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
5424         tri-state comparator.
5425         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
5427 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5429         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
5430         (mergesort): ... here as maximum count for using netsort.
5431         (gcc_qsort): Set nlim to 3 if stable sort is requested.
5432         (gcc_stablesort): New.
5433         * system.h (gcc_stablesort): Declare.
5435 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5437         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
5438         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
5439         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
5441 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
5443         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
5444         lxsdx and stxsdx alternatives.
5445         (*mov<mode>_hardfloat64): Ditto.
5446         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
5448 2018-09-03  Richard Biener  <rguenther@suse.de>
5450         PR tree-optimization/87200
5451         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
5452         simplify result.
5454 2018-09-03  Martin Liska  <mliska@suse.cz>
5456         PR tree-optimization/87201
5457         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5458         Fix parenthesis in an expression.
5460 2018-09-03  Richard Biener  <rguenther@suse.de>
5462         PR tree-optimization/87197
5463         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
5464         visited.  CSE the VN_INFO hashtable lookup.
5466         PR tree-optimization/87169
5467         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
5468         iterating make sure there's no extra backedges from irreducible
5469         regions feeding the header.  Mark the destination block
5470         executable.
5472 2018-09-03  Martin Liska  <mliska@suse.cz>
5474         PR driver/83193
5475         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
5476         * common/common-targhooks.c (default_get_valid_option_values):
5477         New function.
5478         * common/common-targhooks.h (default_get_valid_option_values):
5479         Likewise.
5480         * common/config/i386/i386-common.c: Move processor_target_table
5481         from i386.c.
5482         (ix86_get_valid_option_values): New function.
5483         (TARGET_GET_VALID_OPTION_VALUES): New macro.
5484         * config/i386/i386.c (struct ptt): Move to i386-common.c.
5485         (PTA_*): Move all defined masks into i386-common.c.
5486         (ix86_function_specific_restore): Use new processor_cost_table.
5487         * config/i386/i386.h (struct ptt): Moved from i386.c.
5488         (struct pta): Likewise.
5489         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
5490         * doc/tm.texi.in: Likewise.
5491         * opt-suggestions.c (option_proposer::suggest_option):
5492         Pass prefix to build_option_suggestions.
5493         (option_proposer::get_completions): Likewise.
5494         (option_proposer::build_option_suggestions): Use the new target
5495         hook.
5496         * opts.c (struct option_help_tuple): New struct.
5497         (print_filtered_help): Use the new target hook.
5499 2018-09-03  Martin Liska  <mliska@suse.cz>
5501         PR middle-end/59521
5502         * predict.c (set_even_probabilities): Add likely_edges
5503         argument and handle cases where we have precisely one
5504         likely edge.
5505         (combine_predictions_for_bb): Catch also likely_edges.
5506         (tree_predict_by_opcode): Handle gswitch statements.
5507         * tree-cfg.h (find_case_label_for_value): New declaration.
5508         (find_taken_edge_switch_expr): Likewise.
5509         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5510         Find pivot in decision tree based on probabily, not by number of
5511         nodes.
5513 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
5515         * doc/standards.texi (Standards): Update Objective-C reference.
5517 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5519         * doc/install.texi (Prerequisites): Update link for MPC.
5521 2018-09-01  Michael Matz  <matz@suse.de>
5523         PR tree-optimization/87074
5524         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5525         PHIs for outer-loop uses.
5527 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5529         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5530         * doc/invoke.texi (C Dialect Options): Ditto.
5532 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5534         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5536 2018-08-31  Richard Biener  <rguenther@suse.de>
5538         PR tree-optimization/87168
5539         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5540         (rpo_elim::eliminate_avail): When OP was not visited it must
5541         be available.
5543 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
5545         * tree-vrp.c (copy_value_range): Convert param "from" from
5546         "value_range *" to "const value_range *".
5547         (range_is_null): Likewise for param "vr".
5548         (range_int_cst_p): Likewise.
5549         (range_int_cst_singleton_p): Likewise.
5550         (symbolic_range_p): Likewise.
5551         (value_ranges_intersect_p): Likewise for both params.
5552         (value_range_nonnegative_p): Likewise for param "vr".
5553         (value_range_constant_singleton): Likewise.
5554         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5555         (extract_range_into_wide_ints): Likewise for param "vr".
5556         (extract_range_from_multiplicative_op): Likewise for params "vr0"
5557         and "vr1".
5558         (vrp_can_optimize_bit_op): Likewise.
5559         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5560         "vr1_".
5561         (extract_range_from_unary_expr): Likewise.
5562         (debug_value_range): Likewise for param "vr".
5563         (value_range::dump): Add "const" qualifier.
5564         (vrp_prop::check_array_ref): Convert local "vr" from
5565         "value_range *" to "const value_range *".
5566         (vrp_prop::check_mem_ref): Likewise.
5567         (vrp_prop::visit_stmt): Likewise for local "old_vr".
5568         (vrp_intersect_ranges_1): Likewise for param "vr_1".
5569         (vrp_intersect_ranges): Likewise.
5570         (simplify_stmt_for_jump_threading): Likewise for local "vr".
5571         (vrp_prop::vrp_finalize): Likewise.
5572         * tree-vrp.h (value_range::dump): Add "const" qualifier.
5573         (vrp_intersect_ranges): Add "const" qualifier to params as above.
5574         (extract_range_from_unary_expr): Likewise.
5575         (value_range_constant_singleton): Likewise.
5576         (symbolic_range_p): Likewise.
5577         (copy_value_range): Likewise.
5578         (extract_range_from_binary_expr_1): Likewise.
5579         (range_int_cst_p): Likewise.
5580         (vrp_set_zero_nonzero_bits): Likewise.
5581         (range_int_cst_singleton_p): Likewise.
5583 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
5585         * config/aarch64/arm_neon.h (vabsd_s64): New.
5586         (vnegd_s64): Likewise.
5588 2018-08-31  Martin Jambor  <mjambor@suse.cz>
5590         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5592 2018-08-31  Martin Liska  <mliska@suse.cz>
5594         * ipa-icf.c (sem_item::add_type): Use
5595         sem_item::m_type_hash_cache.
5596         * ipa-icf.h: Move the cache from sem_item_optimizer
5597         to sem_item.
5599 2018-08-31  Nathan Sidwell  <nathan@acm.org>
5601         * doc/extend.texi (Backwards Compatibility): Remove implicit
5602         extern C leeway of () being (...).
5604 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5606         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5608 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
5610         PR target/86684
5611         PR target/87149
5612         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5614 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
5616         PR middle-end/87138
5617         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5618         gen_int_mode.  Formatting fixes.
5620 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
5622         * target.def (custom_function_descriptors): Improve documentation.
5623         * doc/tm.texi.in (Trampolines): Expand discussion of function
5624         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5625         beginning of the section.
5626         * doc/tm.texi: Regenerated.
5628 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
5630         * cfg.h (class auto_edge_flag): Spell out the template-id of the
5631         base class in the initializer list.  This is a workaround for
5632         building with older GCC.
5633         (class auto_bb_flag): Likewise.
5635 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5637         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5638         (altivec_vcmpequ<VI_char>_p): Remove star.
5639         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5640         vector load modes.
5641         (expand_strncmp_vec_sequence): New function.
5642         (emit_final_str_compare_vec): New function.
5643         (expand_strn_compare): Add support for vector strncmp.
5644         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5645         length specification to bytes.
5646         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5647         (vcmpnezb_p): New pattern.
5648         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5649         for option -mstring-compare-inline-limit.
5651 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
5653         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5654         (PTA_SKYLAKE): Add PTA_AES.
5655         (PTA_GOLDMONT): Likewise.
5657 2018-08-29  Jan Hubicka  <jh@suse.cz>
5659         PR lto/86517
5660         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5661         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5663 2018-08-29  Jan Hubicka  <jh@suse.cz>
5665         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5666         TYPE_STUB_DECL.
5667         (hash_tree): Do not visit TYPE_STUB_DECL.
5668         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5669         stream TYPE_STUB_DECL.
5670         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5671         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5672         after free_lang_data.
5673         (type_in_anonymous_namespace_p): Likewise.
5675 2018-08-29  Jan Hubicka  <jh@suse.cz>
5677         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5678         comment that it has to be even number.
5679         (class sreal): Change m_sig type to int32_t.
5680         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5681         int64_t for temporary calculations.
5682         (sreal_verify_basics): Drop one bit from minimum and maximum.
5684 2018-08-30  Richard Biener  <rguenther@suse.de>
5686         PR tree-optimization/87147
5687         * tree-ssa-sccvn.c (SSA_VISITED): New function.
5688         (visit_phi): When the degenerate result is from the backedge and
5689         we didn't visit its definition yet drop to VARYING.
5690         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5692 2018-08-29  Jan Hubicka  <jh@suse.cz>
5694         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5695         DECL_VINDEX.
5696         (hash_tree): Likewise.
5698 2018-08-29  Jan Hubicka  <jh@suse.cz>
5700         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5701         and TYPE_NEXT_REF_TO.
5703 2018-08-29  Jan Hubicka  <jh@suse.cz>
5705         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5706         comment that it has to be even number.
5707         (class sreal): Change m_sig type to int32_t.
5708         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5709         int64_t for temporary calculations.
5710         (sreal_verify_basics): Drop one bit from minimum and maximum.
5712 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
5714         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5716 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
5718         PR middle-end/86995
5719         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5720         if to_add is negative.
5722 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5724         PR middle-end/87053
5725         * builtins.c (c_strlen): Improve range checks.
5727 2018-08-29  Martin Sebor  <msebor@redhat.com>
5728             Jeff Law  <law@redhat.com>
5730         PR tree-optimization/86714
5731         PR tree-optimization/86711
5732         * builtins.c (c_strlen): Add arguments to call to string_constant.
5733         * expr.c (string_constant): Add argument.  Detect missing nul
5734         terminator and outermost declaration it's missing in.
5735         * expr.h (string_constant): Add argument.
5736         * fold-const.c (read_from_constant_string): Add arguments to call to
5737         string_constant.
5738         (c_getstr): Likewise.
5739         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5740         to string_constant.
5741         * tree-ssa-strlen.c (get_stridx): Likewise.
5743 2018-08-29  Jan Hubicka  <jh@suse.cz>
5745         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5746         Do not stream DECL_VINDEX.
5747         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5748         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5749         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5751 2018-08-29  Richard Biener  <rguenther@suse.de>
5753         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5754         virtual operands that are not default defs to honor region
5755         boundaries.
5756         (rpo_vn_valueize): Remove ineffective code here.
5758 2018-08-29  Richard Biener  <rguenther@suse.de>
5760         PR tree-optimization/87132
5761         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5762         when skipping defs reachable over backedges.
5764 2018-08-29  Richard Biener  <rguenther@suse.de>
5766         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5767         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5768         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5769         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5770         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5771         (vn_reference_lookup_3): Remove use of const_parms.
5772         (free_rpo_vn): Do not free const_parms.
5773         (do_rpo_vn): Do not call init_const_parms.
5774         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5775         SSA_NAME_POINTS_TO_READONLY_MEMORY.
5776         (call_may_clobber_ref_p_1): Likewise.
5778 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
5780         PR other/86726
5781         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5782         (-O): Ditto.
5783         (-ftree-scev-cprop): Document.
5785 2018-08-29  Jan Hubicka  <jh@suse.cz>
5787         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5788         parameters.
5789         (sreal constructor): Update.
5790         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5791         sreal:operator/): Update.
5793 2018-08-29  Martin Liska  <mliska@suse.cz>
5795         * tree-switch-conversion.c (switch_conversion::expand):
5796         Strenghten assumption about gswitch statements.
5798 2018-08-29  Richard Biener  <rguenther@suse.de>
5800         PR tree-optimization/87117
5801         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5802         re-value-number released SSA VDEFs.
5804 2018-08-29  Richard Biener  <rguenther@suse.de>
5806         PR tree-optimization/87126
5807         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5809 2018-08-28  Jim Wilson  <jimw@sifive.com>
5811         * config/riscv/pic.md: Rewrite.
5812         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5813         invalid address.
5814         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5815         (SOFTF, default_load, softload, softstore): New.
5817 2018-08-28  Jeff Law  <law@redhat.com>
5819         * fold-const.c (fold_binary_loc): Remove recently added assert.
5821 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
5823         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5824         to OP parmeter of generated function.
5826 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
5828         PR tree-optimization/87009
5829         * match.pd: Add boolean optimizations.
5831 2018-08-28  Martin Sebor  <msebor@redhat.com>
5833         PR middle-end/86631
5834         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5835         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5836         (pass_walloca::gate): Use it.
5837         (alloca_call_type): Same.
5838         (pass_walloca::execute): Same.
5839         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5841 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
5843         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5844         GCC_VERSION for usage of "__gcc_dump_printf__" format from
5845         >= 3005 to >= 9000.
5847 2018-08-28  Richard Biener  <rguenther@suse.de>
5849         PR tree-optimization/87124
5850         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5851         constants before looking up avail.
5853 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
5855         PR middle-end/87099
5856         * calls.c (maybe_warn_nonstring_arg): Punt early if
5857         warn_stringop_overflow is zero.  Don't call get_range_strlen
5858         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5859         Swap comparison operands to have constants on rhs.  Only use
5860         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
5861         increment lenrng[0].
5863 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
5865         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5866         use of tree_to_shwi.  Remove duplicated test for the size being
5867         a whole number of bytes.
5869 2018-08-28  Richard Biener  <rguenther@suse.de>
5871         PR tree-optimization/87117
5872         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5873         Handle removed stmt without LHS (GIMPLE_NOP).
5875 2018-08-28  Richard Biener  <rguenther@suse.de>
5877         PR tree-optimization/87117
5878         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
5879         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
5881 2018-08-28  Richard Biener  <rguenther@suse.de>
5883         PR tree-optimization/87117
5884         * tree-ssa-pre.c (compute_avail): Do not make expressions
5885         with predicated values available.
5886         (get_expr_value_id): Assert we do not run into predicated value
5887         expressions.
5889 2018-08-28  Richard Biener  <rguenther@suse.de>
5891         PR tree-optimization/87117
5892         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
5893         get virtual operands.
5894         (get_expr_operands): Handle STRING_CST like other decls.
5896 2018-08-28  Martin Liska  <mliska@suse.cz>
5898         * tree.h: Update documentation of fndecl_built_in_p
5899         functions.
5902 2018-08-27  Jeff Law  <law@redhat.com>
5903         PR tree-optimization/87110
5904         * tree-ssa-dse.c (compute_trims): Handle non-constant
5905         TYPE_SIZE_UNIT.
5907 2018-08-27  Martin Sebor  <msebor@redhat.com>
5909         PR tree-optimization/86914
5910         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
5912 2018-08-27  Martin Sebor  <msebor@redhat.com>
5914         PR tree-optimization/87112
5915         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
5916         the type of the bound argument.
5918 2018-08-27  Jeff Law  <law@redhat.com>
5920         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
5921         type does not have a TYPE_SIZE_UNIT.
5923 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
5925         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
5926         with include of backend.h.
5928 2018-08-27  Richard Biener  <rguenther@suse.de>
5930         PR tree-optimization/86927
5931         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
5932         use const cond reduction code.
5934 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
5936         PR tree-optimization/85758
5937         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
5939 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5941         PR c++/87091
5942         * diagnostic-show-locus.c (class layout_range): Update for
5943         conversion of show_caret_p to a tri-state.
5944         (layout_range::layout_range): Likewise.
5945         (make_range): Likewise.
5946         (layout::maybe_add_location_range): Likewise.
5947         (layout::should_print_annotation_line_p): Don't show annotation
5948         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
5949         (layout::get_state_at_point): Update for conversion of
5950         show_caret_p to a tri-state.  Bail out early for
5951         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
5952         underlining or source colorization.
5953         (gcc_rich_location::add_location_if_nearby): Update for conversion
5954         of show_caret_p to a tri-state.
5955         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
5956         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
5957         Likewise.
5958         (selftest::test_one_liner_labels): Likewise.
5959         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
5960         conversion of show_caret_p to a tri-state.
5961         * pretty-print.c (text_info::set_location): Likewise.
5962         * pretty-print.h (text_info::set_location): Likewise.
5963         * substring-locations.c (format_warning_n_va): Likewise.
5964         * tree-diagnostic.c (default_tree_printer): Likewise.
5965         * tree-pretty-print.c (newline_and_indent): Likewise.
5967 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5969         PR c++/87091
5970         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
5971         line above for line-insertion fix-it hints.
5972         (selftest::test_fixit_insert_containing_newline): Update the
5973         expected results, and add a test with line-numbering enabled.
5975 2018-08-27  Martin Liska  <mliska@suse.cz>
5977         PR sanitizer/86962
5978         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
5979         params with DECL_HAS_VALUE_EXPR_P.
5981 2018-08-27  Martin Liska  <mliska@suse.cz>
5983         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
5984         selected expansion strategy.
5986 2018-08-27  Martin Liska  <mliska@suse.cz>
5988         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
5989         * builtins.c (is_builtin_fn): Likewise.
5990         * attribs.c (diag_attr_exclusions): Use new function
5991         fndecl_built_in_p and remove check for FUNCTION_DECL if
5992         possible.
5993         (builtin_mathfn_code): Likewise.
5994         (fold_builtin_expect): Likewise.
5995         (fold_call_expr): Likewise.
5996         (fold_builtin_call_array): Likewise.
5997         (fold_call_stmt): Likewise.
5998         (set_builtin_user_assembler_name): Likewise.
5999         (is_simple_builtin): Likewise.
6000         * calls.c (gimple_alloca_call_p): Likewise.
6001         (maybe_warn_nonstring_arg): Likewise.
6002         * cfgexpand.c (expand_call_stmt): Likewise.
6003         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
6004         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
6005         (cgraph_node::verify_node): Likewise.
6006         * cgraphclones.c (build_function_decl_skip_args): Likewise.
6007         (cgraph_node::create_clone): Likewise.
6008         * config/arm/arm.c (arm_insert_attributes): Likewise.
6009         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
6010         * dse.c (scan_insn): Likewise.
6011         * expr.c (expand_expr_real_1): Likewise.
6012         * fold-const.c (operand_equal_p): Likewise.
6013         (fold_binary_loc): Likewise.
6014         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
6015         * gimple-low.c (lower_stmt): Likewise.
6016         * gimple-pretty-print.c (dump_gimple_call): Likewise.
6017         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6018         Likewise.
6019         * gimple.c (gimple_build_call_from_tree): Likewise.
6020         (gimple_call_builtin_p): Likewise.
6021         (gimple_call_combined_fn): Likewise.
6022         * gimplify.c (gimplify_call_expr): Likewise.
6023         (gimple_boolify): Likewise.
6024         (gimplify_modify_expr): Likewise.
6025         (gimplify_addr_expr): Likewise.
6026         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
6027         * ipa-cp.c (determine_versionability): Likewise.
6028         * ipa-fnsummary.c (compute_fn_summary): Likewise.
6029         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
6030         * ipa-split.c (visit_bb): Likewise.
6031         (split_function): Likewise.
6032         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6033         * lto-cgraph.c (input_node): Likewise.
6034         * lto-streamer-out.c (write_symbol): Likewise.
6035         * omp-low.c (setjmp_or_longjmp_p): Likewise.
6036         (lower_omp_1): Likewise.
6037         * predict.c (strip_predict_hints): Likewise.
6038         * print-tree.c (print_node): Likewise.
6039         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
6040         * trans-mem.c (is_tm_irrevocable): Likewise.
6041         (is_tm_load): Likewise.
6042         (is_tm_simple_load): Likewise.
6043         (is_tm_store): Likewise.
6044         (is_tm_simple_store): Likewise.
6045         (is_tm_abort): Likewise.
6046         (tm_region_init_1): Likewise.
6047         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
6048         * tree-cfg.c (verify_gimple_call): Likewise.
6049         (move_stmt_r): Likewise.
6050         (stmt_can_terminate_bb_p): Likewise.
6051         * tree-eh.c (lower_eh_constructs_2): Likewise.
6052         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
6053         * tree-inline.c (remap_gimple_stmt): Likewise.
6054         (copy_bb): Likewise.
6055         (estimate_num_insns): Likewise.
6056         (fold_marked_statements): Likewise.
6057         * tree-sra.c (scan_function): Likewise.
6058         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
6059         (optimize_stack_restore): Likewise.
6060         (pass_fold_builtins::execute): Likewise.
6061         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
6062         (mark_all_reaching_defs_necessary_1): Likewise.
6063         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
6064         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
6065         (pass_forwprop::execute): Likewise.
6066         * tree-ssa-loop-im.c (stmt_cost): Likewise.
6067         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
6068         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
6069         * tree-ssa-strlen.c (get_string_length): Likewise.
6070         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
6071         (find_func_aliases_for_call): Likewise.
6072         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
6073         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
6074         * tree-tailcall.c (find_tail_calls): Likewise.
6075         * tree.c (need_assembler_name_p): Likewise.
6076         (free_lang_data_in_decl): Likewise.
6077         (get_call_combined_fn): Likewise.
6078         * ubsan.c (is_ubsan_builtin_p): Likewise.
6079         * varasm.c (incorporeal_function_p): Likewise.
6080         * tree.h (DECL_BUILT_IN): Remove and replace with
6081         fndecl_built_in_p.
6082         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
6083         (fndecl_built_in_p): New.
6085 2018-08-27  Martin Liska  <mliska@suse.cz>
6087         PR tree-optimization/86847
6088         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
6089         Dump also subtree probability.
6090         (switch_decision_tree::do_jump_if_equal): New function.
6091         (switch_decision_tree::emit_case_nodes): Handle special
6092         situations in balanced tree that can be emitted much simpler.
6093         Fix calculation of probabilities that happen in tree expansion.
6094         * tree-switch-conversion.h (struct cluster): Add
6095         is_single_value_p.
6096         (struct simple_cluster): Likewise.
6097         (struct case_tree_node): Add new function has_child.
6098         (do_jump_if_equal): New.
6100 2018-08-27  Martin Liska  <mliska@suse.cz>
6102         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
6103         Add new argument to bit_test_cluster constructor.
6104         (bit_test_cluster::emit): Set bits really number of values
6105         handlel by a test.
6106         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
6107         probability argument.
6108         * tree-switch-conversion.h (struct bit_test_cluster):
6109         Add m_handles_entire_switch.
6111 2018-08-27  Martin Liska  <mliska@suse.cz>
6113         PR tree-optimization/86702
6114         * tree-switch-conversion.c (jump_table_cluster::emit):
6115         Make probabilities even for values in jump table
6116         according to number of cases handled.
6117         (switch_decision_tree::compute_cases_per_edge): Pass
6118         argument to reset_out_edges_aux function.
6119         (switch_decision_tree::analyze_switch_statement): Likewise.
6120         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
6121         Make it static.
6123 2018-08-27  Martin Liska  <mliska@suse.cz>
6125         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
6126         cfun argument explicitly.
6127         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
6128         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
6129         function gimple_switch_default_bb.
6130         (convert_switch_statements):
6131         (expand_builtins):
6132         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
6133         * stmt.c (label_to_block_fn): Use label_to_block and pass
6134         cfun argument explicitly and use gimple_switch_label_bb.
6135         (expand_case): Likewise.
6136         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
6137         cfun argument explicitly. Likewise.
6138         (make_edges_bb): Likewise.
6139         (make_cond_expr_edges): Likewise.
6140         (get_cases_for_edge): Likewise.
6141         (make_gimple_switch_edges): Likewise.
6142         (label_to_block_fn): Likewise.
6143         (label_to_block): Likewise.
6144         (make_goto_expr_edges): Likewise.
6145         (make_gimple_asm_edges): Likewise.
6146         (main_block_label): Likewise.
6147         (group_case_labels_stmt): Likewise.
6148         (find_taken_edge_computed_goto): Likewise.
6149         (find_taken_edge_switch_expr): Likewise.
6150         (gimple_verify_flow_info): Likewise.
6151         (gimple_redirect_edge_and_branch): Likewise.
6152         (gimple_switch_label_bb): New function.
6153         (gimple_switch_default_bb): Likewise.
6154         (gimple_switch_edge): Likewise.
6155         (gimple_switch_default_edge): Likewise.
6156         * tree-cfg.h (label_to_block_fn): Remove and replace ...
6157         (label_to_block): ... with this.
6158         (gimple_switch_label_bb): New.
6159         (gimple_switch_default_bb): Likewise.
6160         (gimple_switch_edge): Likewise.
6161         (gimple_switch_default_edge): Likewise.
6162         * tree-cfgcleanup.c (convert_single_case_switch): Use
6163         new gimple functions and pass new argument to label_to_block.
6164         (cleanup_control_flow_bb):
6165         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
6166         cfun argument explicitly.
6167         (make_eh_edges): Likewise.
6168         (redirect_eh_dispatch_edge): Likewise.
6169         (lower_resx): Likewise.
6170         (lower_eh_dispatch): Likewise.
6171         (maybe_remove_unreachable_handlers): Likewise.
6172         (unsplit_eh): Likewise.
6173         (cleanup_empty_eh): Likewise.
6174         (verify_eh_edges): Likewise.
6175         (verify_eh_dispatch_edge): Likewise.
6176         * tree-ssa-dom.c (record_edge_info): Likewise.
6177         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
6178         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
6179         (thread_through_normal_block): Likewise.
6180         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
6181         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
6182         * tree-switch-conversion.c (switch_conversion::collect): Use new
6183         gimple functions.
6184         (switch_conversion::check_final_bb): Likewise.
6185         (switch_conversion::gather_default_values): Pass new argument
6186         to label_to_block.
6187         (switch_conversion::build_constructors): Likewise.
6188         (switch_decision_tree::compute_cases_per_edge): Use new
6189         gimple_switch_edge function.
6190         (switch_decision_tree::analyze_switch_statement): Pass new argument
6191         to label_to_block.
6192         (switch_decision_tree::try_switch_expansion): Use
6193         gimple_switch_default_edge.
6194         * tree-vrp.c (find_switch_asserts): Pass new argument
6195         to label_to_block.
6196         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
6197         (vr_values::simplify_switch_using_ranges): Likewise.
6199 2018-08-27  Richard Biener  <rguenther@suse.de>
6201         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
6202         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
6204         * tree-ssa-sccvn.h (struct vn_pval): New structure.
6205         (struct vn_nary_op_s): Add unwind_to member.  Add
6206         predicated_values flag and put result into a union together
6207         with a linked list of vn_pval.
6208         (struct vn_ssa_aux): Add name member to make maintaining
6209         a map of SSA name to vn_ssa_aux possible.  Remove no longer
6210         needed info, dfsnum, low, visited, on_sccstack, use_processed
6211         and range_info_anti_range_p members.
6212         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
6213         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
6214         New functions.
6215         (vn_valueize): New global.
6216         (vn_context_bb): Likewise.
6217         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
6218         VN_INFO_PTR_INFO): Remove.
6219         * tree-ssa-sccvn.c: ... (rewrite)
6220         (pass_fre::execute): For -O2+ initialize loops and run
6221         RPO VN in optimistic mode (iterating).  For -O1 and -Og
6222         run RPO VN in non-optimistic mode.
6223         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
6224         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
6225         * doc/invoke.texi (sccvn-max-scc-size): Remove.
6226         (rpo-vn-max-loop-depth): Document.
6227         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
6228         when valuezing the VUSE signals we walked out of the region.
6229         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
6230         (phi_translate): Set VN context block to use for availability
6231         lookup.
6232         (compute_avail): Likewise.
6233         (pre_valueize): New function.
6234         (pass_pre::execute): Adjust to the RPO VN API.
6236         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
6237         (propagate_constants_for_unrolling): Remove.
6238         (tree_unroll_loops_completely): Perform value-numbering
6239         on the unrolled bodies loop parent.
6241 2018-08-27  Richard Biener  <rguenther@suse.de>
6243         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
6244         for partial antic compute.
6246 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
6248         PR rtl-optimization/87065
6249         * combine.c (simplify_if_then_else): Formatting fix.
6250         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
6251         check.
6252         (known_cond): Don't return const_true_rtx for vector modes.  Use
6253         CONST0_RTX instead of const0_rtx.  Formatting fixes.
6255 2018-08-27  Martin Liska  <mliska@suse.cz>
6257         PR gcov-profile/87069
6258         * gcov.c (process_file): Record files already processed
6259         and warn about a file being processed multiple times.
6261 2018-08-27  Martin Liska  <mliska@suse.cz>
6263         PR driver/83193
6264         * config/aarch64/aarch64.c (aarch64_override_options_internal):
6265         Set default values for x_aarch64_*_string strings.
6266         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
6267         prefix.  For -mabi do not print '=ABI' in help and use
6268         <option_value> format for -msve-vector-bits and -moverride
6269         options.
6271 2018-08-26  Jeff Law  <law@redhat.com>
6273         * config/mips/frame-header-opt.c: Include "backend.h" rather than
6274         "cfg.h"
6276 2018-08-26  Marek Polacek  <polacek@redhat.com>
6278         PR c++/87029, Implement -Wredundant-move.
6279         * doc/invoke.texi: Document -Wredundant-move.
6281 2018-08-25  Martin Sebor  <msebor@redhat.com>
6283         PR tree-optimization/87059
6284         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
6285         to the same type as the other.
6286         * fold-const.c (fold_binary_loc): Assert expectation.
6288 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
6290         * config/darwin.c (machopic_legitimize_pic_address): Clean up
6291         extraneous parentheses, dead code section and formatting.
6293 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
6295         PR c++/87091
6296         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
6297         wide enough for jumps in the line-numbering to be visible.
6298         (layout::print_gap_in_line_numbering): New member function.
6299         (layout::calculate_line_spans): When using line numbering, merge
6300         line spans that are only 1 line apart.
6301         (diagnostic_show_locus): When printing line numbers, show gaps in
6302         line numbering directly, rather than printing headers.
6303         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
6304         line-numbering with multiple line spans.
6305         (selftest::test_fixit_insert_containing_newline_2): Add test of
6306         line-numbering, in which the spans are close enough to be merged.
6308 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6310         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
6311         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
6312         * tree-vrp.c (range_is_nonnull): Remove.
6313         (range_includes_zero_p): Accept value_range instead of min/max.
6314         (extract_range_from_binary_expr_1): Do not early bail on
6315         POINTER_PLUS_EXPR.
6316         Use range_includes_zero_p instead of range_is_nonnull.
6317         (extract_range_from_unary_expr): Use range_includes_zero_p instead
6318         of range_is_nonnull.
6319         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
6320         special case VR_ANTI_RANGE.
6321         (vrp_finalize): Same.
6322         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
6323         instead of min/max.
6324         (range_is_nonnull): Remove.
6325         * vr-values.c (vrp_stmt_computes_nonzero): Use
6326         range_includes_zero_p instead of range_is_nonnull.
6327         (extract_range_basic): Pass value_range to range_includes_zero_p
6328         instead of range_is_nonnull.
6330 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
6332         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
6333         * emit-rtl.h (rtl_data): Remove return_bnd.
6334         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
6335         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
6336         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
6337         (POINTER_BOUNDS_MODE): Remove definition.
6338         (make_pointer_bounds_mode): Remove.
6339         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
6340         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
6341         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
6342         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
6343         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
6344         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
6345         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
6347         * config/i386/i386-modes.def (BND32, BND64): Remove.
6348         * config/i386/i386.c (dbx_register_map): Remove bound registers.
6349         (dbx64_register_map): Ditto.
6350         (svr4_dbx_register_map): Ditto.
6351         (indirect_thunk_bnd_needed): Remove.
6352         (indirect_thunks_bnd_used): Ditto.
6353         (indirect_return_bnd_needed): Ditto.
6354         (indirect_return_via_cx_bnd): Ditto.
6355         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
6356         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
6357         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
6358         (output_indirect_thunk_function): Remove handling of
6359         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
6360         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
6361         (ix86_save_reg): Remove handling of crtl->return_bnd.
6362         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
6363         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
6364         and UNSPEC_BNDLX_ADDR.
6365         (ix86_output_indirect_branch_via_reg): Remove handling of
6366         indirect_thunk_prefix_bnd.
6367         (ix86_output_indirect_branch_via_push): Ditto.
6368         (ix86_output_function_return): Ditto.
6369         (ix86_output_indirect_function_return): Ditto.
6370         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
6371         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
6372         (CALL_USED_REGISTERS): Ditto.
6373         (REG_ALLOC_ORDER): Update for removal of bound registers.
6374         (HI_REGISTER_NAMES): Ditto.
6375         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
6376         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
6377         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
6378         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
6379         (FIRST_PSEUDO_REG): Update.
6380         (BND): Remove mode iterator.
6381         * config/i386/predicates.md (bnd_mem_operator): Remove.
6383 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6385         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
6386         vectors.
6388 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6390         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
6391         the case in which the permute needs only a single element and
6392         repeats for every vector of the result.  Extend that case to
6393         handle variable-length vectors.
6394         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6396 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
6398         PR debug/79342
6399         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
6400         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
6402 2018-08-24  Richard Biener  <rguenther@suse.de>
6404         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
6405         bb_flags_allocated members.
6406         (auto_flag): New RAII class for allocating flags.
6407         (auto_edge_flag): New RAII class for allocating edge flags.
6408         (auto_bb_flag): New RAII class for allocating bb flags.
6409         * cfgloop.c (verify_loop_structure): Allocate temporary edge
6410         flag dynamically.
6411         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
6412         in favor of temporarily allocated BB flag.
6413         * hsa-brig.c: Re-order includes.
6414         * hsa-dump.c: Likewise.
6415         * hsa-regalloc.c: Likewise.
6416         * print-rtl.c: Likewise.
6417         * profile-count.c: Likewise.
6419 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
6421         PR target/86989
6422         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
6423         the TOC register.
6425 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6427         PR 87073/bootstrap
6428         * wide-int-range.cc (wide_int_range_div): Do not ignore result
6429         from wide_int_range_multiplicative_op.
6431 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6433         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
6434         "permutaion".
6436 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
6438         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
6439         to 'expanded'.
6441 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
6443         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
6444         full GENERIC expression used for replacement.
6446 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
6448         * tree-vrp.c (abs_extent_range): Remove.
6449         (extract_range_into_wide_ints): Pass wide ints by reference.
6450         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
6451         Pass wide ints by reference in all calls to
6452         extract_range_into_wide_ints.
6453         * wide-int-range.cc (wide_int_range_div): New.
6454         * wide-int-range.h (wide_int_range_div): New.
6455         (wide_int_range_includes_zero_p): New.
6456         (wide_int_range_zero_p): New.
6458 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
6460         * config/aarch64/aarch64.md (arches): New enum.
6461         (arch): New enum attr.
6462         (arch_enabled): New attr.
6463         (enabled): Now uses arch_enabled only.
6464         (simd, sve, fp16): Removed attribute.
6465         (fp): Attr now defined in terms of 'arch'.
6466         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
6467         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
6468         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
6469         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
6470         attributes into 'arch'.
6471         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
6472         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
6473         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
6474         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
6475         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
6476         'simd' attribute into 'arch'.
6477         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
6478         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
6479         Convert use of 'fp' attribute to 'arch'.
6480         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
6481         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
6482         into 'arch'.
6483         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
6484         (different modes) Merge 'fp' and 'simd' into 'arch'.
6485         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
6486         'simd' into 'arch'.
6488 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
6490         PR rtl-optimization/87026
6491         * expmed.c (canonicalize_comparison): If we can no longer create
6492         pseudoregisters, don't.
6494 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
6496         PR target/86951
6497         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6498         prototype.
6499         * config/arm/arm.c (speculation_barrier_libfunc): New static
6500         variable.
6501         (arm_init_libfuncs): Initialize it.
6502         (arm_emit_speculation_barrier): New function.
6503         * config/arm/arm.md (speculation_barrier): Call
6504         arm_emit_speculation_barrier for architectures that do not have
6505         DSB or ISB.
6506         (speculation_barrier_insn): Only match on Armv7 or later.
6508 2018-08-23  Richard Biener  <rguenther@suse.de>
6510         PR middle-end/87024
6511         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6512         calls.
6514 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6516         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6517         of single-vector TBLs.
6518         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6519         one input is given.
6521 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6523         PR target/85910
6524         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6525         aarch64_evpc_tbl guard.
6527 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6529         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6530         behaviour.
6532 2018-08-22  Martin Sebor  <msebor@redhat.com>
6534         PR middle-end/87052
6535         * tree-pretty-print.c (pretty_print_string): Add argument.
6536         (dump_generic_node): Call to pretty_print_string with string size.
6538 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
6540         PR rtl-optimization/86771
6541         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6542         of two SETs into those two SETs, one to be placed at i2, if that SETs
6543         destination is modified between i2 and i3.
6545 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6547         PR tree-optimization/86725
6548         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6549         function.
6550         (vect_analyze_scalar_cycles_1): Check it.
6552 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6554         PR tree-optimization/86725
6555         * tree-vect-loop.c (vect_is_simple_reduction): When treating
6556         an outer loop phi as a double reduction, make sure that the
6557         single user of the phi result is an inner loop phi.
6559 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6561         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6562         grouped stores with gaps to a strided group.
6564 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6566         * tree-vect-stmts.c (get_group_load_store_type)
6567         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6568         first statement in a group.
6570 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6572         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6573         the sequence used in gcc/gcc.c.
6575 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6577         PR other/704
6578         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6579         building it.
6581 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6583         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6584         Darwin10-specific unwinder-shim.
6585         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6586         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6587         New to cater for Darwin10 Rosetta.
6589 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6591         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6592         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6594 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6596         PR bootstrap/81033
6597         PR target/81733
6598         PR target/52795
6599         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6600         (dwarf2out_switch_text_section): Generate a local label for the
6601         second function sub-section and apply it as the second FDE start
6602         label.
6603         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6604         second sub-section start.
6606 2018-08-22  Richard Biener  <rguenther@suse.de>
6608         PR tree-optimization/86988
6609         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6611 2018-08-22  Richard Biener  <rguenther@suse.de>
6613         PR tree-optimization/86945
6614         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6616 2018-08-22  Alexandre Oliva <oliva@adacore.com>
6618         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6619         a comment about how uses of r2 for .sdata2 come about.
6621 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
6623         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6625 2018-08-21  Marek Polacek  <polacek@redhat.com>
6627         PR c++/86981, Implement -Wpessimizing-move.
6628         * doc/invoke.texi: Document -Wpessimizing-move.
6630 2018-08-21  Jan Hubicka  <jh@suse.cz>
6632         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6633         * tree.h (is_redundant_typedef): Remove.
6634         * dwarf2out.c (is_redundant_typedef): Turn into static function.
6636 2018-08-21  Jan Hubicka  <jh@suse.cz>
6638         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6639         when possible.
6641 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
6643         * expmed.c (extract_low_bits): Reject invalid subregs early.
6645 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6647         PR middle-end/86121
6648         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6649         behaviour.
6651 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6653         * config/vxworks.h: Guard vxworks_asm_out_constructor and
6654         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6655         * config/vxworks.c: Likewise.
6657 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6659         * config/vxworks.c: Set targetm.have_ctors_dtors
6660         if HAVE_INITFINI_ARRAY_SUPPORT.
6661         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6662         if HAVE_INITFINI_ARRAY_SUPPORT.
6664 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6666         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6667         default search path for VxWorks < 7.
6669 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6671         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6672         (get_format_string): Refer to c_getstr.
6674 2018-08-21  Tom de Vries  <tdevries@suse.de>
6676         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6677         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6678         (debuginfo_early_stop): Declare.
6679         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6680         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6681         (debuginfo_early_stop): New function.
6682         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6683         and debuginfo_early_stop.
6684         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6685         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6686         (general_init): Call debuginfo_early_init.
6687         (finalize): Call debuginfo_fini.
6688         (do_compile): Call debuginfo_init.
6689         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6690         -fdump-early-debug.
6691         (@item -fdump-debug, @item -fdump-earlydebug): Add.
6693 2018-08-21  Tom de Vries  <tdevries@suse.de>
6695         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6696         flag_dump_noaddr and flag_dump_unnumbered.
6698 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
6700         * wide-int-range.cc (wide_int_range_abs): New.
6701         (wide_int_range_order_set): Rename from wide_int_range_min_max.
6702         * wide-int-range.h (wide_int_range_abs): New.
6703         (wide_int_range_min_max): New.
6704         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6705         case to call wide_int_range_abs.
6706         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6707         (extract_range_from_abs_expr): Delete.
6709 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
6711         PR target/87033
6712         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6713         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6714         for indexed loads.
6716 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6717             Jeff Law <law@redhat.com>
6719         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6720         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6722 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
6724         PR other/84889
6725         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6726         (decl_attributes): Likewise.
6727         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6728         instance.
6729         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6730         * diagnostic-core.h (class auto_diagnostic_group): New class.
6731         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6732         (diagnostic_report_diagnostic): Handle the first diagnostics within
6733         a group.
6734         (emit_diagnostic): Add auto_diagnostic_group instance.
6735         (inform): Likewise.
6736         (inform_n): Likewise.
6737         (warning): Likewise.
6738         (warning_at): Likewise.
6739         (warning_n): Likewise.
6740         (pedwarn): Likewise.
6741         (permerror): Likewise.
6742         (error): Likewise.
6743         (error_n): Likewise.
6744         (error_at): Likewise.
6745         (sorry): Likewise.
6746         (fatal_error): Likewise.
6747         (internal_error): Likewise.
6748         (internal_error_no_backtrace): Likewise.
6749         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6750         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6751         * diagnostic.h (struct diagnostic_context): Add fields
6752         "diagnostic_group_nesting_depth",
6753         "diagnostic_group_emission_count", "begin_group_cb",
6754         "end_group_cb".
6755         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6756         Add auto_diagnostic_group instance(s).
6757         (find_explicit_erroneous_behavior): Likewise.
6758         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6759         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6760         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6761         (gimplify_va_arg_expr): Likewise.
6762         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6763         (HSA_SORRY_AT): Likewise.
6764         * ipa-devirt.c (compare_virtual_tables): Likewise.
6765         (warn_odr): Likewise.
6766         * multiple_target.c (expand_target_clones): Likewise.
6767         * opts-common.c (cmdline_handle_error): Likewise.
6768         * reginfo.c (globalize_reg): Likewise.
6769         * substring-locations.c (format_warning_n_va): Likewise.
6770         * tree-inline.c (expand_call_inline): Likewise.
6771         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6772         * tree-ssa-loop-niter.c
6773         (do_warn_aggressive_loop_optimizations): Likewise.
6774         * tree-ssa-uninit.c (warn_uninit): Likewise.
6775         * tree.c (warn_deprecated_use): Likewise.
6777 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
6779         PR target/87014
6780         * config/i386/i386.md (eh_return): Always update EH return
6781         address in word_mode.
6783 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
6785         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6786         TARGET_SPLIT_COMPLEX_ARG is defined.
6788 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6790         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6792 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6794         PR target/86984
6795         * expr.c (expand_assignment): Assert that bitpos is positive.
6796         (store_field): Likewise
6797         (expand_expr_real_1): Make sure that bitpos is positive.
6798         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6799         integer overflow.
6801 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6803         * Makefile.in (CPP_ID_DATA_H): Delete.
6804         (CPP_INTERNAL_H): Don't add it.
6805         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6806         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6808 2018-08-20  Richard Biener  <rguenther@suse.de>
6810         PR tree-optimization/78655
6811         * tree-vrp.c (extract_range_from_binary_expr_1): Make
6812         pointer + offset nonnull if either operand is nonnull work.
6814 2018-08-20  Tom de Vries  <tdevries@suse.de>
6816         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6817         unless the referenced die describes the added property using
6818         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
6819         Otherwise, add a DW_AT_location to the referenced die.
6821 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
6823         PR target/86994
6824         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6825         register_operand when calling ix86_set_reg_reg_cost.
6826         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6827         Set *total to 0 for operands that satisfy x86_64_immediate_operand
6828         predicate and to 1 otherwise.
6830 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
6832         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6833         emit a diagnostic that it is not supported and reset the option.
6834         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6835         supported and consume the option.  (ASM_FINAL_SPEC): New.
6837 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6839         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6840         a sentence.
6842 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
6844         C-SKY port: Documentation
6846         * doc/extend.texi (C-SKY Function Attributes): New section.
6847         * doc/invoke.texi (Option Summary): Add C-SKY options.
6848         (C-SKY Options): New section.
6849         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6851 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6852             Huibin Wang  <huibin_wang@c-sky.com>
6853             Sandra Loosemore  <sandra@codesourcery.com>
6854             Chung-Lin Tang  <cltang@codesourcery.com>
6856         C-SKY port: Backend implementation
6858         * config/csky/*: New.
6859         * common/config/csky/*: New.
6861 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6862             Huibin Wang  <huibin_wang@c-sky.com>
6863             Sandra Loosemore  <sandra@codesourcery.com>
6864             Chung-Lin Tang  <cltang@codesourcery.com>
6865             Andrew Jenner  <andrew@codesourcery.com>
6867         C-SKY port: Configury
6869         * config.gcc (csky-*-*): New.
6870         * configure.ac: Add csky to targets for dwarf2 debug_line support.
6871         * configure: Regenerated.
6873 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
6875         * dump-context.h: Include "dumpfile.h".
6876         (dump_context::dump_printf_va): Convert final param from va_list
6877         to va_list *.  Convert from ATTRIBUTE_PRINTF to
6878         ATTRIBUTE_GCC_DUMP_PRINTF.
6879         (dump_context::dump_printf_loc_va): Likewise.
6880         * dumpfile.c: Include "stringpool.h".
6881         (make_item_for_dump_printf_va): Delete.
6882         (make_item_for_dump_printf): Delete.
6883         (class dump_pretty_printer): New class.
6884         (dump_pretty_printer::dump_pretty_printer): New ctor.
6885         (dump_pretty_printer::emit_items): New member function.
6886         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
6887         function.
6888         (dump_pretty_printer::emit_item): New member function.
6889         (dump_pretty_printer::stash_item): New member function.
6890         (dump_pretty_printer::format_decoder_cb): New member function.
6891         (dump_pretty_printer::decode_format): New member function.
6892         (dump_context::dump_printf_va): Reimplement in terms of
6893         dump_pretty_printer.
6894         (dump_context::dump_printf_loc_va): Convert final param from va_list
6895         to va_list *.
6896         (dump_context::begin_scope): Reimplement call to
6897         make_item_for_dump_printf.
6898         (dump_printf): Update for change to dump_printf_va.
6899         (dump_printf_loc): Likewise.
6900         (selftest::test_capture_of_dump_calls): Convert "stmt" from
6901         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
6902         with %T, %E, and %G.
6903         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
6904         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
6905         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
6906         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
6907         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
6908         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
6909         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
6910         within a dump_printf_loc call to "%wu".
6911         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
6912         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
6913         missing space after "=".
6914         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
6915         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
6916         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
6917         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
6918         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
6919         duplicate "vectorized" from message.
6921 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
6923         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
6924         polyNxK_t element's TYPE_STRING_FLAG.
6926 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6928         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
6929         (they were unnamed before).  Fix comments.
6931 2018-08-17  Nathan Sidwell  <nathan@acm.org>
6933         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
6935 2018-08-17  Richard Biener  <rguenther@suse.de>
6937         PR tree-optimization/86841
6938         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6940 2018-08-17  Martin Liska  <mliska@suse.cz>
6942         * common.opt: Remove Warn, Init and Report for options with
6943         Ignore/Deprecated flag. Warning is done automatically for
6944         Deprecated flags.
6945         * config/i386/i386.opt: Likewise.
6946         * config/ia64/ia64.opt: Likewise.
6947         * config/rs6000/rs6000.opt: Likewise.
6948         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6949         Remove usage of flag_check_pointer_bounds.
6950         * lto-wrapper.c (merge_and_complain): Do not handle
6951         OPT_fcheck_pointer_bounds.
6952         (append_compiler_options): Likewise.
6953         * opt-functions.awk: Do not handle Deprecated.
6954         * optc-gen.awk: Check that Var, Report and Init are not
6955         used for an option with Ignore/Deprecated flag.
6956         * opts-common.c (decode_cmdline_option): Do not report
6957         CL_ERR_DEPRECATED.
6958         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
6959         options.
6960         * opts.h (struct cl_option): Remove cl_deprecated flag.
6961         (CL_ERR_DEPRECATED): Remove error enum value.
6963 2018-08-17  Richard Biener  <rguenther@suse.de>
6965         PR middle-end/86505
6966         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
6967         across a va-arg-pack using call adjust its return value accordingly.
6969 2018-08-16  Martin Sebor  <msebor@redhat.com>
6971         PR tree-optimization/86853
6972         * gimple-ssa-sprintf.c (struct format_result): Rename member.
6973         (struct fmtresult): Add member and initialize it in ctors.
6974         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
6975         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
6976         (format_directive): Set POSUNDER4K when MAYFAIL is set.
6977         (parse_directive): Handle %C same as %c.
6978         (sprintf_dom_walker::compute_format_length): Adjust.
6979         (is_call_safe): Adjust.
6981 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6983         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
6984         for determining how to count the elements.
6985         * builtins.h (c_strlen): Adjust prototype.
6986         * expr.c (string_constant): Add new parameter mem_size.
6987         Set *mem_size appropriately.
6988         * expr.h (string_constant): Adjust protoype.
6989         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
6990         * gimple-fold.h (get_range_strlen): Adjust prototype.
6991         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
6992         (format_string): Call get_string_length with eltsize.
6994 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6996         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
6997         to emit the span, rather than setting it as the prefix.
6999 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
7001         * diagnostic-show-locus.c (layout::start_annotation_line): Add
7002         "margin_char" parameter, defaulting to space.  Use it in place
7003         of pp_space for the initial part of the margin.
7004         (layout::print_leading_fixits): Use '+' when filling the margin
7005         of line-insertion fix-it hints.
7007 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7009         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
7010         Delete.
7012 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7014         * config/rs6000/altivec.md: Don't set length attribute to the default
7015         value.
7016         * config/rs6000/darwin.md: Ditto.
7017         * config/rs6000/dfp.md: Ditto.
7018         * config/rs6000/htm.md: Ditto.
7019         * config/rs6000/rs6000.md: Ditto.
7020         * config/rs6000/sync.md: Ditto.
7021         * config/rs6000/vsx.md: Ditto.
7023 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7025         * config/rs6000/altivec.md: Don't set length attribute to the default
7026         value, for branch instructions.
7027         * config/rs6000/darwin.md: Ditto.
7028         * config/rs6000/rs6000.md: Ditto.
7030 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7032         * config/rs6000/rs6000.md (length): Always define as const_int 4.
7033         (unnamed conditional branch define_insn): Set length to 4 or 8
7034         depending on offset.
7035         (<bd>_<mode>): Similar, for alternative 0.
7036         (<bd>tf_<mode>): Ditto.
7038 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7040         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
7042 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
7044         * doc/rtl.texi: Replace old RTX class names with new names.
7047 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
7049         * expmed.h (canonicalize_comparison): New declaration.
7050         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
7051         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
7052         * optabs.c (prepare_cmp_insn): Likewise.
7053         * rtl.h (unsigned_condition_p): New function which checks if a
7054         comparison operator is unsigned.
7056 2018-08-16  Nathan Sidwell  <nathan@acm.org>
7058         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
7059         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
7061 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7063         PR target/84711
7064         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
7065         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
7066         (mov<mov>): ..this and enable unconditionally.
7068 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7070         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
7072 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
7074         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
7075         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
7076         "Common" with "Target".
7078 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
7080         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
7081         * doc/invoke.texi (mmitigate-rop): Remove.
7082         * config/i386/i386.c: Do not include "regrename.h".
7083         (ix86_rop_should_change_byte_p, reg_encoded_number)
7084         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
7085         Remove.
7086         (ix86_reorg): Remove call to ix86_mitigate_rop.
7087         * config/i386/i386.md (attr "modrm_class"): Remove.
7088         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
7089         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
7090         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
7092 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7094         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
7095         allow folding of mergeh() and mergel() for the float and double types.
7096         (fold_mergehl_helper): Rework to handle building a permute tree
7097         for float vectors.
7099 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
7101         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
7102         for TARGET_SSE.
7104 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
7106         * common.opt (fdiagnostics-show-labels): New option.
7107         * diagnostic-show-locus.c (class layout_range): Add field
7108         "m_label".
7109         (class layout): Add field "m_show_labels_p".
7110         (layout_range::layout_range): Add param "label" and use it to
7111         initialize m_label.
7112         (make_range): Pass in NULL for new "label" param of layout_range's
7113         ctor.
7114         (layout::layout): Initialize m_show_labels_p.
7115         (layout::maybe_add_location_range): Pass in loc_range->m_label
7116         when constructing layout_range instances.
7117         (struct line_label): New struct.
7118         (layout::print_any_labels): New member function.
7119         (layout::print_line): Call it if label-printing is enabled.
7120         (selftest::test_one_liner_labels): New test.
7121         (selftest::test_diagnostic_show_locus_one_liner): Call it.
7122         * diagnostic.c (diagnostic_initialize): Initialize
7123         context->show_labels_p.
7124         * diagnostic.h (struct diagnostic_context): Add field
7125         "show_labels_p".
7126         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7127         -fno-diagnostics-show-labels.
7128         * dwarf2out.c (gen_producer_string): Add
7129         OPT_fdiagnostics_show_labels to the ignored options.
7130         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
7131         param.
7132         (gcc_rich_location::maybe_add_expr): Likewise.
7133         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
7134         label" param, defaulting to NULL.
7135         (gcc_rich_location::add_expr): Add "label" param.
7136         (gcc_rich_location::maybe_add_expr): Likewise.
7137         (class text_range_label): New class.
7138         (class range_label_for_type_mismatch): New class.
7139         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
7140         of format_warning_va.
7141         (fmtwarn_n): Likewise for new params of format_warning_n_va.
7142         * lto-wrapper.c (merge_and_complain): Add
7143         OPT_fdiagnostics_show_labels to the "pick one setting" options.
7144         (append_compiler_options): Likewise to the dropped options.
7145         (append_diag_options): Likewise to the passed-on options.
7146         * opts.c (common_handle_option): Handle the new option.
7147         * selftest-diagnostic.c
7148         (test_diagnostic_context::test_diagnostic_context): Enable
7149         show_labels_p.
7150         * substring-locations.c: Include "gcc-rich-location.h".
7151         (format_warning_n_va): Add "fmt_label" and "param_label" params
7152         and use them as appropriate.
7153         (format_warning_va): Add "fmt_label" and "param_label" params,
7154         passing them on to format_warning_n_va.
7155         (format_warning_at_substring): Likewise.
7156         (format_warning_at_substring_n): Likewise.
7157         * substring-locations.h (format_warning_va): Add "fmt_label" and
7158         "param_label" params.
7159         (format_warning_n_va): Likewise.
7160         (format_warning_at_substring): Likewise.
7161         (format_warning_at_substring_n): Likewise.
7162         * toplev.c (general_init): Initialize global_dc->show_labels_p.
7164 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
7166         PR testsuite/86519
7167         * builtins.c (expand_builtin_memcmp): Do not expand the call
7168         when overflow is detected.
7170 2018-08-15  Martin Sebor  <msebor@redhat.com>
7172         PR tree-optimization/71625
7173         * config/aarch64/aarch64-builtins.c
7174         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
7176 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
7178         * config/s390/s390.c (s390_reorg): Remove loop.
7180 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7182         * config/darwin.c
7183          (darwin_function_switched_text_sections): Delete.
7184         * gcc/config/darwin.h
7185          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
7187 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7189         PR target/81685
7190         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
7191         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
7192         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
7194 2018-08-15  Martin Liska  <mliska@suse.cz>
7196         PR tree-optimization/86925
7197         * predict.c (expr_expected_value_1): When taking
7198         later predictor, assign also probability.
7199         Use fold_build2_initializer_loc in order to fold
7200         the expression in -frounding-math.
7202 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
7204         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
7205         patterns.
7206         (expand_vec_perm_1): Try the new method.
7208 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7210         PR target/86547
7211         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
7212         Check whether lra_live_max_point is 0 before dividing.
7214 2018-08-14  Martin Sebor  <msebor@redhat.com>
7216         PR tree-optimization/86650
7217         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
7218         (vrp_prop::check_mem_ref): Same.
7220 2018-08-13  Liu Hao <lh_mouse@126.com>
7222         * pretty-print.c (eat_esc_sequence): Swap the foreground and
7223         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
7224         and clear it thereafter, as it only works for DBCS.
7226 2018-08-13  Liu Hao <lh_mouse@126.com>
7228         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
7229         handle returned by _get_osf_handle().
7231 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7233         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
7234         for folding vec_perm.
7236 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7238         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
7239         Add support for gimple-folding of vec_pack() and vec_unpack()
7240         intrinsics.
7242 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
7244         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
7245         vec_xst variants to the list.
7246         (rs6000_gimple_fold_builtin): Add support for folding unaligned
7247         vector loads and stores.
7249 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
7251         * config.gcc (rs6000-ibm-aix4.x): Delete.
7252         (rs6000-ibm-aix5.1): Delete.
7253         (rs6000-ibm-aix5.2): Delete.
7254         (rs6000-ibm-aix5.3): Delete.
7255         * config/rs6000/aix43.h: Delete.
7256         * config/rs6000/aix51.h: Delete.
7257         * config/rs6000/aix52.h: Delete.
7258         * config/rs6000/t-aix43: Delete.
7260 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
7262         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7263         New function.
7264         (s390_decompose_address): Factor out constant pool ref
7265         decomposition.
7267 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7269         * config/nds32/nds32-predicates.c
7270         (nds32_can_use_bclr_p): Change return type as bool.
7271         (nds32_can_use_bset_p): Ditto.
7272         (nds32_can_use_btgl_p): Ditto.
7273         (nds32_can_use_bitci_p): Ditto.
7274         * config/nds32/nds32-protos.h
7275         (nds32_can_use_bclr_p): Change declaration.
7276         (nds32_can_use_bset_p): Ditto.
7277         (nds32_can_use_btgl_p): Ditto.
7278         (nds32_can_use_bitci_p): Ditto.
7280 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7282         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
7283         Support -msched-prolog-epilog option.
7284         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
7286 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7288         * common/config/nds32/nds32-common.c
7289         (nds32_option_optimization_table): Enalbe -malways-align.
7291 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7293         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
7294         extra_headers.
7295         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
7296         OPT_misr_secure_ case.
7297         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
7298         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
7299         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
7300         secure attribute.
7301         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
7302         (nds32_isr_info): New field security_level.
7303         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
7304         * config/nds32/nds32.md (return_internal): Consider critical attribute.
7305         * config/nds32/nds32.opt (misr-secure): New option.
7306         * config/nds32/nds32_init.inc: New file.
7307         * config/nds32/nds32_isr.h: New file.
7309 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
7311         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
7312         Update comment for atomic instructions.
7313         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
7314         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
7315         Remove.
7316         (atomic_loaddi): Revise fence expansion to only emit fence prior to
7317         load for __ATOMIC_SEQ_CST model.
7318         (atomic_loaddi_1): Remove float register target.
7319         (atomic_storedi): Handle CONST_INT values.
7320         (atomic_storedi_1): Remove float register source.  Add special case
7321         for zero value.
7322         (memory_barrier): New expander and insn.
7324 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
7326         PR tree-optimization/86835
7327         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
7328         new_stmt after def_gsi, make sure to insert new_square_stmt after
7329         that stmt, not 2 stmts before it.
7331 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
7333         PR target/82418
7334         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
7335         instead of SWI48.
7337 2018-08-10  Martin Liska  <mliska@suse.cz>
7339         PR target/83610
7340         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
7341         function type.
7342         * builtins.c (expand_builtin_expect_with_probability):
7343         New function.
7344         (expand_builtin_expect_with_probability): New function.
7345         (build_builtin_expect_predicate): Add new argumnet probability
7346         for BUILT_IN_EXPECT_WITH_PROBABILITY.
7347         (fold_builtin_expect):
7348         (fold_builtin_2):
7349         (fold_builtin_3):
7350         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
7351         * builtins.h (fold_builtin_expect): Set new argument.
7352         * doc/extend.texi: Document __builtin_expect_with_probability.
7353         * doc/invoke.texi: Likewise.
7354         * gimple-fold.c (gimple_fold_call): Pass new argument.
7355         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
7356         also BUILT_IN_EXPECT_WITH_PROBABILITY.
7357         * predict.c (get_predictor_value): New function.
7358         (expr_expected_value): Add new argument probability. Assume
7359         that predictor and probability are always non-null.
7360         (expr_expected_value_1): Likewise.  For __builtin_expect and
7361         __builtin_expect_with_probability set probability.  Handle
7362         combination in binary expressions.
7363         (tree_predict_by_opcode): Simplify code by simply calling
7364         get_predictor_value.
7365         (pass_strip_predict_hints::execute): Add handling of
7366         BUILT_IN_EXPECT_WITH_PROBABILITY.
7367         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
7368         new predictor.
7369         * tree.h (DECL_BUILT_IN_P): New function.
7371 2018-08-10  Martin Liska  <mliska@suse.cz>
7373         PR tree-optimization/85799
7374         * passes.def: Add argument for pass_strip_predict_hints.
7375         * predict.c (class pass_strip_predict_hints): Add new argument
7376         early_p.
7377         (strip_predictor_early): New function.
7378         (pass_strip_predict_hints::execute): Call the function to
7379         strip predictors.
7380         (strip_predict_hints): New function.
7381         * predict.def: Fix comment.
7383 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
7385         * Makefile.in: Clarify which tm.texi to copy over to assert the
7386         right to grant a GFDL license for all.
7388 2018-08-09  Jeff Law  <law@redhat.com>
7390         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
7391         unused variable.
7393 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
7395         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
7396         prototype.
7398 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7400         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
7401         reductions for variable-length vectors.
7403 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
7405         PR other/84889
7406         * common.opt (fdiagnostics-show-line-numbers): New option.
7407         * diagnostic-show-locus.c (class layout): Add fields
7408         "m_show_line_numbers_p" and "m_linenum_width";
7409         (num_digits): New function.
7410         (test_num_digits): New function.
7411         (layout::layout): Initialize new fields.  Update m_x_offset
7412         logic to handle any left margin.
7413         (layout::print_source_line): Print line number when requested.
7414         (layout::start_annotation_line): New member function.
7415         (layout::print_annotation_line): Call it.
7416         (layout::print_leading_fixits): Likewise.
7417         (layout::print_trailing_fixits): Likewise.  Update calls to
7418         move_to_column for new parameter.
7419         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
7420         it to potentially call start_annotation_line.
7421         (layout::show_ruler): Call start_annotation_line.
7422         (selftest::test_line_numbers_multiline_range): New selftest.
7423         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
7424         and selftest::test_line_numbers_multiline_range.
7425         * diagnostic.c (diagnostic_initialize): Initialize
7426         show_line_numbers_p.
7427         * diagnostic.h (struct diagnostic_context): Add field
7428         "show_line_numbers_p".
7429         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7430         -fno-diagnostics-show-line-numbers.
7431         * dwarf2out.c (gen_producer_string): Add
7432         OPT_fdiagnostics_show_line_numbers to the ignored options.
7433         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
7434         one setting" options.
7435         (append_compiler_options): Likewise to the dropped options.
7436         (append_diag_options): Likewise to the passed-on options.
7437         * opts.c (common_handle_option): Handle the new option.
7438         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
7440 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7442         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
7443         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
7444         third argument of type "const signed char" to descriptions of
7445         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
7446         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
7447         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
7448         __builtin_bcdsub_ov functions.
7450 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7452         PR tree-optimization/86858
7453         * tree-vect-loop.c (vect_is_simple_reduction): Restore
7454         flow_bb_inside_loop_p calls.
7456 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7458         PR tree-optimization/86871
7459         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
7460         instead of gimple_assign_lhs.
7462 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
7464         PR target/86887
7465         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
7466         register constraint to operand 0.
7467         (add<mode>3_carryinC): Likewise.
7468         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
7470 2018-08-09  Martin Liska  <mliska@suse.cz>
7472         PR c/86895
7473         * common.opt: Remove extra line.
7475 2018-08-09  Martin Liska  <mliska@suse.cz>
7477         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
7478         at the end of a line, make first letter capital and end up
7479         a sentence with a dot.
7480         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
7481         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
7482         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
7483         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
7484         (PARAM_MAX_ISL_OPERATIONS): Likewise.
7485         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
7486         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
7487         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
7488         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
7489         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
7490         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
7491         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
7492         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7493         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7494         (PARAM_TREE_REASSOC_WIDTH): Likewise.
7495         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7496         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7497         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7499 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
7501         PR target/84332
7502         * config/s390/s390.c (s390_option_override_internal): Reduce the
7503         stack-clash-protection-probe-interval param if it would be too big
7504         for z900.
7506 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
7508         PR target/46179
7509         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7510         * config/m68k/m68k.c (handle_move_double): Don't call
7511         m68k_final_prescan_insn.
7512         (m68k_adjust_decorated_operand): Renamed from
7513         m68k_final_prescan_insn, remove first and third operand and
7514         simplify.
7515         (print_operand): Call it.
7516         (print_operand_address): Call it.
7518 2018-08-08  Nathan Sidwell  <nathan@acm.org>
7520         * diagnostic.c (diagnostic_report_current_module): Use
7521         linemap_included_from & linemap_included_from_linemap.
7523 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
7525         * config/aarch64/aarch64-cores.def: Add phecda core.
7526         * config/aarch64/aarch64-tune.md: Regenerate.
7527         * doc/invoke.texi: Add phecda core.
7529 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7531         PR target/85295
7532         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7533         definitions.
7534         * config/s390/s390.md ("movti"): Add more alternatives for
7535         constant to GPR copies.
7537 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7539         * config/s390/s390.c: Fix whitespace damage throughout the file.
7540         * config/s390/s390.h: Likewise.
7541         * config/s390/tpf.h: Likewise.
7543 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7545         * config/s390/s390.c (s390_loadrelative_operand_p):
7546         Remove TARGET_CPU_ZARCH usages.
7547         (s390_rtx_costs): Likewise.
7548         (s390_legitimate_constant_p): Likewise.
7549         (s390_cannot_force_const_mem): Likewise.
7550         (legitimate_reload_constant_p): Likewise.
7551         (s390_preferred_reload_class): Likewise.
7552         (legitimize_pic_address): Likewise.
7553         (legitimize_tls_address): Likewise.
7554         (s390_split_branches): Removed.
7555         (s390_add_execute): Removed.
7556         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7557         (s390_mainpool_start): Likewise.
7558         (s390_mainpool_finish): Likewise.
7559         (s390_mainpool_cancel): Removed.
7560         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7561         (s390_chunkify_cancel): Likewise.
7562         (s390_return_addr_rtx): Likewise.
7563         (s390_register_info): Remove split_branches_pending_p uages.
7564         (s390_optimize_register_info): Likewise.
7565         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7566         split_branches_pending_p usages.
7567         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7568         (s390_load_got): Likewise.
7569         (s390_expand_split_stack_prologue): Likewise.
7570         (output_asm_nops): Likewise.
7571         (s390_function_profiler): Likewise.
7572         (s390_emit_call): Likewise.
7573         (s390_conditional_register_usage): Likewise.
7574         (s390_optimize_prologue): Likewise.
7575         (s390_reorg): Remove TARGET_CPU_ZARCH and
7576         split_branches_pending_p usages.
7577         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7578         usages.
7579         (s390_output_indirect_thunk_function): Likewise.
7580         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7581         (TARGET_CPU_ZARCH_P): Removed.
7582         (struct machine_function): Remove split_branches_pending_p.
7583         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7585 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7587         * common/config/s390/s390-common.c (processor_flags_table):
7588         Remove flags.
7589         * config.gcc: Remove with_arch/with_tune support.
7590         * config/s390/2064.md: Remove cpu attribute comparisons.
7591         * config/s390/driver-native.c (s390_host_detect_local_cpu):
7592         Remove MTN.
7593         * config/s390/linux.h (ASM_SPEC):
7594         Remove -march support.
7595         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7596         Use a table to get an arch level.
7597         * config/s390/s390-opts.h (enum processor_type):
7598         Remove enum values.
7599         * config/s390/s390.c
7600         (processor_table): Remove entries, add arch_level values.
7601         (s390_issue_rate): Remove cases.
7602         (s390_option_override): Adjust
7603         s390_option_override_internal() call.
7604         (s390_option_override_internal): Remove deprecation warning.
7605         (s390_valid_target_attribute_tree): Adjust
7606         s390_option_override_internal() call.
7607         * config/s390/s390.h (struct s390_processor):
7608         Share with s390-c.c, add arch_level field.
7609         * config/s390/s390.md:
7610         Remove occurrences in cpu attribute.
7611         * config/s390/s390.opt: Remove -march/-mtune support.
7612         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7613         * doc/invoke.texi: Remove deprecation warning.
7615 2018-08-08  Luis Machado  <luis.machado@linaro.org>
7617         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7618         global.
7619         (qdf24xx_tunings): Set vector cost structure to
7620         qdf24xx_vector_cost.
7622         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7623         <register_sextend>: Set to 3.
7625 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
7627         PR target/86838
7628         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7629         * config/aarch64/aarch64-simd.md
7630         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7631         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7632         (aarch64_frecpx<mode>): ...this new pattern.
7633         * config/aarch64/aarch64-simd-builtins.def: Remove comment
7634         about aarch64_frecp<FRECP:frecp_suffix><mode>.
7636 2018-08-07  Martin Liska  <mliska@suse.cz>
7638         PR middle-end/83023
7639         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7640         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7641         * predict.def (PRED_MALLOC_NONNULL): New predictor.
7642         * doc/extend.texi: Document that malloc attribute adds
7643         hit to compiler.
7645 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
7647         PR target/86785
7648         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7649         Define to speculation_safe_value_not_needed.
7651 2018-08-06  Jeff Law  <law@redhat.com>
7653         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7654         the vr_values instance to cprop_into_stmt.
7655         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7656         (cprop_operand): Also query EVRP to determine if OP is a constant.
7658 2018-08-06  Nathan Sidwell  <nathan@acm.org>
7660         * diagnostic.c (diagnostic_report_current_module): Reroll
7661         included-at loop.  Translate text.
7663 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
7665         * function-tests.c (selftest::test_expansion_to_rtl): Call
7666         free_after_compilation.
7668 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7670         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7672 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
7674         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7675         loops with memory block operations from getting unrolled.
7677 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
7679         PR target/86807
7680         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7681         Define to speculation_safe_value_not_needed.
7683 2018-08-06  Jeff Law  <law@redhat.com>
7685         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7686         assert.
7688 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7690         PR target/86662
7691         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7692         with all enabled __intN types.
7694         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7696 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7698         * alias.c (record_set): Check for clobber high.
7699         * cfgexpand.c (expand_gimple_stmt): Likewise.
7700         * combine-stack-adj.c (single_set_for_csa): Likewise.
7701         * combine.c (find_single_use_1): Likewise.
7702         (set_nonzero_bits_and_sign_copies): Likewise.
7703         (get_combine_src_dest): Likewise.
7704         (is_parallel_of_n_reg_sets): Likewise.
7705         (try_combine): Likewise.
7706         (record_dead_and_set_regs_1): Likewise.
7707         (reg_dead_at_p_1): Likewise.
7708         (reg_dead_at_p): Likewise.
7709         * dce.c (deletable_insn_p): Likewise.
7710         (mark_nonreg_stores_1): Likewise.
7711         (mark_nonreg_stores_2): Likewise.
7712         * df-scan.c (df_find_hard_reg_defs): Likewise.
7713         (df_uses_record): Likewise.
7714         (df_get_call_refs): Likewise.
7715         * dwarf2out.c (mem_loc_descriptor): Likewise.
7716         * haifa-sched.c (haifa_classify_rtx): Likewise.
7717         * ira-build.c (create_insn_allocnos): Likewise.
7718         * ira-costs.c (scan_one_insn): Likewise.
7719         * ira.c (equiv_init_movable_p): Likewise.
7720         (rtx_moveable_p): Likewise.
7721         (interesting_dest_for_shprep): Likewise.
7722         * jump.c (mark_jump_label_1): Likewise.
7723         * postreload-gcse.c (record_opr_changes): Likewise.
7724         * postreload.c (reload_cse_simplify): Likewise.
7725         (struct reg_use): Add source expr.
7726         (reload_combine): Check for clobber high.
7727         (reload_combine_note_use): Likewise.
7728         (reload_cse_move2add): Likewise.
7729         (move2add_note_store): Likewise.
7730         * print-rtl.c (print_pattern): Likewise.
7731         * recog.c (decode_asm_operands): Likewise.
7732         (store_data_bypass_p): Likewise.
7733         (if_test_bypass_p): Likewise.
7734         * regcprop.c (kill_clobbered_value): Likewise.
7735         (kill_set_value): Likewise.
7736         * reginfo.c (reg_scan_mark_refs): Likewise.
7737         * reload1.c (maybe_fix_stack_asms): Likewise.
7738         (eliminate_regs_1): Likewise.
7739         (elimination_effects): Likewise.
7740         (mark_not_eliminable): Likewise.
7741         (scan_paradoxical_subregs): Likewise.
7742         (forget_old_reloads_1): Likewise.
7743         * reorg.c (find_end_label): Likewise.
7744         (try_merge_delay_insns): Likewise.
7745         (redundant_insn): Likewise.
7746         (own_thread_p): Likewise.
7747         (fill_simple_delay_slots): Likewise.
7748         (fill_slots_from_thread): Likewise.
7749         (dbr_schedule): Likewise.
7750         * resource.c (update_live_status): Likewise.
7751         (mark_referenced_resources): Likewise.
7752         (mark_set_resources): Likewise.
7753         * rtl.c (copy_rtx): Likewise.
7754         * rtlanal.c (reg_referenced_p): Likewise.
7755         (single_set_2): Likewise.
7756         (noop_move_p): Likewise.
7757         (note_stores): Likewise.
7758         * sched-deps.c (sched_analyze_reg): Likewise.
7759         (sched_analyze_insn): Likewise.
7761 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7763         * cse.c (invalidate_reg): New function extracted from...
7764         (invalidate): ...here.
7765         (canonicalize_insn): Check for clobber high.
7766         (invalidate_from_clobbers): invalidate clobber highs.
7767         (invalidate_from_sets_and_clobbers): Likewise.
7768         (count_reg_usage): Check for clobber high.
7769         (insn_live_p): Likewise.
7770         * cselib.c (cselib_expand_value_rtx_1):Likewise.
7771         (cselib_invalidate_regno): Check for clobber in setter.
7772         (cselib_invalidate_rtx): Pass through setter.
7773         (cselib_invalidate_rtx_note_stores):
7774         (cselib_process_insn): Check for clobber high.
7775         * cselib.h (cselib_invalidate_rtx): Add operand.
7777 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7779         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7780         (mark_not_eliminable): Likewise.
7781         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7782         * lra-lives.c (process_bb_lives): Check for clobber high.
7783         * lra.c (new_insn_reg): Remember clobber highs.
7784         (collect_non_operand_hard_regs): Check for clobber high.
7785         (lra_set_insn_recog_data): Likewise.
7786         (add_regs_to_insn_regno_info): Likewise.
7787         (lra_update_insn_regno_info): Likewise.
7789 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7791         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7792         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7794 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7796         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7797         (copy_insn_1): Likewise.
7798         (gen_hard_reg_clobber_high): New gen function.
7799         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7800         * genemit.c (gen_exp): Likewise.
7801         (gen_emit_seq): Pass through info.
7802         (gen_insn): Check for CLOBBER_HIGH.
7803         (gen_expand): Pass through info.
7804         (gen_split): Likewise.
7805         (output_add_clobbers): Likewise.
7806         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7807         (remove_clobbers): Likewise.
7808         * rtl.h (gen_hard_reg_clobber_high): New declaration.
7810 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7812         * doc/rtl.texi (clobber_high): Add.
7813         (parallel): Add in clobber high
7814         * rtl.c (rtl_check_failed_code3): Add function.
7815         * rtl.def (CLOBBER_HIGH): Add expression.
7816         * rtl.h (RTL_CHECKC3): Add macro.
7817         (rtl_check_failed_code3): Add declaration.
7818         (XC3EXP): Add macro.
7820 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
7822         PR target/86386
7823         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7824         cfun->machine->max_used_stack_alignment if needed.
7826 2018-08-04  Martin Sebor  <msebor@redhat.com>
7828         PR tree-optimization/86571
7829         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7830         NaN output to 4.
7832 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
7834         PR target/86799
7835         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7836         Define.
7838 2018-08-03  Jeff Law  <law@redhat.com>
7840         PR target/86795
7841         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7842         Define to speculation_safe_value_not_needed.
7844 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
7846         * doc/gcov.texi (-x): Remove duplicate "to".
7847         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7848         (-Wif-not-aligned): Remove duplicate "is".
7849         (-flto): Remove duplicate "the".
7850         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7851         duplicate "v5.00.b".
7852         (MSP430 Options): Remove duplicate "and" from the description
7853         of "-mgprel-sec=regexp".
7854         (x86 Options): Remove duplicate copies of "vmldLog102" and
7855         vmlsLog104 from description of "-mveclibabi=type".
7857 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
7859         * internal-fn.h (first_commutative_argument): Declare.
7860         * internal-fn.c (first_commutative_argument): New function.
7861         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7862         restrictions for pattern statements.  Use first_commutative_argument
7863         to look for commutative operands in calls to internal functions.
7865 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
7867         * Makefile.in (wide-int-range.o): New.
7868         * tree-vrp.c: Move all the wide_int_* functions to...
7869         * wide-int-range.cc: ...here.
7870         * tree-vrp.h: Move all the wide_int_* prototypes to...
7871         * wide-int-range.h: ...here.
7873 2018-08-03  Tom de Vries  <tdevries@suse.de>
7875         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
7876         UI_NONE.
7877         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
7878         * except.c (output_function_exception_table): Do early exit if
7879         targetm_common.except_unwind_info (&global_options) == UI_NONE.
7881 2018-08-03  Martin Liska  <mliska@suse.cz>
7883         * predict.c (dump_prediction): Change to 2 digits
7884         in fraction part.
7886 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7888         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
7889         neon_dup_q to...
7890         (falkor_am_1_gtov_gtov): ... a new insn reservation.
7892 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
7894         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
7895         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
7896         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
7897         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
7898         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
7900 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7902         * diagnostic-show-locus.c (diagnostic_show_locus): Use
7903         pp_take_prefix when saving the existing prefix.
7904         * diagnostic.c (diagnostic_append_note): Likewise.
7905         * langhooks.c (lhd_print_error_function): Likewise.
7906         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
7907         param's type.  Free the existing prefix.
7908         (pp_take_prefix): New function.
7909         (pretty_printer::pretty_printer): Drop the prefix parameter.
7910         Rename the length parameter to match the comment.
7911         (pretty_printer::~pretty_printer): Free the prefix.
7912         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
7913         parameter.
7914         (struct pretty_printer): Drop the "const" from "prefix" field's
7915         type and clarify memory management.
7916         (pp_set_prefix): Drop the "const" from the 2nd param.
7917         (pp_take_prefix): New decl.
7919 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
7921         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
7922         for word_mode_ok here instead of passing as argument.
7923         (expand_block_compare): Change select_block_compare_mode() call.
7924         (expand_strncmp_gpr_sequence): New function.
7925         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
7927 2018-08-02  Jeff Law  <law@redhat.com>
7929         PR target/86790
7930         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7931         Define to speculation_safe_value_not_needed.
7933         PR target/86784
7934         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7935         Define to speculation_safe_value_not_needed.
7937 2018-08-02  Tom de Vries  <tdevries@suse.de>
7939         PR target/86660
7940         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
7941         function.  Return UI_TARGET unconditionally.
7942         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
7943         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
7945 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7947         * genemit.c (print_overload_test): Fix typo.
7949 2018-08-02  Richard Biener  <rguenther@suse.de>
7951         PR tree-optimization/86816
7952         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
7953         which checks for value availability before querying it.
7954         (gvn_uses_equal): Use it.
7955         (same_succ_hash): Likewise.
7956         (gimple_equal_p): Likewise.
7958 2018-08-02  Nick Clifton  <nickc@redhat.com>
7960         PR target/86813
7961         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7962         Define to speculation_safe_value_not_needed.
7964         PR target/86810
7965         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7966         Define to speculation_safe_value_not_needed.
7968         PR target/86810
7969         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7970         Define to speculation_safe_value_not_needed.
7972         PR target/86803
7973         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7974         Define to speculation_safe_value_not_needed.
7976         PR target/86797
7977         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7978         Define to speculation_safe_value_not_needed.
7980         PR target/86791
7981         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7982         Define to speculation_safe_value_not_needed.
7984         PR target/86789
7985         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7986         Define to speculation_safe_value_not_needed.
7988         PR target/86787
7989         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7990         Define to speculation_safe_value_not_needed.
7992         PR target/86782
7993         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7994         speculation_safe_value_not_needed.
7996         PR target/86781
7997         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
7998         to speculation_safe_value_not_needed.
8000 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8002         * doc/md.texi: Expand the documentation of instruction names
8003         to mention port-local uses.  Document '@' in pattern names.
8004         * read-md.h (overloaded_instance, overloaded_name): New structs.
8005         (mapping): Declare.
8006         (md_reader::handle_overloaded_name): New member function.
8007         (md_reader::get_overloads): Likewise.
8008         (md_reader::m_first_overload): New member variable.
8009         (md_reader::m_next_overload_ptr): Likewise.
8010         (md_reader::m_overloads_htab): Likewise.
8011         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
8012         m_next_overload_ptr and m_overloads_htab.
8013         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
8014         (get_mode_token, get_code_token, get_int_token): New functions.
8015         (map_attr_string): Add an optional argument that passes back
8016         the associated iterator.
8017         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
8018         (md_reader::handle_overloaded_name, add_overload_instance): New
8019         functions.
8020         (apply_iterators): Handle '@' names.  Report an error if '@'
8021         is used without iterators.
8022         (initialize_iterators): Initialize the new iterator_group fields.
8023         * genopinit.c (handle_overloaded_code_for)
8024         (handle_overloaded_gen): New functions.
8025         (main): Use them to print declarations of maybe_code_for_* and
8026         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
8027         * genemit.c (print_overload_arguments, print_overload_test)
8028         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
8029         (main): Use it to print definitions of maybe_code_for_* and
8030         maybe_gen_* functions.
8031         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
8032         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
8033         instead of explicit mode checks.
8034         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
8035         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
8036         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
8037         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
8038         (aarch64_expand_compare_and_swap): Likewise
8039         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
8040         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
8041         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
8042         (aarch64_constant_pool_reload_icode): Delete.
8043         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
8044         instead of aarch64_constant_pool_reload_icode.  Use
8045         code_for_aarch64_reload_mov instead of explicit mode checks.
8046         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
8047         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
8048         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
8049         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
8050         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
8051         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
8052         (aarch64_atomic_load_op_code): Delete.
8053         (aarch64_emit_atomic_load_op): Likewise.
8054         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
8055         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
8056         instead of aarch64_emit_atomic_load_op.
8057         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
8058         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
8059         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
8060         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
8061         character before the pattern name.
8062         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8063         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
8064         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
8065         (aarch64_frecps<mode>): Likewise.
8066         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
8067         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
8068         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
8069         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
8070         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
8072 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8074         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
8075         Allow HFmode constants if TARGET_FP_F16INST.
8077 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
8079         PR target/86014
8080         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
8081         No longer check last store for clobber of address register.
8083 2018-08-02  Martin Liska  <mliska@suse.cz>
8085         PR gcov-profile/86817
8086         * gcov.c (process_all_functions): New function.
8087         (main): Call it.
8088         (process_file): Move functions processing to
8089         process_all_functions.
8091 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
8093         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
8094         "const" to the "gimple *" and "rtx_insn *" parameters.
8095         * dumpfile.h (dump_user_location_t::dump_user_location_t):
8096         Likewise.
8097         (dump_location_t::dump_location_t): Likewise.
8099 2018-08-01  Martin Sebor  <msebor@redhat.com>
8101         PR tree-optimization/86650
8102         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
8103         rather than a "gcall *".  Directly pass the data of interest
8104         to percent_K_format, rather than building a temporary CALL_EXPR
8105         to hold it.
8106         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
8107         (gimple_fold_builtin_strncat): Adjust.
8108         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
8109         gcall* argument with gimple*.
8110         * gimple-ssa-warn-restrict.c (check_call): Same.
8111         (wrestrict_dom_walker::before_dom_children): Same.
8112         (builtin_access::builtin_access): Same.
8113         (check_bounds_or_overlap): Same
8114         (maybe_diag_overlap): Same.
8115         (maybe_diag_offset_bounds): Same.
8116         * tree-diagnostic.c (default_tree_printer): Move usage of
8117         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
8118         to this callsite.
8119         * tree-pretty-print.c (percent_K_format): Add argument.
8120         * tree-pretty-print.h: Add argument.
8121         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
8122         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
8123         (maybe_diag_stxncpy_trunc): Same.
8124         (handle_builtin_stxncpy): Same.
8125         (handle_builtin_strcat): Same.
8127 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8129         * match.pd: Optimise pointer range checks.
8131 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8133         PR tree-optimization/86758
8134         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
8135         to remove pattern statements.
8137 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8139         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
8140         result of dfs_enumerate_from when constructing stmt_vec_infos,
8141         instead of additionally calling get_loop_body.
8143 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8145         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
8146         parameter.
8147         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
8148         When creating an iv, assert that the step is not known to be zero.
8149         (vect_setup_realignment): Update call accordingly.
8150         * tree-vect-stmts.c (vectorizable_store): Likewise.
8151         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
8153 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8155         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
8156         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
8157         (vectorizable_reduction): Likewise.
8158         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8159         (vect_detect_hybrid_slp_stmts): Likewise.
8160         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8162 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
8164         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
8165         (wide_int_set_zero_nonzero_bits): ...this.
8166         (zero_nonzero_bits_from_vr): Rename to...
8167         (vrp_set_zero_nonzero_bits): ...this.
8168         (extract_range_from_multiplicative_op_1): Abstract wide int
8169         code...
8170         (wide_int_range_multiplicative_op): ...here.
8171         (extract_range_from_binary_expr_1): Extract wide int binary
8172         operations into their own functions.
8173         (wide_int_range_lshift): New.
8174         (wide_int_range_can_optimize_bit_op): New.
8175         (wide_int_range_shift_undefined_p): New.
8176         (wide_int_range_bit_xor): New.
8177         (wide_int_range_bit_ior): New.
8178         (wide_int_range_bit_and): New.
8179         (wide_int_range_trunc_mod): New.
8180         (extract_range_into_wide_ints): New.
8181         (vrp_shift_undefined_p): New.
8182         (extract_range_from_multiplicative_op): New.
8183         (vrp_can_optimize_bit_op): New.
8184         * tree-vrp.h (value_range::dump): New.
8185         (wide_int_range_multiplicative_op): New.
8186         (wide_int_range_lshift):New.
8187         (wide_int_range_shift_undefined_p): New.
8188         (wide_int_range_bit_xor): New.
8189         (wide_int_range_bit_ior): New.
8190         (wide_int_range_bit_and): New.
8191         (wide_int_range_trunc_mod): New.
8192         (zero_nonzero_bits_from_bounds): Rename to...
8193         (wide_int_set_zero_nonzero_bits): ...this.
8194         (zero_nonzero_bits_from_vr): Rename to...
8195         (vrp_set_zero_nonzero_bits): ...this.
8196         (range_easy_mask_min_max): Rename to...
8197         (wide_int_range_can_optimize_bit_op): this.
8198         * vr-values.c (simplify_bit_ops_using_ranges): Rename
8199         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
8201 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8203         * tree-vectorizer.h (vect_orig_stmt): New function.
8204         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
8205         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
8206         (vect_create_epilog_for_reduction): Likewise.
8207         (vectorizable_live_operation): Likewise.
8208         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8209         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
8210         * tree-vect-stmts.c (vectorizable_call): Likewise.
8211         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
8213 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8215         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
8216         argument.
8217         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
8218         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
8219         (vect_transform_loop): Likewise.
8220         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8222 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8224         * tree-vectorizer.h (vect_schedule_slp): Return void.
8225         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8226         (vect_schedule_slp): Likewise.
8228 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8230         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
8231         argument.
8232         (vect_transform_loop): Update calls accordingly.  Schedule SLP
8233         instances before the main loop, if any exist.
8235 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8237         PR tree-optimization/86749
8238         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
8239         If the lhs is used in a COND_EXPR, check that it is being used
8240         as the "then" or "else" value.
8242 2018-08-01  Tom de Vries  <tdevries@suse.de>
8244         PR target/86800
8245         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8246         speculation_safe_value_not_needed.
8248 2018-08-01  Richard Biener  <rguenther@suse.de>
8250         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
8251         as base and offset.
8253 2018-08-01  Martin Liska  <mliska@suse.cz>
8255         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
8256         format how successful transformation is dumped.
8257         (gimple_mod_pow2_value_transform): Likewise.
8258         (gimple_mod_subtract_transform): Likewise.
8259         (gimple_stringops_transform): Likewise.
8261 2018-08-01  Martin Liska  <mliska@suse.cz>
8263         PR value-prof/35543
8264         * value-prof.c (interesting_stringop_to_profile_p):
8265         Simplify the code and add BUILT_IN_MEMMOVE.
8266         (gimple_stringops_transform): Likewise.
8268 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
8270         * config/aarch64/aarch64-simd.md
8271         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
8272         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
8273         use GPI iterator instead of SI mode.
8275 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8277         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
8278         rs6000_speculation_barrier.
8279         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
8280         new barrier pattern name.
8282 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8284         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
8285         (speculation_barrier): New insn.
8287 2018-08-01  Richard Biener  <rguenther@suse.de>
8289         PR bootstrap/86724
8290         * graphite.h: Include isl/id.h and isl/space.h to allow build
8291         with ISL 0.20.
8293 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
8295         PR target/86651
8296         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8297         mode for COFF targets.
8298         * defaults.h (TARGET_COFF): Define.
8299         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
8300         TARGET_COFF): Define.
8301         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
8302         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
8303         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
8305 2018-07-31  Alexandre Oliva <oliva@adacore.com>
8307         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
8308         * gimple-streamer-out.c (output_bb): Save it.
8309         * lto-streamer-in.c (input_struct_function_base): Restore
8310         instance discriminator if available.  Create map on demand.
8311         * lto-streamer-out.c (output_struct_function_base): Save it if
8312         available.
8313         * final.c (decl_to_instance_map): Document LTO strategy.
8315 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
8316             Olivier Hainque  <hainque@adacore.com>
8318         * debug.h (decl_to_instance_map_t): New type.
8319         (decl_to_instance_map): Declare.
8320         (maybe_create_decl_to_instance_map): New inline function.
8321         * final.c (bb_discriminator, last_bb_discriminator): New statics,
8322         to track basic block discriminators.
8323         (final_start_function_1): Initialize them.
8324         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
8325         bb_discriminator.
8326         (decl_to_instance_map): New variable.
8327         (map_decl_to_instance, maybe_set_discriminator): New functions.
8328         (notice_source_line): Set discriminator.
8330 2018-07-31  Ian Lance Taylor  <iant@golang.org>
8332         * targhooks.c (default_have_speculation_safe_value): Add
8333         ATTRIBUTE_UNUSED.
8335 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8337         * dump-context.h: Include "pretty-print.h".
8338         (dump_context::refresh_dumps_are_enabled): New decl.
8339         (dump_context::emit_item): New decl.
8340         (class dump_context): Add fields "m_test_pp" and
8341         "m_test_pp_flags".
8342         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8343         (temp_dump_context::get_dumped_text): New decl.
8344         (class temp_dump_context): Add field "m_pp".
8345         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
8346         (dump_context::refresh_dumps_are_enabled): ...and add a test for
8347         m_test_pp.
8348         (set_dump_file): Update for above change.
8349         (set_alt_dump_file): Likewise.
8350         (dump_loc): New overload, taking a pretty_printer *.
8351         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
8352         to any test pretty-printer.
8353         (make_item_for_dump_gimple_stmt): New function, adapted from
8354         optinfo::add_gimple_stmt.
8355         (dump_context::dump_gimple_stmt): Call it, and use the result,
8356         eliminating the direct usage of dump_file and alt_dump_file in
8357         favor of indirectly using them via emit_item.
8358         (make_item_for_dump_gimple_expr): New function, adapted from
8359         optinfo::add_gimple_expr.
8360         (dump_context::dump_gimple_expr): Call it, and use the result,
8361         eliminating the direct usage of dump_file and alt_dump_file in
8362         favor of indirectly using them via emit_item.
8363         (make_item_for_dump_generic_expr): New function, adapted from
8364         optinfo::add_tree.
8365         (dump_context::dump_generic_expr): Call it, and use the result,
8366         eliminating the direct usage of dump_file and alt_dump_file in
8367         favor of indirectly using them via emit_item.
8368         (make_item_for_dump_printf_va): New function, adapted from
8369         optinfo::add_printf_va.
8370         (make_item_for_dump_printf): New function.
8371         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
8372         and use the result, eliminating the direct usage of dump_file and
8373         alt_dump_file in favor of indirectly using them via emit_item.
8374         (make_item_for_dump_dec): New function.
8375         (dump_context::dump_dec): Call it, and use the result,
8376         eliminating the direct usage of dump_file and alt_dump_file in
8377         favor of indirectly using them via emit_item.
8378         (make_item_for_dump_symtab_node): New function, adapted from
8379         optinfo::add_symtab_node.
8380         (dump_context::dump_symtab_node): Call it, and use the result,
8381         eliminating the direct usage of dump_file and alt_dump_file in
8382         favor of indirectly using them via emit_item.
8383         (dump_context::begin_scope): Reimplement, avoiding direct usage
8384         of dump_file and alt_dump_file in favor of indirectly using them
8385         via emit_item.
8386         (dump_context::emit_item): New member function.
8387         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8388         Set up test pretty-printer on the underlying context.  Call
8389         refresh_dumps_are_enabled.
8390         (temp_dump_context::~temp_dump_context): Call
8391         refresh_dumps_are_enabled.
8392         (temp_dump_context::get_dumped_text): New member function.
8393         (selftest::verify_dumped_text): New function.
8394         (ASSERT_DUMPED_TEXT_EQ): New macro.
8395         (selftest::test_capture_of_dump_calls): Run all tests twice, with
8396         and then without optinfo enabled.  Add uses of
8397         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
8398         * dumpfile.h: Update comment for the dump_* API.
8399         * optinfo-emit-json.cc
8400         (selftest::test_building_json_from_dump_calls): Update for new
8401         param for temp_dump_context ctor.
8402         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
8403         and "m_owned" field.
8404         (optinfo_item::~optinfo_item): Likewise.
8405         (optinfo::add_item): New member function.
8406         (optinfo::emit): Update comment.
8407         (optinfo::add_string): Delete.
8408         (optinfo::add_printf): Delete.
8409         (optinfo::add_printf_va): Delete.
8410         (optinfo::add_gimple_stmt): Delete.
8411         (optinfo::add_gimple_expr): Delete.
8412         (optinfo::add_tree): Delete.
8413         (optinfo::add_symtab_node): Delete.
8414         (optinfo::add_dec): Delete.
8415         * optinfo.h (class dump_context): New forward decl.
8416         (optinfo::add_item): New decl.
8417         (optinfo::add_string): Delete.
8418         (optinfo::add_printf): Delete.
8419         (optinfo::add_printf_va): Delete.
8420         (optinfo::add_gimple_stmt): Delete.
8421         (optinfo::add_gimple_expr): Delete.
8422         (optinfo::add_tree): Delete.
8423         (optinfo::add_symtab_node): Delete.
8424         (optinfo::add_dec): Delete.
8425         (optinfo::add_poly_int): Delete.
8426         (optinfo_item::optinfo_item): Remove "owned" param.
8427         (class optinfo_item): Remove field "m_owned".
8429 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
8431         PR middle-end/86705
8432         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
8433         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
8434         requested variable alignment.
8435         (expand_one_ssa_partition): Likewise.
8436         (expand_one_var): Likewise.
8438 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8440         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8441         to speculation_safe_value_not_needed.
8443 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8445         * targhooks.h (speculation_safe_value_not_needed): New prototype.
8446         * targhooks.c (speculation_safe_value_not_needed): New function.
8447         * target.def (have_speculation_safe_value): Update documentation.
8448         * doc/tm.texi: Regenerated.
8450 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8452         * config/aarch64/iterators.md (ALLI_TI): New iterator.
8453         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
8454         expand.
8455         (despeculate_copy<ALLI:mode>_insn): New insn.
8456         (despeculate_copyti_insn): New insn.
8457         (despeculate_simple<ALLI:mode>): New insn
8458         (despeculate_simpleti): New insn.
8459         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
8460         function.
8461         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
8462         aarch64_speculation_safe_value.
8463         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
8465 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8467         * config/aarch64/aarch64-speculation.cc: New file.
8468         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
8469         before pass_reorder_blocks.
8470         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
8471         prototype.
8472         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
8473         X14 and X15 when tracking speculation.
8474         * config/aarch64/aarch64.md (register name constants): Add
8475         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
8476         (unspec): Add UNSPEC_SPECULATION_TRACKER.
8477         (speculation_barrier): New insn attribute.
8478         (cmp<mode>): Allow SP in comparisons.
8479         (speculation_tracker): New insn.
8480         (speculation_barrier): Add speculation_barrier attribute.
8481         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
8482         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
8483         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
8485 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8487         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
8488         aarch64_track_speculation is true.
8489         (tb<optab><mode>1): Likewise.
8490         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
8491         generate CB[N]Z when tracking speculation.
8492         (aarch64_split_compare_and_swap): Likewise.
8493         (aarch64_split_atomic_op): Likewise.
8495 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8497         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8499 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8501         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8502         (speculation_barrier): New insn.
8504 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8506         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8507         * config/arm/arm.md (speculation_barrier): New expand.
8508         (speculation_barrier_insn): New pattern.
8510 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8512         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8513         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8514         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8515         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8516         list.
8517         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8518         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8519         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8520         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8521         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8522         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8523         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8524         * builtins.c (expand_speculation_safe_value): New function.
8525         (expand_builtin): Call it.
8526         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8527         * doc/extend.texi: Document __builtin_speculation_safe_value.
8528         * doc/md.texi: Document "speculation_barrier" pattern.
8529         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8530         TARGET_HAVE_SPECULATION_SAFE_VALUE.
8531         * doc/tm.texi: Regenerated.
8532         * target.def (have_speculation_safe_value, speculation_safe_value): New
8533         hooks.
8534         * targhooks.c (default_have_speculation_safe_value): New function.
8535         (default_speculation_safe_value): New function.
8536         * targhooks.h (default_have_speculation_safe_value): Add prototype.
8537         (default_speculation_safe_value): Add prototype.
8539 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8541         * dump-context.h (dump_context::dump_loc): New decl.
8542         * dumpfile.c (dump_context::dump_loc): New member function.
8543         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8544         and dump_gimple_stmt.
8545         (dump_context::dump_gimple_expr_loc): Likewise, using
8546         dump_gimple_expr.
8547         (dump_context::dump_generic_expr_loc): Likewise, using
8548         dump_generic_expr.
8549         (dump_context::dump_printf_loc_va): Likewise, using
8550         dump_printf_va.
8551         (dump_context::begin_scope): Explicitly using the global function
8552         "dump_loc", rather than the member function.
8554 2018-07-31  Martin Sebor  <msebor@redhat.com>
8556         PR tree-optimization/86741
8557         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8559 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
8561         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8562         depend on whether prefetch instructions will be emitted or not.
8563         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8564         will be emitted or not.
8565         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8566         (TARGET_SETMEM_PFD): New macros.
8568 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8570         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8571         (NULL_STMT_VEC_INFO): Delete.
8572         (stmt_vec_info::operator*): Likewise.
8573         (stmt_vec_info::operator gimple *): Likewise.
8574         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8575         of NULL_STMT_VEC_INFO.
8576         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8577         (vect_reassociating_reduction_p): Likewise.
8578         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8579         (vectorizable_store): Likewise.
8580         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8581         (vec_info::free_stmt_vec_infos): Likewise.
8583 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8585         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8586         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8588 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8590         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8591         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8592         (vec_info::free_stmt_vec_info): New private member functions.
8593         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8594         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8595         * tree-parloops.c (gather_scalar_reductions): Remove calls to
8596         set_stmt_vec_info_vec and free_stmt_vec_infos.
8597         * tree-vect-loop.c (_loop_vec_info): Remove call to
8598         set_stmt_vec_info_vec.
8599         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8600         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8601         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8602         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8603         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
8604         assignments in {vec_info::,}new_stmt_vec_info that are redundant
8605         with the clearing in the xcalloc.
8606         (stmt_vec_info_vec): Delete.
8607         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8608         (vectorize_loops): Likewise.
8609         (vec_info::~vec_info): Remove argument from call to
8610         free_stmt_vec_infos.
8611         (vec_info::add_stmt): Remove vinfo argument from call to
8612         new_stmt_vec_info.
8614 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8616         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8617         rather than a gimple stmt.
8618         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
8619         information for pattern statements when passed the original
8620         statement; instead wait to be passed the pattern statement itself.
8621         Don't call set_vinfo_for_stmt here.
8622         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8623         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8624         stmt_vec_infos here.
8625         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8626         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8627         stmt_vec_infos entry.
8629 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8631         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8632         * tree-vectorizer.c (vec_info::replace_stmt): New function.
8633         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8634         * tree-vect-stmts.c (vectorizable_call): Likewise.
8635         (vectorizable_simd_clone_call): Likewise.
8637 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8639         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8640         * tree-vectorizer.c (vec_info::remove_stmt): New function.
8641         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8642         * tree-vect-loop.c (vect_transform_loop): Likewise.
8643         * tree-vect-slp.c (vect_schedule_slp): Likewise.
8644         * tree-vect-stmts.c (vect_remove_stores): Likewise.
8646 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8648         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8649         (vect_dr_stmt): Delete.
8650         * tree-vectorizer.c (vec_info::lookup_dr): New function.
8651         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8652         of DR_VECT_AUX.
8653         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8654         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8655         (vect_verify_datarefs_alignment, vect_peeling_supportable)
8656         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8657         (vect_analyze_data_refs): Likewise.
8658         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
8659         argument.
8660         (vect_find_same_alignment_drs): Likewise.
8661         (vect_slp_analyze_node_dependences): Update calls accordingly.
8662         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8663         instead of DR_VECT_AUX.
8664         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8665         of a vector data references.  Use vec_info::lookup_dr instead of
8666         DR_VECT_AUX.
8667         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8668         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8669         instead of DR_VECT_AUX.
8671 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8673         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8674         dr_vec_info.
8675         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8676         accordingly.
8677         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8678         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8679         (vect_gen_prolog_loop_niters): Likewise.
8681 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8683         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8684         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8685         (vect_known_alignment_in_bytes, vect_dr_behavior)
8686         (vect_get_scalar_dr_size): Take references as dr_vec_infos
8687         instead of data_references.  Update calls to other routines for
8688         which the same change has been made.
8689         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8690         dr_vec_infos instead of stmt_vec_infos.
8691         (vect_analyze_data_ref_dependence): Update call accordingly.
8692         (vect_slp_analyze_data_ref_dependence)
8693         (vect_record_base_alignments): Use DR_VECT_AUX.
8694         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8695         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8696         (vector_alignment_reachable_p, vect_get_data_access_cost)
8697         (vect_peeling_supportable, vect_analyze_group_access_1)
8698         (vect_analyze_group_access, vect_analyze_data_ref_access)
8699         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8700         (vect_compile_time_alias, vect_small_gap_p)
8701         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8702         (vect_supportable_dr_alignment): Take references as dr_vec_infos
8703         instead of data_references.  Update calls to other routines for
8704         which the same change has been made.
8705         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8706         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8707         (vect_slp_analyze_and_verify_node_alignment)
8708         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8709         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8710         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
8711         above changes.
8712         (_vect_peel_info::dr): Replace with...
8713         (_vect_peel_info::dr_info): ...this new field.
8714         (vect_peeling_hash_get_most_frequent)
8715         (vect_peeling_hash_choose_best_peeling): Update accordingly.
8716         (vect_peeling_hash_get_lowest_cost):
8717         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
8718         routines for which the same change has been made.
8719         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
8720         data_reference.
8721         * tree-vect-loop-manip.c (get_misalign_in_elems)
8722         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
8723         above changes.
8724         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8725         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8726         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8727         (get_group_load_store_type, get_negative_load_store_type)
8728         (vect_get_data_ptr_increment, vectorizable_store)
8729         (vectorizable_load): Likewise.
8730         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8731         Update calls to other routines for which the same change has been made.
8733 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8735         * tree-vectorizer.h (vec_info::move_dr): New member function.
8736         (dataref_aux): Rename to...
8737         (dr_vec_info): ...this and add "dr" and "stmt" fields.
8738         (_stmt_vec_info::dr_aux): Update accordingly.
8739         (_stmt_vec_info::data_ref_info): Delete.
8740         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8741         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8742         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8743         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8744         of data_ref.
8745         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
8746         (STMT_VINFO_DR_INFO): New macro.
8747         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8748         (set_dr_misalignment): Update after rename of dataref_aux.
8749         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
8750         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8751         initialization of STMT_VINFO_DATA_REF.
8752         * tree-vectorizer.c (vec_info::move_dr): New function.
8753         * tree-vect-patterns.c (vect_recog_bool_pattern)
8754         (vect_recog_mask_conversion_pattern)
8755         (vect_recog_gather_scatter_pattern): Use it.
8756         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8757         the "dr" and "stmt" fields of dr_vec_info instead of
8758         STMT_VINFO_DATA_REF.
8760 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8762         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8763         (is_pattern_stmt_p): Use it.
8764         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8765         on pattern statements.
8767 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8769         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8770         original stmt as a stmt_vec_info rather than a gimple stmt.
8771         (vect_pattern_recog_1): Take the statement directly as a
8772         stmt_vec_info, rather than via a gimple_stmt_iterator.
8773         Update call to vect_mark_pattern_stmts.
8774         (vect_pattern_recog): Update calls accordingly.
8776 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8778         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8779         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8780         a vect_def_type for the first argument.
8781         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8782         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
8783         operand if it isn't defined by a vectorized statement.
8784         (vect_build_gather_load_calls): Remove the mask_dt argument and
8785         update calls to vect_get_vec_def_for_stmt_copy.
8786         (vectorizable_bswap): Likewise the dt argument.
8787         (vectorizable_call): Update calls to vectorizable_bswap and
8788         vect_get_vec_def_for_stmt_copy.
8789         (vectorizable_simd_clone_call, vectorizable_assignment)
8790         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8791         (vectorizable_comparison): Update calls to
8792         vect_get_vec_def_for_stmt_copy.
8793         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
8794         vect_is_simple_use.
8795         (vect_get_loop_based_defs): Remove dt argument and update call
8796         to vect_get_vec_def_for_stmt_copy.
8797         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8798         and vect_get_vec_def_for_stmt_copy.
8799         (vectorizable_load): Update calls to vect_build_gather_load_calls
8800         and vect_get_vec_def_for_stmt_copy.
8801         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8802         (vectorizable_reduction, vectorizable_live_operation): Update calls
8803         to vect_get_vec_def_for_stmt_copy.
8805 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8807         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8808         and gimple stmt arguments with a stmt_vec_info.
8809         (vect_record_base_alignments): Update calls accordingly.
8810         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8811         and gimple stmt arguments with a stmt_vec_info.
8812         (vect_build_slp_tree_1): Remove vinfo argument and update call
8813         to vect_record_max_nunits.
8814         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8815         and vect_record_max_nunits.
8817 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8819         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8820         file and take a stmt_vec_info instead of a gimple stmt.
8821         (supportable_widening_operation, vect_finish_replace_stmt)
8822         (vect_finish_stmt_generation, vect_get_store_rhs)
8823         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8824         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8825         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8826         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8827         (vect_create_data_ref_ptr, bump_vector_ptr)
8828         (vect_permute_store_chain, vect_setup_realignment)
8829         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8830         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8831         (vectorizable_reduction, vectorizable_induction)
8832         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8833         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8834         than gimple stmts as arguments.
8835         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8836         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8837         (can_group_stmts_p, vect_check_gather_scatter)
8838         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8839         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8840         (vect_permute_load_chain, vect_shift_permute_load_chain)
8841         (vect_transform_grouped_load)
8842         (vect_record_grouped_load_vectors): Likewise.
8843         * tree-vect-loop.c (vect_fixup_reduc_chain)
8844         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8845         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8846         (vectorizable_reduction, vectorizable_induction)
8847         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8848         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8849         (vect_get_load_store_mask): Likewise.
8850         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8851         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8852         * tree-vect-stmts.c (vect_mark_relevant)
8853         (is_simple_and_all_uses_invariant)
8854         (exist_non_indexing_operands_for_use_p, process_use)
8855         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8856         (vect_get_vec_def_for_operand, vect_get_vec_defs)
8857         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8858         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8859         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8860         (get_negative_load_store_type, get_load_store_type)
8861         (vect_check_load_store_mask, vect_check_store_rhs)
8862         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8863         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8864         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8865         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8866         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8867         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8868         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8869         (supportable_widening_operation): Likewise.
8871 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8873         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8874         a stmt_vec_info instead of a gcall.
8875         (vect_check_gather_scatter): Update call accordingly.
8876         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
8877         of a gphi.
8878         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
8879         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
8880         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
8881         instead of a gimple stmt.
8882         (vect_transform_loop): Update calls accordingly.
8883         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
8884         stmt_vec_infos instead of gimple stmts.
8885         (vect_analyze_slp_instance): Update use accordingly.
8886         * tree-vect-stmts.c (read_vector_array, write_vector_array)
8887         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
8888         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
8889         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
8890         (vect_gen_widened_results_half, vect_get_loop_based_defs)
8891         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
8892         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
8893         down to subroutines.
8895 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8897         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
8898         of the worklist from a vector of gimple stmts to a vector of
8899         stmt_vec_infos.
8900         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8901         (vect_mark_stmts_to_be_vectorized): Likewise
8903 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8905         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
8906         statement before passing it to vect_analyze_stmt.
8907         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
8908         the chain of phi vector definitions.  Track the exit phi via its
8909         stmt_vec_info.
8910         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
8911         STMT_VINFO_REDUC_DEF.
8912         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
8913         stmt_vec_infos to handle the statement chains.
8914         (vect_get_slp_defs): Record the first statement in the node
8915         using a stmt_vec_info.
8916         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
8917         statements here and pass their stmt_vec_info down to subroutines.
8918         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
8919         down to vect_finish_stmt_generation.
8920         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
8921         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
8922         stmt_vec_infos to subroutines.
8923         (vect_remove_stores): Use stmt_vec_infos to handle the statement
8924         chains.
8926 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8928         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
8929         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
8930         (vect_permute_store_chain, vect_setup_realignment)
8931         (vect_permute_load_chain, vect_shift_permute_load_chain)
8932         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
8933         stmts internally, and when passing values to other vectorizer routines.
8934         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8935         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
8936         (vect_analyze_loop_operations, get_initial_def_for_reduction)
8937         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
8938         (vectorizable_reduction, vectorizable_induction)
8939         (vectorizable_live_operation, vect_transform_loop_stmt)
8940         (vect_transform_loop): Likewise.
8941         * tree-vect-patterns.c (vect_reassociating_reduction_p)
8942         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
8943         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
8944         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8945         (vect_slp_analyze_node_operations_1): Likewise.
8946         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8947         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
8948         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
8949         (vect_finish_stmt_generation_1, get_group_load_store_type)
8950         (get_load_store_type, vect_build_gather_load_calls)
8951         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8952         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8953         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8954         (vectorizable_store, vectorizable_load, vectorizable_condition)
8955         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
8956         (supportable_widening_operation): Likewise.
8957         (vect_get_vector_types_for_stmt): Likewise.
8958         * tree-vectorizer.h (vect_dr_behavior): Likewise.
8960 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8962         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8963         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
8964         (vect_permute_store_chain, vect_permute_load_chain)
8965         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
8966         repeated stmt_vec_info lookups.
8967         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8968         (vect_update_ivs_after_vectorizer): Likewise.
8969         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
8970         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8971         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
8972         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8973         (vect_bb_slp_scalar_cost): Likewise.
8974         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
8976 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8978         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
8979         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
8980         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
8981         to get gassigns and gcalls, rather than operating on generc gimple
8982         stmts.
8983         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
8984         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
8985         (vectorizable_load, vect_analyze_stmt): Likewise.
8986         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
8988 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8990         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
8991         return stmt_vec_infos rather than gimple stmts.  Do not accept
8992         null arguments.
8993         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8994         of a gimple stmt.
8995         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8996         Update use of get_later_stmt.
8997         (vect_get_constant_vectors): Update call accordingly.
8998         (vect_schedule_slp_instance): Likewise
8999         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
9000         (vect_slp_analyze_instance_dependence): Likewise.
9001         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
9003 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9005         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
9006         (stmt_info_for_cost::stmt_info): ...this new field.
9007         (add_stmt_costs): Update accordingly.
9008         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9009         (vect_get_known_peeling_cost): Likewise.
9010         (vect_estimate_min_profitable_iters): Likewise.
9011         * tree-vect-stmts.c (record_stmt_cost): Likewise.
9013 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9015         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
9016         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9017         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
9018         accordingly.
9019         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
9021 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9023         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
9024         a gimple stmt to a stmt_vec_info.
9025         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
9027 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9029         * tree-vectorizer.h (vec_info::grouped_stores): Change from
9030         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9031         (_loop_vec_info::reduction_chains): Likewise.
9032         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
9033         accordingly.
9034         * tree-vect-slp.c (vect_analyze_slp): Likewise.
9036 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9038         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
9039         a gimple stmt to a stmt_vec_info.
9040         (_stmt_vec_info::next_element): Likewise.
9041         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
9042         (vect_slp_analyze_and_verify_node_alignment)
9043         (vect_analyze_group_access_1, vect_analyze_group_access)
9044         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
9045         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
9046         (vect_supportable_dr_alignment): Update accordingly.
9047         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
9048         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
9049         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
9050         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9051         * tree-vect-slp.c (vect_build_slp_tree_1)
9052         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
9053         (vect_split_slp_store_group, vect_analyze_slp_instance)
9054         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
9055         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
9056         (get_group_load_store_type, get_load_store_type)
9057         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
9058         (vect_transform_stmt, vect_remove_stores): Likewise.
9060 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9062         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
9063         than a gimple stmt.
9064         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9065         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
9066         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
9067         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
9068         (vector_alignment_reachable_p, vect_get_data_access_cost)
9069         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
9070         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
9071         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
9072         (vect_analyze_group_access_1, vect_analyze_group_access)
9073         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
9074         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
9075         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
9076         result of vect_dr_stmt and use the stmt_vec_info instead of
9077         the associated gimple stmt.
9078         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9079         (vect_gen_prolog_loop_niters): Likewise.
9080         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9082 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9084         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
9085         to a vec<stmt_vec_info>.
9086         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
9087         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
9088         vec<stmt_vec_info>.
9089         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
9090         to a vec<stmt_vec_info>.
9091         (bst_traits::value_type, bst_traits::value_type): Likewise.
9092         (bst_traits::hash): Update accordingly.
9093         (vect_get_and_check_slp_defs): Change the stmts parameter from
9094         a vec<gimple *> to a vec<stmt_vec_info>.
9095         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
9096         (vect_build_slp_tree): Likewise.
9097         (vect_build_slp_tree_2): Likewise.  Update uses of
9098         SLP_TREE_SCALAR_STMTS.
9099         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
9100         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
9101         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
9102         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
9103         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
9104         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
9105         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
9106         (vect_get_constant_vectors, vect_get_slp_defs)
9107         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
9108         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
9109         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
9110         instead of gimple stmts.
9111         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
9112         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
9113         (vect_slp_analyze_instance_dependence): Update uses of
9114         SLP_TREE_SCALAR_STMTS.
9115         (vect_slp_analyze_and_verify_node_alignment): Likewise.
9116         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
9117         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
9118         (get_initial_defs_for_reduction): Likewise.
9119         (vect_create_epilog_for_reduction): Likewise.
9120         (vectorize_fold_left_reduction): Likewise.
9121         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
9122         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
9123         (can_vectorize_live_stmts): Likewise.
9125 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9127         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
9128         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9129         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
9130         than gimple stmts.
9131         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
9132         of a gimple stmt.
9133         (gather_scalar_reductions): Update after above interface changes.
9134         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
9135         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
9136         than gimple stmts.
9137         (vect_force_simple_reduction): Likewise.
9138         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
9139         LOOP_VINFO_REDUCTIONS.
9140         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9142 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9144         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
9145         a gimple stmt to a stmt_vec_info.
9146         * tree-vect-loop.c (vect_active_double_reduction_p)
9147         (vect_force_simple_reduction, vectorizable_reduction): Update
9148         accordingly.
9150 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9152         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
9153         vec<gimple *> to a vec<stmt_vec_info>.
9154         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
9155         the reduction_phis argument from a vec<gimple *> to a
9156         vec<stmt_vec_info>.
9157         (vectorizable_reduction): Likewise the phis local variable that
9158         is passed to vect_create_epilog_for_reduction.  Update for new type
9159         of SLP_TREE_VEC_STMTS.
9160         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
9161         (vectorizable_live_operation): Likewise.
9162         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
9163         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
9165 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9167         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
9168         a gimple stmt to a stmt_vec_info.
9169         (vectorizable_condition, vectorizable_live_operation)
9170         (vectorizable_reduction, vectorizable_induction): Pass back the
9171         vectorized statement as a stmt_vec_info.
9172         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
9173         use of STMT_VINFO_VEC_STMT.
9174         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
9175         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
9176         as stmt_vec_infos rather than gimple stmts.
9177         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
9178         to a stmt_vec_info.
9179         (vectorizable_live_operation): Likewise.
9180         (vectorizable_reduction, vectorizable_induction): Likewise,
9181         updating use of STMT_VINFO_VEC_STMT.
9182         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
9183         of STMT_VINFO_VEC_STMT.
9184         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
9185         (vectorizable_simd_clone_call, vectorizable_conversion)
9186         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9187         (vectorizable_store, vectorizable_load, vectorizable_condition)
9188         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
9189         from a gimple stmt to a stmt_vec_info.
9190         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
9191         pointer to a stmt_vec_info to the vectorizable_* routines.
9193 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9195         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
9196         a gimple stmt to a stmt_vec_info.
9197         (is_pattern_stmt_p): Update accordingly.
9198         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
9199         (vect_record_grouped_load_vectors): Likewise.
9200         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
9201         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
9202         (vect_model_reduction_cost): Likewise.
9203         (vect_create_epilog_for_reduction): Likewise.
9204         (vectorizable_reduction, vectorizable_induction): Likewise.
9205         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9206         Return the stmt_vec_info for the pattern statement.
9207         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
9208         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
9209         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9210         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
9211         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
9212         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
9213         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
9214         (free_stmt_vec_info, vect_is_simple_use): Likewise.
9216 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9218         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
9219         (vect_finish_stmt_generation): Likewise.
9220         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9221         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
9222         (vect_build_gather_load_calls): Use the return value of the above
9223         functions instead of a separate call to vinfo_for_stmt.  Use narrow
9224         scopes for the input gimple stmt and wider scopes for the associated
9225         stmt_vec_info.  Use vec_info::lookup_def when setting these
9226         stmt_vec_infos from an SSA_NAME definition.
9227         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9228         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9229         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9230         (vectorizable_store, vectorizable_load, vectorizable_condition)
9231         (vectorizable_comparison): Likewise.
9232         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
9233         (vectorizable_reduction): Likewise.
9235 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9237         * tree-vectorizer.h (vect_is_simple_use): Add an optional
9238         stmt_vec_info * parameter before the optional gimple **.
9239         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9240         (process_use, vect_get_vec_def_for_operand_1): Update callers.
9241         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
9242         * tree-vect-loop.c (vectorizable_reduction): Likewise.
9243         (vectorizable_live_operation): Likewise.
9244         * tree-vect-patterns.c (type_conversion_p): Likewise.
9245         (vect_look_through_possible_promotion): Likewise.
9246         (vect_recog_rotate_pattern): Likewise.
9247         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9249 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9251         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
9252         a typedef to a wrapper class.
9253         (NULL_STMT_VEC_INFO): New macro.
9254         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
9255         (stmt_vec_info::operator*): New function.
9256         (stmt_vec_info::operator gimple *): Likewise.
9257         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
9258         (add_stmt_costs): Likewise.
9259         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
9260         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9261         (vect_get_known_peeling_cost): Likewise.
9262         (vect_estimate_min_profitable_iters): Likewise.
9263         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9264         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
9265         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9266         (vectorizable_store, free_stmt_vec_infos): Likewise.
9267         (new_stmt_vec_info): Change return type of xcalloc to
9268         _stmt_vec_info *.
9270 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9272         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
9273         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
9274         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
9275         a single_imm_use-based sequence.
9276         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
9278 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9280         * tree-vectorizer.h (vec_info::lookup_def): Declare.
9281         * tree-vectorizer.c (vec_info::lookup_def): New function.
9282         * tree-vect-patterns.c (vect_get_internal_def): Use it.
9283         (vect_widened_op_tree): Likewise.
9284         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9285         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
9286         (vectorizable_reduction): Likewise.
9287         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
9288         of a gimple *.
9289         (vect_is_slp_reduction): Update calls accordingly.  Use
9290         vec_info::lookup_def.
9291         (vect_is_simple_reduction): Likewise
9292         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
9294 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9296         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
9297         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
9298         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
9299         of vinfo_for_stmt.
9300         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
9301         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
9302         (vect_update_vf_for_slp, vect_analyze_loop_operations)
9303         (vect_is_slp_reduction, vectorizable_induction)
9304         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
9305         * tree-vect-patterns.c (vect_init_pattern_stmt):
9306         (vect_determine_min_output_precision_1, vect_determine_precisions)
9307         (vect_pattern_recog): Likewise.
9308         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
9309         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
9310         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
9311         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9312         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
9313         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
9314         info field from a loop to a loop_vec_info.
9316 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9318         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
9319         (vec_info::add_stmt): Declare.
9320         * tree-vectorizer.c (vec_info::add_stmt): New function.
9321         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
9322         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
9323         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9324         (vectorizable_induction): Likewise.
9325         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
9326         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9327         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
9328         (vectorizable_load): Likewise.
9329         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9330         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
9331         (vect_recog_gather_scatter_pattern): Likewise.
9332         (append_pattern_def_seq): Likewise.  Remove a check that is
9333         performed by add_stmt itself.
9335 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9337         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
9338         which make_ssa_name was called with new_stmt before new_stmt
9339         had been created.
9341 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9343         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
9344         split out from...
9345         (vect_is_slp_reduction): ...here...
9346         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
9347         that are already known to be false.
9349 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9351         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
9352         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
9353         STMT_VINFO_NUM_SLP_USES when it's true.
9354         (vect_free_slp_instance): Add a final_p parameter and pass it to
9355         vect_free_slp_tree.
9356         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
9357         (vect_analyze_slp_instance): Likewise.
9358         (vect_slp_analyze_operations): Likewise.
9359         (vect_slp_analyze_bb_1): Likewise.
9360         * tree-vectorizer.c (vec_info): Likewise.
9361         * tree-vect-loop.c (vect_transform_loop): Likewise.
9363 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9365         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9366         function is not called for second and subsequent members of
9367         a reduction group.
9369 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9371         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
9372         cases for nested loops from here to ...
9373         (vect_create_epilog_for_reduction): ...here.  Only call
9374         vect_is_simple_use for inner-loop reductions.
9376 2018-07-31  Martin Liska  <mliska@suse.cz>
9378         PR gcov-profile/85338
9379         PR gcov-profile/85350
9380         PR gcov-profile/85372
9381         * profile.c (struct location_triplet): New.
9382         (struct location_triplet_hash): Likewise.
9383         (output_location): Do not output a BB that
9384         is already recorded for a line.
9385         (branch_prob): Use streamed_locations.
9387 2018-07-31  Martin Liska  <mliska@suse.cz>
9389         PR gcov-profile/85370
9390         * coverage.c (coverage_begin_function): Do not mark target
9391         clones as artificial functions.
9393 2018-07-31  Martin Liska  <mliska@suse.cz>
9395         PR gcov-profile/83813
9396         PR gcov-profile/84758
9397         PR gcov-profile/85217
9398         PR gcov-profile/85332
9399         * profile.c (branch_prob): Do not record GOTO expressions
9400         for GIMPLE statements which locations are already streamed.
9402 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9404         * gcc.c (handle_spec_function): Accept a soft_matched_part
9405         argument, as do_spec_1.  Pass it down to ...
9406         (eval_spec_function): Accept a soft_matched_part argument,
9407         and pass it down to ...
9408         (do_spec_2): Accept a soft_matched_part argument, and pass
9409         it down to do_spec_1.
9410         (do_spec_1): Pass soft_matched_part to handle_spec_function.
9411         (handle_braces): Update call to handle_spec_function.
9412         (driver::set_up_specs): Update calls to do_spec_2.
9413         (compare_debug_dump_opt_spec_function): Likewise.
9414         (compare_debug_self_opt_spec_function): Likewise.
9416 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9418         * common.opt (nolibc): New option.
9419         * doc/invoke.texi (Link Options): Document it.
9420         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
9421         * config/alpha/linux.h: Likewise.
9422         * config/arc/elf.h: Likewise.
9423         * config/arm/uclinux-elf.h: Likewise.
9424         * config/arm/unknown-elf.h: Likewise.
9425         * config/avr/avrlibc.h: Likewise.
9426         * config/bfin/bfin.h: Likewise.
9427         * config/bfin/linux.h: Likewise.
9428         * config/bfin/uclinux.h: Likewise.
9429         * config/darwin.h: Likewise.
9430         * config/darwin10.h: Likewise.
9431         * config/darwin12.h: Likewise.
9432         * config/gnu-user.h: Likewise.
9433         * config/lm32/uclinux-elf.h: Likewise.
9434         * config/pa/pa-hpux11.h: Likewise.
9435         * config/pa/pa64-hpux.h: Likewise.
9436         * config/sparc/sparc.h: Likewise.
9438 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9440         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
9441         undefined variables.
9443 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9445         PR target/86640
9446         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
9447         instead of GEN_INT.
9449 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9451         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9452         terminated string literal.
9454 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9456         PR rtl-optimization/85160
9457         * combine.c (is_just_move): New function.
9458         (try_combine): Allow combining two instructions into two if neither of
9459         the original instructions was a move.
9461 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
9463         PR target/86673
9464         * doc/extend.texi (Global Register Variables): Discourage use of type
9465         qualifiers.
9466         (Local Register Variables): Likewise.
9468 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
9470         PR tree-optimization/86506
9471         * hwint.h (ceil_log2): Resync with hwint.c implementation.
9473 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
9475         PR target/86547
9476         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
9477         hard_regno, make sure no insn between `from` and `to` clobbers it.
9479 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
9480             Tom de Vries  <tdevries@suse.de>
9482         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
9483         (PTX_DEFAULT_RUNTIME_DIM): ... this.
9484         (nvptx_goacc_validate_dims): Set default worker and gang dims to
9485         PTX_DEFAULT_RUNTIME_DIM.
9486         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
9488 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
9490         * config/pa/pa.c (pa_output_addr_vec): Align address table.
9491         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
9492         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9494 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
9496         * config/rs6000/constraints.md (wG constraint): Delete, no longer
9497         used.
9498         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9499         predicate to reflect toc fusion has been deleted.
9500         (toc_fusion_mem_raw): Delete, no longer used.
9501         (toc_fusion_mem_wrapped): Likewise.
9502         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9503         fusion mask bit.
9504         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9505         Delete, no longer used.
9506         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9507         meant to be used for toc fusion.
9508         (rs6000_debug_print_mode): Delete toc fusion debugging.
9509         (rs6000_debug_reg_global): Likewise.
9510         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9511         fusion and secondary reload support that were never used.
9512         (rs6000_option_override_internal): Delete TOC fusion, that was only
9513         partially defined, and it did not work unless you also used the
9514         -mcmodel= switch.
9515         (rs6000_legitimate_address_p): Delete TOC fusion support.
9516         (rs6000_opt_masks): Likewise.
9517         (fusion_wrap_memory_address): Delete function, no longer used.
9518         (fusion_split_address); Delete TOC fusion support.
9519         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9520         longer used with toc fusion being deleted.
9521         (TARGET_TOC_FUSION_FP): Likewise.
9522         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9523         UNSPEC.
9524         (toc fusion spliter): Delete TOC fusion support.
9525         (toc_fusionload_<mode>): Likewise.
9526         (toc_fusionload_di): Likewise.
9527         (fusion_gpr_load_<mode>): Delete generator function, this insn no
9528         longer needs to be named.  Rename predicate to delete TOC fusion.
9529         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9530         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9531         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9532         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9533         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9535 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9537         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9538         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
9539         __int128 in built-in function prototypes.
9540         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9541         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9543 2018-07-27  Martin Sebor  <msebor@redhat.com>
9545         PR tree-optimization/86696
9546         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9547         types, including enums.
9548         (handle_char_store): Be prepared for the above function to fail.
9550 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
9552         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9553         when optimization level is lower than 2 or optimize for size.
9555 2018-07-26  Martin Sebor  <msebor@redhat.com>
9557         PR tree-optimization/86043
9558         PR tree-optimization/86042
9559         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9560         (get_string_cst_length): Rename...
9561         (get_min_string_length): ...to this.  Add argument.
9562         (handle_char_store): Extend to handle multi-character stores by
9563         MEM_REF.
9564         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
9565         * tree.h (initializer_zerop): Add argument.
9567 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
9569         PR middle-end/86660
9570         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9571         declare target to variables if they have always,{to,from,tofrom} map
9572         kinds.
9574 2018-07-26  Martin Liska  <mliska@suse.cz>
9576         PR lto/86548
9577         * lto-wrapper.c: Add linker_output as prefix
9578         for ltrans_output_file.
9580 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
9582         PR rtl-optimization/85805
9583         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9584         value for hard registers if that was written in the same mode.
9586 2018-07-26  Martin Liska  <mliska@suse.cz>
9588         PR gcov-profile/86536
9589         * gcov.c (format_gcov): Use printf format %.*f directly
9590         and do not handle special values.
9592 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9594         * common/config/arc/arc-common.c (arc_option_optimization_table):
9595         Update default optimizations for size.
9597 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9599         * config/arc/arc.md (movsf_insn): Add short instruction selection.
9600         * config/arc/constraints.md (CfZ): New constraint.
9601         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9602         (subsf3_fpu): Likewise.
9603         (cmpsf_fpu): Likewise.
9604         (cmpsf_fpu_uneq): Likewise.
9606 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9608         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9609         accesses as well.
9610         (arc_is_uncached_mem_p): uncached applies to both the variable and
9611         the pointer.
9613 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9615         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9616         register names.
9618 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
9620         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9621         field "m_scopes" from vec to auto_vec.
9623 2018-07-25  Martin Liska  <mliska@suse.cz>
9625         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9626         return type.
9628 2018-07-25  Richard Biener  <rguenther@suse.de>
9630         PR debug/86654
9631         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9632         special wrt context_die late.
9633         (gen_subprogram_die): Re-use DIEs in local scope.
9635 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
9637         PR tree-optimization/86644
9638         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
9640 2018-07-25  Martin Liska  <mliska@suse.cz>
9642         PR middle-end/86645
9643         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9644         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9646 2018-07-25  Martin Liska  <mliska@suse.cz>
9648         PR sanitizer/79635
9649         * params.def: Explain ASan abbreviation and provide
9650         a documentation link.
9652 2018-07-24  Martin Sebor  <msebor@redhat.com>
9654         PR tree-optimization/86622
9655         PR tree-optimization/86532
9656         * builtins.h (string_length): Declare.
9657         * builtins.c (c_strlen): Correct handling of non-constant offsets.
9658         (check_access): Be prepared for non-constant length ranges.
9659         (string_length): Make extern.
9660         * expr.c (string_constant): Only handle the minor non-constant
9661         array index.  Use string_constant to compute the length of
9662         a generic string constant.
9664 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
9666         PR tree-optimization/86618
9667         * tree-vect-stmts.c (vectorizable_call): Don't take the address
9668         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9670 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
9672         PR tree-optimization/86636
9673         * json.cc (json::object::set): Fix comment.  Add assertions.
9674         (json::array::append): Move here from json.h.  Add comment and an
9675         assertion.
9676         (json::string::string): Likewise.
9677         * json.h (json::array::append): Move to json.cc.
9678         (json::string::string): Likewise.
9679         * optinfo-emit-json.cc
9680         (optrecord_json_writer::impl_location_to_json): Assert that we
9681         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9682         wrapper around it.  Expand the location once, rather than three
9683         times.
9684         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9685         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9686         wrappers.
9687         (optrecord_json_writer::optinfo_to_json): Likewise, in four
9688         places.  Fix some overlong lines.
9690 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
9692         * config/aarch64/aarch64-simd.md
9693         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9694         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9695         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9696         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9697         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9698         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9699         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9700         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9701         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9703 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
9705         PR middle-end/86627
9706         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9707         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
9708         and abs_d == d, do the power of two handling if profitable.
9710 2018-07-24  Richard Biener  <rguenther@suse.de>
9712         * match.pd: Add BIT_FIELD_REF canonicalizations.
9714 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9716         PR c/86617
9717         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9719 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9721         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9722         terminated STRING_CST object.
9724 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9726         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9728 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9730         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9731         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9732         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9733         the elements into a register.
9734         (rs6000_split_v4si_init_di_reg): Delete.
9735         (rs6000_split_v4si_init): Delete.
9736         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9737         (vsx_init_v4si): Rewrite as a define_expand.
9739 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9741         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9742         zero_extend argument from memory): New.
9744 2018-07-22  Martin Sebor  <msebor@redhat.com>
9746         PR bootstrap/86621
9747         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9748         diagnosing calls with unknown arguments unless -Walloca-larger-than
9749         is restricted to less than PTRDIFF_MAX bytes.
9751 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
9753         * doc/gcov.texi (Invoking Gcov): Editorial changes.
9755 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9757         * pretty-print.c (text_info::set_location): Remove redundant
9758         "line_table" parameter from call to rich_location::set_range.
9760 2018-07-20  Martin Sebor  <msebor@redhat.com>
9762         PR middle-end/82063
9763         * builtins.c (expand_builtin_alloca): Adjust.
9764         * calls.c (alloc_max_size): Simplify.
9765         * cgraphunit.c (cgraph_node::expand): Adjust.
9766         * common.opt (larger_than_size, warn_frame_larger_than): Remove
9767         variables.
9768         (frame_larger_than_size): Same.
9769         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9770         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9771         Initialize.
9772         * doc/invoke.texi (GCC Command Options): Document option arguments.
9773         Explain byte-size arguments and suffixes.
9774         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9775         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9776         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9777         * doc/options.texi (UInteger): Expand.
9778         (Host_Wide_Int, ByteSize): Document new properties.
9779         * final.c (final_start_function_1): Include sizes in an error message.
9780         * function.c (frame_offset_overflow): Same.
9781         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9782         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9783         Diagnose unbounded alloca calls only for limits of less than
9784         PTRDIFF_MAX.
9785         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
9786         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
9787         for alloca(0).
9788         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
9789         only for limits of less than PTRDIFF_MAX.
9790         * langhooks-def.h (lhd_handle_option): Change function argument
9791         to HOST_WIDE_INT.
9792         * langhooks.c (lhd_handle_option): Same.
9793         * langhooks.h (handle_option): Same.
9794         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9795         ByteSize flags.
9796         (var_type, var_type_struct): Same.
9797         (var_set): Handle ByteSize flag.
9798         * optc-gen.awk: Add comments to output to ease debugging.  Make
9799         use of HOST_WIDE_INT where appropriate.
9800         * opts-gen-save.awk:  Use %lx to format unsigned long.
9801         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9802         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9803         arguments.  Parse bytes-size suffixes.
9804         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9805         (enum_value_to_arg): Same.
9806         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
9807         (handle_option): Adjust.
9808         (generate_option): Change function argument to HOST_WIDE_INT.
9809         (cmdline_handle_error): Adjust.
9810         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9811         (set_option): Change function argument to HOST_WIDE_INT.
9812         (option_enabled): Handle cl_host_wide_int.
9813         (get_option_state): Handle CLVC_SIZE.
9814         (control_warning_option): Same.
9815         * opts.c (common_handle_option): Change function argument to
9816         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
9817         OPT_Wvla_larger_than_.
9818         * opts.h (enum cl_var_type): Add an enumerator.
9819         * stor-layout.c (layout_decl): Print a more meaningful warning.
9820         * toplev.c (output_stack_usage): Adjust.
9822 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
9824         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9825         call to inline_expand_builtin_string_cmp.
9826         (expand_builtin_strcmp): Likewise.
9827         (expand_builtin_strncmp): Likewise.
9828         (inline_string_cmp): Delete the last parameter, change char_type_node
9829         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9830         two operands.
9831         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9832         the inlining expansion on target where the type of the call has same or
9833         narrower precision than unsigned char.
9835 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9837         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9838         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9839         * common.opt (fsave-optimization-record): New option.
9840         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9841         * doc/invoke.texi (-fsave-optimization-record): New option.
9842         * dumpfile.c: Include "optinfo-emit-json.h".
9843         (struct kv_pair): Move to coretypes.h.
9844         (optgroup_options): Make non-static.
9845         (dump_context::end_scope): Call
9846         optimization_records_maybe_pop_dump_scope.
9847         * dumpfile.h (optgroup_options): New decl.
9848         * json.cc: New file.
9849         * json.h: New file.
9850         * optinfo-emit-json.cc: New file.
9851         * optinfo-emit-json.h: New file.
9852         * optinfo.cc: Include "optinfo-emit-json.h".
9853         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9854         (optinfo_enabled_p): Check optimization_records_enabled_p.
9855         (optinfo_wants_inlining_info_p): Likewise.
9856         * optinfo.h: Update comment.
9857         * profile-count.c (profile_quality_as_string): New function.
9858         * profile-count.h (profile_quality_as_string): New decl.
9859         (profile_count::quality): New accessor.
9860         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9861         and optinfo_emit_json_cc_tests.
9862         * selftest.h (selftest::json_cc_tests): New decl.
9863         (selftest::optinfo_emit_json_cc_tests): New decl.
9864         * toplev.c: Include "optinfo-emit-json.h".
9865         (compile_file): Call optimization_records_finish.
9866         (do_compile): Call optimization_records_start.
9867         * tree-ssa-live.c: Include optinfo.h.
9868         (remove_unused_scope_block_p): Retain inlining information if
9869         optinfo_wants_inlining_info_p returns true.
9871 2018-07-20  Richard Biener  <rguenther@suse.de>
9873         PR debug/86585
9874         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9875         to cover -flto-partition=none.
9877 2018-07-20  Martin Liska  <mliska@suse.cz>
9879         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
9880         (get_decl_source_range): Remove unused function.
9882 2018-07-20  Richard Biener  <rguenther@suse.de>
9884         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
9885         (struct vn_phi_s): Likewise.
9886         (struct vn_reference_s): Likewise.
9887         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
9888         for searching the slot of an entry known to be in the hash itself.
9889         (vn_phi_hasher::equal): Likewise.
9890         (vn_reference_hasher::equal): Likewise.
9891         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
9892         globals.
9893         (optimistic_info, current_info): Remove, keeping only valid_info.
9894         (vn_reference_lookup_1): Remove fallback lookup.
9895         (vn_reference_lookup_2): Likewise.
9896         (vn_nary_op_lookup_1): Likewise.
9897         (vn_phi_lookup): Likewise.
9898         (vn_nary_build_or_lookup_1): Make sure to not chain the built
9899         hash element.
9900         (vn_reference_insert): Adjust, chain the inserted hash element
9901         at last_inserted_ref.
9902         (vn_reference_insert_pieces): Likewise.
9903         (visit_reference_op_call): Likewise.
9904         (vn_nary_op_insert_into): Chain the inserted hash element at
9905         last_inserted_nary.
9906         (vn_nary_op_insert_pieces): Adjust.
9907         (vn_nary_op_insert): Likewise.
9908         (vn_nary_op_insert_stmt): Likewise.
9909         (vn_phi_insert): Adjust, chain the inserted hash element at
9910         last_inserted_phi.
9911         (process_scc): Remove clearing and copying the optimistic
9912         table.  Instead remove elements inserted during an optimistic
9913         iteration from the single table we maintain.
9914         (init_scc_vn): Adjust.
9915         (free_scc_vn): Likewise.
9916         (sccvn_dom_walker::record_cond): Likewise.
9917         (sccvn_dom_walker::after_dom_children): Likewise.
9919 2018-07-19  Martin Sebor  <msebor@redhat.com>
9921         PR tree-optimization/84047
9922         PR tree-optimization/83776
9923         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
9924         (check_array_bounds): Call it.
9926 2018-07-19  Martin Sebor  <msebor@redhat.com>
9928         * align.h (align_flags): Use member initialization.
9930 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
9932         * Makefile.in (OBJS): Add optinfo.o.
9933         * coretypes.h (class symtab_node): New forward decl.
9934         (struct cgraph_node): New forward decl.
9935         (class varpool_node): New forward decl.
9936         * dump-context.h: New file.
9937         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
9938         "tree-pass.h".
9939         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
9940         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
9941         (set_alt_dump_file): Likewise.
9942         (dump_context::~dump_context): New dtor.
9943         (dump_gimple_stmt): Move implementation to...
9944         (dump_context::dump_gimple_stmt): ...this new member function.
9945         Add the stmt to any pending optinfo, creating one if need be.
9946         (dump_gimple_stmt_loc): Move implementation to...
9947         (dump_context::dump_gimple_stmt_loc): ...this new member function.
9948         Start a new optinfo and add the stmt to it.
9949         (dump_gimple_expr): Move implementation to...
9950         (dump_context::dump_gimple_expr): ...this new member function.
9951         Add the stmt to any pending optinfo, creating one if need be.
9952         (dump_gimple_expr_loc): Move implementation to...
9953         (dump_context::dump_gimple_expr_loc): ...this new member function.
9954         Start a new optinfo and add the stmt to it.
9955         (dump_generic_expr): Move implementation to...
9956         (dump_context::dump_generic_expr): ...this new member function.
9957         Add the tree to any pending optinfo, creating one if need be.
9958         (dump_generic_expr_loc): Move implementation to...
9959         (dump_context::dump_generic_expr_loc): ...this new member
9960         function.  Add the tree to any pending optinfo, creating one if
9961         need be.
9962         (dump_printf): Move implementation to...
9963         (dump_context::dump_printf_va): ...this new member function.  Add
9964         the text to any pending optinfo, creating one if need be.
9965         (dump_printf_loc): Move implementation to...
9966         (dump_context::dump_printf_loc_va): ...this new member function.
9967         Start a new optinfo and add the stmt to it.
9968         (dump_dec): Move implementation to...
9969         (dump_context::dump_dec): ...this new member function.  Add the
9970         value to any pending optinfo, creating one if need be.
9971         (dump_context::dump_symtab_node): New member function.
9972         (dump_context::get_scope_depth): New member function.
9973         (dump_context::begin_scope): New member function.
9974         (dump_context::end_scope): New member function.
9975         (dump_context::ensure_pending_optinfo): New member function.
9976         (dump_context::begin_next_optinfo): New member function.
9977         (dump_context::end_any_optinfo): New member function.
9978         (dump_context::s_current): New global.
9979         (dump_context::s_default): New global.
9980         (dump_scope_depth): Delete global.
9981         (dumpfile_ensure_any_optinfo_are_flushed): New function.
9982         (dump_symtab_node): New function.
9983         (get_dump_scope_depth): Reimplement in terms of dump_context.
9984         (dump_begin_scope): Likewise.
9985         (dump_end_scope): Likewise.
9986         (selftest::temp_dump_context::temp_dump_context): New ctor.
9987         (selftest::temp_dump_context::~temp_dump_context): New dtor.
9988         (selftest::verify_item): New function.
9989         (ASSERT_IS_TEXT): New macro.
9990         (ASSERT_IS_TREE): New macro.
9991         (ASSERT_IS_GIMPLE): New macro.
9992         (selftest::test_capture_of_dump_calls): New test.
9993         (selftest::dumpfile_c_tests): Call it.
9994         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
9995         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
9996         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
9997         descriptive comment.
9998         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
9999         (dump_node, dump_bb): Move these unrelated decls.
10000         (class dump_manager): Add leading comment.
10001         * optinfo.cc: New file.
10002         * optinfo.h: New file.
10004 2018-07-19  Michael Collison  <michael.collison@arm.com>
10005             Richard Henderson <rth@redhat.com>
10007         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
10008         (subti): Handle op1 zero.
10009         (subvti4, usub4ti4): New.
10010         (*sub<GPI>3_compare1_imm): New.
10011         (sub<GPI>3_carryinCV): New.
10012         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
10013         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
10015 2018-07-19  Michael Collison  <michael.collison@arm.com>
10016             Richard Henderson <rth@redhat.com>
10018         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
10019         (addti3): Create simpler code if low part is already known to be 0.
10020         (addvti4, uaddvti4): New.
10021         (*add<GPI>3_compareC_cconly_imm): New.
10022         (*add<GPI>3_compareC_cconly): New.
10023         (*add<GPI>3_compareC_imm): New.
10024         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
10025         handle constants within this pattern..
10026         (*add<GPI>3_compareV_cconly_imm): New.
10027         (*add<GPI>3_compareV_cconly): New.
10028         (*add<GPI>3_compareV_imm): New.
10029         (add<GPI>3_compareV): New.
10030         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
10031         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
10032         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
10033         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
10034         with 'comparison' operator.
10035         (*add<GPI>3_compareV_cconly_imm): Ditto.
10036         (*add<GPI>3_compareV_cconly): Ditto.
10037         (*add<GPI>3_compareV_imm): Ditto.
10038         (add<GPI>3_compareV): Ditto.
10039         (add<mode>3_carryinC): Ditto.
10040         (*add<mode>3_carryinC_zero): Ditto.
10041         (*add<mode>3_carryinC): Ditto.
10042         (add<mode>3_carryinV): Ditto.
10043         (*add<mode>3_carryinV_zero): Ditto.
10044         (*add<mode>3_carryinV): Ditto.
10046 2018-07-19  Michael Collison  <michael.collison@arm.com>
10047             Richard Henderson <rth@redhat.com>
10049         * config/aarch64/aarch64-modes.def (CC_V): New.
10050         * config/aarch64/aarch64-protos.h
10051         (aarch64_addti_scratch_regs): Declare
10052         (aarch64_subvti_scratch_regs): Declare.
10053         (aarch64_expand_subvti): Declare.
10054         (aarch64_gen_unlikely_cbranch): Declare
10055         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
10056         for signed overflow using CC_Vmode.
10057         (aarch64_get_condition_code_1): Handle CC_Vmode.
10058         (aarch64_gen_unlikely_cbranch): New function.
10059         (aarch64_addti_scratch_regs): New function.
10060         (aarch64_subvti_scratch_regs): New function.
10061         (aarch64_expand_subvti): New function.
10063 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10065         * config/aarch64/aarch64-option-extensions.def: New entry for profile
10066         extension.
10067         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
10068         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
10069         extension.
10071 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10073         PR target/83009
10074         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10075         address check not strict.
10077 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10079         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
10080         Umq with Umn.
10081         (store_pair_lanes<mode>): Likewise.
10082         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
10083         enum value 'ADDR_QUERY_LDP_STP_N'.
10084         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
10085         (aarch64_print_address_internal): Add declaration.
10086         (aarch64_print_ldpstp_address): Remove.
10087         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
10088         (aarch64_print_operand): Change printing of 'y'.
10089         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
10090         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
10091         'true' rather than '1'.
10092         * config/aarch64/constraints.md (Uml): Likewise.
10093         (Uml): Rename to Umn.
10094         (Umq): Remove.
10096 2018-07-19  Richard Biener  <rguenther@suse.de>
10098         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
10099         a trailing array.
10100         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
10101         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
10102         (vn_reference_hasher): Likewise.
10103         (struct vn_tables_s): Remove obstack and alloc-pool members.
10104         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
10105         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
10106         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
10107         (vn_reference_insert_pieces): Likewise.
10108         (alloc_vn_nary_op_noinit): Adjust.
10109         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
10110         (vn_phi_eq): Adjust.
10111         (shared_lookup_phiargs): Remove.
10112         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
10113         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
10114         (visit_reference_op_call): Likewise.
10115         (copy_nary, copy_phi, copy_reference): Remove.
10116         (process_scc): Rewind the obstack when iterating.  Do not
10117         copy the elements to valid_info but just move them from one
10118         hashtable to the other.
10119         (allocate_vn_table): Adjust.
10120         (free_vn_table): Likewise.
10121         (init_scc_vn): Likewise.
10122         (free_scc_vn): Likewise.
10124 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
10126         PR target/86560
10127         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
10128         indirect_return as function type attribute.
10129         (ix86_attribute_table): Change indirect_return to function
10130         type attribute.
10131         * doc/extend.texi: Update indirect_return attribute.
10133 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
10135         * wide-int.h (widest2_int): New.
10136         * gimple-fold.c (arith_overflowed_p): Use it.
10137         * tree.h (widest2_int_cst): New.
10138         * tree-vrp.c (wide_int_binop_overflow): Rename from
10139         vrp_int_const_binop.
10140         Rewrite to work on trees.
10141         (extract_range_from_multiplicative_op_1): Abstract code to...
10142         (wide_int_range_min_max): ...here.
10143         (wide_int_range_cross_product): ...and here.
10144         (extract_range_from_binary_expr_1): Abstract overflow code to...
10145         (wide_int_range_mult_wrapping): ...here.
10146         * tree-vrp.h (wide_int_range_cross_product): New.
10147         (wide_int_range_mult_wrapping): New.
10149 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
10150             Julia Koval  <julia.koval@intel.com>
10152         * config/i386/x86-tune-costs.h (skylake_memcpy,
10153         skylake_memset): Replace rep_prefix with unrolling for size 512.
10155 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
10157         PR middle-end/86544
10158         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
10159         comparision with EQ_EXPR in last stmt.
10161 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10163         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
10164         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
10165         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
10166         previously known as "PowerPC AltiVec Built-in Functions".  Move
10167         some material to new subsubsections "PowerPC AltiVec Built-in
10168         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
10169         ISA 2.07".
10170         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
10171         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
10172         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
10173         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
10175 2018-07-18  Richard Biener  <rguenther@suse.de>
10177         PR tree-optimization/86557
10178         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
10179         EXACT_DIV_EXPR.
10181 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
10183         * config/s390/s390.c (s390_function_profiler): Generate CFI.
10185 2018-07-17  Jeff Law  <law@redhat.com>
10187         * config/arm/arm.c (get_label_padding): Update for recent
10188         changes to label_to_alignment.
10190         PR tree-optimization/86010
10191         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
10193         * config/mips/mips.c (vr4130_align_insns): Update for recent
10194         changes to label_to_alignment.
10196         * config/frv/frv.c (frv_label_align): Update for recent changes
10197         to label_to_alignment.
10199         * config/nios2/nios2.c (nios2_label_align): Update for recent
10200         changes which dropped ALIGN_LABELS_LOG.
10202 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
10204         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
10205         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
10207 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
10209         * config/arc/arc.c (arc_label_align): Use align_labels instead of
10210         deprecated align_labels_log.
10212 2018-07-17  Richard Biener  <rguenther@suse.de>
10214         PR lto/86456
10215         * dwarf2out.c (init_sections_and_labels): Always generate
10216         a debug_line_str_section for early LTO debug.
10217         (dwarf2out_finish): Reset debug_line_str_hash output early.
10218         Bump counter for extra dwarf5 .debug_loc labels to not conflict
10219         with fat LTO part.
10220         (dwarf2out_early_finish): Output debug_line_str.
10222 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10224         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
10225         index register on z196 or later.
10227 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10229         * config/s390/s390.c (s390_default_align): Set default function
10230         alignment to 16.
10231         (s390_override_options_after_change): Call s390_default align.
10232         (s390_option_override_internal): Call s390_default_align.
10233         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
10235 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10237         PR middle-end/86542
10238         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
10239         to _looptemp_ clauses, other than the first two.
10241 2018-07-17  Martin Liska  <mliska@suse.cz>
10243         * opts.c: Do not enable OPT_falign_* for -Os.
10245 2018-07-17  Martin Liska  <mliska@suse.cz>
10247         * align.h (MAX_CODE_ALIGN): New.
10248         (MAX_CODE_ALIGN_VALUE): New.
10249         * common/config/i386/i386-common.c (ix86_handle_option):
10250         (MAX_CODE_ALIGN): Moved to align.h.
10251         * final.c (MAX_CODE_ALIGN): Likewise.
10252         * opts.c (parse_and_check_align_values):
10253         (MAX_CODE_ALIGN): Likewise.
10254         (MAX_CODE_ALIGN_VALUE): Likewise.
10256 2018-07-17  Martin Liska  <mliska@suse.cz>
10258         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
10259         in order to fulfil coding style.
10260         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
10261         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10262         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10263         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
10264         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
10265         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
10266         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
10267         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
10268         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10270 2018-07-17  Martin Liska  <mliska@suse.cz>
10272         * align.h: New file.
10273         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
10274         directly.
10275         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
10276         align_flags of label_to_alignment.
10277         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
10278         align_flags class.
10279         * config/m68k/m68k.c: Do not use removed align_labels_value and
10280         align_loops_value.
10281         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
10282         (LOOP_ALIGN): Likewise.
10283         (LABEL_ALIGN): Likewise.
10284         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10285         Remove not used macro.
10286         (rs6000_loop_align): Change return type to align_flags.
10287         (rs6000_loop_align_max_skip): Remove.
10288         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
10289         Change return type to align_flags.
10290         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10291         Remove not used macro.
10292         (rs6000_loop_align):  Change return type to align_flags.
10293         (rs6000_loop_align_max_skip): Remove.
10294         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
10295         * config/rx/rx-protos.h (rx_align_for_label): Make it
10296         static function.
10297         * config/rx/rx.c (rx_align_for_label): Change return type
10298         to align_flags.
10299         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
10300         macro definitions.
10301         into align_flags class.
10302         (LABEL_ALIGN): Likewise.
10303         (LOOP_ALIGN): Likewise.
10304         * config/s390/s390.c (s390_label_align): Use align_flags
10305         class member.
10306         (s390_asm_output_function_label): Likewise.
10307         * config/sh/sh.c (sh_override_options_after_change):
10308         Use align_flags class directly without macros.
10309         (find_barrier): Likewise.
10310         (barrier_align): Likewise.
10311         (sh_loop_align): Likewise.
10312         * config/spu/spu.c (spu_option_override):
10313         Use align_flags_tuple::get_value instead of removed macros.
10314         (spu_sched_init): Likewise.
10315         * config/spu/spu.h (GTY): Likewise.
10316         * config/visium/visium.c (visium_option_override):
10317         Set "8" as default secondary alignment.
10318         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
10319         in order to guarantee secondary alignment of 8.
10320         * coretypes.h: Include align.h header file.
10321         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
10322         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
10323         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
10324         * doc/tm.texi.in: Likewise.
10325         * final.c (struct label_alignment): Remove not used structure.
10326         (LABEL_ALIGN): Change type to align_flags.
10327         (LOOP_ALIGN): Likewise.
10328         (JUMP_ALIGN): Likewise.
10329         (default_loop_align_max_skip): Remove.
10330         (default_label_align_max_skip): Likewise.
10331         (default_jump_align_max_skip): Likewise.
10332         (default_label_align_after_barrier_max_skip):
10333         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
10334         (LABEL_TO_MAX_SKIP): Remove.
10335         (label_to_alignment): Return align_flags type instead of integer.
10336         (label_to_max_skip): Remove.
10337         (align_fuzz): Use align_flags type.
10338         (compute_alignments): Use align_flags type and use align_flags::max
10339         to combine multiple alignments.
10340         (grow_label_align): Grow vec instead of C array.
10341         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
10342         (shorten_branches):  Use align_flags type and use align_flags::max
10343         to combine multiple alignments.
10344         (final_scan_insn_1): Remove usage of secondary alignment that comes
10345         from label alignment, but instead use proper secondary alignment
10346         which is computed in grow_label_align.
10347         * flags.h (struct align_flags_tuple): Move to align.h.
10348         (struct align_flags): Likewise.
10349         (state_align_loops): Rename to align_loops.
10350         (state_align_jumps): Rename to align_jumps.
10351         (state_align_labels): Rename to align_labels.
10352         (state_align_functions): Rename to align_functions.
10353         (align_loops_log): Remove.
10354         (align_jumps_log): Remove.
10355         (align_labels_log): Remove.
10356         (align_functions_log): Remove.
10357         (align_loops_max_skip): Remove.
10358         (align_jumps_max_skip): Remove.
10359         (align_labels_max_skip): Remove.
10360         (align_functions_max_skip): Remove.
10361         (align_loops_value): Remove.
10362         (align_jumps_value): Remove.
10363         (align_labels_value): Remove.
10364         (align_functions_value): Remove.
10365         * output.h (label_to_alignment): Change return type to align_flags.
10366         (label_to_max_skip): Remove.
10367         * target.def: Remove loop_align_max_skip, label_align_max_skip,
10368         jump_align_max_skip macros.
10369         * targhooks.h (default_loop_align_max_skip): Remove.
10370         (default_label_align_max_skip): Likewise.
10371         (default_jump_align_max_skip): Likewise.
10372         (default_label_align_after_barrier_max_skip): Remove.
10373         * toplev.c (read_log_maxskip): Use ::normalize function.
10374         (parse_N_M): Remove not used argument and also call ::normalize.
10375         (parse_alignment_opts): Do not pass unused arguments.
10376         * varasm.c (assemble_start_function): Use directly align_functions
10377         instead of removed macros.
10378         * system.h: Do not poison removed macros.
10380 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10382         PR middle-end/86539
10383         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
10384         and cond temporaries don't have reference type if iterator has
10385         pointer type.  For init use &for_pre_body instead of pre_p if
10386         for_pre_body is non-empty.
10388 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10390         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
10391         double-double modes to SFmode directly directly.
10392         (trunc<mode>sf2_fprs): Delete.
10394 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10396         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
10397         for conversions between IFmode and the decimal floating point modes.
10398         (init_float128_ieee): Use the correct names for conversions between
10399         KFmode and the decimal floating point modes.
10401 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10403         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
10404         for the conversions between TDmode and IFmode.
10405         (init_float128_ieee): Use more correct names for the conversions
10406         between TDmode and KFmode.
10408 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
10410         PR tree-optimization/86526
10411         * builtins.c (expand_builtin_memcmp): Formatting fixes.
10412         (inline_expand_builtin_string_cmp): Likewise.
10413         (inline_string_cmp): Likewise.  Use c_readstr instead of
10414         builtin_memcpy_read_str.  Add unit_mode temporary.
10416 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10418         PR middle-end/86528
10419         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
10420         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
10422 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10424         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
10425         Alphabetize prototypes of built-in functions, separating out
10426         built-in functions that are listed in this section but should be
10427         described elsewhere.
10429 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
10431         PR target/86511
10432         * expmed.c (emit_store_flag): Do not emit setcc followed by a
10433         conditional move when trapping comparison was split to a
10434         non-trapping one (and vice versa).
10436 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10438         * config/s390/s390.c (s390_function_profiler): Generate nops
10439         instead of profiler call sequences.
10440         * config/s390/s390.opt: Add the new option.
10442 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10444         * config/s390/s390.c (s390_function_profiler): Generate
10445         __mcount_loc section.
10446         * config/s390/s390.opt: Add the new option.
10448 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10450         * common.opt: Add the new warning.
10451         * config/s390/s390.c (s390_function_profiler): Emit "brasl
10452         %r0,__fentry__" when -mfentry is specified.
10453         (s390_option_override_internal): Disallow -mfentry for 31-bit
10454         CPUs.
10455         * config/s390/s390.opt: Add the new option.
10457 2018-07-16  Richard Biener  <rguenther@suse.de>
10459         PR lto/86523
10460         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
10461         for function-local FUNCTION_DECL and RESULT_DECL immediately.
10463 2018-07-16  Martin Liska  <mliska@suse.cz>
10465         PR ipa/86529
10466         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
10467         to ::get_create.
10469 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10471         * config/arc/arcHS.md: Update ARCHS scheduling rules.
10473 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10475         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
10476         for ARCHS4x.
10477         * config/arc/arc-cpus.def (hs4x): New cpu.
10478         (hs4xd): Likewise.
10479         * config/arc/arc-tables.opt: Regenerate.
10480         * config/arc/arc.c (arc_sched_issue_rate): New function.
10481         (TARGET_SCHED_ISSUE_RATE): Define.
10482         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
10483         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
10484         fpu_cvt.
10485         (attr tune): Add ARCHS4x tune values.
10486         (attr tune_dspmpy): Define.
10487         (*tst): Correct instruction type.
10488         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
10489         * config/arc/arcHS4x.md: New file.
10490         * config/arc/fpu.md: Update instruction type attributes.
10491         * config/arc/t-multilib: Regenerate.
10493 2018-07-16  Tom de Vries  <tdevries@suse.de>
10495         PR debug/86455
10496         * var-tracking.c (vt_initialize): Fix pre_dec handling.
10498 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
10500         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10501         early clobber.
10503 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
10505         PR tree-optimization/86514
10506         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10507         conversion to a boolean type from a type with greater precision.
10509 2018-07-16  Tom de Vries  <tdevries@suse.de>
10511         * var-tracking.c (vt_initialize): Print adjusted insn slim if
10512         dump_flags request TDF_SLIM.
10514 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
10516         * fold-const.c (int_const_binop_1): Abstract...
10517         (wide_int_binop): ...wide int code here.
10518         (poly_int_binop): ...poly int code here.
10519         Abstract the rest of int_const_binop_1 into int_const_binop.
10520         * fold-const.h (wide_int_binop): New.
10521         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10522         Remove useless PLUS/MINUS_EXPR case.
10523         (zero_nonzero_bits_from_vr): Move wide int code...
10524         (zero_nonzero_bits_from_bounds): ...here.
10525         (extract_range_from_binary_expr_1): Move mask optimization code...
10526         (range_easy_mask_min_max): ...here.
10527         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10528         (range_easy_mask_min_max): New.
10530 2018-07-15  Jeff Law  <law@redhat.com>
10532         PR target/85993
10533         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10534         block.
10536 2018-07-14  Jim Wilson  <jimw@sifive.com>
10538         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10540 2018-07-14  Paul Koning  <ni1d@arrl.net>
10542         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10544 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
10546         * lto-streamer-out.c (copy_function_or_variable): Dump info about
10547         copying section.
10549 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
10550             Steve Munroe  <munroesj52@gmail.com>
10552         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10553         (_mm_andnot_si128): Likewise.
10554         (_mm_or_si128): Likewise.
10555         (_mm_xor_si128): Likewise.
10557 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
10559         PR middle-end/78809
10560         * builtins.c (expand_builtin_memcmp): Inline the calls first
10561         when result_eq is false.
10562         (expand_builtin_strcmp): Inline the calls first.
10563         (expand_builtin_strncmp): Likewise.
10564         (inline_string_cmp): New routine. Expand a string compare
10565         call by using a sequence of char comparison.
10566         (inline_expand_builtin_string_cmp): New routine. Inline expansion
10567         a call to str(n)cmp/memcmp.
10568         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10569         New option.
10570         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10572 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
10574         * config/arm/driver-arm.c: Include arm-native.h.
10575         (host_detect_local_cpu): Use auto-generated data tables.
10576         (vendors, arm_cpu_table): Delete.  Move part information to ...
10577         * config/arm/arm-cpus.in: ... here.
10578         * config/arm/parsecpu.awk (gen_native): New function.
10579         (vendor, part): New CPU fields.
10580         (END): Add support for building the native CPU detection tables.
10581         * config/arm/t-arm (arm-native.h): Add build rule.
10582         (driver-arm.o): Add dependency on arm-native.h.
10584 2018-07-13  Richard Biener  <rguenther@suse.de>
10586         PR middle-end/85974
10587         * match.pd (addr1 - addr2): Allow either of the operand to
10588         have a conversion.
10590 2018-07-13  Tom de Vries  <tdevries@suse.de>
10592         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10593         in remap_ssa_name.
10595 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
10597         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10598         arrays instead of numbered variables.
10600 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
10602         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10603         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10605 2018-07-13  Richard Biener  <rguenther@suse.de>
10607         PR debug/86452
10608         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10609         instead of get_context_die.
10611 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
10612             Richard Biener  <rguenther@suse.de>
10614         PR middle-end/86489
10615         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10616         that the loop latch destination where phi is defined.
10618 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
10620         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10621         (riscv_expand_epilogue): Add assertion to check interrupt mode.
10622         (riscv_set_current_function): Extract getting interrupt type to new
10623         function.
10624         (riscv_get_interrupt_type): New function.
10625         (riscv_merge_decl_attributes): New function, checking interrupt type is
10626         same.
10627         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10629 2018-07-12  Paul Koning  <ni1d@arrl.net>
10631         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10632         directive.
10634 2018-07-12  Paul Koning  <ni1d@arrl.net>
10636         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10637         zero reference, add doloop_end instead.
10638         * doc/md.texi (decrement_and_branch_until_zero): Remove.
10639         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
10640         detail for doloop_end.
10642 2018-07-12  Martin Sebor  <msebor@redhat.com>
10644         PR c/86453
10645         * attribs.c (decl_attributes): Reject conflicting attributes before
10646         calling attribute handlers.
10648 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10650         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10651          parameter.
10652         (gcc::dump_manager::get_dump_file_name): likewise.
10653         (dump_begin): Likewise.
10654         * dumpfile.h (dump_begin): Update prototype.
10655         (gcc::dump_manager::get_dump_file_name,
10656         gcc::dump_manager::get_dump_file_name): Update prototype.
10658 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10660         * internal-fn.h (vectorizable_internal_fn_p): New function.
10661         * tree-vect-slp.c (compatible_calls_p): Likewise.
10662         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
10663         to internal functions.
10664         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10666 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10668         * fold-const.h (inverse_conditions_p): Declare.
10669         * fold-const.c (inverse_conditions_p): New function.
10670         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
10671         that test the inverse condition of a conditional internal function.
10672         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10673         * internal-fn.c (internal_fn_mask_index): Handle conditional
10674         internal functions.
10675         (vectorized_internal_fn_supported_p): New function.
10676         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10677         (any_pred_load_store): Replace with...
10678         (need_to_predicate): ...this new variable.
10679         (redundant_ssa_names): New variable.
10680         (ifcvt_can_use_mask_load_store): Move initial checks to...
10681         (ifcvt_can_predicate): ...this new function.  Handle tree codes
10682         for which a conditional internal function exists.
10683         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10684         instead of ifcvt_can_use_mask_load_store.  Update after variable
10685         name change.
10686         (predicate_load_or_store): New function, split out from
10687         predicate_mem_writes.
10688         (check_redundant_cond_expr): New function.
10689         (value_available_p): Likewise.
10690         (predicate_rhs_code): Likewise.
10691         (predicate_mem_writes): Rename to...
10692         (predicate_statements): ...this.  Use predicate_load_or_store
10693         and predicate_rhs_code.
10694         (combine_blocks, tree_if_conversion): Update after above name changes.
10695         (ifcvt_local_dce): Handle redundant_ssa_names.
10696         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10697         general conditional functions.
10698         * tree-vect-stmts.c (vectorizable_call): Likewise.
10700 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10701             Alan Hayward  <alan.hayward@arm.com>
10702             David Sherwood  <david.sherwood@arm.com>
10704         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10705         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10706         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10707         plus and minus and convert them into IFN_COND_FMA-based sequences.
10708         (convert_mult_to_fma): Handle conditional plus and minus.
10710 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10712         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10713         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10714         (cond_fnms_optab): New optabs.
10715         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10716         internal functions.
10717         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10718         * internal-fn.h (get_conditional_internal_fn): Declare.
10719         (get_unconditional_internal_fn): Likewise.
10720         * internal-fn.c (cond_ternary_direct): New macro.
10721         (expand_cond_ternary_optab_fn): Likewise.
10722         (direct_cond_ternary_optab_supported_p): Likewise.
10723         (FOR_EACH_COND_FN_PAIR): Likewise.
10724         (get_conditional_internal_fn): New function.
10725         (get_unconditional_internal_fn): Likewise.
10726         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10727         (gimple_match_op::gimple_match_op): Add a new overload for 5
10728         operands.
10729         (gimple_match_op::set_op): Likewise.
10730         (gimple_resimplify5): Declare.
10731         * genmatch.c (decision_tree::gen): Generate simplifications for
10732         5 operands.
10733         * gimple-match-head.c (gimple_simplify): Define an overload for
10734         5 operands.  Handle calls with 5 arguments in the top-level overload.
10735         (convert_conditional_op): Handle conversions from unconditional
10736         internal functions to conditional ones.
10737         (gimple_resimplify5): New function.
10738         (build_call_internal): Pass a fifth operand.
10739         (maybe_push_res_to_seq): Likewise.
10740         (try_conditional_simplification): Try converting conditional
10741         internal functions to unconditional internal functions.
10742         Handle 3-operand unconditional forms.
10743         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10744         Define ternary equivalents of the current rules for binary conditional
10745         internal functions.
10746         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10747         ternary operations.
10748         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10749         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10750         (optab): Handle them.
10751         (SVE_COND_FP_TERNARY): New int iterator.
10752         (sve_fmla_op, sve_fmad_op): New int attributes.
10753         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10754         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10755         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10757 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10759         * target.def (preferred_else_value): New target hook.
10760         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10761         * doc/tm.texi: Regenerate.
10762         * targhooks.h (default_preferred_else_value): Declare.
10763         * targhooks.c (default_preferred_else_value): New function.
10764         * internal-fn.h (conditional_internal_fn_code): Declare.
10765         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10766         (get_conditional_internal_fn): Use it.
10767         (conditional_internal_fn_code): New function.
10768         * gimple-match.h (gimple_match_cond): New struct.
10769         (gimple_match_op): Add a cond member function.
10770         (gimple_match_op::gimple_match_op): Update all forms to take a
10771         gimple_match_cond.
10772         * genmatch.c (expr::gen_transform): Use the same condition as res_op
10773         for the suboperation, but don't specify a particular else_value.
10774         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10775         (visit_nary_op, visit_reference_op_load): Pass
10776         gimple_match_cond::UNCOND to the gimple_match_op constructor.
10777         * gimple-match-head.c: Include tree-eh.h
10778         (convert_conditional_op): New function.
10779         (maybe_resimplify_conditional_op): Likewise.
10780         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10781         (gimple_resimplify2): Likewise.
10782         (gimple_resimplify3): Likewise.
10783         (gimple_resimplify4): Likewise.
10784         (maybe_push_res_to_seq): Return null for conditional operations.
10785         (try_conditional_simplification): New function.
10786         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
10787         constructor.
10788         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10789         IFN_COND_* call.
10790         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10791         function.
10792         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10794 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10796         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10797         DECL_FCONTEXT
10798         (hash_tree): Do not hash DECL_FCONTEXT
10799         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10800         Do not stream DECL_FCONTEXT.
10801         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10802         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10804 2018-07-12  Richard Biener  <rguenther@suse.de>
10806         PR debug/86462
10807         * dwarf2out.c (gen_block_die): Only output blocks when they have
10808         at least one !DECL_IGNORED_P variable.
10810 2018-07-12  Richard Biener  <rguenther@suse.de>
10812         PR target/84829
10813         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10814         Remove -mieee-fp handling.
10816 2018-07-12  Richard Biener  <rguenther@suse.de>
10818         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10819         left-over from last patch.
10821 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
10823         PR tree-optimization/86492
10824         * gimple-ssa-store-merging.c
10825         (imm_store_chain_info::coalesce_immediate_stores): Call
10826         check_no_overlap even for the merge_overlapping case.  Formatting fix.
10828 2018-07-12  Richard Biener  <rguenther@suse.de>
10830         PR middle-end/86479
10831         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10832         move possibly trapping operations into the conditional.
10834 2018-07-12  Richard Biener  <rguenther@suse.de>
10836         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10837         (vn_lookup_simplify_result): Remove recursion limit applied
10838         here.
10839         (vn_nary_build_or_lookup_1): Adjust.
10840         (try_to_simplify): Likewise.
10841         * gimple-match-head.c (gimple_resimplify1): Instead apply one
10842         here.
10843         (gimple_resimplify2): Likewise.
10844         (gimple_resimplify3): Likewise.
10845         (gimple_resimplify4): Likewise.
10847 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
10849         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10850         Use __mmask64 type instead of __mmask8 for __M argument.
10851         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10852         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10853         __U argument.
10854         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10855         __mmask16 for __M argument.
10856         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10857         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10858         to __mmask16 instead of __mmask8.
10859         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10860         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10861         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10862         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10863         instead of __mmask16 for __U argument.
10864         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10865         __mmask16 instead of __mmask8 for __U argument.
10866         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10867         __U argument.
10868         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10869         __mmask16.
10870         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10871         argument.
10872         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10873         __U argument.
10874         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10875         __mmask16.
10876         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
10877         of __mmask16.
10878         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
10879         __U argument.
10880         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
10881         __U argument.
10882         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
10883         __U argument.
10884         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
10885         __U argument.
10886         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
10887         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
10888         return type as well as __M argument type and all casts from __mmask8
10889         to __mmask32.
10890         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
10891         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
10892         return type as well as __M argument type and all casts from __mmask8
10893         to __mmask16.
10894         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
10895         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
10896         return type as well as __M argument type and all casts from __mmask8
10897         to __mmask32.
10898         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
10899         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
10900         return type as well as __M argument type and all casts from __mmask8
10901         to __mmask16.
10902         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
10903         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
10904         __mmask16.
10906 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
10908         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
10909         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
10910         for __U argument.
10912 2018-07-11  Paul Koning  <ni1d@arrl.net>
10914         * doc/md.texi (define_subst): Document how multiple occurrences of
10915         the same argument in the replacement pattern are handled.
10917 2018-07-11  Paul Koning  <ni1d@arrl.net>
10919         * doc/extend.texi (Common Variable Attributes): Move "mode" into
10920         alphabetical order.
10921         (Common Type Attributes): Add "mode" attribute.
10923 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
10925         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
10926         stream DECL_ORIGINAL_TYPE.
10927         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
10928         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
10929         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
10930         Do not walk original type.
10931         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
10932         external decls.
10933         (write_ts_decl_non_common_tree_pointers): Do not stream
10934         DECL_ORIGINAL_TYPE
10935         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
10936         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
10938 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10940         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
10941         thread twice from the same starting edge.
10943 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10945         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
10946         * gimple.c (gimple_call_nonnull_result_p): ...here...
10947         (gimple_call_nonnull_arg): ...and here.
10948         * gimple.h (gimple_call_nonnull_result_p): New.
10949         (gimple_call_nonnull_arg): New.
10951 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
10953         * config/arm/arm-cpus.in: Move information from fpu field of each
10954         cpu definition to the isa field.
10955         * config/arm/parsecpu.awk (fpu): Delete match rule.
10956         (gen_comm_data): Don't add bits from the CPU's FPU entry.
10958 2018-07-11  Richard Biener  <rguenther@suse.de>
10960         PR debug/86457
10961         * dwarf2out.c (init_sections_and_labels): Use
10962         output_asm_line_debug_info consistently.
10963         (dwarf2out_early_finish): Likewise.
10964         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
10965         type units.
10967 2018-07-11  Richard Biener  <rguenther@suse.de>
10969         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
10970         Rework father_bb setting in a way to avoid propagating constants
10971         multiple times on a loop body.
10973 2018-07-10  Mark Wielaard  <mark@klomp.org>
10975         PR debug/86459
10976         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
10978 2018-07-10  Richard Biener  <rguenther@suse.de>
10980         * hash-map.h (hash_map::iterator::operator*): Return
10981         references to key and value.
10983 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
10985         PR c++/86443
10986         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10987         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
10988         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
10989         (gimplify_omp_for): For composite loops, move outer
10990         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
10991         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
10992         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
10993         TREE_LIST for both the original class iterator and the "last" helper
10994         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
10995         loop, remember has_decl_expr from outer composite loops for the
10996         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
10998 2018-07-09  Martin Sebor  <msebor@redhat.com>
11000         PR middle-end/77357
11001         PR middle-end/86428
11002         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
11003         accessing implicitly initialized array elements.
11004         * expr.c (string_constant): Handle string initializers of
11005         character arrays within aggregates.
11006         * gimple-fold.c (fold_array_ctor_reference): Add argument.
11007         Store element offset.  As a special case, handle zero size.
11008         (fold_nonarray_ctor_reference): Same.
11009         (fold_ctor_reference): Add argument.  Store subobject offset.
11010         * gimple-fold.h (fold_ctor_reference): Add argument.
11012 2018-07-09  Paul Koning  <ni1d@arrl.net>
11014         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
11015         (pdp11_insn_cost): New function.
11016         (pdp11_md_asm_adjust): New function.
11017         (TARGET_INVALID_WITHIN_DOLOOP): Define.
11018         (pdp11_rtx_costs): Update to match machine better.
11019         (output_addr_const_pdp11): Correct format mismatch warnings.
11020         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
11021         * config/pdp11/pdp11.md: General change to add base_cost and/or
11022         length attributes for use by new pdp11_insn_cost function.
11023         (MIN_BRANCH): Correct definition.
11024         (MIN_SOB): Ditto.
11025         (doloop_end): Use standard pattern name for looping pattern.
11026         (doloop_end_nocc): New.
11027         (movsf): Add another constraint alternative.
11028         (zero_extendqihi2): Add constraint alternatives for not in place
11029         extend.
11030         (zero_extendhisi2): Remove.
11031         (shift patterns): Add CC handling variants.
11032         (bswaphi2): New.
11033         (bswapsi2): New.
11034         (rothi3): New.
11035         (define_peephole2): New peephole to recognize mov that sets CC for
11036         subsequent test.
11038 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11040         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
11041         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
11042         wi::add.
11044 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
11046         PR c/86420
11047         * real.c (real_nextafter): Return true if result is denormal.
11049 2018-07-09  Martin Liska  <mliska@suse.cz>
11051         * common.opt: Add back wrongly removed attribute.
11053 2018-07-09  Richard Biener  <rguenther@suse.de>
11055         PR debug/86413
11056         * dwarf2out.c (gen_block_die): For an early generated DIE
11057         always output high/low PC attributes.
11059 2018-07-09  Tom de Vries  <tdevries@suse.de>
11061         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
11062         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
11063         onto VAR_DECL with abstract origin.
11065 2018-07-07  Jim Wilson  <jimw@sifive.com>
11067         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
11069 2018-07-07  Tom de Vries  <tdevries@suse.de>
11071         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
11073 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
11075         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
11076         overflow_type.
11077         (combine_bound): Use wide-int overflow calculation instead of
11078         rolling our own.
11079         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
11080         overflow_type.
11081         * fold-const.c (int_const_binop_2): Same.
11082         (extract_muldiv_1): Same.
11083         (fold_div_compare): Same.
11084         (fold_abs_const): Same.
11085         * match.pd: Same.
11086         * poly-int.h (add): Same.
11087         (sub): Same.
11088         (neg): Same.
11089         (mul): Same.
11090         * predict.c (predict_iv_comparison): Same.
11091         * profile-count.c (slow_safe_scale_64bit): Same.
11092         * simplify-rtx.c (simplify_const_binary_operation): Same.
11093         * tree-chrec.c (tree_fold_binomial): Same.
11094         * tree-data-ref.c (split_constant_offset_1): Same.
11095         * tree-if-conv.c (idx_within_array_bound): Same.
11096         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
11097         * tree-ssa-phiopt.c (minmax_replacement): Same.
11098         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
11099         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
11100         * vr-values.c (vr_values::adjust_range_with_scev): Same.
11101         * wide-int.cc (wi::add_large): Same.
11102         (wi::mul_internal): Same.
11103         (wi::sub_large): Same.
11104         (wi::divmod_internal): Same.
11105         * wide-int.h: Change overflow type to overflow_type for neg, add,
11106         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
11107         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
11108         mul_internal, divmod_internal.
11109         (overflow_type): New enum.
11110         (accumulate_overflow): New.
11112 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
11114         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
11115         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
11117 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11119         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
11120         argument is checked for zero before entering loop, avoid checking again.
11122 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11124         * gimplify.h (generic_expr_could_trap_p): Set as global function.
11125         * gimplify.h (generic_expr_could_trap_p): Likwise.
11126         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
11128 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
11130         PR tree-optimization/86401
11131         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
11132         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
11133         (fold_bit_and_mask): ... here.  New helper function for match.pd.
11134         * fold-const.h (fold_bit_and_mask): Declare.
11135         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
11137 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
11139         PR target/86324
11140         * target.def (translate_mode_attribute): New hook.
11141         * targhooks.h (default_translate_mode_attribute): Declare.
11142         * targhooks.c (default_translate_mode_attribute): New function.
11143         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
11144         * doc/tm.texi: Regenerate.
11145         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
11146         (rs6000_translate_mode_attribute): New function.
11148 2018-07-06  Paul Koning  <ni1d@arrl.net>
11150         * doc/md.texi (define_split): Document DONE and FAIL.
11151         (define_peephole2): Ditto.
11153 2018-07-05  Jeff Law  <law@redhat.com>
11155         PR tree-optimization/86010
11156         * tree-ssa-dse.c (compute_trims): More aggressively trim at
11157         both the head and tail of mem* and str* calls.
11159 2018-07-05  Jim Wilson  <jimw@sifive.com>
11161         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
11163 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
11165         * config/aarch64/aarch64-simd.md: correct flags text for
11166         MIN_EXPR replacement.
11168 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
11170         * configure: Regenerated.
11172 2018-07-05  Carl Love  <cel@us.ibm.com>
11174         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
11175         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
11176         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
11177         VSX_BUILTIN_DOUBLEL_V4SF.
11179 2018-07-05  Martin Sebor  <msebor@redhat.com>
11181         PR c++/86400
11182         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
11183         than its domain to compute its the upper bound of a char array.
11185 2018-07-05  Nathan Sidwell  <nathan@acm.org>
11187         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
11188         * doc/cpp.texi: Update comment.
11189         * doc/tm.texi: Rebuilt.
11190         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
11191         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
11192         * doc/extend.texi (Backwards Compatibility): Clarify it is system
11193         headers affected by extern "C".
11194         * system.h: Poison NO_IMPLICIT_EXTERN_C.
11195         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
11196         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
11197         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
11198         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
11199         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
11200         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
11201         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
11202         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
11203         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
11204         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
11205         config/sparc/sp64-elf.h, config/spu/spu.h,
11206         config/stormy16/stormy16.h, config/v850/v850.h,
11207         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
11208         define NO_IMPLICIT_EXTERN_C.
11209         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
11211 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11213         PR target/84711
11214         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11215         instead of GET_MODE_SIZE when comparing Units.
11217 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11219         PR target/84711
11220         * rtlanal.c (set_noop_p): Constrain on mode change,
11221         include hard-reg-set.h
11223 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11225         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
11227 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
11229         Revert
11230         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11232         PR sanitizer/84250
11233         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11234         libasan.
11235         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11237 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11239         PR sanitizer/84250
11240         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11241         libasan.
11242         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11244 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
11246         PR middle-end/86380
11247         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
11249 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
11251         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
11252         neg_*_op* variables.
11254 2018-07-04  Martin Liska  <mliska@suse.cz>
11256         * tree-switch-conversion.c: Define
11257         max_ratio_for_speed and max_ratio_for_size constants.
11259 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11260             Martin Liska  <mliska@suse.cz>
11262         PR middle-end/66240
11263         PR target/45996
11264         PR c/84100
11265         * common.opt: Rename align options with 'str_' prefix.
11266         * common/config/i386/i386-common.c (set_malign_value): New
11267         function.
11268         (ix86_handle_option): Use it to set -falign-* options/
11269         * config/aarch64/aarch64-protos.h (struct tune_params): Change
11270         type from int to string.
11271         * config/aarch64/aarch64.c: Update default values from int
11272         to string.
11273         * config/alpha/alpha.c (alpha_override_options_after_change):
11274         Likewise.
11275         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
11276         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11277         max skip conditionally.
11278         * config/i386/freebsd.h (SUBALIGN_LOG): New.
11279         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11280         max skip conditionally.
11281         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11282         max skip conditionally.
11283         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
11284         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11285         max skip conditionally.
11286         * config/i386/i386.c (struct ptt): Change type from int to
11287         string.
11288         (ix86_default_align): Set default values.
11289         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
11290         max skip conditionally.
11291         * config/i386/iamcu.h (SUBALIGN_LOG): New.
11292         (ASM_OUTPUT_MAX_SKIP_ALIGN):
11293         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
11294         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11295         max skip conditionally.
11296         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
11297         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
11298         * config/i386/x86-64.h (SUBALIGN_LOG): New.
11299         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11300         max skip conditionally.
11301         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11302         * config/ia64/ia64.c (ia64_option_override): Set default values
11303         for alignment options.
11304         * config/m68k/m68k.c: Handle new str_align_* options.
11305         * config/mips/mips.c (mips_set_compression_mode): Change
11306         type of constants.
11307         (mips_option_override): Set default values for options.
11308         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
11309         Likewise.
11310         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11311         Likewise.
11312         * config/rx/rx.c (rx_option_override): Likewise.
11313         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
11314         (LABEL_ALIGN): Use align_labels_log.
11315         (LOOP_ALIGN): Use align_loops_align.
11316         * config/s390/s390.c (s390_asm_output_function_label): Use new
11317         macros.
11318         * config/sh/sh.c (sh_override_options_after_change):
11319         Change type of constants.
11320         * config/spu/spu.c (spu_sched_init): Likewise.
11321         * config/sparc/sparc.c (sparc_option_override): Set default
11322         values for options.
11323         * config/visium/visium.c (visium_option_override): Likewise.
11324         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
11325         emit p2align format with last argument if it's not needed.
11326         * doc/invoke.texi: Document extended format of -falign-*.
11327         * final.c: Use align_labels alignment.
11328         * flags.h (struct target_flag_state): Change type to use
11329         align_flags.
11330         (struct align_flags_tuple): New.
11331         (struct align_flags): Likewise.
11332         (align_loops_log): Redefine macro to use new types.
11333         (align_loops_max_skip): Redefine macro to use new types.
11334         (align_jumps_log): Redefine macro to use new types.
11335         (align_jumps_max_skip): Redefine macro to use new types.
11336         (align_labels_log): Redefine macro to use new types.
11337         (align_labels_max_skip): Redefine macro to use new types.
11338         (align_functions_log): Redefine macro to use new types.
11339         (align_loops): Redefine macro to use new types.
11340         (align_jumps): Redefine macro to use new types.
11341         (align_labels): Redefine macro to use new types.
11342         (align_functions): Redefine macro to use new types.
11343         (align_functions_max_skip): Redefine macro to use new types.
11344         (align_loops_value): New macro.
11345         (align_jumps_value): New macro.
11346         (align_labels_value): New macro.
11347         (align_functions_value): New macro.
11348         * function.c (invoke_set_current_function_hook): Propagate
11349         alignment values from flags to global variables default in
11350         topleev.h.
11351         * ipa-icf.c (sem_function::equals_wpa): Use
11352         cl_optimization_option_eq instead of memcmp.
11353         * lto-streamer.h (cl_optimization_stream_out): Support streaming
11354         of string types.
11355         (cl_optimization_stream_in): Likewise.
11356         * optc-save-gen.awk: Support strings in cl_optimization.
11357         * opth-gen.awk: Likewise.
11358         * opts.c (finish_options): Remove error checking of invalid
11359         value ranges.
11360         (MAX_CODE_ALIGN): Remove.
11361         (MAX_CODE_ALIGN_VALUE): Likewise.
11362         (parse_and_check_align_values): New function.
11363         (check_alignment_argument): Likewise.
11364         (common_handle_option): Use check_alignment_argument.
11365         * opts.h (parse_and_check_align_values): Declare.
11366         * toplev.c (init_alignments): Remove.
11367         (read_log_maxskip): New.
11368         (parse_N_M): Likewise.
11369         (parse_alignment_opts): Likewise.
11370         (backend_init_target): Remove usage of init_alignments.
11371         * toplev.h (parse_alignment_opts): Declare.
11372         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
11373         argument.
11374         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
11375         * tree.c (cl_option_hasher::equal): New.
11376         * varasm.c: Use new global macros.
11378 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11380         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
11381         Use a simpler align directive also if MAXSKIP = ALIGN-1.
11382         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11383         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11384         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11385         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11386         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
11387         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
11388         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
11389         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11390         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11391         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11392         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11394 2018-07-04  Martin Liska  <mliska@suse.cz>
11395             Jonathan Wakely  <jwakely@redhat.com>
11397         * coverage.c: Use correct type.
11398         * doc/invoke.texi: Language correction.
11400 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
11402         PR target/85620
11403         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
11404         ENDBRANCH for non-tail call which may return via indirect branch.
11405         * doc/extend.texi: Document indirect_return attribute.
11407 2018-07-03  Martin Sebor  <msebor@redhat.com>
11409         PR tree-optimization/86274
11410         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
11411         precondition.
11412         (format_floating): Correct handling of infinities and NaNs.
11414 2018-07-03  Martin Sebor  <msebor@redhat.com>
11416         * print-tree.c (print_real_cst): New function.
11417         (print_node_brief): Call it.
11418         (print_node): Ditto.
11420 2018-07-03  Jeff Law  <law@redhat.com>
11422         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
11423         into a single pattern.
11425         * config/h8300/h8300.md (ors code_iterator): New.
11426         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
11427         a single pattern and single splitter.
11428         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
11429         (iorqi3_1, xorqi3_1): Likewise.
11430         (iorqi3, xorqi3 expanders): Similarly.
11432         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
11433         (movmd_internal) into a single pattern using the P mode iterator.
11434         (movmd splitters): Similarly.
11435         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
11436         (movsd splitters): Similarly.
11438         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
11439         ADDB, ADDW and ADDL into a single ADD attribute which selects the
11440         right table based on the size of the operand.
11441         * config/h8300/h8300.md (length_table): Corresponding changes. All
11442         references to "addb", "addw" and "addl" changed to "add".
11443         (btst patterns): Merge two variants into a single pattern.
11444         (tstqi, tsthi): Likewise.
11445         (addhi3_incdec, addsi3_incdec): Likewise.
11446         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
11447         (mulhi3, mulsi3): Likewise.
11448         (udivhi3, udivsi3): Likewise.
11449         (divhi3, divsi3): Likewise.
11450         (andorqi3, andorhi3, andorsi3): Likewise.
11452 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
11454         PR target/85694
11455         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
11456         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
11458 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11460         PR tree-optimization/85694
11461         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
11462         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
11463         UNSPEC_URHADD.
11464         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
11465         (<u>avg<mode>3_ceil): New patterns.
11467 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
11469         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
11470         scan-tree-dump directive.
11471         * gcc.dg/vect/slp-perm-2.c: Likewise.
11472         * gcc.dg/vect/slp-perm-3.c: Likewise.
11473         * gcc.dg/vect/slp-perm-5.c: Likewise.
11474         * gcc.dg/vect/slp-perm-6.c: Likewise.
11475         * gcc.dg/vect/slp-perm-7.c: Likewise.
11476         * gcc.dg/vect/slp-perm-8.c: Likewise.
11478 2018-07-03  Marek Polacek  <polacek@redhat.com>
11480         PR middle-end/86202
11481         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
11483 2018-07-03  Richard Biener  <rguenther@suse.de>
11485         PR ipa/86389
11486         * tree-ssa-structalias.c (find_func_clobbers): Properly
11487         handle indirect calls.
11489 2018-07-03  Jeff Law  <law@redhat.com>
11491         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
11492         (shifts): New code iterator.
11493         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11494         expander.  Fix HImode handling on H8/SX.
11495         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11496         (subqi3, subhi3, subsi3 expanders): Likewise.
11497         (andqi3, andhi3, andsi3 expanders): Likewise.
11498         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11499         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11500         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11501         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11502         (zero_extendqihi2, zero_extendqisi2): Likewise.
11503         (extendqihi2, extendqisi2): Likewise.
11504         (rotlqi3, rotlhi3, rotlsi3): Likewise.
11505         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11506         (rotlqi3_1, rotlhi3_1): Likewise.
11507         (logicalhi3_sn, logicalsi3_sn): Likewise.
11508         (logicalhi3, logicalsi3): Likewise.
11510 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11512         * tree-vect-patterns.c (vect_recog_rotate_pattern)
11513         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11514         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11515         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11516         type to append_pattern_def_seq instead of creating a stmt_vec_info
11517         directly.
11518         (build_mask_conversion): Likewise.  Remove vinfo argument.
11519         (vect_add_conversion_to_patterm): Likewise, renaming to...
11520         (vect_add_conversion_to_pattern): ...this.
11521         (vect_recog_mask_conversion_pattern): Update call to
11522         build_mask_conversion.  Pass the vector type to
11523         append_pattern_def_seq here too.
11524         (vect_recog_gather_scatter_pattern): Update call to
11525         vect_add_conversion_to_pattern.
11527 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11529         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11530         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11531         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11532         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11533         STMT_VINFO_PATTERN_DEF_SEQ to null here.
11534         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11535         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11536         append_pattern_def_seq instead of new_pattern_def_seq.
11537         (vect_recog_divmod_pattern): Do both of the above.
11538         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11539         is null.
11541 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11543         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11544         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11545         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11546         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11547         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11548         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11549         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11550         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11551         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11552         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11553         parameter with a single stmt_vec_info.
11554         (vect_recog_func_ptr): Likewise.
11555         (vect_recog_gather_scatter_pattern): Likewise, folding in...
11556         (vect_try_gather_scatter_pattern): ...this.
11557         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11558         the stmt_vec_info of the statement to be matched.  Don't clear
11559         STMT_VINFO_RELATED_STMT.
11560         (vect_pattern_recog): Update call accordingly.
11562 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11564         PR tree-optimization/85694
11565         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11566         (uavgM3_ceil): Document new optabs.
11567         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11568         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11569         functions.
11570         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11571         (savg_ceil_optab): New optabs.
11572         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11573         (vect_vect_recog_func_ptrs): Add it.
11574         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11575         constant directly from the associated lhs.
11577 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11579         * tree-vect-patterns.c (vect_split_statement): New function.
11580         (vect_convert_input): Use it to try to split an existing cast.
11582 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11584         * poly-int.h (print_hex): New function.
11585         * dumpfile.h (dump_dec, dump_hex): Declare.
11586         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11587         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11588         min_input_precision, operation_precision and operation_sign.
11589         * tree-vect-patterns.c (vect_get_range_info): New function.
11590         (vect_same_loop_or_bb_p, vect_single_imm_use)
11591         (vect_operation_fits_smaller_type): Delete.
11592         (vect_look_through_possible_promotion): Add an optional
11593         single_use_p parameter.
11594         (vect_recog_over_widening_pattern): Rewrite to use new
11595         stmt_vec_info infomration.  Handle one operation at a time.
11596         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11597         (vect_truncatable_operation_p, vect_set_operation_type)
11598         (vect_set_min_input_precision): New functions.
11599         (vect_determine_min_output_precision_1): Likewise.
11600         (vect_determine_min_output_precision): Likewise.
11601         (vect_determine_precisions_from_range): Likewise.
11602         (vect_determine_precisions_from_users): Likewise.
11603         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11604         (vect_vect_recog_func_ptrs): Put over_widening first.
11605         Add cast_forwprop.
11606         (vect_pattern_recog): Call vect_determine_precisions.
11608 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11610         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11611         statements that have been replaced by further pattern statements.
11612         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11614 2018-07-03  Richard Biener  <rguenther@suse.de>
11616         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11617         always set *dt.  Dump vectype in vectype overload.
11618         * dumpfile.h (dump_gimple_expr): New function.
11619         (dump_gimple_expr_loc): Likewise.
11620         * dumpfile.c (dump_gimple_expr): New function.
11621         (dump_gimple_expr_loc): Likewise.
11623 2018-07-02  Jeff Law  <law@redhat.com>
11625         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11626         the H8/300, H8/300H and H8/S variants into a single pattern.
11627         (movhi_h8300, movqi_h8300hs): Similarly.
11628         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11629         (QHI mode iterator): New.
11631         * config/h8300/h8300.md: Remove trailing whitespace.
11633 2018-07-02  Jim Wilson  <jimw@sifive.com>
11635         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11636         instead of emit_insn for interrupt returns.
11637         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11638         (riscv_sret, riscv_uret): Likewise.
11640 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11642         * pretty-print.c (selftest::test_pp_format): Move save and restore
11643         of quotes to class auto_fix_quotes, and add an instance.
11644         * selftest.c: Include "intl.h".
11645         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11646         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11647         * selftest.h (selftest::auto_fix_quotes): New class.
11649 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
11651         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11652         (aarch64_sve_prepare_conditional_op): Remove.
11653         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11654         Allow aarch64_simd_reg_or_zero as select operand; remove
11655         the aarch64_sve_prepare_conditional_op call.
11656         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11657         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11658         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11659         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11660         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11661         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11662         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11663         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11664         and a splitters to match all of the *_any patterns.
11665         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11667         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11668         (SVE_COND_FP_BINARY_REV): Remove.
11669         (sve_int_op_rev, sve_fp_op_rev): New.
11670         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11671         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11672         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11673         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11674         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11675         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11676         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11677         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11678         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11680         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11681         Remove match_dup 1 from the inner unspec.
11682         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11684         * config/aarch64/aarch64.md (movprfx): New attr.
11685         (length): Default movprfx to 8.
11686         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11687         (*madd<SVE_I>, *msub<SVE_I): Likewise.
11688         (*<su>mul<SVE_I>3_highpart): Likewise.
11689         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11690         (*v<ASHIFT><SVE_I>3): Likewise.
11691         (*<su><MAXMIN><SVE_I>3): Likewise.
11692         (*<su><MAXMIN><SVE_F>3): Likewise.
11693         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11694         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11695         (*div<SVE_F>4): Likewise.
11697 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
11699         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11700         in dump string.
11702 2018-07-02  Richard Biener  <rguenther@suse.de>
11704         PR tree-optimization/86363
11705         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11706         memset argument refers to a non-variable address.
11708 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11710         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11711         {PLUS,MINUS}_EXPR code to...
11712         (adjust_symbolic_bound): ...here,
11713         (combine_bound): ...here,
11714         (set_value_range_with_overflow): ...and here.
11716 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11718         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11719         code...
11720         (extract_range_from_abs_expr): ...here.
11722 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
11724         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11725         -fno-omit-frame-pointer when not optimizing.
11727 2018-07-02  Martin Liska  <mliska@suse.cz>
11729         PR ipa/86279
11730         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11731         (propagate_nothrow): Likewise.
11733 2018-07-02  Martin Liska  <mliska@suse.cz>
11735         PR ipa/86323
11736         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11738 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11740         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11741         function in r262149, changing "loc" param from source_location to
11742         const dump_location_t &.
11743         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11744         declaration, as above.
11746 2018-07-01  Paul Koning  <ni1d@arrl.net>
11748         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11749         -munit-asm, -mgnu-asm, -mdec-asm.
11750         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11751         (pdp11_output_labelref): New.
11752         (pdp11_output_def): New.
11753         (pdp11_output_addr_vec_elt): New.
11754         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
11755         %# and %@ format codes.
11756         (pdp11_option_override): New.
11757         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11758         (pdp11_output_ident): New.
11759         (pdp11_asm_named_section): New.
11760         (pdp11_asm_init_sections): New.
11761         (pdp11_file_start): New.
11762         (pdp11_file_end): New.
11763         (output_ascii): Use .ascii/.asciz for -mdec-asm.
11764         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
11765         %o, like %c but octal.
11766         (pdp11_option_override): New.
11767         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11768         -mdec-asm.
11769         (DATA_SECTION_ASM_OP): Ditto.
11770         (READONLY_DATA_SECTION_ASM_OP): New.
11771         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11772         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11773         (ASM_OUTPUT_LABELREF): Ditto.
11774         (ASM_OUTPUT_DEF): Ditto.
11775         (ASM_OUTPUT_EXTERNAL): New.
11776         (ASM_OUTPUT_SOURCE_FILENAME): New.
11777         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11778         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11779         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
11780         %# and %@ format codes.
11781         * config/pdp11/pdp11.opt (mgnu-asm): New.
11782         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11783         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11784         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11786 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
11788         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11789         dereferencing path[] beyond its length.
11790         (debug_path): New.
11791         (debug_all_paths): New.
11792         (rewire_first_differing_edge): New.
11793         (adjust_paths_after_duplication): New.
11794         (duplicate_thread_path): Call adjust_paths_after_duplication.
11795         Add new argument.
11796         (thread_through_all_blocks): Add new argument to
11797         duplicate_thread_path.
11799 2018-06-30  Jim Wilson  <jimw@sifive.com>
11801         * config/riscv/predicates.md (p2m1_shift_operand): New.
11802         (high_mask_shift_operand): New.
11803         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11804         pattern using p2m1_shift_operand.
11805         (lshsi3_zero_extend_3+2): New combiner pattern using
11806         high_mask_shift_operand.
11808 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11810         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11811         split out from...
11812         (vect_recog_rotate_pattern): ...here.
11813         (vect_convert_input): Try to insert casts of invariants in the
11814         preheader.
11815         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11816         preheader to be empty.
11818 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11820         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11821         vector type.  If given, install it in the new statement's
11822         STMT_VINFO_VECTYPE.
11823         (vect_element_precision): New function.
11824         (vect_unpromoted_value): New struct.
11825         (vect_unpromoted_value::vect_unpromoted_value): New function.
11826         (vect_unpromoted_value::set_op): Likewise.
11827         (vect_look_through_possible_promotion): Likewise.
11828         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11829         (vect_widened_op_tree, vect_convert_input): Likewise.
11830         (vect_convert_inputs, vect_convert_output): Likewise.
11831         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11832         to handle the optional cast of the multiplication result and
11833         vect_widened_op_tree to detect the widened multiplication itself.
11834         Do not require the input and output of promotion casts to have
11835         the same sign, but base the signedness of the operation on the
11836         input rather than the result.  If the pattern includes two
11837         promotions, check that those promotions have the same sign.
11838         Do not restrict the MULT_EXPR handling to a double-width result;
11839         handle quadruple-width results and wider.  Use vect_convert_inputs
11840         to convert the inputs to the common type.
11841         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
11842         to handle the optional cast of the ABS result.  Also allow a sign
11843         change or a sign extension between the ABS and MINUS.
11844         Use vect_widened_op_tree to detect the widened subtraction and use
11845         vect_convert_inputs to convert the inputs to the common type.
11846         (vect_handle_widen_op_by_const): Delete.
11847         (vect_recog_widen_op_pattern): New function.
11848         (vect_recog_widen_mult_pattern): Use it.
11849         (vect_recog_widen_shift_pattern): Likewise.
11850         (vect_recog_widen_sum_pattern): Use
11851         vect_look_through_possible_promotion to handle the promoted
11852         PLUS_EXPR operand.
11854 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11856         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11857         the containing gimple_seq *.
11858         * gimple-iterator.h (gsi_for_stmt): Declare it.
11859         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11860         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11861         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11862         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11863         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11864         checks.
11865         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11866         split out from...
11867         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
11868         statement being replaced is part of an existing pattern
11869         definition sequence, inserting the new pattern statements before
11870         the original one.
11871         (vect_pattern_recog_1): Don't return a bool.  If the statement
11872         is already part of a pattern, instead apply pattern matching
11873         to the pattern definition statements.  Don't clear the
11874         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11875         (vect_pattern_recog): Don't break after the first match;
11876         continue processing the pattern definition statements instead.
11877         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
11879 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11881         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11882         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11883         (vect_recog_widen_sum_pattern): Use it.
11885 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11887         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11888         phi is not a pattern statement and has not been replaced by
11889         a pattern statement.
11890         * tree-vect-patterns.c (type_conversion_p): Don't check
11891         STMT_VINFO_IN_PATTERN_P.
11892         (vect_recog_vector_vector_shift_pattern): Likewise.
11893         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
11894         the pattern statement rather than the original statement; check
11895         directly for a WIDEN_MULT_EXPR here.
11896         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
11897         vect_is_simple_use to return the pattern statement rather
11898         than the original statement; use is_pattern_stmt_p to check
11899         for such a pattern statement.
11900         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
11901         to return the pattern statement rather than the original statement;
11902         don't do the same transformation here.
11903         (vect_is_simple_use): If the defining statement has been replaced
11904         by a pattern statement, return the pattern statement instead.
11905         Remove the corresponding (local) transformation from the vectype
11906         overload.
11908 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11910         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
11911         end and default to null.
11912         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11913         (vectorizable_reduction): Update calls accordingly, dropping the
11914         gimple ** argument if the passed-back statement isn't needed.
11915         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
11916         (vect_recog_rotate_pattern): Likewise.
11917         (vect_recog_mask_conversion_pattern): Likewise.
11918         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11919         (vect_mask_constant_operand_p): Likewise.
11920         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
11921         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
11922         (get_group_load_store_type, get_load_store_type): Likewise.
11923         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
11924         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
11925         (vectorizable_conversion, vectorizable_assignment): Likewise.
11926         (vectorizable_shift, vectorizable_operation): Likewise.
11927         (vectorizable_store, vect_is_simple_cond): Likewise.
11928         (vectorizable_condition, vectorizable_comparison): Likewise.
11929         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
11930         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
11931         and move it to the end.  Cope with null def_stmt_outs.
11933 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11935         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
11937 2018-06-29  Jeff Law  <law@redhat.com>
11939         * config/v850/v850.c (v850_legitimate_address_p): Handle large
11940         displacements for TARGET_V850E2V3 and newer.
11941         (TARGET_LRA_P): Remove.  Defaults to LRA now.
11942         * config/v850/v850.md (sign23byte_load): Remove.
11943         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
11944         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
11946 2018-06-29  Martin Liska  <mliska@suse.cz>
11948         PR lto/85759
11949         * coverage.c (coverage_init): Mangle full path name.
11950         * doc/invoke.texi: Document the change.
11951         * gcov-io.c (mangle_path): New.
11952         * gcov-io.h (mangle_path): Likewise.
11953         * gcov.c (mangle_name): Use mangle_path for path mangling.
11955 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11957         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
11958         if starting source register is not even.
11960 2018-06-29  Martin Liska  <mliska@suse.cz>
11962         PR tree-optimization/86263
11963         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
11964         Make edge redirection.
11966 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
11968         * dumpfile.c (dump_loc): Add indentation based on scope depth.
11969         (dump_scope_depth): New variable.
11970         (get_dump_scope_depth): New function.
11971         (dump_begin_scope): New function.
11972         (dump_end_scope): New function.
11973         * dumpfile.h (get_dump_scope_depth): New declaration.
11974         (dump_begin_scope): New declaration.
11975         (dump_end_scope): New declaration.
11976         (class auto_dump_scope): New class.
11977         (AUTO_DUMP_SCOPE): New macro.
11978         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
11979         AUTO_DUMP_SCOPE.
11981 2018-06-29  Richard Biener  <rguenther@suse.de>
11983         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
11984         compute_all_dependences succeeds.
11985         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
11986         exceed --param loop-max-datarefs-for-datadeps.
11988 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
11990         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
11992 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
11994         PR target/86348
11995         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
11996         alternative 0 in preferred_for_speed attribute.
11998 2018-06-28  Paul Koning  <ni1d@arrl.net>
12000         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
12001         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
12002         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
12003         * config/pdp11/pdp11.md: Correct "length" attribute calculation
12004         for shift insn patterns.
12006 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
12008         * cgraph.c (cgraph_node::get_body): Replace assignments to
12009         "dump_file" with calls to set_dump_file.
12010         * dumpfile.c (alt_dump_file): Make static, and group with...
12011         (alt_flags): ...this definition.
12012         (dumps_are_enabled): New variable.
12013         (refresh_dumps_are_enabled): New function.
12014         (set_dump_file): New function.
12015         (set_alt_dump_file): New function.
12016         (gcc::dump_manager::dump_start): Replace assignments to
12017         "dump_file" and "alt_dump_file" with calls to set_dump_file and
12018         set_alt_dump_file.
12019         (gcc::dump_manager::dump_finish): Likewise.
12020         * dumpfile.h (alt_dump_file): Delete decl.
12021         (dumps_are_enabled): New variable decl.
12022         (set_dump_file): New function decl.
12023         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
12024         global.
12025         * tree-nested.c (lower_nested_functions): Replace assignments to
12026         "dump_file" with calls to set_dump_file.
12028 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
12030         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
12031         goto_locus of each outgoing edge of each basic block.
12033 2018-06-28  Richard Biener  <rguenther@suse.de>
12035         * dwarf2out.c (decl_scope_table): Remove.
12036         (push_decl_scope): Likewise.
12037         (pop_decl_scope): Likewise.
12038         (gen_type_die_for_member): Do not call push/pop_decl_scope.
12039         (gen_struct_or_union_type_die): Likewise.
12040         (gen_tagged_type_die): Likewise.
12041         (dwarf2out_init): Do not initialize decl_scope_table.
12042         (dwarf2out_c_finalize): Do not free it.
12044 2018-06-28  Richard Biener  <rguenther@suse.de>
12046         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
12047         deciding whether to not re-use a DIE.
12049 2018-06-28  Richard Biener  <rguenther@suse.de>
12051         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
12052         DW_AT_abstract_origin attribute.
12054 2018-06-28  Martin Liska  <mliska@suse.cz>
12056         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12057         Use newly introduced constants.
12058         * tree-switch-conversion.h (struct jump_table_cluster):
12059         Define max_ratio_for_size and max_ratio_for_speed.
12061 2018-06-28  Martin Liska  <mliska@suse.cz>
12063         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12064         Add new checking assert to catch invalid state.
12065         (jump_table_cluster::can_be_handled): Handle single case
12066         clusters.
12067         (jump_table_cluster::is_beneficial): Bail out for such case.
12068         (bit_test_cluster::find_bit_tests):
12069         Add new checking assert to catch invalid state.
12070         (bit_test_cluster::can_be_handled): Handle single case
12071         clusters.
12072         (bit_test_cluster::is_beneficial): Bail out for such case.
12073         (switch_decision_tree::analyze_switch_statement):
12074         Fix comment.
12076 2018-06-28  Martin Liska  <mliska@suse.cz>
12078         * common.opt: Introduce -completion option.
12079         * gcc.c (driver_handle_option): Handle it.
12080         (driver::main): Print completions if completion
12081         is set.
12082         * opt-suggestions.c (option_proposer::get_completions):
12083         New function.
12084         (option_proposer::suggest_completion): Likewise.
12085         (option_proposer::find_param_completions): Likewise.
12086         (verify_autocompletions): Likewise.
12087         (test_completion_valid_options): Likewise.
12088         (test_completion_valid_params): Likewise.
12089         (in_completion_p): Likewise.
12090         (empty_completion_p): Likewise.
12091         (test_completion_partial_match): Likewise.
12092         (test_completion_garbage): Likewise.
12093         (opt_proposer_c_tests): Likewise.
12094         * opt-suggestions.h: Declare new functions.
12095         * opts.c (common_handle_option): Handle OPT__completion_.
12096         * selftest-run-tests.c (selftest::run_tests): Add
12097         opt_proposer_c_tests.
12098         * selftest.c (assert_str_startswith): New.
12099         * selftest.h (assert_str_startswith): Likewise.
12100         (opt_proposer_c_tests): New.
12101         (ASSERT_STR_STARTSWITH): Likewise.
12103 2018-06-28  Martin Liska  <mliska@suse.cz>
12105         * Makefile.in: Add opt-suggestions.o.
12106         * gcc-main.c: Include opt-suggestions.h.
12107         * gcc.c (driver::driver): Likewise.
12108         (driver::~driver): Remove m_option_suggestions.
12109         (driver::build_option_suggestions): Moved to option_proposer.
12110         (driver::suggest_option): Likewise.
12111         (driver::handle_unrecognized_options): Use option_proposer.
12112         * gcc.h (class driver): Add new memver m_option_proposer.
12113         * opt-suggestions.c: New file.
12114         * opt-suggestions.h: New file.
12116 2018-06-28  Martin Liska  <mliska@suse.cz>
12118         * vec.h (class auto_string_vec): New (moved from auto_argvec).
12119         (auto_string_vec::~auto_string_vec): Likewise.
12121 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
12123         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
12124         prevent_decl_creation_for_types fields up and add reset_location field.
12125         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
12126         statement if id->reset_location is true.
12127         (copy_edges_for_bb): Do not set goto_locus on the new edges if
12128         id->reset_location is true.
12129         (copy_phis_for_bb): Force input_location on the arguments if
12130         id->reset_location is true.
12131         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
12132         is set on the function to be inlined.
12134 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
12136         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
12138 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
12140         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
12141         registers for Pmode.
12142         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
12143         hard registers for the clobbered pseudo.
12145 2018-06-27  Paul Koning  <ni1d@arrl.net>
12147         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12148         mutually exclusive options.
12149         * config/pdp11/constraints.md (h): New constraint.
12150         (O): Update definition to match shift code generation.
12151         (D): New constraint.
12152         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
12153         (CCFP): Remove.
12154         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
12155         function.
12156         (output_jump): Change arguments.
12157         (pdp11_fixed_cc_regs): New function.
12158         (pdp11_cc_mode): Ditto.
12159         (pdp11_expand_shift): Ditto.
12160         (pdp11_assemble_shift): Ditto.
12161         (pdp11_small_shift): Ditto.
12162         (pdp11_branch_cost): Remove.
12163         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
12164         from output.
12165         (pdp11_register_move_cost): Update for CC registers.
12166         (pdp11_rtx_costs): Add case for LSHIFTRT.
12167         (pdp11_output_jump): Add CCNZ mode conditional branches.
12168         (notice_update_cc_on_set): Remove.
12169         (pdp11_cc_mode): New function.
12170         (simple_memory_operand): Correct pre/post decrement case.
12171         (no_side_effect_operand): New function.
12172         (pdp11_regno_reg_class): Add CC_REGS class.
12173         (pdp11_fixed_cc_regs): New function.
12174         (pdp11_small_shift): New function.
12175         (pdp11_expand_shift): New function to expand shift insns.
12176         (pdp11_assemble_shift): New function to output shifts.
12177         (pdp11_branch_cost): Remove.
12178         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
12179         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
12180         (WCHAR_TYPE): Ditto.
12181         (PTRDIFF_TYPE): Ditto.
12182         (ADJUST_INSN_LENGTH): New macro.
12183         (FIXED_REGISTERS): Add CC registers.
12184         (CALL_USED_REGISTERS): Ditto.
12185         (reg_class): Ditto.
12186         (REG_CLASS_NAMES): Ditto.
12187         (REG_CLASS_CONTENTS): Ditto.
12188         (SELECT_CC_MODE): Use new function.
12189         (TARGET_FLAGS_REGNUM): New macro.
12190         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
12191         (cc0_reg_rtx): Remove.
12192         (CC_STATUS_MDEP): Remove.
12193         (CC_STATUS_MDEFP_INIT): Remove.
12194         (CC_IN_FPU): Remove.
12195         (NOTICE_UPDATE_CC): Remove.
12196         (REGISTER_NAMES): Add CC registers.
12197         (BRANCH_COST): Change to constant 1.
12198         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
12199         handling.
12200         * config/pdp11/pdp11.opt (mbcopy): Remove.
12201         (mbcopy-builtin): Remove.
12202         (mbranch-cheap): Remove.
12203         (mbranch-expensive): Remove.
12204         * config/pdp11/predicates.md (expand_shift_operand): Update to
12205         match shift code generation.
12206         (ccnz_operator): New predicate.
12207         * doc/invoke.texi (PDP-11 Options): Remove deleted options
12208         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
12209         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
12210         exclusive options.
12211         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
12212         description of O constraint.
12214 2018-06-27  Jeff Law  <law@redhat.com>
12215             Austin Law  <austinklaw@gmail.com>
12217         * config/v850/v850.md (addsi3_set_flags): New pattern.
12218         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
12219         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
12220         (zero_extendhisi2_v850_set_flags): Likewise.
12221         (zero_extendqisi2_v850_set_flags): Likewise.
12222         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
12223         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
12224         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
12226         * config/v850/v850-protos.h (notice_update_cc): Remove.
12227         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
12228         (v850_print_operand): Handle 'D' and "d".
12229         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
12230         Add handling of arithmetic/logical operations compared against zero.
12231         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
12232         Do not look at v850_compare_op, instead get mode from last argument.
12233         (v850_gen_compare): Remove
12234         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
12235         after reload for prologue insns.
12236         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
12237         patterns.
12238         (construct_save_jarl): Likewise.
12239         (TARGET_FLAGS_REGNUM): Define.
12240         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
12241         (NOTICE_UPDATE_CC): Remove.
12242         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
12243         than cc0.  Conditionalize on reload_completed.
12244         (cmpsi_insn, setfcc_insn): Likewise.
12245         (tst1 splitter): Turn into define_and_split which sets the flags
12246         after reload.
12247         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
12248         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
12249         (cstoresf4, cstoredf4): Clobber the flags.
12250         (cmpsi, cmpsf, cmpdf): Remove expanders.
12251         (setf_insn): Remove pattern.
12252         (addsi3): Turn into define_and_split which clobbers the flags after
12253         reload and a suitable pattern (addsi3_clobber_flags) for use after
12254         reload.
12255         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
12256         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
12257         (ashrsi3, ashrsi3_v850e2): Likewise.
12258         (bins): Clobber the flags.
12259         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
12260         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
12261         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
12262         (call_value_internal_short, call_value_internal_long): Likewise.
12263         (callt_save_interrupt, callt_return_interrupt): Likewise.
12264         (save_interrupt, return_interrupt): Likewise.
12265         (callt_save_all_interrupt, save_all_interrupt): Likewise.
12266         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
12267         (restore_all_interrupt, _restore_all_interrupt): Likewise.
12268         (All FP comparisons): Only allow after reload has completed.
12269         (trfsr): Likewise.
12270         (divh, divhu): Tweak output template.
12271         (branch_z_normal, branch_z_invert): Remove
12272         (branch_nz_normal, branch_nz_invert): Likewise.
12273         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
12275         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
12276         * config/v850/v850.c (notice_update_cc): Remove.
12277         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
12278         (CC_NO_CARRY): Likewise.
12279         (NOTICE_UPDATE_CC): Define to nothing.
12280         * config/v850/v850.md: Remove block comment on cc0 handling
12281         Remove "cc" attribute from all patterns.  Remove cc_status handling
12282         from all patterns.  Minor formatting fixes.
12284 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12286         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
12287         (cortex-a76.cortex-a55): Likewise.
12288         * config/aarch64/aarch64-tune.md: Regenerate.
12289         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
12290         cortex-a76.cortex-a55.
12292 2018-06-27  Jeff Law  <law@redhat.com>
12294         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
12295         (MULTILIB_DIRNAMES): Similarly.
12297 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
12299         * gimple.h (gimple_return_retbnd): Delete.
12300         (gimple_return_set_retbnd): Likewise.
12301         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
12302         gimple_return_set_retbnd.
12303         * gimple-pretty-print.c (dump_gimple_return): Remove call to
12304         gimple_return_retbnd and adjust.
12305         * tree-inline.h (struct copy_body_data): Remove retbnd field.
12306         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
12307         Explicitly return NULL in a couple more cases.  Move assertion
12308         on debug statements and remove unreachable code.
12309         (reset_debug_binding): Do not test id->retbnd.
12310         (expand_call_inline): Do not set it.
12312 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
12314         * configure.ac: Add --disable-gcov option.
12315         * configure: Regenerate.
12316         * Makefile.in: Honour @enable_gcov@.
12317         * doc/install.texi: Document --disable-gcov.
12319 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12321         * config/arm/arm-cpus.in (cortex-a76): New entry.
12322         (cortex-a76.cortex-a55): Likewise.
12323         * config/arm/arm-tables.opt: Regenerate.
12324         * config/arm/arm-tune.md: Likewise.
12325         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
12326         * doc/invoke.texi (ARM Options): Document cortex-a76 and
12327         cortex-a76.cortex-a55.
12329 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
12331         PR target/85769
12332         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
12334 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
12336         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
12337         comment.
12338         (EPILOGUE_USES): Likewise.
12340 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
12342         * tree-inline.c (remap_location): New function extracted from...
12343         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
12344         (copy_phis_for_bb): ...here.  Call remap_location.
12345         (copy_cfg_body): Adjust call to copy_edges_for_bb.
12347 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
12349         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
12350         unaligned vsx for 16B memset.
12352 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
12354         PR target/86285
12355         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
12356         ieee128_float_type_node to long_double_type_node unless
12357         TARGET_LONG_DOUBLE_128 is set.
12359 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
12361         * cfgloop.c (get_loop_location): Convert return type from
12362         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
12363         by implicit construction from rtx_insn *, and using
12364         dump_user_location_t::from_function_decl for the fallback case.
12365         * cfgloop.h (get_loop_location): Convert return type from
12366         location_t to dump_user_location_t.
12367         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
12368         dump_printf_loc to pass in a dump_location_t rather than a
12369         location_t, via the gimple stmt.
12370         * coverage.c (get_coverage_counts): Update calls to
12371         dump_printf_loc to pass in dump_location_t rather than a
12372         location_t.
12373         * doc/optinfo.texi (Dump types): Convert example of
12374         dump_printf_loc from taking "locus" to taking "insn".  Update
12375         description of the "_loc" calls to cover dump_location_t.
12376         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
12377         "selftest.h".
12378         (dump_user_location_t::dump_user_location_t): New constructors,
12379         from gimple *stmt and rtx_insn *.
12380         (dump_user_location_t::from_function_decl): New function.
12381         (dump_loc): Make static.
12382         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
12383         const dump_location_t &.
12384         (dump_generic_expr_loc): Delete.
12385         (dump_printf_loc): Convert param "loc" from location_t to
12386         const dump_location_t &.
12387         (selftest::test_impl_location): New function.
12388         (selftest::dumpfile_c_tests): New function.
12389         * dumpfile.h: Include "profile-count.h".
12390         (class dump_user_location_t): New class.
12391         (struct dump_impl_location_t): New struct.
12392         (class dump_location_t): New class.
12393         (dump_printf_loc): Convert 2nd param from source_location to
12394         const dump_location_t &.
12395         (dump_generic_expr_loc): Delete.
12396         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
12397         const dump_location_t &.
12398         * gimple-fold.c (fold_gimple_assign): Update call to
12399         dump_printf_loc to pass in a dump_location_t rather than a
12400         location_t, via the gimple stmt.
12401         (gimple_fold_call): Likewise.
12402         * gimple-loop-interchange.cc
12403         (loop_cand::analyze_iloop_reduction_var): Update for change to
12404         check_reduction_path.
12405         (tree_loop_interchange::interchange): Update for change to
12406         find_loop_location.
12407         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
12408         change in return-type of find_loop_location.
12409         (graphite_regenerate_ast_isl): Likewise.
12410         * graphite-optimize-isl.c (optimize_isl): Likewise.
12411         * graphite.c (graphite_transform_loops): Likewise.
12412         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
12413         pass in a dump_location_t rather than a location_t, via the
12414         gimple stmt.
12415         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12416         * ipa.c (walk_polymorphic_call_targets): Likewise.
12417         * loop-unroll.c (report_unroll): Convert "locus" param from
12418         location_t to dump_location_t.
12419         (decide_unrolling): Update for change to get_loop_location's
12420         return type.
12421         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
12422         location_t to dump_user_location_t.
12423         (grid_find_single_omp_among_assignments_1): Updates calls to
12424         dump_printf_loc to pass in a dump_location_t rather than a
12425         location_t, via the gimple stmt.
12426         (grid_parallel_clauses_gridifiable): Convert "tloc" from
12427         location_t to dump_location_t.  Updates calls to dump_printf_loc
12428         to pass in a dump_location_t rather than a location_t, via the
12429         gimple stmt.
12430         (grid_inner_loop_gridifiable_p): Likewise.
12431         (grid_dist_follows_simple_pattern): Likewise.
12432         (grid_gfor_follows_tiling_pattern): Likewise.
12433         (grid_target_follows_gridifiable_pattern): Likewise.
12434         (grid_attempt_target_gridification): Convert initialization
12435         of local "grid" from memset to zero-initialization; FIXME: does
12436         this require C++11?  Update call to dump_printf_loc to pass in a
12437         optinfo_location rather than a location_t, via the gimple stmt.
12438         * profile.c (read_profile_edge_counts): Updates call to
12439         dump_printf_loc to pass in a dump_location_t rather than a
12440         location_t
12441         (compute_branch_probabilities): Likewise.
12442         * selftest-run-tests.c (selftest::run_tests): Call
12443         dumpfile_c_tests.
12444         * selftest.h (dumpfile_c_tests): New decl.
12445         * tree-loop-distribution.c (pass_loop_distribution::execute):
12446         Update for change in return type of find_loop_location.
12447         * tree-parloops.c (parallelize_loops): Likewise.
12448         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
12449         "locus" from location_t to dump_user_location_t.
12450         (canonicalize_loop_induction_variables): Likewise.
12451         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
12452         for change in return type of find_loop_location.
12453         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
12454         to dump_printf_loc to pass in a dump_location_t rather than a
12455         location_t, via the stmt.
12456         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
12457         Likewise.
12458         * tree-vect-loop-manip.c (find_loop_location): Convert return
12459         type from source_location to dump_user_location_t.
12460         (vect_do_peeling): Update for above change.
12461         (vect_loop_versioning): Update for change in type of
12462         vect_location.
12463         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
12464         from location_t to dump_user_location_t.
12465         (vect_estimate_min_profitable_iters): Update for change in type
12466         of vect_location.
12467         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
12468         location_t to dump_location_t.
12469         (vect_slp_bb): Update for change in type of vect_location.
12470         * tree-vectorizer.c (vect_location): Convert from source_location
12471         to dump_user_location_t.
12472         (try_vectorize_loop_1): Update for change in vect_location's type.
12473         (vectorize_loops): Likewise.
12474         (increase_alignment): Likewise.
12475         * tree-vectorizer.h (vect_location): Convert from source_location
12476         to dump_user_location_t.
12477         (find_loop_location): Convert return type from source_location to
12478         dump_user_location_t.
12479         (check_reduction_path): Convert 1st param from location_t to
12480         dump_user_location_t.
12481         * value-prof.c (check_counter): Update call to dump_printf_loc to
12482         pass in a dump_user_location_t rather than a location_t; update
12483         call to error_at for change in type of "locus".
12484         (check_ic_target): Update call to dump_printf_loc to
12485         pass in a dump_user_location_t rather than a location_t, via the
12486         call_stmt.
12488 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12490         * config/s390/s390.h (enum processor_flags): Do not use
12491         default tune parameter when -march was specified.
12493 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
12495         PR target/86314
12496         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12497         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12498         operands.
12500 2018-06-26  Richard Biener  <rguenther@suse.de>
12502         PR tree-optimization/86287
12503         PR bootstrap/86316
12504         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12505         (vect_analyze_loop): Initialize n_stmts.
12507 2018-06-26  Richard Biener  <rguenther@suse.de>
12509         PR middle-end/86271
12510         * fold-const.c (fold_convertible_p): Pointer extension
12511         isn't valid.
12513 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
12515         PR debug/86064
12516         * dwarf2out.c (loc_list_has_views): Adjust comments.
12517         (dw_loc_list): Split single cross-partition range with
12518         nonzero locview.
12520 2018-06-25  Jeff Law  <law@redhat.com>
12522         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12523         on -mbig-switch by default.
12525         * config/v850/predicates.md (const_float_1_operand): Fix match_code
12526         test.
12527         (const_float_0_operand): Remove unused predicate.
12528         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12529         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12530         (recipsf2): New expander.  Original pattern now called
12531         (recipsf2_insn).
12532         (recipdf2, recipdf2_insn): Similarly.
12533         (rsqrtsf2, rsqrtsf2_insn): Similarly
12534         (rsqrtdf2, rsqrtdf2_insn): Similarly
12536 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
12538         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12539         Simplify logic for FreeBSD (twice).
12541 2018-06-25  Martin Sebor  <msebor@redhat.com>
12543         PR tree-optimization/86204
12544         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12545         a strnlen result if it's less than the length of the string.
12547 2018-06-25  Martin Sebor  <msebor@redhat.com>
12549         PR tree-optimization/85700
12550         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12551         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12552         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12554 2018-06-25  Martin Sebor  <msebor@redhat.com>
12556         * doc/extend.texi (Zero-length arrays): Update and clarify.
12558 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
12560         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12561         added IEEE/IBM long double multilib support on PowerPC little
12562         endian Linux systems.
12563         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12564         (MULTILIB_DEFAULTS): Likewise.
12565         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12566         Likewise.
12567         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12568         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12569         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12571 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
12573         PR middle-end/86311
12574         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12575         (REORDER_45): Likewise.
12577 2018-06-25  Jeff Law  <law@redhat.com>
12579         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12580         dividend to 32 bits.  Adjust length.
12581         (udivmodhi4): Cleanup output template.  Fix length.
12583 2018-06-25  Carl Love  <cel@us.ibm.com>
12585         * config/rs6000/vsx.md: Change word selector to prefered location.
12587 2018-06-25  Richard Biener  <rguenther@suse.de>
12589         PR tree-optimization/86304
12590         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12591         epilogue-if-converted loops as well.
12593 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
12595         * lto-section-out.c (lto_begin_section): Do not print section
12596         name for noaddr and unnumbered dumps.
12598 2018-06-25  Richard Biener  <rguenther@suse.de>
12600         * tree-vectorizer.h (struct vec_info_shared): New structure
12601         with parts split out from struct vec_info and loop_nest from
12602         struct _loop_vec_info.
12603         (struct vec_info): Adjust accordingly.
12604         (struct _loop_vec_info): Likewise.
12605         (LOOP_VINFO_LOOP_NEST): Adjust.
12606         (LOOP_VINFO_DATAREFS): Likewise.
12607         (LOOP_VINFO_DDRS): Likewise.
12608         (struct _bb_vec_info): Likewise.
12609         (BB_VINFO_DATAREFS): Likewise.
12610         (BB_VINFO_DDRS): Likewise.
12611         (struct _stmt_vec_info): Add dr_aux member.
12612         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12613         (DR_MISALIGNMENT_UNINITIALIZED): New.
12614         (set_dr_misalignment): Adjust.
12615         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12616         (vect_analyze_loop): Adjust prototype.
12617         (vect_analyze_loop_form): Likewise.
12618         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12619         Compute dependences lazily.
12620         (vect_record_base_alignments): Use shared datarefs/ddrs.
12621         (vect_verify_datarefs_alignment): Likewise.
12622         (vect_analyze_data_refs_alignment): Likewise.
12623         (vect_analyze_data_ref_accesses): Likewise.
12624         (vect_analyze_data_refs): Likewise.
12625         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12626         constructor parameter for shared part.
12627         (vect_analyze_loop_form): Pass in shared part and adjust.
12628         (vect_analyze_loop_2): Pass in storage for the number of
12629         stmts.  Move loop nest finding to the caller.  Compute
12630         datarefs lazily.
12631         (vect_analyze_loop): Pass in shared part.
12632         (vect_transform_loop): Verify shared datarefs are unchanged.
12633         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12634         constructor parameter for shared part.
12635         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12636         (vect_slp_bb): Verify shared datarefs are unchanged before
12637         transform.
12638         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12639         change.
12640         (new_stmt_vec_info): Initialize DR_AUX misalignment to
12641         DR_MISALIGNMENT_UNINITIALIZED.
12642         * tree-vectorizer.c (vec_info::vec_info): Add constructor
12643         parameter for shared part.
12644         (vec_info::~vec_info): Adjust.
12645         (vec_info_shared::vec_info_shared): New.
12646         (vec_info_shared::~vec_info_shared): Likewise.
12647         (vec_info_shared::save_datarefs): Likewise.
12648         (vec_info_shared::check_datarefs): Likewise.
12649         (try_vectorize_loop_1): Construct shared part live for analyses
12650         of a single loop for multiple vector sizes.
12651         * tree-parloops.c (gather_scalar_reductions): Adjust.
12653 2018-06-25  Richard Biener  <rguenther@suse.de>
12655         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12656         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12657         (vect_analyze_data_refs): Remove similar code from here and
12658         simplify accordingly.
12660 2018-06-25  Richard Biener  <rguenther@suse.de>
12662         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12663         for reverse storage order accesses rather than asserting
12664         they cannot happen here.
12666 2018-06-25  Tom de Vries  <tdevries@suse.de>
12668         PR debug/86257
12669         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12670         Use data16 instead of .byte for insn prefix.
12672 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
12674         PR C++/86082
12675         * parser.c (make_char_string_pack): Pass this literal chars
12676         through cpp_interpret_string.
12677         (cp_parser_userdef_numeric_literal): Check the result of
12678         make_char_string_pack.
12680 2018-06-24  Maya Rashish  <coypu@sdf.org>
12682         * ginclude/stddef.h: Simplify conditions around avoiding
12683         re-definition of __size_t.
12685 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12687         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12688         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12690 2018-06-22  Maya Rashish  <coypu@sdf.org>
12692         * doc/invoke.texi (mno-fancy-math-387): Update for changes
12693         made to OpenBSD and NetBSD through the years.
12695 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12697         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12698         behavior of vec_pack (vector double, vector double) to match
12699         behavior of vec_float2 (vector double, vector double).
12701 2018-06-22  Olivier Hainque  <hainque@adacore.com>
12703         * gimplify.c (gimplify_function_tree): Prevent creation
12704         of a trampoline for the address of the current function
12705         passed to entry/exit instrumentation hooks.
12707 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
12709         PR target/86222
12710         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12711         correctly.
12713 2018-06-22  Martin Liska  <mliska@suse.cz>
12715         PR tree-optimization/86263
12716         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12717         Bail out if is_enabled is false.
12718         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12719         New declaration.
12720         (jump_table_cluster::is_enabled): New function.
12722 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12724         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12725         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12726         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12727         (lto_input_ts_binfo_tree_pointers): Likewise.
12728         * tree-streamer-out.c (streamer_write_tree_bitfields,
12729         write_ts_binfo_tree_pointers): Likewise.
12730         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12732 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12734         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12736 2018-06-22  Martin Liska  <mliska@suse.cz>
12738         * symbol-summary.h (get): Make it pure and inline move
12739         functionality from ::get function.
12740         (get): Remove and inline into ::get and ::get_create.
12741         (get_create): Move code from ::get function.
12743 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12745         PR target/85994
12746         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12747         -x assembler-with-cpp.
12749 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12751         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12752         _FILE_OFFSET_BITS=64 for C++.
12754 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
12756         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12757         conversion insn that shows up when pr85657-3.c is compiled using
12758         IEEE 128-bit long double.
12759         (neg<mode>2_internal): Use the correct mode to check whether the
12760         mode is IBM extended.
12761         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12762         multiply and divide external functions from being created more
12763         than once.
12765 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12767         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12768         functions.
12769         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
12770         the edge can be forwarded.
12771         (cfg_layout_merge_blocks): Likewise.
12773 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12775         * except.c (finish_eh_generation): Commit edge insertions only after
12776         the EH edges have been redirected from post-landing to landing pads.
12778 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12780         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12781         create_tmp_var_for to create the FRAME decl.
12782         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12784 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12786         * tree-inline.c (copy_edges_for_bb): Minor tweak.
12787         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12788         debug statement when resetting its value.
12789         (expand_call_inline): Copy the locus of the call onto the assignment
12790         of the return value, if any.  Use local variable in more cases.
12792 2018-06-21  Martin Liska  <mliska@suse.cz>
12794         * ipa-pure-const.c (propagate_nothrow): Use
12795         funct_state_summaries->get.
12796         (dump_malloc_lattice): Likewise.
12797         (propagate_malloc): Likewise.
12799 2018-06-21  Richard Biener  <rguenther@suse.de>
12801         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12802         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12803         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12804         comment.
12805         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12806         BLOCK_ABSTRACT_ORIGIN unconditionally.
12808 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
12810         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12811         deleting it.
12812         * ipa-reference.c (ipa_reference_c_finalize): Delete
12813         ipa_ref_opt_sum_summaries and set it to NULL.
12815 2018-06-21  Tom de Vries  <tdevries@suse.de>
12817         PR tree-optimization/85859
12818         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12819         test with comment from bb_no_side_effects_p.
12821 2018-06-21  Richard Biener  <rguenther@suse.de>
12823         PR tree-optimization/86232
12824         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12825         max for constant niter.
12827 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12829         * config/aarch64/aarch64-simd.md
12830         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12832 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12834         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12835         Make opernads of the unspec commutative.
12837 2018-06-21  Richard Biener  <rguenther@suse.de>
12839         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12840         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12841         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12842         (vect_analyze_data_ref_dependence): Re-order checks to deal with
12843         NULL DR_STEP.
12844         (vect_record_base_alignments): Do not record base alignment
12845         for gathers or scatters.
12846         (vect_compute_data_ref_alignment): Drop return value that is always
12847         true.  Bail out early for gathers or scatters.
12848         (vect_enhance_data_refs_alignment): Bail out early for gathers
12849         or scatters.
12850         (vect_find_same_alignment_drs): Likewise.
12851         (vect_analyze_data_refs_alignment): Remove dead code.
12852         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12853         (vect_analyze_data_refs): For possible gathers or scatters do
12854         not create an alternate DR, just check their possible validity
12855         and mark them.  Adjust DECL_NONALIASED handling to not rely
12856         on DR_BASE_ADDRESS.
12857         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12858         update inits of gathers or scatters.
12859         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12860         Also copy gather/scatter flag to pattern vinfo.
12862 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12864         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12865         behavior of vec_packsu (vector unsigned long long, vector unsigned
12866         long long) to match behavior of vec_packs with same signature.
12868 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
12869             Thomas Schwinge <thomas@codesourcery.com>
12870             Cesar Philippidis  <cesar@codesourcery.com>
12872         * gimplify.c (gimplify_scan_omp_clauses): Add support for
12873         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12874         (gimplify_adjust_omp_clauses): Likewise.
12875         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
12876         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
12877         (gimplify_omp_target_update): Update handling of acc update and
12878         enter/exit data.
12879         * omp-low.c (install_var_field): Remove unused parameter
12880         base_pointers_restrict.
12881         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
12882         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
12883         FINALIZE}
12884         (omp_target_base_pointers_restrict_p): Delete.
12885         (scan_omp_target): Update call to scan_sharing_clauses.
12886         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
12887         FINALIZE}.
12888         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
12889         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12890         (convert_local_omp_clauses): Likewise.
12891         * tree-pretty-print.c (dump_omp_clause): Likewise.
12892         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
12893         FINALIZE}.
12894         (omp_clause_code_name): Likewise.
12896 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12898         PR debug/86194
12899         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
12900         be narrowed.
12902         PR tree-optimization/86231
12903         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
12904         anti-range don't overwrite *vr0min before using it to compute *vr0max.
12906 2018-06-20  Tom de Vries  <tdevries@suse.de>
12908         PR tree-optimization/86097
12909         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
12910         iv type if signedness of iv type is not the same as that of *nit.
12912 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12914         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
12915         EDGE_EH edges, verify they are all EDGE_EH.
12917 2018-06-20  Maya Rashish  <coypu@sdf.org>
12919         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
12921 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12923         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
12924         * config/aarch64/aarch64.c (xgene1_tunings): Add
12925         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
12926         (aarch64_mode_valid_for_sched_fusion_p):
12927         Allow 16-byte modes.
12928         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
12929         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
12930         128-bit modes.
12931         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
12932         New pattern.
12933         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
12934         * config/aarch64/iterators.md (VQ2): New mode iterator.
12936 2018-06-20  Martin Liska  <mliska@suse.cz>
12938         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12939         Change default ratio from 10 to 8.
12941 2018-06-20  Martin Liska  <mliska@suse.cz>
12943         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12944         New.
12945         (bit_test_cluster::find_bit_tests): Likewise.
12946         (switch_decision_tree::analyze_switch_statement): Find clusters.
12947         * tree-switch-conversion.h (struct jump_table_cluster): Document
12948         hierarchy.
12950 2018-06-20  Martin Liska  <mliska@suse.cz>
12952         * tree-switch-conversion.c (switch_conversion::collect):
12953         Record m_uniq property.
12954         (switch_conversion::expand): Bail out for special conditions.
12955         (group_cluster::~group_cluster): New.
12956         (group_cluster::group_cluster): Likewise.
12957         (group_cluster::dump): Likewise.
12958         (jump_table_cluster::emit): New.
12959         (switch_decision_tree::fix_phi_operands_for_edges): New.
12960         (struct case_node): Remove struct.
12961         (jump_table_cluster::can_be_handled): New.
12962         (case_values_threshold): Moved to header.
12963         (reset_out_edges_aux): Likewise.
12964         (jump_table_cluster::is_beneficial): New.
12965         (bit_test_cluster::can_be_handled): Likewise.
12966         (add_case_node): Remove.
12967         (bit_test_cluster::is_beneficial): New.
12968         (case_bit_test::cmp): New.
12969         (bit_test_cluster::emit): New.
12970         (expand_switch_as_decision_tree_p): Remove.
12971         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
12972         (fix_phi_operands_for_edge): Likewise.
12973         (switch_decision_tree::analyze_switch_statement): New.
12974         (compute_cases_per_edge): Move ...
12975         (switch_decision_tree::compute_cases_per_edge): ... here.
12976         (try_switch_expansion): Likewise.
12977         (switch_decision_tree::try_switch_expansion): Likewise.
12978         (record_phi_operand_mapping): Likewise.
12979         (switch_decision_tree::record_phi_operand_mapping): Likewise.
12980         (emit_case_decision_tree): Likewise.
12981         (switch_decision_tree::emit): Likewise.
12982         (balance_case_nodes): Likewise.
12983         (switch_decision_tree::balance_case_nodes): Likewise.
12984         (dump_case_nodes): Likewise.
12985         (switch_decision_tree::dump_case_nodes): Likewise.
12986         (emit_jump): Likewise.
12987         (switch_decision_tree::emit_jump): Likewise.
12988         (emit_cmp_and_jump_insns): Likewise.
12989         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
12990         (emit_case_nodes): Likewise.
12991         (switch_decision_tree::emit_case_nodes): Likewise.
12992         (conditional_probability): Remove.
12993         * tree-switch-conversion.h (enum cluster_type): New.
12994         (PRINT_CASE): New.
12995         (struct cluster): Likewise.
12996         (cluster::cluster): Likewise.
12997         (struct simple_cluster): Likewise.
12998         (simple_cluster::simple_cluster): Likewise.
12999         (struct group_cluster): Likewise.
13000         (struct jump_table_cluster): Likewise.
13001         (struct bit_test_cluster): Likewise.
13002         (struct min_cluster_item): Likewise.
13003         (struct case_tree_node): Likewise.
13004         (case_tree_node::case_tree_node): Likewise.
13005         (jump_table_cluster::case_values_threshold): Likewise.
13006         (struct case_bit_test): Likewise.
13007         (struct switch_decision_tree): Likewise.
13008         (struct switch_conversion): Likewise.
13009         (switch_decision_tree::reset_out_edges_aux): Likewise.
13011 2018-06-20  Martin Liska  <mliska@suse.cz>
13013         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
13014         (hoist_edge_and_branch_if_true): Likewise.
13015         (expand_switch_using_bit_tests_p): Likewise.
13016         (struct case_bit_test): Likewise.
13017         (case_bit_test_cmp): Likewise.
13018         (emit_case_bit_tests): Likewise.
13019         (switch_conversion::switch_conversion): New class.
13020         (struct switch_conv_info): Remove old struct.
13021         (collect_switch_conv_info): More to ...
13022         (switch_conversion::collect): ... this.
13023         (check_range): Likewise.
13024         (switch_conversion::check_range): Likewise.
13025         (check_all_empty_except_final): Likewise.
13026         (switch_conversion::check_all_empty_except_final): Likewise.
13027         (check_final_bb): Likewise.
13028         (switch_conversion::check_final_bb): Likewise.
13029         (create_temp_arrays): Likewise.
13030         (switch_conversion::create_temp_arrays): Likewise.
13031         (free_temp_arrays): Likewise.
13032         (gather_default_values): Likewise.
13033         (switch_conversion::gather_default_values): Likewise.
13034         (build_constructors): Likewise.
13035         (switch_conversion::build_constructors): Likewise.
13036         (constructor_contains_same_values_p): Likewise.
13037         (switch_conversion::contains_same_values_p): Likewise.
13038         (array_value_type): Likewise.
13039         (switch_conversion::array_value_type): Likewise.
13040         (build_one_array): Likewise.
13041         (switch_conversion::build_one_array): Likewise.
13042         (build_arrays): Likewise.
13043         (switch_conversion::build_arrays): Likewise.
13044         (gen_def_assigns): Likewise.
13045         (switch_conversion::gen_def_assigns): Likewise.
13046         (prune_bbs): Likewise.
13047         (switch_conversion::prune_bbs): Likewise.
13048         (fix_phi_nodes): Likewise.
13049         (switch_conversion::fix_phi_nodes): Likewise.
13050         (gen_inbound_check): Likewise.
13051         (switch_conversion::gen_inbound_check): Likewise.
13052         (process_switch): Use the newly created class.
13053         (switch_conversion::expand): New.
13054         (switch_conversion::~switch_conversion): New.
13055         * tree-switch-conversion.h: New file.
13057 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13059         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
13060         tree-vect-patterns.c.
13061         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
13062         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
13063         (vect_recog_sad_pattern): Likewise.
13064         (vect_recog_widen_sum_pattern): Likewise.
13065         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
13066         (vect_recog_widen_shift_pattern): Remove the type_in argument.
13067         (vect_recog_rotate_pattern): Likewise.
13068         (vect_recog_mult_pattern): Likewise.
13069         (vect_recog_vector_vector_shift_pattern): Likewise.
13070         (vect_recog_divmod_pattern): Likewise.
13071         (vect_recog_mixed_size_cond_pattern): Likewise.
13072         (vect_recog_bool_pattern): Likewise.
13073         (vect_recog_mask_conversion_pattern): Likewise.
13074         (vect_try_gather_scatter_pattern): Likewise.
13075         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
13076         (vect_recog_over_widening_pattern): Likewise.
13077         (vect_recog_gather_scatter_pattern): Likewise.
13078         (vect_recog_func_ptr): Move from tree-vectorizer.h
13079         (vect_vect_recog_func_ptrs): Move further down the file.
13080         (vect_recog_func): Likewise.  Remove the third argument.
13081         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
13082         (vect_pattern_recog_1): Expect the pattern function to do any
13083         necessary target tests.  Also expect it to provide a vector type.
13084         Remove the type_in handling.
13086 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13088         * tree-vect-patterns.c (vect_pattern_detected): New function.
13089         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
13090         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
13091         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
13092         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
13093         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13094         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13095         (vect_recog_mask_conversion_pattern)
13096         (vect_try_gather_scatter_pattern): Likewise.
13098 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13100         * tree-vect-patterns.c (vect_get_internal_def): New function.
13101         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13102         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
13103         (search_type_for_mask_1): Use it.
13105 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13107         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13108         redundant WIDEN_SUM_EXPR handling.
13109         (vect_recog_sad_pattern): Likewise.
13111 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13113         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13114         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
13115         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
13116         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
13117         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
13119 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13121         * tree-vect-stmts.c (vectorizable_call): Make sure that we
13122         use the stmt_vec_info of the original bb statement for the
13123         new zero assignment, even if the call is part of a pattern.
13125 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13127         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
13128         that the sequence is attached to the original statement rather
13129         than the pattern statement.
13130         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
13131         PATTERN_DEF_SEQ from the original statement rather than
13132         the main pattern statement.
13133         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
13134         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
13135         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
13137 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13139         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
13140         definition statements before the early exit for statements that aren't
13141         live or relevant.
13142         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
13143         split out from...
13144         (vect_transform_loop): ...here.  Process pattern definition
13145         statements without first checking whether the main pattern
13146         statement is live or relevant.
13148 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13150         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
13151         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
13153 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
13155         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
13156         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
13157         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
13158         (expand_block_compare): Change select_block_compare_mode call.
13159         (expand_strncmp_align_check): Use new functions, fix comment.
13160         (emit_final_str_compare_gpr): New function.
13161         (expand_strn_compare): Refactor and clean up code.
13162         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
13164 2018-06-19  Tony Reix  <tony.reix@atos.com>
13165             Damien Bergamini  <damien.bergamini@atos.com>
13166             David Edelsohn  <dje.gcc@gmail.com>
13168         * collect2.c (static_obj): New variable.
13169         (static_libs): New variable.
13170         (is_in_list): Uncomment declaration.
13171         (main): Track AIX libraries linked statically.
13172         (is_in_list): Uncomment definition.
13173         (scan_prog_file): Don't add AIX shared libraries initializer
13174         to constructor list if linking statically.
13176 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
13178         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
13179         constant.
13180         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
13182 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
13184         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
13185         blocks.
13187 2018-06-19  Martin Liska  <mliska@suse.cz>
13189         * config/i386/i386.c (ix86_can_inline_p): Do not use
13190         ipa_fn_summaries::get_create.
13191         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
13192         get.
13193         (devirtualization_time_bonus): Likewise.
13194         (ipcp_propagate_stage): Likewise.
13195         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13196         (edge_set_predicate): Likewise.
13197         (evaluate_conditions_for_known_args): Likewise.
13198         (evaluate_properties_for_edge): Likewise.
13199         (ipa_call_summary::reset): Tranform to ...
13200         (ipa_call_summary::~ipa_call_summary): ... this.
13201         (ipa_fn_summary::reset): Transform to ...
13202         (ipa_fn_summary::~ipa_fn_summary): ... this.
13203         (ipa_fn_summary_t::remove): Rename to ...
13204         (ipa_fn_summary_t::remove_callees): ... this.
13205         (ipa_fn_summary_t::duplicate): Use placement new
13206         instead of memory copy.
13207         (ipa_call_summary_t::duplicate): Likewise.
13208         (ipa_call_summary_t::remove): Remove.
13209         (dump_ipa_call_summary): Change get_create to get.
13210         (ipa_dump_fn_summary): Dump only when summary exists.
13211         (analyze_function_body): Use symbol_summary::get instead
13212         of get_create.
13213         (compute_fn_summary): Likewise.
13214         (estimate_edge_devirt_benefit): Likewise.
13215         (estimate_edge_size_and_time): Likewise.
13216         (inline_update_callee_summaries): Likewise.
13217         (remap_edge_change_prob): Likewise.
13218         (remap_edge_summaries): Likewise.
13219         (ipa_merge_fn_summary_after_inlining): Likewise.
13220         (write_ipa_call_summary): Likewise.
13221         (ipa_fn_summary_write): Likewise.
13222         (ipa_free_fn_summary): Likewise.
13223         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
13224         (struct ipa_call_summary): Likewise.
13225         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
13226         of get_create.
13227         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
13228         (estimate_size_after_inlining): Likewise.
13229         (estimate_growth): Likewise.
13230         (growth_likely_positive): Likewise.
13231         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13232         (inline_call): Likewise.
13233         * ipa-inline.c (caller_growth_limits): Likewise.
13234         (can_inline_edge_p): Likewise.
13235         (can_inline_edge_by_limits_p): Likewise.
13236         (compute_uninlined_call_time): Likewise.
13237         (compute_inlined_call_time): Likewise.
13238         (want_inline_small_function_p): Likewise.
13239         (edge_badness): Likewise.
13240         (update_caller_keys): Likewise.
13241         (update_callee_keys): Likewise.
13242         (inline_small_functions): Likewise.
13243         (inline_to_all_callers_1): Likewise.
13244         (dump_overall_stats): Likewise.
13245         (early_inline_small_functions): Likewise.
13246         (early_inliner): Likewise.
13247         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13248         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13249         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13250         * ipa-split.c (execute_split_functions): Likewise.
13251         * symbol-summary.h: Likewise.
13252         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13254 2018-06-19  Richard Biener  <rguenther@suse.de>
13256         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
13257         (vectorize_loops): ... here.  Fix dbgcnt handling.
13258         (try_vectorize_loop): Wrap try_vectorize_loop_1.
13260 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
13262         PR target/86197
13263         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
13264         ieee128 argument takes up only one (vector) register, not two (floating
13265         point) registers.
13267 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13269         * gimplify.c (gimplify_init_constructor): Really never clear for an
13270         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
13272 2018-06-19  Richard Biener  <rguenther@suse.de>
13274         PR tree-optimization/86179
13275         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
13276         after failed recognition.
13278 2018-06-18  Martin Sebor  <msebor@redhat.com>
13280         PR middle-end/85602
13281         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
13282         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
13283         Handle integer subtraction.
13284         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
13285         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
13287 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13289         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
13290         param from rtx to rtx_insn *.
13291         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
13292         param.
13293         (frv_ifcvt_modify_insn): Likwise.
13294         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
13295         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
13296         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
13297         as_a <rtx_insn *> cast to local "unprotected_region" once
13298         it's been established that it's not NULL or pc_rtx.
13299         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
13300         param "sethi" from rtx to rtx_insn *.
13301         (nds32_group_float_insns): Likewise for param "insn".
13302         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
13303         param.
13304         (vax_output_int_subtract): Likewise.
13305         * config/vax/vax.c (vax_output_int_add): Likewise for param
13306         "insn".
13307         (vax_output_int_subtract): Likewise.
13308         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
13309         (emit_pattern_after): Likewise for param "after".
13310         (emit_insn_after): Likewise.
13311         (emit_jump_insn_after): Likewise.
13312         (emit_call_insn_after): Likewise.
13313         (emit_debug_insn_after): Likewise.
13314         (emit_pattern_before): Likewise for param "before".
13315         (emit_insn_before): Likewise.
13316         (emit_jump_insn_before): Likewise.
13317         * final.c (get_insn_template): Likewise for param "insn", removing
13318         a cast.
13319         * output.h (get_insn_template): Likewise for 2nd param.
13320         * rtl.h (emit_insn_before): Likewise.
13321         (emit_jump_insn_before): Likewise.
13322         (emit_debug_insn_before_noloc): Likewise.
13323         (emit_insn_after): Likewise.
13324         (emit_jump_insn_after): Likewise.
13325         (emit_call_insn_after): Likewise.
13326         (emit_debug_insn_after): Likewise.
13327         (set_insn_deleted): Likewise for param.
13329 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
13331         PR target/85358
13332         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
13333         floating point modes, so that IFmode is numerically greater than
13334         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
13335         to declare the ordering.  This prevents IFmode from being
13336         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
13337         machine.  Include rs6000-modes.h to share the fractional values
13338         between genmodes* and the rest of the compiler.
13339         (IFmode): Likewise.
13340         (KFmode): Likewise.
13341         (TFmode): Likewise.
13342         * config/rs6000/rs6000-modes.h: New file.
13343         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
13344         meaning of rs6000_long_double_size so that 126..128 selects an
13345         appropriate 128-bit floating point type.
13346         (rs6000_option_override_internal): Likewise.
13347         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
13348         (TARGET_LONG_DOUBLE_128): Change the meaning of
13349         rs6000_long_double_size so that 126..128 selects an appropriate
13350         128-bit floating point type.
13351         (LONG_DOUBLE_TYPE_SIZE): Update comment.
13352         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
13353         source and destination to match the standard usage.
13354         (truncifkf2): Likewise.
13355         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
13356         ISA 2.07 to use an explicit clobber, instead of passing in a
13357         temporary.
13358         (copysign<mode>3_soft): Likewise.
13360 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13362         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13363         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
13364         (vect_slp_analyze_instance_dependence): Likewise.
13365         (vect_enhance_data_refs_alignment): Likewise.
13366         (vect_analyze_data_refs_alignment): Likewise.
13367         (vect_slp_analyze_and_verify_instance_alignment
13368         (vect_analyze_data_ref_accesses): Likewise.
13369         (vect_prune_runtime_alias_test_list): Likewise.
13370         (vect_analyze_data_refs): Likewise.
13371         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
13372         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13373         (vect_analyze_scalar_cycles_1): Likewise.
13374         (vect_get_loop_niters): Likewise.
13375         (vect_analyze_loop_form_1): Likewise.
13376         (vect_update_vf_for_slp): Likewise.
13377         (vect_analyze_loop_operations): Likewise.
13378         (vect_analyze_loop): Likewise.
13379         (vectorizable_induction): Likewise.
13380         (vect_transform_loop): Likewise.
13381         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
13382         * tree-vect-slp.c (vect_analyze_slp): Likewise.
13383         (vect_make_slp_decision): Likewise.
13384         (vect_detect_hybrid_slp): Likewise.
13385         (vect_slp_analyze_operations): Likewise.
13386         (vect_slp_bb): Likewise.
13387         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
13388         (vectorizable_bswap): Likewise.
13389         (vectorizable_call): Likewise.
13390         (vectorizable_simd_clone_call): Likewise.
13391         (vectorizable_conversion): Likewise.
13392         (vectorizable_assignment): Likewise.
13393         (vectorizable_shift): Likewise.
13394         (vectorizable_operation): Likewise.
13395         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
13397 2018-06-18  Martin Sebor  <msebor@redhat.com>
13399         PR tree-optimization/81384
13400         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
13401         * builtins.c (expand_builtin_strnlen): New function.
13402         (expand_builtin): Call it.
13403         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
13404         * builtins.def (BUILT_IN_STRNLEN): New.
13405         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
13406         Warn for bounds in excess of maximum object size.
13407         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
13408         single-value ranges.  Handle strnlen.
13409         (handle_builtin_strlen): Handle strnlen.
13410         (strlen_check_and_optimize_stmt): Same.
13411         * doc/extend.texi (Other Builtins): Document strnlen.
13413 2018-06-18  Maya Rashish  <coypu@sdf.org>
13415         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
13416         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
13417         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
13419         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
13420         here to ...
13421         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
13423 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13425         * tree.c (escaped_string::escape): Replace cast to char * by
13426         const_cast<char *> (unescaped).
13428 2018-06-18  Nick Clifton  <nickc@redhat.com>
13430         PR 84195
13431         * tree.c (escaped_string): New class.  Converts an unescaped
13432         string into its escaped equivalent.
13433         (warn_deprecated_use): Use the new class to convert the
13434         deprecation message, if present.
13435         (test_escaped_strings): New self test.
13436         (test_c_tests): Add test_escaped_strings.
13437         * doc/extend.texi (deprecated): Add a note that the
13438         deprecation message is affected by the -fmessage-length
13439         option, and that control characters will be escaped.
13440         (#pragma GCC error): Document this pragma.
13441         (#pragma GCC warning): Likewise.
13442         * doc/invoke.texi (-fmessage-length): Document this option's
13443         effect on the #warning and #error preprocessor directives and
13444         the deprecated attribute.
13446 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
13448         * tree.c (decl_value_expr_lookup): Revert latest change.
13449         (decl_value_expr_insert): Likewise.
13451 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
13453         * gimplify.c (nonlocal_vlas): Delete.
13454         (nonlocal_vla_vars): Likewise.
13455         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
13456         referenced VLAs.
13457         (gimplify_body): Do not create and destroy nonlocal_vlas.
13458         * tree-nested.c: Include diagnostic.h.
13459         (use_pointer_in_frame): Tweak.
13460         (lookup_field_for_decl): Add assertion and declare the transformation.
13461         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
13462         internal error when the reference is in a wrong context.  Do not
13463         create a debug decl by default.
13464         (note_nonlocal_block_vlas): Delete.
13465         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
13466         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
13467         create a debug decl by default.
13468         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
13469         call is in a wrong context.
13470         (fixup_vla_decls): New function.
13471         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
13472         debug variables were created.
13473         * tree.c (decl_value_expr_lookup): Add checking assertion.
13474         (decl_value_expr_insert): Likewise.
13476 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13478         PR middle-end/82479
13479         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
13480         * tree-scalar-evolution.c (interpret_expr): Likewise.
13481         (expression_expensive_p): Likewise.
13482         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
13483         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
13484         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
13485         (ssa_defined_by_minus_one_stmt_p): New.
13487 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13489         PR middle-end/64946
13490         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
13491         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
13492         * dojump.c (do_jump): Likewise.
13493         * expr.c (expand_expr_real_2): Check operand type's sign.
13494         * fold-const.c (const_unop): Handle ABSU_EXPR.
13495         (fold_abs_const): Likewise.
13496         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13497         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13498         (strip_sign_op_1): Likesise.
13499         * match.pd: Add new pattern to generate ABSU_EXPR.
13500         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13501         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13502         * tree-eh.c (operation_could_trap_helper_p): Likewise.
13503         * tree-inline.c (estimate_operator_cost): Likewise.
13504         * tree-pretty-print.c (dump_generic_node): Likewise.
13505         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13506         * tree.def (ABSU_EXPR): New.
13508 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
13510         PR middle-end/86095
13511         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13512         documented as preserved for backward compatibility only.
13513         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13515         PR rtl-optimization/86108
13516         * bb-reorder.c (create_forwarder_block): Renamed to ...
13517         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
13518         jump from new landing pad to the second part.
13519         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13520         Adjust callers.
13522 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
13524         PR middle-end/85878
13525         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13526         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13527         Only call store_expr for halves if the mode is the same.
13529         PR middle-end/86123
13530         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13531         Fix up comment formatting.
13533 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13535         * typed-splay-tree.h (typed_splay_tree::remove): New function.
13536         (typed_splay_tree::closure,
13537         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13538         (typed_splay_tree::typed_splay_tree,
13539         typed_splay_tree::operator =): Declared private.
13540         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13541         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13542         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13543         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13544         typed_splay_tree::splay_tree_splay,
13545         typed_splay_tree::splay_tree_foreach_helper,
13546         typed_splay_tree::splay_tree_insert,
13547         typed_splay_tree::splay_tree_remove,
13548         typed_splay_tree::splay_tree_lookup,
13549         typed_splay_tree::splay_tree_predecessor,
13550         typed_splay_tree::splay_tree_successor,
13551         typed_splay_tree::splay_tree_min,
13552         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13553         (typed_splay_tree::root, typed_splay_tree::comp,
13554         typed_splay_tree::delete_key,
13555         typed_splay_tree::delete_value): New data members.
13556         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13557         typed_splay_tree::remove.
13559 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
13561         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13562         -mginv and -mno-ginv to the assembler.
13563         * config/mips/mips.opt (-mcrc): New option.
13564         (-mginv): Likewise.
13565         * doc/invoke.text (-mcrc): Document.
13566         (-mginv): Likewise.
13568 2018-06-15  Nick Clifton  <nickc@redhat.com>
13570         PR 84195
13571         * tree.c (escaped_string): New class.  Converts an unescaped
13572         string into its escaped equivalent.
13573         (warn_deprecated_use): Use the new class to convert the
13574         deprecation message, if present.
13575         (test_escaped_strings): New self test.
13576         (test_c_tests): Add test_escaped_strings.
13577         * doc/extend.texi (deprecated): Add a note that the
13578         deprecation message is affected by the -fmessage-length
13579         option, and that control characters will be escaped.
13580         (#pragma GCC error): Document this pragma.
13581         (#pragma GCC warning): Likewise.
13582         * doc/invoke.texi (-fmessage-length): Document this option's
13583         effect on the #warning and #error preprocessor directives and
13584         the deprecated attribute.
13586 2018-06-15  Richard Biener  <rguenther@suse.de>
13588         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13589         here, also noting vector size used.
13590         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
13591         size used in MSG_OPTIMIZED_LOCATIONS dump.
13592         (pass_slp_vectorize::execute): Adjust.
13594 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
13596         PR target/85968
13597         * config/arc/arc.c (arc_return_address_register): Fix
13598         if-condition.
13600 2018-06-15  Richard Biener  <rguenther@suse.de>
13602         PR middle-end/86159
13603         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13604         leave useless conversion stripping to force_gimple_operand_gsi.
13605         (gimplify_build2): Likewise.
13606         (gimplify_build1): Likewise.
13608 2018-06-15  Richard Biener  <rguenther@suse.de>
13610         PR middle-end/86076
13611         * tree-cfg.c (move_stmt_op): unshare invariant addresses
13612         before adjusting their block.
13614 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13616         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13617         multilibs for *-*-rtems*.
13618         * config/riscv/t-rtems: New file.
13620 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
13622         PR middle-end/86122
13623         * match.pd ((A +- CST1) +- CST2): Punt if last resort
13624         unsigned_type_for returns NULL.
13626         PR target/85945
13627         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13628         subregs of multi-word pseudos unless the float mode has word size.
13630 2018-06-14  Richard Biener  <rguenther@suse.de>
13632         PR middle-end/86139
13633         * tree-vect-generic.c (build_word_mode_vector_type): Remove
13634         duplicate and harmful type_hash_canon.
13635         * tree.c (type_hash_canon): Assert we didn't find ourselves.
13637 2018-06-14  Richard Biener  <rguenther@suse.de>
13639         PR ipa/86124
13640         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13641         NULL cgraph_node.
13643 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13645         * config/rtems.h (STDINT_LONG32): Define.
13647 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
13648             Prachi Godbole  <prachi.godbole@imgtec.com>
13650         * config/mips/mips-cpus.def: Define P6600.
13651         * config/mips/mips-tables.opt: Regenerate.
13652         * config/mips/mips.c (mips_ucbranch_type): New enum.
13653         (mips_rtx_cost_data): Add support for P6600.
13654         (mips_issue_rate): Likewise.
13655         (mips_multipass_dfa_lookahead): Likewise.
13656         (mips_avoid_hazard): Likewise.
13657         (mips_reorg_process_insns): Likewise.
13658         (mips_classify_branch_p6600): New function.
13659         * config/mips/mips.h (TUNE_P6600): New define.
13660         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13661         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13662         * config/mips/mips.md: Include p6600.md.
13663         (processor): Add p6600.
13664         * config/mips/p6600.md: New file.
13665         * doc/invoke.texi: Add p6600 to supported architectures.
13667 2018-06-13  Martin Sebor  <msebor@redhat.com>
13669         PR tree-optimization/86114
13670         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13671         of integer types.
13672         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13674 2018-06-13  Richard Biener  <rguenther@suse.de>
13676         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13677         Properly set vector type of the intermediate stmt.
13678         * tree-vect-stmts.c (vectorizable_operation): The destination
13679         var always has vectype_out type.
13681 2018-06-13  Jeff Law  <law@redhat.com>
13683         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13684         integer 0 for argument to print_rtl_with_bb.
13685         (rl78_reorg): Likewise.
13687 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
13689         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13690         from rtx to rtx_insn *.
13691         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13692         "label".
13693         (add_sched_insns_for_speculation): Likewise for local "target",
13694         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13695         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13696         from rtx_insn ** to rtx_code_label **.
13697         (reorg_emit_nops): Likewise.
13698         (c6x_reorg): Likewise for local "call_labels".
13699         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13700         rtx to rtx_insn *.
13701         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13702         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13703         the loops over LABEL_REFS.
13704         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13705         braf_label.
13706         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13707         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13708         (split_branches): Strengthen local "olabel" from rtx to
13709         rtx_insn *, adding a safe_as_a cast.
13710         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13711         to "rtx_insn *".
13712         (add_insn_after): Likewise for first two params.
13713         (add_insn_before): Likewise.
13714         (remove_insn): Likewise for param.
13715         (emit_pattern_before_noloc): Likewise for second and third params.
13716         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13717         (emit_call_insn_before_noloc): Likewise.
13718         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13719         to "rtx_insn *".
13720         (emit_barrier_before): Likewise.
13721         (emit_label_before): Strengthen "label" param from "rtx" to
13722         "rtx_code_label *".  Strengthen "before" param from "rtx" to
13723         "rtx_insn *".
13724         (emit_insn_after_1): Strengthen "after" param from "rtx" to
13725         "rtx_insn *".
13726         (emit_pattern_after_noloc): Likewise.
13727         (emit_insn_after_noloc): Likewise.
13728         (emit_jump_insn_after_noloc): Likewise.
13729         (emit_call_insn_after_noloc): Likewise.
13730         (emit_debug_insn_after_noloc): Likewise.
13731         (emit_barrier_after): Likewise.
13732         (emit_label_after): Likewise for both params.
13733         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
13734         "loc" param from "int" to "location_t".
13735         (emit_insn_after_setloc): Likewise.
13736         (emit_jump_insn_after_setloc): Likewise.
13737         (emit_call_insn_after_setloc): Likewise.
13738         (emit_debug_insn_after_setloc): Likewise.
13739         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
13740         "loc" param from "int" to "location_t".
13741         (emit_pattern_before): Convert NULL_RTX to NULL.
13742         (emit_insn_before_setloc): Convert "loc" param from "int" to
13743         "location_t".
13744         (emit_jump_insn_before_setloc): Likewise.
13745         (emit_call_insn_before_setloc): Likewise.
13746         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13747         rtx_insn *.  Convert "loc" param from "int" to "location_t".
13748         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13749         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13750         Convert 3rd param from "int" to "location_t".
13751         (emit_barrier_before, emit_barrier_after, next_real_insn):
13752         Strengthen param from rtx to rtx_insn *.
13753         (emit_label_before): Strengthen 1st param from "rtx" to
13754         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
13755         "rtx_insn *".
13756         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13757         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13758         Strengthen 2nd param from "rtx" to "rtx_insn *".
13759         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13760         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13761         Likewise. Convert 3rd param from "int" to "location_t".
13762         (emit_label_after): Strengthen 1st param from "rtx" to
13763         "rtx_code_label *".
13764         (next_real_insn, remove_insn): Strengthen param from "rtx" to
13765         "rtx_insn *".
13766         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13767         from "rtx" to "rtx_insn *".
13769 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
13771         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13772         bodies streamed in with -Q.
13773         * dumpfile.c (dump_files): Add lto-stream-out dump file.
13774         * dumpfile.h (tree_dump_index): Add lto_stream_out.
13775         * gimple-streamer-out.c: Include gimple-pretty-print.h
13776         (output_bb): Dump stmts streamed.
13777         * lto-section-out.c: Include print-tree.h
13778         (lto_begin_section): Dump sections created.
13779         (lto_output_decl_index): Dump decl encoded.
13780         * lto-streamer-out.c: Include print-tree.h
13781         (create_output_block): Dump output block created.
13782         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13783         (output_function): Dump function output.
13784         (output_constructor): Dump constructor streamed.
13785         (write_global_stream): Output indexes encoded.
13786         (produce_asm_for_decls): Dump streams encoded.
13787         * lto-streamer.c (streamer_dump_file): New global var.
13788         * lto-streamer.h (streamer_dump_file): Declare.
13789         * passes.c (ipa_write_summaries): Initialize streamer dump.
13790         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13791         in.
13793 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
13795         PR target/86048
13796         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13797         offsets for register save directives.  Emit a second batch of save
13798         directives, if need be, when the function accesses prior frames.
13800 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13802         * config/arc/fpu.md (fmasf4): Force operand to register.
13803         (fnmasf4): Likewise.
13805 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13807         * config/arc/arc-protos.h (arc_pad_return): Remove.
13808         * config/arc/arc.c (machine_function): Remove force_short_suffix
13809         and size_reason.
13810         (arc_print_operand): Adjust printing of '&'.
13811         (arc_verify_short): Remove conditional printing of short suffix.
13812         (arc_final_prescan_insn): Remove reference to size_reason.
13813         (pad_return): New function.
13814         (arc_reorg): Call pad_return.
13815         (arc_pad_return): Remove.
13816         (arc_init_machine_status): Remove reference to force_short_suffix.
13817         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13818         (attr length): When attribute iscompact is true force to 2
13819         regardless; in the case of maybe check if we want to force the
13820         instruction to have 4 bytes length.
13821         (nopv): Change it to generate 4 byte long nop as well.
13822         (blockage): New pattern.
13823         (simple_return): Remove call to arc_pad_return.
13824         (p_return_i): Likewise.
13826 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13828         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13830 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13832         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13833         ARC cores.
13835 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13837         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13838         for ARC700 and ARCv2.
13840 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
13842         PR target/86076
13843         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13844         operands[2] instead of operands[1].
13847 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13849         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13850         case, check whether the outer register overlaps an unallocatable
13851         register, not just whether it fits the required class.
13853 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13855         * poly-int.h (can_div_trunc_p): Add new overload in which all values
13856         are poly_ints.
13857         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13858         (memrefs_conflict_p): Likewise.
13859         (init_alias_analysis): Likewise.
13860         * cfgexpand.c (expand_debug_expr): Likewise.
13861         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13862         * cse.c (fold_rtx): Likewise.
13863         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13864         * expr.c (emit_block_move_hints): Likewise.
13865         (clear_storage_hints, push_block, emit_push_insn): Likewise.
13866         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13867         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13868         (emit_group_store): Likewise.
13869         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
13870         to read the PRE/POST_MODIFY increment.
13871         * calls.c (store_one_arg): Use strip_offset.
13872         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13873         poly_int_rtx_p.
13874         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13875         by a VEC_SELECT.
13876         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
13877         (simplify_binary_operation_1): Extend CONST_INT handling to
13878         poly_int_rtx_p.
13879         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
13880         than a HOST_WIDE_INT.
13881         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
13882         poly_int64.
13883         (adjust_mems, add_stores): Update accodingly.
13884         (vt_canonicalize_addr): Track polynomial offsets.
13885         (emit_note_insn_var_location): Likewise.
13886         (vt_add_function_parameter): Likewise.
13887         (vt_initialize): Likewise.
13889 2018-06-12  Jeff Law  <law@redhat.com>
13891         * config.gcc (alpha*-*-freebsd*): Remove.
13892         * config/alpha/freebsd.h: Remove.
13894 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
13896         PR other/69968
13897         * spellcheck-tree.c (levenshtein_distance): Rename to...
13898         (get_edit_distance): ...this, and update for underlying renaming.
13899         * spellcheck-tree.h (levenshtein_distance): Rename to...
13900         (get_edit_distance): ...this.
13901         * spellcheck.c (levenshtein_distance): Rename to...
13902         (get_edit_distance): ...this.  Convert from Levenshtein distance
13903         to Damerau-Levenshtein distance by supporting transpositions of
13904         adjacent characters.  Rename "v1" to "v_next" and "v0" to
13905         "v_one_ago".
13906         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
13907         (selftest::test_edit_distance_unit_test_oneway): ...this, and
13908         update for underlying renaming.
13909         (selftest::levenshtein_distance_unit_test): Rename to...
13910         (selftest::test_get_edit_distance_unit): ...this, and update for
13911         underlying renaming.
13912         (selftest::test_find_closest_string): Add example from PR 69968
13913         where transposition helps
13914         (selftest::test_metric_conditions): Update for renaming.
13915         (selftest::test_metric_conditions): Likewise.
13916         (selftest::spellcheck_c_tests): Likewise.
13917         * spellcheck.h (levenshtein_distance): Rename both overloads to...
13918         (get_edit_distance): ...this.
13919         (best_match::consider): Update for renaming.
13921 2018-06-12  Martin Sebor  <msebor@redhat.com>
13923         PR tree-optimization/85259
13924         * builtins.c (compute_objsize): Handle constant offsets.
13925         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
13926         true iff a warning has been issued.
13927         * gimple.h (gimple_nonartificial_location): New function.
13928         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
13929         gimple_nonartificial_location and handle -Wno-system-headers.
13930         (handle_builtin_stxncpy): Same.
13932 2018-06-12  Martin Sebor  <msebor@redhat.com>
13934         PR c/85931
13935         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
13937 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
13939         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
13940         BUILTIN_VEC_XST entries for pointer to double and long long.
13942 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
13944         PR target/85990
13945         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13946         Update comments.
13947         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13948         Likewise.
13950 2018-06-12  Martin Liska  <mliska@suse.cz>
13952         * doc/options.texi: Document IntegerRange.
13954 2018-06-12  Martin Liska  <mliska@suse.cz>
13956         * config/i386/i386.opt: Make MPX-related options as Deprecated.
13957         * opt-functions.awk: Handle Deprecated flag.
13958         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
13959         and report error.
13960         (read_cmdline_option): Report warning for a deprecated option.
13961         * opts.h (struct cl_option): Add new field cl_deprecated.
13962         (CL_ERR_DEPRECATED): New.
13964 2018-06-12  Martin Liska  <mliska@suse.cz>
13966         * doc/options.texi: Document Deprecated option flag.
13968 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13970         * config/arc/arc-arch.h (arc_extras): New enum.
13971         (arc_cpu_t):Add field extra.
13972         (arc_cpu_types): Consider the extras.
13973         * config/arc/arc-cpus.def: Add extras info.
13974         * config/arc/arc-opts.h (processor_type): Consider extra field.
13975         * config/arc/arc.c (arc_override_options): Handle extra field.
13977 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13979         * config/arc/arc-arch.h: Update ARC_OPTX macro.
13980         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
13981         field.
13982         * config/arc/arc.c (arc_init): Update pic warning.
13983         (irq_range): Update irq range parsing warnings.
13984         (arc_override_options): Update various warning messages.
13985         (arc_handle_aux_attribute): Likewise.
13987 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
13989         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
13991 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13993         * doc/sourcebuild.texi: Document usage of line number 0 in verify
13994         compiler messages directives.
13996 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
13998         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
13999         * config/mips/mips-tables.opt: Regenerate.
14000         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
14001         mips64r6.
14002         * doc/invoke.texi: Document -march=i6500.
14004 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
14006         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
14007         (i6400_gpmul): Add cpu_unit.
14008         (i6400_gpdiv): Likewise.
14009         (i6400_msa_add_d): Update reservations.
14010         (i6400_msa_int_add) Likewise.
14011         (i6400_msa_short_logic3) Likewise.
14012         (i6400_msa_short_logic2) Likewise.
14013         (i6400_msa_short_logic) Likewise.
14014         (i6400_msa_move) Likewise.
14015         (i6400_msa_cmp) Likewise.
14016         (i6400_msa_short_float2) Likewise.
14017         (i6400_msa_div_d) Likewise.
14018         (i6400_msa_long_logic1) Likewise.
14019         (i6400_msa_long_logic2) Likewise.
14020         (i6400_msa_mult) Likewise.
14021         (i6400_msa_long_float2) Likewise.
14022         (i6400_msa_long_float4) Likewise.
14023         (i6400_msa_long_float5) Likewise.
14024         (i6400_msa_long_float8) Likewise.
14025         (i6400_fpu_fadd): Include frint type.
14026         (i6400_fpu_store): New define_insn_reservation.
14027         (i6400_fpu_load): Likewise.
14028         (i6400_fpu_move): Likewise.
14029         (i6400_fpu_fcmp): Likewise.
14030         (i6400_fpu_fmadd): Likewise.
14031         (i6400_int_mult): Include imul3nc type and update reservation.
14032         (i6400_int_div): Include idiv3 type and update reservation.
14033         (i6400_int_load): Update to check type not move_type.
14034         (i6400_int_store): Likewise.
14035         (i6400_int_prefetch): Set zero latency.
14037 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
14039         * gcc.c: Document new %@{...} sequence.
14040         (LINK_COMMAND_SPEC): Use it for the -L switches.
14041         (cpp_unique_options): Use it for the -I switches.
14042         (at_file_argbuf): New global variable.
14043         (in_at_file): Likewise.
14044         (alloc_args): Create at_file_argbuf.
14045         (clear_args): Truncate at_file_argbuf.
14046         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
14047         (open_at_file): New function.
14048         (close_at_file): Likewise.
14049         (create_at_file): Delete.
14050         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
14051         <'o'>: Likewise.
14052         <'@'>: New case.
14053         (validate_switches_from_spec): Deal with %@{...} sequence.
14054         (validate_switches): Likewise.
14055         (driver::finalize): Call clear_args.
14057 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
14059         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
14061 2018-06-11  Martin Sebor  <msebor@redhat.com>
14063         * doc/invoke.texi (-Wall): List -Wc++17-compat.
14064         (Wno-class-memaccess): Add @opindex.
14065         (Wno-templates, Wno-multiple-inheritance): Same.
14066         (Wno-virtual-inheritance, Wno-namespaces): Same.
14067         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
14068         (Wno-format-overflow, Wno-format-truncation): Same.
14069         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
14070         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
14071         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
14072         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
14073         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
14074         (Wno-misspelled-isr): Same.
14076 2018-06-11  Martin Sebor  <msebor@redhat.com>
14078         * PR tree-optimization/86083
14079         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
14081 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
14083         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
14085 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
14087         PR target/85755
14088         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
14089         on the correct operand.
14090         (*movdi_internal64): Ditto.
14092 2018-06-11  Martin Liska  <mliska@suse.cz>
14094         PR tree-optimization/86089
14095         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
14097 2018-06-11  Julia Koval  <julia.koval@intel.com>
14099         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
14100         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
14101         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
14103 2018-06-11  Olivier Hainque  <hainque@adacore.com>
14105         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
14106         for Ada with strict dwarf2.
14108 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
14110         PR target/85755
14111         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
14112         addresses.
14114 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
14116         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
14118 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
14120         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
14121         TARGET_ELF.
14123 2018-06-08  Martin Liska  <mliska@suse.cz>
14125         * tree-cfg.h (debug_function): Fix argument type to match
14126         implementation.
14128 2018-06-08  Martin Liska  <mliska@suse.cz>
14130         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
14131         Remove usage of MPX-related (and removed) fields.
14132         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
14134 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
14136         * cfg.c (debug): Use TDF_NONE rather than 0.
14137         * cfghooks.c (debug): Likewise.
14138         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
14139         (struct dump_option_value_info): Convert to...
14140         (struct kv_pair): ...this template type.
14141         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
14142         rather than 0.
14143         (optinfo_verbosity_options): Likewise.
14144         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
14145         OPTGROUP_NONE.
14146         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
14147         than int for "optgroup_flags" param.
14148         (dump_generic_expr_loc): Use dump_flags_t rather than int for
14149         "dump_kind" param.
14150         (dump_dec): Likewise.
14151         (dump_finish): Use TDF_NONE rather than 0.
14152         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
14153         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
14154         than 0.  Update for change to option_ptr.
14155         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
14156         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
14157         0.  Update for changes to optinfo_verbosity_options and
14158         optgroup_options.
14159         (opt_info_switch_p): Convert optgroup_flags from int to
14160         optgroup_flags_t.
14161         (dump_basic_block): Use dump_flags_t rather than int
14162         for "dump_kind" param.
14163         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
14164         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
14165         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
14166         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
14167         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
14168         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
14169         TDF_NONE): Convert from macros to...
14170         (enum dump_flag): ...this new enum.
14171         (dump_flags_t): Update to use enum.
14172         (operator|, operator&, operator~, operator|=, operator&=):
14173         Implement for dump_flags_t.
14174         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
14175         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
14176         Convert from macros to...
14177         (enum optgroup_flag): ...this new enum.
14178         (optgroup_flags_t): New typedef.
14179         (operator|, operator|=): Implement for optgroup_flags_t.
14180         (struct dump_file_info): Convert field "alt_flags" to
14181         dump_flags_t.  Convert field "optgroup_flags" to
14182         optgroup_flags_t.
14183         (dump_basic_block): Use dump_flags_t rather than int for param.
14184         (dump_generic_expr_loc): Likewise.
14185         (dump_dec): Likewise.
14186         (dump_register): Convert param "optgroup_flags" to
14187         optgroup_flags_t.
14188         (opt_info_enable_passes): Likewise.
14189         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
14190         than 0.
14191         * gimple-pretty-print.c (debug): Likewise.
14192         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
14193         (merged_store_group::apply_stores): Likewise.
14194         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
14195         * gimple.c (verify_gimple_pp): Likewise.
14196         * graphite-poly.c (print_pbb_body): Likewise.
14197         * passes.c (pass_manager::register_one_dump_file): Convert
14198         local "optgroup_flags" to optgroup_flags_t.
14199         * print-tree.c (print_node): Use TDF_NONE rather than 0.
14200         (debug): Likewise.
14201         (debug_body): Likewise.
14202         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
14203         to optgroup_flags_t.
14204         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
14205         than 0.
14206         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
14207         (convert_mult_to_fma): Likewise.
14208         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
14209         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
14210         * tree-vect-data-refs.c (dump_lower_bound): Convert param
14211         "dump_kind" to dump_flags_t.
14213 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14215         * config/rs6000/rs6000.c (min, max): Delete.
14217 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14219         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
14220         -mabi=spe and -mabi=no-spe.
14222 2018-06-08  Martin Liska  <mliska@suse.cz>
14224         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
14225         where we expect an existing summary.
14227 2018-06-08  Martin Liska  <mliska@suse.cz>
14229         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
14230         * ipa-inline.h (estimate_edge_growth): Likewise.
14232 2018-06-08  Martin Liska  <mliska@suse.cz>
14234         * cgraph.c (function_version_hasher::hash): Use
14235         cgraph_node::get_uid ().
14236         (function_version_hasher::equal):
14237         * cgraph.h (cgraph_node::get_uid): New method.
14238         * ipa-inline.c (update_caller_keys): Use
14239         cgraph_node::get_uid ().
14240         (update_callee_keys): Likewise.
14241         * ipa-utils.c (searchc): Likewise.
14242         (ipa_reduced_postorder): Likewise.
14243         * lto-cgraph.c (input_node): Likewise.
14244         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
14245         * symbol-summary.h (symtab_insertion): Likewise.
14246         (symtab_removal): Likewise.
14247         (symtab_duplication): Likewise.
14248         * tree-pretty-print.c (dump_function_header): Likewise.
14249         * tree-sra.c (convert_callers_for_node): Likewise.
14251 2018-06-08  Martin Liska  <mliska@suse.cz>
14253         * cgraph.c (symbol_table::create_edge): Always assign a new
14254         unique number.
14255         (symbol_table::free_edge): Do not recycle numbers.
14256         * cgraph.h (cgraph_edge::get): New method.
14257         * symbol-summary.h (symtab_removal): Use it.
14258         (symtab_duplication): Likewise.
14259         (call_summary::hashable_uid): Remove.
14261 2018-06-08  Martin Liska  <mliska@suse.cz>
14263         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
14264         (initialize_growth_caches): Remove.
14265         (free_growth_caches): Likewise.
14266         (do_estimate_edge_time): Use edge_growth_cache.
14267         (do_estimate_edge_size): Likewise.
14268         (do_estimate_edge_hints): Likewise.
14269         * ipa-inline.c (reset_edge_caches): Likewise.
14270         (recursive_inlining): Likewise.
14271         (inline_small_functions): Likewise.
14272         * ipa-inline.h (initialize_growth_caches): Remove.
14273         (estimate_edge_size): Likewise.
14274         (estimate_edge_time): Likewise.
14275         (estimate_edge_hints): Likewise.
14276         (reset_edge_growth_cache): Likewise.
14277         * symbol-summary.h (call_summary::remove): New method.
14279 2018-06-08  Martin Liska  <mliska@suse.cz>
14281         * ipa-cp.c (class edge_clone_summary): New summary.
14282         (grow_edge_clone_vectors): Remove.
14283         (ipcp_edge_duplication_hook): Remove.
14284         (class edge_clone_summary_t): New call_summary class.
14285         (ipcp_edge_removal_hook): Remove.
14286         (edge_clone_summary_t::duplicate): New function.
14287         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
14288         (create_specialized_node): Likewise.
14289         (ipcp_driver): Initialize edge_clone_summaries and do not
14290         register hooks.
14292 2018-06-08  Martin Liska  <mliska@suse.cz>
14294         * symbol-summary.h (get): New function.
14295         (call_summary::m_initialize_when_cloning): New class member.
14297 2018-06-08  Martin Liska  <mliska@suse.cz>
14299         * cgraph.c (cgraph_node::remove): Do not recycle uid.
14300         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
14301         (symbol_table::allocate_cgraph_symbol): Do not set uid.
14302         * passes.c (uid_hash_t): Record removed_nodes by their uids.
14303         (remove_cgraph_node_from_order): Use the removed_nodes set.
14304         (do_per_function_toporder): Likwise.
14305         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
14306         instead of summary_uid.
14307         (symtab_removal): Likewise.
14308         (symtab_duplication): Likewise.
14310 2018-06-08  Martin Liska  <mliska@suse.cz>
14312         * ipa-cp.c (ipcp_store_bits_results): Use
14313         ipcp_transformation_sum.
14314         (ipcp_store_vr_results): Likewise.
14315         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
14316         to ...
14317         (ipcp_transformation_initialize): ... this.
14318         (ipa_set_node_agg_value_chain):
14319         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
14320         (write_ipcp_transformation_info): Likewise.
14321         (read_ipcp_transformation_info): Likewise.
14322         (ipcp_update_bits): Likewise.
14323         (ipcp_update_vr): Likewise.
14324         (ipcp_transform_function): Likewise.
14325         * ipa-prop.h: Rename ipcp_transformation_summary to
14326         ipcp_transformation.
14327         (class ipcp_transformation_t): New function summary.
14328         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
14329         (ipa_get_agg_replacements_for_node): Likewise.
14331 2018-06-08  Martin Liska  <mliska@suse.cz>
14333         * ipa-pure-const.c (struct funct_state_d): Do it class instead
14334         of struct.
14335         (class funct_state_summary_t): New function_summary class.
14336         (has_function_state): Remove.
14337         (get_function_state): Likewise.
14338         (set_function_state): Likewise.
14339         (add_new_function): Likewise.
14340         (funct_state_summary_t::insert): New function.
14341         (duplicate_node_data): Remove.
14342         (remove_node_data): Remove.
14343         (funct_state_summary_t::duplicate): New function.
14344         (register_hooks): Create new funct_state_summaries.
14345         (pure_const_generate_summary): Use it.
14346         (pure_const_write_summary): Likewise.
14347         (pure_const_read_summary): Likewise.
14348         (propagate_pure_const): Likewise.
14349         (propagate_nothrow): Likewise.
14350         (dump_malloc_lattice): Likewise.
14351         (propagate_malloc): Likewise.
14352         (execute): Do not register hooks, just remove summary
14353         instead.
14354         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
14355         constructor.
14357 2018-06-08  Martin Liska  <mliska@suse.cz>
14359         * ipa-reference.c (remove_node_data): Remove.
14360         (duplicate_node_data): Likewise.
14361         (class ipa_ref_var_info_summary_t): New class.
14362         (class ipa_ref_opt_summary_t): Likewise.
14363         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
14364         (get_reference_optimization_summary): Use
14365         ipa_ref_opt_sum_summaries.
14366         (set_reference_vars_info): Remove.
14367         (set_reference_optimization_summary): Likewise.
14368         (ipa_init): Create summaries.
14369         (init_function_info): Use function summary.
14370         (ipa_ref_opt_summary_t::duplicate): New function.
14371         (ipa_ref_opt_summary_t::remove): New function.
14372         (get_read_write_all_from_node): Fix GNU coding style.
14373         (propagate): Use function summary.
14374         (write_node_summary_p): Fix GNU coding style.
14375         (stream_out_bitmap): Likewise.
14376         (ipa_reference_read_optimization_summary): Use function summary.
14377         (ipa_reference_c_finalize): Do not release hooks.
14379 2018-06-08  Martin Liska  <mliska@suse.cz>
14381         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
14382         (analyze_function_body): Extract multiple calls of get_create.
14383         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14384         * ipa-inline.c (recursive_inlining): Use ::get method.
14385         * ipa-inline.h (estimate_edge_growth): Likewise.
14387 2018-06-08  Martin Liska  <mliska@suse.cz>
14389         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
14390         HSA_INVALID.
14391         (hsa_function_summary::hsa_function_summary): Use the new enum
14392         value.
14393         (hsa_gpu_implementation_p): Use hsa_summaries::get.
14394         * hsa-gen.c (hsa_get_host_function): Likewise.
14395         (get_brig_function_name): Likewise.
14396         * ipa-hsa.c (process_hsa_functions): Likewise.
14397         (ipa_hsa_write_summary): Likewise.
14398         * symbol-summary.h (symtab_duplication): Use ::get function/
14399         (get): New function.
14401 2018-06-08  Martin Liska  <mliska@suse.cz>
14403         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
14404         of get.
14405         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
14406         (hsa_register_kernel): Likewise.
14407         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
14408         * hsa-gen.c (hsa_get_host_function): Likewise.
14409         (get_brig_function_name): Likewise.
14410         (generate_hsa): Likewise.
14411         (pass_gen_hsail::execute): Likewise.
14412         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
14413         (devirtualization_time_bonus): Likewise.
14414         (ipcp_propagate_stage): Likewise.
14415         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14416         (edge_set_predicate): Likewise.
14417         (evaluate_conditions_for_known_args): Likewise.
14418         (evaluate_properties_for_edge): Likewise.
14419         (ipa_fn_summary::reset): Likewise.
14420         (ipa_fn_summary_t::duplicate): Likewise.
14421         (dump_ipa_call_summary): Likewise.
14422         (ipa_dump_fn_summary): Likewise.
14423         (analyze_function_body): Likewise.
14424         (compute_fn_summary): Likewise.
14425         (estimate_edge_devirt_benefit): Likewise.
14426         (estimate_edge_size_and_time): Likewise.
14427         (estimate_calls_size_and_time): Likewise.
14428         (estimate_node_size_and_time): Likewise.
14429         (inline_update_callee_summaries): Likewise.
14430         (remap_edge_change_prob): Likewise.
14431         (remap_edge_summaries): Likewise.
14432         (ipa_merge_fn_summary_after_inlining): Likewise.
14433         (ipa_update_overall_fn_summary): Likewise.
14434         (read_ipa_call_summary): Likewise.
14435         (inline_read_section): Likewise.
14436         (write_ipa_call_summary): Likewise.
14437         (ipa_fn_summary_write): Likewise.
14438         (ipa_free_fn_summary): Likewise.
14439         * ipa-hsa.c (process_hsa_functions): Likewise.
14440         (ipa_hsa_write_summary): Likewise.
14441         (ipa_hsa_read_section): Likewise.
14442         * ipa-icf.c (sem_function::merge): Likewise.
14443         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14444         (do_estimate_edge_time): Likewise.
14445         (estimate_size_after_inlining): Likewise.
14446         (estimate_growth): Likewise.
14447         (growth_likely_positive): Likewise.
14448         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14449         (inline_call): Likewise.
14450         * ipa-inline.c (caller_growth_limits): Likewise.
14451         (can_inline_edge_p): Likewise.
14452         (can_inline_edge_by_limits_p): Likewise.
14453         (compute_uninlined_call_time): Likewise.
14454         (compute_inlined_call_time): Likewise.
14455         (want_inline_small_function_p): Likewise.
14456         (edge_badness): Likewise.
14457         (update_caller_keys): Likewise.
14458         (update_callee_keys): Likewise.
14459         (recursive_inlining): Likewise.
14460         (inline_small_functions): Likewise.
14461         (inline_to_all_callers_1): Likewise.
14462         (dump_overall_stats): Likewise.
14463         (early_inline_small_functions): Likewise.
14464         (early_inliner): Likewise.
14465         * ipa-inline.h (estimate_edge_growth): Likewise.
14466         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14467         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14468         * ipa-prop.h (IPA_NODE_REF): Likewise.
14469         (IPA_EDGE_REF): Likewise.
14470         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14471         (propagate_malloc): Likewise.
14472         * ipa-split.c (execute_split_functions): Likewise.
14473         * symbol-summary.h: Rename get to get_create.
14474         (get): Likewise.
14475         (get_create): Likewise.
14476         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14478 2018-06-08  Martin Liska  <mliska@suse.cz>
14480         * symbol-summary.h (release): Move definition out of class
14481         declaration.
14482         (symtab_removal): Likewise.
14483         (symtab_duplication): Likewise.
14485 2018-06-08  Martin Liska  <mliska@suse.cz>
14487         * symbol-summary.h (function_summary): Move constructor
14488         implementation out of class declaration.
14489         (release): Likewise.
14490         (symtab_insertion): Likewise.
14491         (symtab_removal): Likewise.
14492         (symtab_duplication): Likewise.
14493         (get): Likewise.
14495 2018-06-08  Martin Liska  <mliska@suse.cz>
14497         * Makefile.in: Remove support for MPX (macros, related functions,
14498         fields in cgraph_node, ...).
14499         * builtin-types.def (BT_BND): Likewise.
14500         (BT_FN_BND_CONST_PTR): Likewise.
14501         (BT_FN_CONST_PTR_BND): Likewise.
14502         (BT_FN_VOID_PTR_BND): Likewise.
14503         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14504         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14505         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14506         (expand_builtin_mempcpy_with_bounds): Likewise.
14507         (expand_builtin_memset_with_bounds): Likewise.
14508         (expand_builtin_memset_args): Likewise.
14509         (std_expand_builtin_va_start): Likewise.
14510         (expand_builtin): Likewise.
14511         (expand_builtin_with_bounds): Likewise.
14512         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14513         (DEF_LIB_BUILTIN_CHKP): Likewise.
14514         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14515         (DEF_CHKP_BUILTIN): Likewise.
14516         (BUILT_IN_MEMCPY): Likewise.
14517         (BUILT_IN_MEMMOVE): Likewise.
14518         (BUILT_IN_MEMPCPY): Likewise.
14519         (BUILT_IN_MEMSET): Likewise.
14520         (BUILT_IN_STPCPY): Likewise.
14521         (BUILT_IN_STRCAT): Likewise.
14522         (BUILT_IN_STRCHR): Likewise.
14523         (BUILT_IN_STRCPY): Likewise.
14524         (BUILT_IN_STRLEN): Likewise.
14525         (BUILT_IN_MEMCPY_CHK): Likewise.
14526         (BUILT_IN_MEMMOVE_CHK): Likewise.
14527         (BUILT_IN_MEMPCPY_CHK): Likewise.
14528         (BUILT_IN_MEMSET_CHK): Likewise.
14529         (BUILT_IN_STPCPY_CHK): Likewise.
14530         (BUILT_IN_STRCAT_CHK): Likewise.
14531         (BUILT_IN_STRCPY_CHK): Likewise.
14532         * calls.c (store_bounds): Likewise.
14533         (emit_call_1): Likewise.
14534         (special_function_p): Likewise.
14535         (maybe_warn_nonstring_arg): Likewise.
14536         (initialize_argument_information): Likewise.
14537         (finalize_must_preallocate): Likewise.
14538         (compute_argument_addresses): Likewise.
14539         (expand_call): Likewise.
14540         * cfgexpand.c (expand_call_stmt): Likewise.
14541         (expand_return): Likewise.
14542         (expand_gimple_stmt_1): Likewise.
14543         (pass_expand::execute): Likewise.
14544         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14545         (cgraph_node::remove): Likewise.
14546         (cgraph_node::dump): Likewise.
14547         (cgraph_node::verify_node): Likewise.
14548         * cgraph.h (chkp_function_instrumented_p): Likewise.
14549         (symtab_node::get_alias_target): Likewise.
14550         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14551         (cgraph_local_p): Likewise.
14552         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14553         (cgraph_edge::rebuild_references): Likewise.
14554         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14555         (walk_polymorphic_call_targets): Likewise.
14556         (cgraph_node::expand_thunk): Likewise.
14557         (symbol_table::output_weakrefs): Likewise.
14558         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14559         (ix86_handle_option): Likewise.
14560         * config/i386/constraints.md: Likewise.
14561         * config/i386/i386-builtin-types.def (BND): Likewise.
14562         (VOID): Likewise.
14563         (PVOID): Likewise.
14564         (ULONG): Likewise.
14565         * config/i386/i386-builtin.def (BDESC_END): Likewise.
14566         (BDESC_FIRST): Likewise.
14567         (BDESC): Likewise.
14568         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14569         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14570         * config/i386/i386.c (enum reg_class): Likewise.
14571         (ix86_target_string): Likewise.
14572         (ix86_option_override_internal): Likewise.
14573         (ix86_conditional_register_usage): Likewise.
14574         (ix86_valid_target_attribute_inner_p): Likewise.
14575         (ix86_set_indirect_branch_type): Likewise.
14576         (ix86_set_current_function): Likewise.
14577         (ix86_function_arg_regno_p): Likewise.
14578         (init_cumulative_args): Likewise.
14579         (ix86_function_arg_advance): Likewise.
14580         (ix86_function_arg): Likewise.
14581         (ix86_pass_by_reference): Likewise.
14582         (ix86_function_value_regno_p): Likewise.
14583         (ix86_function_value_1): Likewise.
14584         (ix86_function_value_bounds): Likewise.
14585         (ix86_return_in_memory): Likewise.
14586         (ix86_setup_incoming_vararg_bounds): Likewise.
14587         (ix86_va_start): Likewise.
14588         (indirect_thunk_need_prefix): Likewise.
14589         (print_reg): Likewise.
14590         (ix86_print_operand): Likewise.
14591         (ix86_expand_call): Likewise.
14592         (ix86_output_function_return): Likewise.
14593         (reg_encoded_number): Likewise.
14594         (BDESC_VERIFYS): Likewise.
14595         (ix86_init_mpx_builtins): Likewise.
14596         (ix86_init_builtins): Likewise.
14597         (ix86_emit_cmove): Likewise.
14598         (ix86_emit_move_max): Likewise.
14599         (ix86_expand_builtin): Likewise.
14600         (ix86_builtin_mpx_function): Likewise.
14601         (ix86_get_arg_address_for_bt): Likewise.
14602         (ix86_load_bounds): Likewise.
14603         (ix86_store_bounds): Likewise.
14604         (ix86_load_returned_bounds): Likewise.
14605         (ix86_store_returned_bounds): Likewise.
14606         (ix86_class_likely_spilled_p): Likewise.
14607         (ix86_hard_regno_mode_ok): Likewise.
14608         (x86_order_regs_for_local_alloc): Likewise.
14609         (ix86_mitigate_rop): Likewise.
14610         (ix86_bnd_prefixed_insn_p): Likewise.
14611         (ix86_mpx_bound_mode): Likewise.
14612         (ix86_make_bounds_constant): Likewise.
14613         (ix86_initialize_bounds): Likewise.
14614         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14615         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14616         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14617         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14618         (TARGET_CHKP_BOUND_MODE): Likewise.
14619         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14620         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14621         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14622         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14623         * config/i386/i386.h (TARGET_MPX): Likewise.
14624         (TARGET_MPX_P): Likewise.
14625         (VALID_BND_REG_MODE): Likewise.
14626         (FIRST_BND_REG): Likewise.
14627         (LAST_BND_REG): Likewise.
14628         (enum reg_class): Likewise.
14629         (BND_REG_P): Likewise.
14630         (BND_REGNO_P): Likewise.
14631         (BNDmode): Likewise.
14632         (ADJUST_INSN_LENGTH): Likewise.
14633         * config/i386/i386.md: Likewise.
14634         * config/i386/i386.opt: Likewise.
14635         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14636         (defined): Likewise.
14637         (LINK_MPX): Likewise.
14638         (MPX_SPEC): Likewise.
14639         (LIBMPX_SPEC): Likewise.
14640         (LIBMPXWRAPPERS_SPEC): Likewise.
14641         (CHKP_SPEC): Likewise.
14642         * config/i386/predicates.md: Likewise.
14643         * dbxout.c (dbxout_type): Likewise.
14644         * doc/extend.texi: Likewise.
14645         * doc/invoke.texi: Likewise.
14646         * doc/md.texi: Likewise.
14647         * doc/tm.texi: Likewise.
14648         * doc/tm.texi.in: Likewise.
14649         * dwarf2out.c (is_base_type): Likewise.
14650         (gen_formal_types_die): Likewise.
14651         (gen_subprogram_die): Likewise.
14652         (gen_type_die_with_usage): Likewise.
14653         (gen_decl_die): Likewise.
14654         (dwarf2out_late_global_decl): Likewise.
14655         * expr.c (expand_assignment): Likewise.
14656         (emit_storent_insn): Likewise.
14657         (store_expr_with_bounds): Likewise.
14658         (store_expr): Likewise.
14659         (expand_expr_real_1): Likewise.
14660         * expr.h (store_expr_with_bounds): Likewise.
14661         * function.c (use_register_for_decl): Likewise.
14662         (struct bounds_parm_data): Likewise.
14663         (assign_parms_augmented_arg_list): Likewise.
14664         (assign_parm_find_entry_rtl): Likewise.
14665         (assign_parm_is_stack_parm): Likewise.
14666         (assign_parm_load_bounds): Likewise.
14667         (assign_bounds): Likewise.
14668         (assign_parms): Likewise.
14669         (expand_function_start): Likewise.
14670         * gcc.c (CHKP_SPEC): Likewise.
14671         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14672         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14673         (wrestrict_dom_walker::check_call): Likewise.
14674         * gimple.c (gimple_build_call_from_tree): Likewise.
14675         * gimple.h (enum gf_mask): Likewise.
14676         (gimple_call_with_bounds_p): Likewise.
14677         (gimple_call_set_with_bounds): Likewise.
14678         * gimplify.c (gimplify_init_constructor): Likewise.
14679         * ipa-cp.c (initialize_node_lattices): Likewise.
14680         (propagate_constants_across_call): Likewise.
14681         (find_more_scalar_values_for_callers_subset): Likewise.
14682         * ipa-hsa.c (process_hsa_functions): Likewise.
14683         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14684         * ipa-icf.c (sem_function::merge): Likewise.
14685         * ipa-inline.c (early_inliner): Likewise.
14686         * ipa-pure-const.c (warn_function_noreturn): Likewise.
14687         (warn_function_cold): Likewise.
14688         (propagate_pure_const): Likewise.
14689         * ipa-ref.h (enum GTY): Likewise.
14690         * ipa-split.c (find_retbnd): Likewise.
14691         (consider_split): Likewise.
14692         (split_function): Likewise.
14693         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14694         * ipa.c (walk_polymorphic_call_targets): Likewise.
14695         (symbol_table::remove_unreachable_nodes): Likewise.
14696         (process_references): Likewise.
14697         (cgraph_build_static_cdtor_1): Likewise.
14698         * lto-cgraph.c (lto_output_node): Likewise.
14699         (output_refs): Likewise.
14700         (compute_ltrans_boundary): Likewise.
14701         (input_overwrite_node): Likewise.
14702         (input_node): Likewise.
14703         (input_cgraph_1): Likewise.
14704         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14705         * passes.c (pass_manager::execute_early_local_passes): Likewise.
14706         (class pass_chkp_instrumentation_passes): Likewise.
14707         (make_pass_chkp_instrumentation_passes): Likewise.
14708         * passes.def: Likewise.
14709         * rtl.h (struct GTY): Likewise.
14710         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14711         * stor-layout.c (layout_type): Likewise.
14712         * symtab.c: Likewise.
14713         * target.def: Likewise.
14714         * targhooks.c (default_chkp_bound_type): Likewise.
14715         (default_chkp_bound_mode): Likewise.
14716         (default_builtin_chkp_function): Likewise.
14717         (default_chkp_function_value_bounds): Likewise.
14718         (default_chkp_make_bounds_constant): Likewise.
14719         (default_chkp_initialize_bounds): Likewise.
14720         * targhooks.h (default_chkp_bound_type): Likewise.
14721         (default_chkp_bound_mode): Likewise.
14722         (default_builtin_chkp_function): Likewise.
14723         (default_chkp_function_value_bounds): Likewise.
14724         (default_chkp_make_bounds_constant): Likewise.
14725         (default_chkp_initialize_bounds): Likewise.
14726         * toplev.c (compile_file): Likewise.
14727         (process_options): Likewise.
14728         * tree-core.h (DEF_BUILTIN): Likewise.
14729         (DEF_BUILTIN_CHKP): Likewise.
14730         * tree-inline.c (declare_return_variable): Likewise.
14731         (remap_gimple_stmt): Likewise.
14732         (copy_bb): Likewise.
14733         (initialize_inlined_parameters): Likewise.
14734         (expand_call_inline): Likewise.
14735         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14736         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14737         (make_pass_ipa_chkp_produce_thunks): Likewise.
14738         (make_pass_chkp): Likewise.
14739         (make_pass_chkp_opt): Likewise.
14740         (make_pass_chkp_instrumentation_passes): Likewise.
14741         * tree-pretty-print.c (dump_generic_node): Likewise.
14742         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14743         * tree-ssa-dce.c (propagate_necessity): Likewise.
14744         (eliminate_unnecessary_stmts): Likewise.
14745         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14746         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14747         * tree-ssa-sccvn.h: Likewise.
14748         * tree-ssa-strlen.c (get_string_length): Likewise.
14749         (valid_builtin_call): Likewise.
14750         (adjust_last_stmt): Likewise.
14751         (handle_builtin_strchr): Likewise.
14752         (handle_builtin_strcpy): Likewise.
14753         (handle_builtin_stxncpy): Likewise.
14754         (handle_builtin_memcpy): Likewise.
14755         (handle_builtin_strcat): Likewise.
14756         (strlen_check_and_optimize_stmt): Likewise.
14757         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14758         * tree-streamer-in.c: Likewise.
14759         * tree-streamer.c (record_common_node): Likewise.
14760         * tree.c (tree_code_size): Likewise.
14761         (wide_int_to_tree_1): Likewise.
14762         (type_contains_placeholder_1): Likewise.
14763         (build_common_tree_nodes): Likewise.
14764         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14765         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14766         (POINTER_BOUNDS_P): Likewise.
14767         (BOUNDED_TYPE_P): Likewise.
14768         (BOUNDED_P): Likewise.
14769         (CALL_WITH_BOUNDS_P): Likewise.
14770         (pointer_bounds_type_node): Likewise.
14771         * value-prof.c (gimple_ic): Likewise.
14772         * var-tracking.c (vt_add_function_parameters): Likewise.
14773         * varasm.c (make_decl_rtl): Likewise.
14774         (assemble_start_function): Likewise.
14775         (output_constant): Likewise.
14776         (maybe_assemble_visibility): Likewise.
14777         * varpool.c (ctor_for_folding): Likewise.
14778         * chkp-builtins.def: Remove.
14779         * ipa-chkp.c: Remove.
14780         * ipa-chkp.h: Remove.
14781         * rtl-chkp.c: Remove.
14782         * rtl-chkp.h: Remove.
14783         * tree-chkp-opt.c: Remove.
14784         * tree-chkp.c: Remove.
14785         * tree-chkp.h: Remove.
14787 2018-06-07  Carl Love  <cel@us.ibm.com>
14789         * config/rs6000/vsx.md (vextract_fp_from_shorth,
14790         vextract_fp_from_shortl): Add BE support.
14792 2018-06-07  Paul Koning  <ni1d@arrl.net>
14794         * compare-elim.c (try_merge_compare): Don't merge compare if
14795         address contains a side effect.
14796         (try_eliminate_compare): Likewise.
14798 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
14800         * config.gcc: Support "tremont".
14801         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14802         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14803         PROCESSOR_TREMONT.
14804         * config/i386/i386.c (m_TREMONT): Define.
14805         (processor_target_table): Add "tremont".
14806         (PTA_TREMONT): Define.
14807         (ix86_lea_outperforms): Add TARGET_TREMONT.
14808         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14809         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14810         and M_INTEL_GOLDMONT_PLUS.
14811         (fold_builtin_cpu): Add "tremont".
14812         (ix86_add_stmt_cost): Add TARGET_TREMONT.
14813         (ix86_option_override_internal): Add "tremont".
14814         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14815         (processor_type): Add PROCESSOR_TREMONT.
14816         * config/i386/x86-tune.def: Add m_TREMONT.
14817         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14819 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14821         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14822         symbol defined for msp430i* devices to be lower case.
14824 2018-06-07  Richard Biener  <rguenther@suse.de>
14826         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14827         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
14828         Properly wrap signed arithmetic if overflow wraps.
14830 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
14832         PR tree-optimization/69615
14833         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14834         of a cast from a same precision integral SSA_NAME in a bb dominated
14835         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
14836         cast to utype if rhs2 has already a compatible type.
14838 2018-06-07  Richard Biener  <rguenther@suse.de>
14840         PR tree-optimization/85935
14841         * graphite-scop-detection.c (find_params_in_bb): Analyze
14842         condition operands with respect to the correct loop.  Assert
14843         the analysis doesn't fail.
14845 2018-06-04  Carl Love  <cel@us.ibm.com>
14847         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14848         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
14849         as it is slightly cheaper.
14850         (first_match_or_eos_index_<mode>):
14851         Calculate index using natural element order.
14852         (first_match_index_<mode>):
14853         Calculate index using natural element order.
14854         (first_match_or_eos_index_<mode>):
14855         Calculate index using natural order.
14856         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14857         for BE and LE modes.
14858         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14859         P9V_BUILTIN_VCLZLSBB_V16QI.
14860         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14861         specific.
14863 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14865         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14866         indentation and line wrap for many prototypes.  Add missing
14867         @smallexample directives around block of prototypes for vec_xl and
14868         vec_xst.
14870 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
14872         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14873         track if we pass or return IEEE 128-bit floating point.
14874         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14875         C++ mangling that is compatible with GCC 8.1.
14876         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
14877         (init_cumulative_args): Note if we pass or return IEEE 128-bit
14878         floating point types.
14879         (rs6000_function_arg_advance_1): Likewise.
14880         (rs6000_mangle_type): Optionally generate mangled names that match
14881         what GCC 8.1 generated for IEEE 128-bit floating point types.
14882         (rs6000_globalize_decl_name): If we have an external function that
14883         passes or returns IEEE 128-bit types, generate a weak reference
14884         from the mangled name used in GCC 8.1 to the current mangled
14885         name.
14886         (rs6000_init_builtins): Make __ibm128 use the long double type if
14887         long double is IBM extended double.  Make __float128 use the long
14888         double type if long double is IEEE 128-bit.
14890         PR target/85657
14891         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
14892         macro for __ibm128 built-in functions.
14893         (PACK_IF): Add __ibm128 pack/unpack functions.
14894         (UNPACK_IF): Likewise.
14895         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
14896         enable long double built-in functions if long double is IEEE
14897         128-bit floating point.
14898         (rs6000_invalid_builtin): Update long double built-in function
14899         error message.
14900         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
14901         functions, adjust the built-in function to use the long double
14902         built-in function if __ibm128 and long double are the same type.
14903         * doc/extend.texi (PowerPC builtins): Update documention for
14904         __builtin_{,un}pack_longdouble.  Add documentation for
14905         __builtin_{,un}pack_ibm128.
14907 2018-06-06  Jim Wilson  <jimw@sifive.com>
14909         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
14910         (struct machine_function): New field interrupt_mode.
14911         (riscv_handle_type_attribute): New function.  Add forward declaration.
14912         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
14913         (riscv_expand_epilogue): Check interrupt_mode field.
14914         (riscv_set_current_function): Check interrupt attribute args and
14915         set interrupt_mode field.
14916         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
14917         (riscv_sret, riscv_uret): New.
14918         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
14919         new arguments to interrupt attribute.
14921 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
14923         PR target/63177
14924         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
14925         Don't handle -mcpu=power8 if -mpower9-vector is also used.
14927 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14929         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
14930         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
14931         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14932         several redundant entries.
14934 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
14936         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
14937         type from "rtx" to "rtx_insn *".
14938         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
14939         for local "call_insn", removing cast.
14940         (ix86_expand_call): Likewise, introducing a "call_insn" local.
14942 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
14944         PR tree-optimization/86066
14945         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
14946         for BIT_INSERT_EXPR stores.
14948 2018-06-06  Richard Biener  <rguenther@suse.de>
14950         PR tree-optimization/86062
14951         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14952         component refs ontop
14953         of to be offsetted base.
14955 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14957         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
14958         to be static and remove check on interrupt attribute name.
14960 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14962         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14963         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
14965 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
14967         PR target/79924
14968         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
14969         second argument.
14970         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
14971         Remove second argument, change how error is called.
14972         (aarch64_layout_arg): Remove second argument from
14973         aarch64_err_no_fpadvsimd call.
14974         (aarch64_init_cumulative_args): Ditto.
14975         (aarch64_gimplify_va_arg_expr): Ditto.
14976         * config/aarch64/aarch64.md (mov<mode>): Ditto.
14978 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
14980         * config/i386/i386.md (simple_return_indirect_internal): New expander.
14981         (*simple_return_indirect_internal<mode>): Rename from
14982         simple_return_indirect_internal.  Use W mode iterator.
14983         (rstorssp): New expander.
14984         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
14985         (clrssbsy): New expander.
14986         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
14988 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14990         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
14991         __typeof__.
14992         (cmse_check_pointed_object): Likewise.
14994 2018-06-05  Martin Liska  <mliska@suse.cz>
14996         PR gcov-profile/47618
14997         * doc/invoke.texi: Document how -fprofile-dir format
14998         is extended.
15000 2018-06-05  Richard Biener  <rguenther@suse.de>
15002         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
15003         removal pretend DOM info isn't available so we do not update
15004         it and only remove edges, not dominated blocks.  Actually free
15005         DOM info in case we removed something.  Remove unreachable blocks.
15006         (mfb_keep_latches): Work with either DOM info or marked backedges.
15007         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
15008         first.  Mark backedges if DOM info isn't available.
15009         (Re-)compute DOM info after cleanup_control_flow_pre.
15011 2018-06-05  Richard Biener  <rguenther@suse.de>
15013         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
15014         (locus_discrim_hasher::hash): Adjust.
15015         (locus_discrim_hasher::equal): Likewise.
15016         (next_discriminator_for_locus): Work on line directly.
15017         (same_line_p): Pass in expanded locus1 as well.
15018         (assign_discriminators): Avoid redundant location expansions.
15020 2018-06-05  Richard Biener  <rguenther@suse.de>
15022         PR tree-optimization/86046
15023         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
15024         if required after clearing TREE_ADDRESSABLE.
15026 2018-06-05  Richard Biener  <rguenther@suse.de>
15028         PR tree-optimization/86047
15029         * tree-ssa-loop.c (for_each_index): Glob handling of all
15030         decls and constants and really handle all of them.
15032 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15034         PR target/81497
15035         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
15036         qualifier_void_pointer and qualifier_const_void_pointer.
15037         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
15038         (arm_init_builtins): Handle the above.
15039         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
15040         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
15041         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
15042         void intrinsics.
15044 2018-06-05  Martin Liska  <mliska@suse.cz>
15046         * auto-profile.c (read_autofdo_file): Do not use
15047         gcov_ctr_summary struct.
15048         (afdo_callsite_hot_enough_for_early_inline): Likewise.
15049         * coverage.c (struct counts_entry): Likewise.
15050         (read_counts_file): Read just single summary entry.
15051         (get_coverage_counts): Use gcov_summary struct.
15052         * coverage.h (get_coverage_counts): Likewise.
15053         * gcov-dump.c (dump_working_sets): Likewise.
15054         (tag_summary): Dump just single summary.
15055         * gcov-io.c (gcov_write_summary): Write just histogram
15056         summary.
15057         (gcov_read_summary): Read just single summary.
15058         (compute_working_sets): Use gcov_summary struct.
15059         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
15060         of GCOV_COUNTERS_SUMMABLE.
15061         (GCOV_COUNTERS_SUMMABLE): Remove.
15062         (GCOV_FIRST_VALUE_COUNTER): Replace with
15063         GCOV_COUNTER_V_INTERVAL.
15064         (struct gcov_ctr_summary): Remove.
15065         (struct gcov_summary): Directly use fields of former
15066         gcov_ctr_summary.
15067         (compute_working_sets): Use gcov_summary struct.
15068         * gcov.c (read_count_file): Do not use ctrs fields.
15069         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
15070         struct.
15071         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
15072         struct.
15073         * profile.c: Likewise.
15074         * profile.h: Likewise.
15076 2018-06-05  Martin Liska  <mliska@suse.cz>
15078         PR gcov-profile/84846
15079         * gcov.c (output_lines): Print working directory only
15080         in intermediate format.
15082 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
15084         * config/s390/s390-builtin-types.def: Add void function type.
15085         * config/s390/s390-builtins.def: Use the function type for the
15086         tbeginc builtin.
15088 2018-06-04  Jim Wilson  <jimw@sifive.com>
15090         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
15091         to int.
15092         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
15093         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
15094         handle EH_RETURN_DATA_REGNO registers properly.
15095         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
15096         (riscv_expand_epilogue): Update comment.  Change argument name and
15097         type.  Update code to use new name and type.  Pass new args to
15098         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
15099         EXCEPTION_RETURN.
15100         * config/riscv/riscv.md (NORMAL_RETURN): New.
15101         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
15102         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
15103         (eh_return): Call gen_eh_return_internal and emit barrier.
15104         (eh_return_internal): Call riscv_expand_epilogue.
15106 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
15108         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
15109         bit_insertion field and declare can_be_merged_into method.
15110         (merged_store_group::can_be_merged_into): New method.
15111         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
15112         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
15113         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
15115 2018-06-04  Richard Biener  <rguenther@suse.de>
15117         PR tree-optimization/85955
15118         * builtins.c (fold_builtin_sincos): Convert pointers to
15119         destination to appropriate type before dereferencing.
15121 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
15123         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
15125 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
15127         * expr.c (expand_expr_real_1): Force the operand into memory if
15128         its TYPE_MODE is BLKmode and if there is no integer mode for
15129         the number of bits being extracted.
15131 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15133         PR target/85832
15134         PR target/86036
15135         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15136         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
15138 2018-06-04  Richard Biener  <rguenther@suse.de>
15140         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
15141         (cleanup_tree_cfg_noloop): ... single caller.  Do
15142         start_recording_case_labels later.
15144 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
15146         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
15147         to _IMMINTRIN_H_INCLUDED.
15148         * config/i386/pconfigintrin.h: Ditto.
15149         * config/i386/waitpkgintrin.h: Ditto.
15150         * config/i386/immintrin.h: Add includes for sgxintrin.h,
15151         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
15152         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
15153         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
15154         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
15155         waitpkgintrin.h and cldemoteintrin.h.
15157 2018-06-04  Richard Biener  <rguenther@suse.de>
15159         PR tree-optimization/86038
15160         * tracer.c (find_best_successor): Check probability for
15161         being initialized, bail out if not.
15163 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
15165         PR target/86003
15166         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
15167         of bits to ignore when comparing architectures.
15169 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15171         PR tree-optimization/69615
15172         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
15173         maximum or minimum of the type, try to merge it also as if
15174         range1 is + [-, x - 1] or + [x + 1, -].
15176         PR c++/86025
15177         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
15179 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
15181         PR tree-optimization/86034
15182         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
15183         the unsigned bitfield type in a bit insertion sequence if it does not
15184         have a larger precision than the bitfield size.
15185         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
15187 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
15189         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
15191 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15193         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
15194         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
15195         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
15196         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
15198 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15200         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
15201         Disable -fdelete-null-pointer-checks for ELF toolchain.
15203 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15204             Kito Cheng  <kito.cheng@gmail.com>
15206         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
15207         (nds32le-*-*, nds32be-*-*): Integrate checking process.
15208         (nds32*-*-*): Add glibc and uclibc conditions.
15209         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
15210         (TARGET_EXCEPT_UNWIND_INFO): Define.
15211         * config/nds32/elf.h: New file.
15212         * config/nds32/linux.h: New file.
15213         * config/nds32/nds32-elf.opt: New file.
15214         * config/nds32/nds32-linux.opt: New file.
15215         * config/nds32/nds32-fp-as-gp.c
15216         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
15217         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
15218         TARGET_LINUX_ABI.
15219         (nds32_asm_file_end): Ditto.
15220         (nds32_print_operand): Ditto.
15221         (nds32_insert_attributes): Ditto.
15222         (nds32_init_libfuncs): New function.
15223         (TARGET_HAVE_TLS): Define.
15224         (TARGET_INIT_LIBFUNCS): Define.
15225         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
15226         spec content.
15227         (TARGET_ELF): Apply different mcmodel setting.
15228         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
15229         been migrated into elf.h and linux.h files.
15230         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
15231         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
15232         (mcmodel): The content has been migrated into nds32-elf.opt and
15233         nds32-linux.opt files.
15234         * config/nds32/t-elf: New file.
15235         * config/nds32/t-linux: New file.
15237 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15238             Shiva Chen  <shiva0217@gmail.com>
15240         * config/nds32/constants.md (unspec_volatile_element): Add
15241         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
15242         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
15243         optimization.
15244         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
15245         (make_pass_nds32_fp_as_gp): Declare.
15246         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
15247         optmization pass.
15248         (nds32_asm_function_end_prologue): Remove unused asm output.
15249         (nds32_asm_function_begin_epilogue): Remove unused asm output.
15250         (nds32_asm_file_start): Output necessary fp_as_gp information.
15251         (nds32_option_override): Adjust register usage.
15252         (nds32_expand_prologue): Consider fp_as_gp situation.
15253         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
15254         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
15255         (epilogue): Ditto.
15256         (return): Ditto.
15257         (simple_return): Ditto.
15258         (omit_fp_begin): Output special directive for fp_as_gp.
15259         (omit_fp_end): Output special directive for fp_as_gp.
15260         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
15261         mforbid-fp-as-gp): New options.
15263 2018-06-01  Mark Wielaard  <mark@klomp.org>
15265         * dwarf2out.c (dwarf2out_finish): Remove generation of
15266         DW_AT_loclists_base.
15268 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
15270         * gimple-ssa-store-merging.c: Include gimple-fold.h.
15271         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
15272         (struct merged_store_group): Add bit_insertion field.
15273         (dump_char_array): Use standard hexadecimal format.
15274         (merged_store_group::merged_store_group): Set bit_insertion to false.
15275         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
15276         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
15277         also print the mask in the dump file.
15278         (pass_store_merging::gate): Minor tweak.
15279         (imm_store_chain_info::coalesce_immediate): Fix wrong association
15280         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
15281         stores with INTEGER_CST stores.
15282         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
15283         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
15284         and use it throughout.  Generate bit insertion sequences if need be.
15285         (pass_store_merging::process_store): Remove redundant condition.
15286         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
15288 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
15290         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
15291         the 128-bit floating point types.  Fix function comment.
15293 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15295         * config/aarch64/aarch64-simd.md
15296         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
15297         mnemonics.
15298         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
15299         mnemonics.
15301 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15303         PR tree-optimization/85989
15304         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
15305         variable.
15306         (backprop::intersect_uses): Check it when deciding whether this
15307         is a backedge reference.
15308         (backprop::process_block): Add each phi to m_visited_phis
15309         after visiting it, then clear it at the end.
15311 2018-06-01  Richard Biener  <rguenther@suse.de>
15313         * tree-vectorizer.h (vect_dr_stmt): New function.
15314         (vect_get_load_cost): Adjust.
15315         (vect_get_store_cost): Likewise.
15316         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
15317         Use vect_dr_stmt instead of DR_SMTT.
15318         (vect_record_base_alignments): Likewise.
15319         (vect_calculate_target_alignment): Likewise.
15320         (vect_compute_data_ref_alignment): Likewise and make static.
15321         (vect_update_misalignment_for_peel): Likewise.
15322         (vect_verify_datarefs_alignment): Likewise.
15323         (vector_alignment_reachable_p): Likewise.
15324         (vect_get_data_access_cost): Likewise.  Pass down
15325         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
15326         (vect_get_peeling_costs_all_drs): Likewise.
15327         (vect_peeling_hash_get_lowest_cost): Likewise.
15328         (vect_enhance_data_refs_alignment): Likewise.
15329         (vect_find_same_alignment_drs): Likewise.
15330         (vect_analyze_data_refs_alignment): Likewise.
15331         (vect_analyze_group_access_1): Likewise.
15332         (vect_analyze_group_access): Likewise.
15333         (vect_analyze_data_ref_access): Likewise.
15334         (vect_analyze_data_ref_accesses): Likewise.
15335         (vect_vfa_segment_size): Likewise.
15336         (vect_small_gap_p): Likewise.
15337         (vectorizable_with_step_bound_p): Likewise.
15338         (vect_prune_runtime_alias_test_list): Likewise.
15339         (vect_analyze_data_refs): Likewise.
15340         (vect_supportable_dr_alignment): Likewise.
15341         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
15342         (vect_gen_prolog_loop_niters): Likewise.
15343         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
15344         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
15345         modify DR_STMT.
15346         (vect_recog_mask_conversion_pattern): Likewise.
15347         (vect_try_gather_scatter_pattern): Likewise.
15348         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
15349         to vect_get_store_cost.
15350         (vect_get_store_cost): Get stmt_info instead of DR.
15351         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
15352         (vect_get_load_cost): Get stmt_info instead of DR.
15354 2018-06-01  Richard Biener  <rguenther@suse.de>
15356         PR middle-end/86017
15357         * gimple-fold.c (var_decl_component_p): Also allow offsetted
15358         vars wrapped in MEM_REFs.
15360 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15362         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15363         Fix subreg tests so that we only return a choice between
15364         GENERAL_REGS and FP_REGS if the original classes included both.
15366 2018-06-01  Richard Biener  <rguenther@suse.de>
15368         PR ipa/85960
15369         * tree-ssa-structalias.c (get_function_part_constraint):
15370         Handle NULL fi->decl.
15371         (find_func_aliases_for_call): Properly handle indirect
15372         fi from direct call.
15373         (find_func_clobbers): Likewise.
15374         (ipa_pta_execute): Likewise.
15375         (create_variable_info_for): For functions that are ifunc_resolver
15376         resolve to a varinfo that contains the result of the resolver call.
15377         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
15378         aliases.
15380 2018-05-31  Michael Collison  <michael.collison@arm.com>
15382         * config/aarch64/aarch64.md:
15383         (*fix_to_zero_extenddfdi2): New pattern.
15384         * gcc.target/aarch64/fix_extend1.c: New testcase.
15386 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
15388         PR middle-end/78809
15389         PR middle-end/83026
15390         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
15391         and BUILT_IN_STRNCMP_EQ.
15392         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
15393         BUILT_IN_STRNCMP_EQ.
15394         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
15395         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15396         (gimple_fold_builtin): Likewise.
15397         * tree-ssa-strlen.c (compute_string_length): New function.
15398         (determine_min_obsize): New function.
15399         (handle_builtin_string_cmp): New function to handle calls to
15400         string compare functions.
15401         (strlen_optimize_stmt): Add handling to builtin string compare
15402         calls.
15403         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
15404         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15405         * tree.c (build_common_builtin_nodes): Add new defines of
15406         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
15408 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
15410         PR target/85984
15411         * bb-reorder.c (pass_partition_blocks::gate): Return false for
15412         functions with naked attribute.
15414 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
15416         * config/i386/sse.md (avx_vec_concat<mode>):
15417         Substitute concat_tg_mode mode attribute with xtg_mode.
15418         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
15419         (concat_tg_mode): Remove mode attribute.
15421 2018-05-31  Martin Sebor  <msebor@redhat.com>
15423         PR c/82063
15424         * calls.c (alloc_max_size): Correct a logic error/typo.
15425         Treat excessive arguments as infinite.  Warn for invalid arguments.
15426         * doc/invoke.texi (-Walloc-size-larger-than): Update.
15428 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
15430         PR target/85829
15431         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
15432         and movx for Haswell.
15434 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
15435             Cesar Philippidis  <cesar@codesourcery.com>
15437         PR middle-end/85879
15438         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
15439         when emitting error on private/firstprivate reductions.
15440         * omp-low.c (lower_omp_target): Avoid reference-type processing
15441         on pointers for firstprivate clause.
15443 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15445         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15446         (st1x2): Likewise.
15447         (st1x3): Likewise.
15448         * config/aarch64/aarch64-simd.md
15449         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
15450         (aarch64_ld1_x3_<mode>): Likewise
15451         (aarch64_st1x2<VALLDIF:mode>): Likewise
15452         (aarch64_st1_x2_<mode>): Likewise
15453         (aarch64_st1x3<VALLDIF:mode>): Likewise
15454         (aarch64_st1_x3_<mode>): Likewise
15455         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
15456         (vld1_s8_x3): Likewise.
15457         (vld1_u16_x3): Likewise.
15458         (vld1_s16_x3): Likewise.
15459         (vld1_u32_x3): Likewise.
15460         (vld1_s32_x3): Likewise.
15461         (vld1_u64_x3): Likewise.
15462         (vld1_s64_x3): Likewise.
15463         (vld1_f16_x3): Likewise.
15464         (vld1_f32_x3): Likewise.
15465         (vld1_f64_x3): Likewise.
15466         (vld1_p8_x3): Likewise.
15467         (vld1_p16_x3): Likewise.
15468         (vld1_p64_x3): Likewise.
15469         (vld1q_u8_x3): Likewise.
15470         (vld1q_s8_x3): Likewise.
15471         (vld1q_u16_x3): Likewise.
15472         (vld1q_s16_x3): Likewise.
15473         (vld1q_u32_x3): Likewise.
15474         (vld1q_s32_x3): Likewise.
15475         (vld1q_u64_x3): Likewise.
15476         (vld1q_s64_x3): Likewise.
15477         (vld1q_f16_x3): Likewise.
15478         (vld1q_f32_x3): Likewise.
15479         (vld1q_f64_x3): Likewise.
15480         (vld1q_p8_x3): Likewise.
15481         (vld1q_p16_x3): Likewise.
15482         (vld1q_p64_x3): Likewise.
15483         (vst1_s64_x2): Likewise.
15484         (vst1_u64_x2): Likewise.
15485         (vst1_f64_x2): Likewise.
15486         (vst1_s8_x2): Likewise.
15487         (vst1_p8_x2): Likewise.
15488         (vst1_s16_x2): Likewise.
15489         (vst1_p16_x2): Likewise.
15490         (vst1_s32_x2): Likewise.
15491         (vst1_u8_x2): Likewise.
15492         (vst1_u16_x2): Likewise.
15493         (vst1_u32_x2): Likewise.
15494         (vst1_f16_x2): Likewise.
15495         (vst1_f32_x2): Likewise.
15496         (vst1_p64_x2): Likewise.
15497         (vst1q_s8_x2): Likewise.
15498         (vst1q_p8_x2): Likewise.
15499         (vst1q_s16_x2): Likewise.
15500         (vst1q_p16_x2): Likewise.
15501         (vst1q_s32_x2): Likewise.
15502         (vst1q_s64_x2): Likewise.
15503         (vst1q_u8_x2): Likewise.
15504         (vst1q_u16_x2): Likewise.
15505         (vst1q_u32_x2): Likewise.
15506         (vst1q_u64_x2): Likewise.
15507         (vst1q_f16_x2): Likewise.
15508         (vst1q_f32_x2): Likewise.
15509         (vst1q_f64_x2): Likewise.
15510         (vst1q_p64_x2): Likewise.
15511         (vst1_s64_x3): Likewise.
15512         (vst1_u64_x3): Likewise.
15513         (vst1_f64_x3): Likewise.
15514         (vst1_s8_x3): Likewise.
15515         (vst1_p8_x3): Likewise.
15516         (vst1_s16_x3): Likewise.
15517         (vst1_p16_x3): Likewise.
15518         (vst1_s32_x3): Likewise.
15519         (vst1_u8_x3): Likewise.
15520         (vst1_u16_x3): Likewise.
15521         (vst1_u32_x3): Likewise.
15522         (vst1_f16_x3): Likewise.
15523         (vst1_f32_x3): Likewise.
15524         (vst1_p64_x3): Likewise.
15525         (vst1q_s8_x3): Likewise.
15526         (vst1q_p8_x3): Likewise.
15527         (vst1q_s16_x3): Likewise.
15528         (vst1q_p16_x3): Likewise.
15529         (vst1q_s32_x3): Likewise.
15530         (vst1q_s64_x3): Likewise.
15531         (vst1q_u8_x3): Likewise.
15532         (vst1q_u16_x3): Likewise.
15533         (vst1q_u32_x3): Likewise.
15534         (vst1q_u64_x3): Likewise.
15535         (vst1q_f16_x3): Likewise.
15536         (vst1q_f32_x3): Likewise.
15537         (vst1q_f64_x3): Likewise.
15538         (vst1q_p64_x3): Likewise.
15540 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15542         * config/msp430/msp430.c (msp430_output_labelref): Prepend
15543         user_label_prefix to name.
15545         * tree-core.h: Update comment about the format of NAME string
15546         passed to handler in attribute_spec.
15548         * config/msp430/msp430.md: Remove erroneous subreg expression from
15549         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15550         zero_extend{q,h}isi2.
15552 2018-05-30  Borislav Petkov  <bp@suse.de>
15554         * doc/extend.texi: Document some architecture specific
15555         constraints and sort entries.
15557 2018-05-30  Martin Sebor  <msebor@redhat.com>
15559         PR middle-end/85369
15560         * builtins.c (expand_builtin_stpcpy_1): New function.
15561         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15562         only if the former succeeds.
15564 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15566         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15567         in saphira.
15569 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15571         * doc/invoke.texi (-flinker-output): Document
15573 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15575         * passes.c (ipa_write_summaries): Only modify statements if body
15576         is in memory.
15577         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15578         incrementally linking.
15579         (ipa_passes): Likewise.
15580         * lto-cgraph.c (lto_output_node): When incrementally linking do not
15581         pass down resolution info.
15582         * common.opt (flag_incremental_link): Update info.
15583         * gcc.c (plugin specs): Turn flinker-output=* to
15584         -plugin-opt=-linker-output-known
15585         * toplev.c (compile_file): Also cut compilation when doing incremental
15586         link.
15587         * flag-types. (enum lto_partition_model): Add
15588         LTO_LINKER_OUTPUT_NOLTOREL.
15589         (invoke.texi): Add -flinker-output docs.
15590         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15591         link same way as WPA; do not stream in dead initializers.
15593         * dwarf2out.c (dwarf2out_die_ref_for_decl,
15594         darf2out_register_external_decl): Support incremental link.
15596 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15598         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15600 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15602         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15603         it down to simple_object_copy_lto_debug_sections.
15604         (run_gcc): Determine incremental LTO link time and configure
15605         lto1 into non-wpa mode, disable renaming of debug sections.
15607 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15609         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15610         descriptions of various incorrectly documented functions.
15612 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15614         Revert:
15615         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15616         address check not strict.
15618 2018-05-30  Richard Biener  <rguenther@suse.de>
15620         PR tree-optimization/85964
15621         * tracer.c (better_p): Drop initialized count check, we only
15622         call the function with initialized counts now.
15623         (find_best_successor): Do find a best edge if one
15624         has uninitialized count.
15625         (find_best_predecessor): Likewise.  Do BB frequency check only
15626         if count is initialized.
15628 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
15630         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15631         (aarch64_ldrstr_offset_compare): New.
15632         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15633         load/store orderings.
15634         (aarch64_gen_adjusted_ldpstp): Likewise.
15636 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
15638         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15639         Check for subset of GENERAL_REGS and FP_REGS.
15640         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15641         r=w alternative.
15643 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
15645         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15646         and wi::to_poly_offset.  Add the current offset and then check
15647         whether the sum fits, rather than using an unchecked addition of
15648         a checked term.  Check for a shwi rather than a uhwi.
15649         * expr.c (get_bit_range): Use tree_to_poly_uint64.
15650         (store_constructor): Use poly_int_tree_p.
15651         (expand_expr_real_1): Likewise.
15652         * function.c (assign_temp): Likewise.
15653         * fold-const.c (const_binop): Use poly_int_tree_p and
15654         wi::to_poly_offset.
15655         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
15656         division.
15657         * ipa-icf-gimple.c (func_checker::compare_operand): Use
15658         to_poly_offset for MEM offsets.
15659         * ipa-icf.c (sem_variable::equals): Likewise.
15660         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15661         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15662         wi::to_poly_offset for BIT_FIELD_REF offsets.
15663         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15664         wi::to_poly_offset.
15665         * var-tracking.c (emit_note_insn_var_location): Use
15666         tree_to_poly_uint64.
15668 2018-05-29  Jim Wilson  <jimw@sifive.com>
15670         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15672 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
15674         PR target/85950
15675         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15676         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15677         sequence.
15678         (sse4_1_round<mode>2): Use nonimmediate_operand
15679         for operand 1 predicate.
15681 2018-05-29  Martin Sebor  <msebor@redhat.com>
15682             Richard Biener  <rguenther@suse.de>
15684         PR testsuite/85888
15685         * calls.c (get_size_range): Call determine_value_range instead
15686         of get_value_range..
15687         * tree-vrp.h (determine_value_range): Declared new function.
15688         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15690 2018-05-29  Richard Biener  <rguenther@suse.de>
15692         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15693         sure to use non-pattern stmts for get_earlier_stmt arguments.
15694         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15695         called on pattern stmts.
15696         (get_later_stmt): Likewise.
15698 2018-05-29  Martin Liska  <mliska@suse.cz>
15700         PR gcov-profile/85759
15701         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15702         env variables.
15704 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
15706         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15707         VEC_UNPACK_*_EXPR.
15708         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15709         VEC_PACK_*_EXPR.
15711         PR target/85918
15712         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15713         VEC_PACK_FLOAT_EXPR): New tree codes.
15714         * tree-pretty-print.c (op_code_prio): Handle
15715         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15716         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15717         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15718         * tree-inline.c (estimate_operator_cost): Likewise.
15719         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15720         * fold-const.c (const_binop): Likewise.
15721         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15722         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15723         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15724         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15725         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15726         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15727         * expr.c (expand_expr_real_2): Likewise.
15728         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15729         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15730         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15731         optabs.
15732         * optabs.c (expand_widen_pattern_expr): For
15733         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15734         sign from result type rather than operand's type.
15735         (expand_binop_directly): For vec_packu_float_optab and
15736         vec_packs_float_optab allow result type to be different from operand's
15737         type.
15738         * optabs-tree.c (optab_for_tree_code): Handle
15739         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15740         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
15741         * tree-vect-generic.c (expand_vector_operations_1):  Handle
15742         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15743         VEC_PACK_FLOAT_EXPR.
15744         * tree-vect-stmts.c (supportable_widening_operation): Handle
15745         FIX_TRUNC_EXPR.
15746         (supportable_narrowing_operation): Handle FLOAT_EXPR.
15747         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15748         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15749         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
15750         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15751         mode attributes.
15752         (vec_pack<floatprefix>_float_<mode>): New expander.
15753         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15754         attributes.
15755         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15756         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15757         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15758         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15759         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15760         Document.
15761         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15762         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15763         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15764         VEC_PACK_FLOAT_EXPR): Document.
15766 2018-05-29  Richard Biener  <rguenther@suse.de>
15768         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15769         member.
15770         (stmt_vec_info_vec): Make pointer.
15771         (init_stmt_vec_info_vec): Remove.
15772         (free_stmt_vec_info_vec): Likewise.
15773         (set_stmt_vec_info_vec): New function.
15774         (free_stmt_vec_infos): Likewise.
15775         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15776         (set_vinfo_for_stmt): Likewise.
15777         (get_earlier_stmt): Likewise.
15778         (get_later_stmt): Likewise.
15779         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15780         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15781         (vec_info::~vec_info): Free stmt_vec_infos.
15782         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15783         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15784         (pass_slp_vectorize::execute): Likewise.
15785         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15786         (free_stmt_vec_info_vec): Likewise.
15787         (set_stmt_vec_info_vec): New function.
15788         (free_stmt_vec_infos): Likewise.
15789         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15790         the global stmt_vec_info_vec.
15791         * tree-parloops.c (gather_scalar_reductions): Use
15792         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15793         vector.
15795 2018-05-29  Richard Biener  <rguenther@suse.de>
15797         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15799 2018-05-29  Martin Liska  <mliska@suse.cz>
15800             David Malcolm  <dmalcolm@redhat.com>
15802         * vec.c (test_reverse): New.
15803         (vec_c_tests): Add new test.
15804         * vec.h (vl_ptr>::reverse): New function.
15806 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
15808         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15810         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15811         and later.
15813 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15815         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15817 2018-05-28  Richard Biener  <rguenther@suse.de>
15819         PR tree-optimization/85933
15820         * tree-vect-data-refs.c (vect_record_base_alignments): Only
15821         look at stmts marked as vectorizable.
15823 2018-05-28  Richard Biener  <rguenther@suse.de>
15825         PR tree-optimization/85934
15826         * tree-vect-generic.c (expand_vector_operations_1): Hoist
15827         vector boolean check before scalar optimization.
15829 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
15831         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15832         for armv5te.
15834 2018-05-28  Mark Wielaard  <mark@klomp.org>
15836         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15837         if it is an expression containing a minus sign.
15839 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
15841         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15843 2018-05-27  Paul Koning  <ni1d@arrl.net>
15845         * config/pdp11/pdp11.md (truncsihi2): Remove.
15847 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
15848             Chung-Ju Wu  <jasonwucj@gmail.com>
15850         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15851         implementation.
15852         (unaligned_store_dw): Ditto.
15853         * config/nds32/nds32-memory-manipulation.c
15854         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15855         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15856         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15857         (emit_setmem_word_loop): Rename to ...
15858         (emit_setmem_doubleword_loop): ... this.
15859         (nds32_gen_dup_4_byte_to_word_value): New function.
15860         (nds32_gen_dup_8_byte_to_double_word_value): New function.
15861         (nds32_expand_setmem_loop): Refine implementation.
15862         (nds32_expand_setmem_loop_v3m): Ditto.
15863         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15864         pattern.
15866 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15868         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15870 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15872         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15873         (nds32_init_machine_status): Initialize machine->attr_naked_p and
15874         machine->attr_no_prologue_p.
15875         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
15876         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
15877         (nds32_expand_epilogue): Consider attr_naked_p.
15878         (nds32_expand_epilogue_v3pop): Likewise.
15879         (nds32_can_use_return_insn): Likewise.
15880         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
15881         attr_no_prologue_p fields.
15882         * config/nds32/nds32.opt (mret-in-naked-func): New option.
15884 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
15886         PR target/85918
15887         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
15888         attributes.
15889         * config/i386/sse.md
15890         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
15891         Rename to ...
15892         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
15893         ... this.
15894         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
15895         Rename to ...
15896         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
15897         ... this.
15898         (*<floatsuffix>floatv2div2sf2): Rename to ...
15899         (*float<floatunssuffix>v2div2sf2): ... this.
15900         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
15901         (float<floatunssuffix>v2div2sf2_mask): ... this.
15902         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
15903         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
15904         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
15905         to ...
15906         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
15907         ... this.
15908         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15909         Rename to ...
15910         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15911         ... this.
15912         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15913         Rename to ...
15914         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15915         ... this.
15916         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
15917         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
15918         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
15919         gen_ufix_truncv8dfv8si2.
15920         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
15921         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
15922         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
15923         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
15924         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
15925         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
15926         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
15927         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
15929 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15931         PR target/85900
15932         PR target/85345
15933         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
15935 2018-05-25  Jim Wilson  <jimw@sifive.com>
15937         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
15938         * config/riscv/riscv.c (struct machine_function): Add
15939         interrupt_handler_p and attribute_checked_p fields.
15940         (riscv_attribute_table): Add interrupt.
15941         (riscv_interrupt_type_p): New.
15942         (riscv_save_reg_p): Save extra regs for interrupt handler.
15943         (riscv_use_save_libcall): Return false  for interrupt handler.
15944         (riscv_first_stack_step): Add forward declaration.
15945         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
15946         for interrupt handler with large frame.  Use it for saved reg list.
15947         (riscv_expand_prologue): Move flag_stack_usage_info support to
15948         eliminate duplication.
15949         (riscv_expand_epilogue): Generate mret for interrupt handler.
15950         (riscv_epilogue_uses): New.
15951         (riscv_can_use_return_insn): Return false for interrupt handler.
15952         (riscv_function_ok_for_sibcall): Likewise.
15953         (riscv_set_current_function): Add interrupt handler support.
15954         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
15955         * config/riscv/riscv.md (UNSPECV_MRET): New.
15956         (GP_REGNUM): New.
15957         (riscv_frflags, riscv_fsflags): Use tab after opcode.
15958         (riscv_mret): New.
15959         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
15961 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
15963         PR tree-optimization/85712
15964         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
15965         this candidate has already been replaced in-situ by a copy.
15967 2018-05-25  Jason Merrill  <jason@redhat.com>
15969         PR c++/80485 - inline function non-zero address.
15970         * symtab.c (nonzero_address): Check DECL_COMDAT.
15972 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
15974         PR target/83628
15975         * config/alpha/alpha.md (ashlsi3): New insn pattern.
15976         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
15977         extension of SImode operation.  Use const123_operand predicate.
15978         (*saddsi_1): Remove.
15979         (*saddl_se_1): Ditto.
15980         (*ssubsi_1): Ditto.
15981         (*ssubl_se_1): Ditto.
15982         * config/alpha/predicates.md (const123_operand): New predicate.
15983         * config/alpha/constraints.md (P): Use IN_RANGE.
15985 2018-05-25  Richard Biener  <rguenther@suse.de>
15987         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
15988         defaulted to true.
15989         (ref_maybe_used_by_stmt_p): Likewise.
15990         (stmt_may_clobber_ref_p): Likewise.
15991         (stmt_may_clobber_ref_p_1): Likewise.
15992         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
15993         and pass it along.
15994         (ref_maybe_used_by_stmt_p): Likewise.
15995         (stmt_may_clobber_ref_p): Likewise.
15996         (stmt_may_clobber_ref_p_1): Likewise.
15997         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
15998         the alias oracle to disambiguate DRs with stmts DR analysis
15999         couldn't handle.
16000         (vect_analyze_data_refs): Do not give up on not analyzable
16001         DRs for BB vectorization.  Remove code truncating the dataref
16002         vector.
16004 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
16006         PR target/85832
16007         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16008         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
16009         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
16011 2018-05-25  Richard Biener  <rguenther@suse.de>
16013         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
16014         function, combining stmt data ref gathering and fatal analysis
16015         parts.
16016         (vect_analyze_data_refs): Remove now redudnant code and simplify.
16017         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
16018         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
16019         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
16020         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
16022 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16024         PR tree-optimization/85720
16025         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
16026         SCC if all partitions are builtins.
16027         (version_loop_by_alias_check): New parameter.  Generate cancelable
16028         runtime alias check if all partitions are builtins.
16029         (distribute_loop): Update call to above function.
16031 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16033         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
16034         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
16035         (parm_default_def_partition_arg): Ditto.
16036         (set_parm_default_def_partition): Ditto.
16037         (get_parm_default_def_partitions): Ditto and make it static.
16038         (get_undefined_value_partitions): Ditto and make it static.
16039         (remove_ssa_form): Refactor call to init_var_map here.
16040         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
16041         computation for loop region.
16042         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
16043         (register_default_def): Delete.
16044         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
16045         (parm_default_def_partition_arg): Ditto.
16046         (set_parm_default_def_partition): Ditto.
16047         (get_parm_default_def_partitions): Ditto and make it static.
16048         (get_undefined_value_partitions): Ditto and make it static.
16049         (coalesce_with_default, coalesce_with_default): Update comment.
16050         (create_coalesce_list_for_region): New func factored out from
16051         create_outofssa_var_map.
16052         (populate_coalesce_list_for_outofssa): New func factored out from
16053         create_outofssa_var_map and coalesce_ssa_name.
16054         (create_outofssa_var_map): Delete.
16055         (coalesce_ssa_name): Refactor to support live range computation.
16056         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
16057         (get_parm_default_def_partitions): Delete.
16058         (get_undefined_value_partitions): Ditto.
16059         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
16060         computation for loop region.
16061         (new_tree_live_info, loe_visit_block): Ditto.
16062         (live_worklist, set_var_live_on_entry): Ditto.
16063         (calculate_live_on_exit, verify_live_on_entry): Ditto.
16064         * tree-ssa-live.h (struct _var_map): New fields.
16065         (init_var_map): Change decl.
16066         (region_contains_p): New.
16068 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16070         * tree-ssa-live.h (live_merge_and_clear): Delete.
16072 2018-05-25  Richard Biener  <rguenther@suse.de>
16074         PR c++/85912
16075         * tree-dump.c (dequeue_and_dump): Remove access to removed
16076         operand 2 of a SWITCH_EXPR.
16078 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16080         * doc/sourcebuild.texi (vect_double_cond_arith): Include
16081         multiplication and division.
16082         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
16083         (cond_udiv@var{m}, cond_umod@var{m}): Document.
16084         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
16085         (cond_udiv_optab, cond_umod_optab): New optabs.
16086         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
16087         (IFN_COND_RDIV): New internal functions.
16088         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
16089         TRUNC_MOD_EXPR and RDIV_EXPR.
16090         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
16091         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
16092         New unspecs.
16093         (SVE_INT_BINARY): Include mult.
16094         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
16095         (optab, sve_int_op): Handle mult.
16096         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
16097         UNSPEC_COND_DIV.
16098         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
16099         for SVE_INT_BINARY_SD.
16101 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16103         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
16104         (optab, sve_int_op): Handle div and udiv.
16105         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
16106         for SVE_INT_BINARY_SD.
16107         (*<optab><mode>3): New insn for the same.
16109 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16111         * tree-vect-patterns.c: Include predict.h.
16112         (vect_recog_divmod_pattern): Restrict check for division support
16113         to when optimizing for size.
16115 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16117         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
16118         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
16119         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
16120         (gimple_match_op::set_op): Likewise.
16121         (gimple_resimplify4): Declare.
16122         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
16123         (expr::gen_transform): Likewise.
16124         (decision_tree::gen): Generate a simplification routine for 4 operands.
16125         * gimple-match-head.c (gimple_simplify): Add an overload for
16126         4 operands.  In the top-level function, handle up to 4 call
16127         arguments and call gimple_resimplify4.
16128         (gimple_resimplify4): New function.
16129         (build_call_internal): Pass a fourth operand.
16130         (maybe_push_to_seq): Likewise.
16131         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
16132         Fold VEC_COND_EXPRs of an operation and a default value into
16133         an IFN_COND_* function if possible.
16134         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
16135         New unspecs.
16136         (SVE_COND_FP_BINARY): Include them.
16137         (optab, sve_fp_op): Handle them.
16138         (SVE_INT_BINARY_REV): New code iterator.
16139         (SVE_COND_FP_BINARY_REV): New int iterator.
16140         (commutative): New int attribute.
16141         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
16142         Declare.
16143         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
16144         function.
16145         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
16146         (*cond_<optab><mode>): New patterns for reversed operands.
16148 2018-05-25  Richard Biener  <rguenther@suse.de>
16150         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
16151         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
16152         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
16153         (STMT_VINFO_GROUPED_ACCESS): Adjust.
16154         * tree-vect-data-refs.c (everywhere): Adjust users.
16155         * tree-vect-loop.c (everywhere): Likewise.
16156         * tree-vect-slp.c (everywhere): Likewise.
16157         * tree-vect-stmts.c (everywhere): Likewise.
16158         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
16160 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16162         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
16163         Rename to...
16164         (gcc_cv_as_section_exclude): ... this.
16165         Try Solaris as #exclude syntax.
16166         * configure: Regenerate.
16167         * config.in: Regenerate.
16168         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
16169         SECTION_EXCLUDE.
16170         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
16171         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
16173         * varasm.c (default_elf_asm_named_section): Don't check if
16174         HAVE_GAS_SECTION_EXCLUDE is defined.
16176 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16178         * doc/md.texi: Update the documentation of the cond_* optabs
16179         to mention the new final operand.  Fix GET_MODE_NUNITS call.
16180         Describe the scalar case too.
16181         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
16182         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
16183         instead of 2.
16184         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
16185         (get_conditional_internal_fn): Update comment.
16186         * tree-vect-loop.c (vectorizable_reduction): Pass the original
16187         accumulator value as a final argument to conditional functions.
16188         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
16189         a define_expand and add an "else" operand.  Assert for now that
16190         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
16191         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
16192         (*cond_<optab><mode>): New patterns.
16193         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
16194         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16195         (UNSPEC_COND_EOR): Delete.
16196         (optab): Remove associated mappings.
16197         (SVE_INT_BINARY): New code iterator.
16198         (sve_int_op): Remove int attribute and add "minus" to the code
16199         attribute.
16200         (SVE_COND_INT_OP): Delete.
16201         (SVE_COND_FP_OP): Rename to...
16202         (SVE_COND_FP_BINARY): ...this.
16204 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16206         * optabs.c (can_reuse_operands_p): New function.
16207         (maybe_legitimize_operands): Try to reuse the results for
16208         earlier operands.
16210 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16212         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
16213         Add {q} suffix to insn mnemonic.
16215 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16217         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
16218         (msp430_warn_func_return): New.
16220 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
16222         * fold-const.c (tree_nonzero_bits): New function.
16223         * fold-const.h (tree_nonzero_bits): Likewise.
16224         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
16225         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
16227 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
16229         PR target/85900
16230         PR target/85345
16231         * varasm.c (assemble_alias): Check ifunc_resolver only on
16232         FUNCTION_DECL.
16234 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16236         PR target/85903
16237         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
16238         when memory input operand is handled.
16240 2018-05-24  Luis Machado  <luis.machado@linaro.org>
16242         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
16243         global.
16244         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
16246 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16248         * match.pd: Delay FMA folds until after vectorization.
16250 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16252         PR target/83009
16253         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16254         address check not strict.
16256 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16258         * gimple-match.h (gimple_match_op): New class.
16259         (mprts_hook): Replace parameters with a gimple_match_op *.
16260         (maybe_build_generic_op): Likewise.
16261         (gimple_simplified_result_is_gimple_val): Replace parameters with
16262         a const gimple_match_op *.
16263         (gimple_simplify): Replace code_helper * and tree * parameters with
16264         a gimple_match_op * parameter.
16265         (gimple_resimplify1): Replace code_helper *, tree and tree *
16266         parameters with a gimple_match_op * parameter.
16267         (gimple_resimplify2): Likewise.
16268         (gimple_resimplify3): Likewise.
16269         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
16270         parameters with a gimple_match_op * parameter.
16271         * gimple-match-head.c (gimple_simplify): Change prototypes of
16272         auto-generated functions to take a gimple_match_op * instead of
16273         separate code_helper * and tree * parameters.  Make the same
16274         change in the top-level overload and update calls to the
16275         gimple_resimplify routines.  Update calls to the auto-generated
16276         functions and to maybe_push_res_to_seq in the publicly-facing
16277         operation-specific gimple_simplify overloads.
16278         (gimple_match_op::MAX_NUM_OPS): Define.
16279         (gimple_resimplify1): Replace rcode and ops with a single res_op
16280         parameter.  Update call to gimple_simplify.
16281         (gimple_resimplify2): Likewise.
16282         (gimple_resimplify3): Likewise.
16283         (mprts_hook): Replace parameters with a gimple_match_op *.
16284         (maybe_build_generic_op): Likewise.
16285         (build_call_internal): Replace type, nargs and ops with
16286         a gimple_match_op *.
16287         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
16288         with a single gimple_match_op *.  Update calls to mprts_hook,
16289         build_call_internal and gimple_simplified_result_is_gimple_val.
16290         Factor out code that is common to the tree_code and combined_fn cases.
16291         * genmatch.c (expr::gen_transform): Replace tem_code and
16292         tem_ops with a gimple_match_op called tem_op.  Update calls
16293         to the gimple_resimplify functions and maybe_push_res_to_seq.
16294         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
16295         res_ops.  Update call to the gimple_resimplify functions.
16296         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
16297         (decision_tree::gen): Make the functions take a gimple_match_op *
16298         called res_op instead of separate res_code and res_ops parameters.
16299         Update call accordingly.
16300         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
16301         and ops with a single res_op parameter.  Update calls to
16302         maybe_build_generic_op and maybe_push_res_to_seq.
16303         (fold_stmt_1): Update calls to gimple_simplify and
16304         replace_stmt_with_simplification.
16305         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
16306         and gimple_simplified_result_is_gimple_val.
16307         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
16308         gimple_simplify.
16309         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
16310         with a gimple_match_op *.
16311         (vn_nary_build_or_lookup): Likewise.  Update call to
16312         vn_nary_build_or_lookup_1.
16313         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
16314         gimple_match_op *.  Update calls to the gimple_resimplify routines
16315         and to gimple_simplified_result_is_gimple_val.
16316         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
16317         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
16318         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
16319         (visit_nary_op): Likewise.
16320         (visit_reference_op_load): Likewise.
16322 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16324         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
16325         modifier for printing the step amount.
16327 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
16329         PR target/78849
16330         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
16331         types.
16333 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
16335         * doc/sourcebuild.texi (Endianness): New subsubsection.
16337 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16339         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16340         <prefetch_dynamic_strides>: New const bool field.
16341         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16342         prefetch_dynamic_strides.
16343         (exynosm1_prefetch_tune): Likewise.
16344         (thunderxt88_prefetch_tune): Likewise.
16345         (thunderx_prefetch_tune): Likewise.
16346         (thunderx2t99_prefetch_tune): Likewise.
16347         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
16348         false.
16349         (aarch64_override_options_internal): Update to set
16350         PARAM_PREFETCH_DYNAMIC_STRIDES.
16351         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16352         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16353         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16354         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16355         prefetch-dynamic-strides setting.
16357 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16359         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16360         <minimum_stride>: New const int field.
16361         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16362         minimum_stride field defaulting to -1.
16363         (exynosm1_prefetch_tune): Likewise.
16364         (thunderxt88_prefetch_tune): Likewise.
16365         (thunderx_prefetch_tune): Likewise.
16366         (thunderx2t99_prefetch_tune): Likewise.
16367         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
16368         <default_opt_level>: Set to 3.
16369         (aarch64_override_options_internal): Update to set
16370         PARAM_PREFETCH_MINIMUM_STRIDE.
16371         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16372         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16373         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16374         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16375         stride is constant and is below the minimum stride threshold.
16377 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16379         * config/arm/arm-cpus.in (mode26): Delete.
16380         (armv4): Delete mode26 reference.
16381         * config/arm/arm.c (arm_configure_build_target): Delete use of
16382         isa_bit_mode26.
16384 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16386         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
16387         New insn pattern.
16388         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
16389         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
16390         for non-SSE modes.
16391         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
16392         (floatunsdidf2): Ditto.
16394 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16396         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
16397         (fixuns_trunc<mode>si2_avx512f): Ditto.
16398         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
16399         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
16400         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
16402 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16404         PR rtl-optimization/79985
16405         * df-scan.c (df_insn_refs_collect): Remove special case for
16406         global registers and asm statements.
16408 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16410         * extend.texi (Global Register Variables): Rewrite the bullet list.
16411         Note that the register is available for allocation. Note that access
16412         via inline asm must use constraints. Add note about async-signal
16413         handlers. Remove paragraph about automagic register selection.
16415 2018-05-23  Richard Biener  <rguenther@suse.de>
16417         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
16418         of fixed offset from memset VN.
16420 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
16422         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
16423         first_interp field.
16424         (alloc_cand_and_find_basis): Initialize first_interp field.
16425         (slsr_process_mul): Modify first_interp field.
16426         (slsr_process_add): Likewise.
16427         (slsr_process_cast): Modify first_interp field for each new
16428         interpretation.
16429         (slsr_process_copy): Likewise.
16430         (dump_candidate): Dump first_interp field.
16431         (replace_mult_candidate): Process all interpretations, not just
16432         subsequent ones.
16433         (replace_rhs_if_not_dup): Likewise.
16434         (replace_one_candidate): Likewise.
16436 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
16438         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
16439         Add new boolean.
16440         (aarch64_needs_frame_chain): New function.
16441         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
16443 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
16445         PR target/84882
16446         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
16447         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
16448         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
16449         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
16450         as true for strict-align.
16451         (aarch64_can_inline_p): Perform checks even when callee has no
16452         attributes to check for strict alignment.
16453         * doc/extend.texi (AArch64 Function Attributes): Document
16454         no-strict-align.
16455         * doc/invoke.texi: (AArch64 Options): Likewise.
16457 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
16459         PR tree-optimization/85853
16460         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
16461         the handling of the root of the node to...
16462         (vect_slp_analyze_node_operations_1): ...this new function,
16463         and run the whole thing with the child nodes' def types
16464         set according to their SLP node's def type.
16466 2018-05-23  Richard Biener  <rguenther@suse.de>
16468         PR middle-end/85874
16469         * tree-data-ref.c (create_runtime_alias_checks): Defer
16470         and ignore overflow warnings.
16472 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
16474         PR tree-optimization/85822
16475         * tree-vrp.c (is_masked_range_test): Fix handling of negative
16476         constants.
16478 2018-05-23  Richard Biener  <rguenther@suse.de>
16480         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16481         memset constants via native_interpret_expr.
16483 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
16485         PR target/85345
16486         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
16487         attribute.
16488         (cgraph_node::create_alias): Likewise.
16489         (cgraph_node::get_availability): Check ifunc_resolver instead
16490         of looking up ifunc attribute.
16491         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
16492         * varasm.c (do_assemble_alias): Likewise.
16493         (assemble_alias): Likewise.
16494         (default_binds_local_p_3): Likewise.
16495         * cgraph.h (cgraph_node): Add ifunc_resolver.
16496         (cgraph_node::only_called_directly_or_aliased_p): Return false
16497         for IFUNC resolver.
16498         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16499         attribute.
16500         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16501         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16502         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16503         instead of looking up ifunc attribute.
16505 2018-05-22  Luis Machado  <luis.machado@linaro.org>
16507         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16509 2018-05-22  Martin Sebor  <msebor@redhat.com>
16511         PR middle-end/85359
16512         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16513         only when expasion succeeds.
16514         (expand_builtin_strcmp): Same.
16515         (expand_builtin_strncmp): Same.
16517 2018-05-22  Martin Sebor  <msebor@redhat.com>
16519         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16521 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16522             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16524         * config/aarch64/aarch64-ldpstp.md: Replace uses of
16525         aarch64_mem_pair_operand with memory_operand and delete operand swapping
16526         code.
16527         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16528         Add check for legitimate_address.
16529         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16530         (aarch64_swap_ldrstr_operands): New.
16531         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16532         Define prototype.
16534 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16535             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16537         * config/aarch64/aarch64.md: New patterns to generate stp
16538         and ldp.
16539         (store_pair_sw, store_pair_dw): New patterns to generate stp for
16540         single words and double words.
16541         (load_pair_sw, load_pair_dw): Likewise.
16542         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16543         Delete.
16544         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16545         Delete.
16546         * config/aarch64/aarch64-ldpstp.md: Modify peephole
16547         for different mode ldpstp and add peephole for merged zero stores.
16548         Likewise for loads.
16549         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16550         Add size check.
16551         (aarch64_gen_store_pair): Rename calls to match new patterns.
16552         (aarch64_gen_load_pair): Rename calls to match new patterns.
16553         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16554         (load_pair<DREG:mode><DREG2:mode>): ... This.
16555         (store_pair<mode>): Rename to...
16556         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16557         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16558         New mode iterators.
16559         (V_INT_EQUIV): Handle SImode.
16560         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16561         New predicate.
16563 2018-05-22  Martin Sebor  <msebor@redhat.com>
16565         PR c/85623
16566         * calls.c (maybe_warn_nonstring_arg): Use string length to set
16567         or ajust the presumed bound on an operation to avoid unnecessary
16568         warnings.
16570 2018-05-22  Martin Sebor  <msebor@redhat.com>
16572         PR tree-optimization/85826
16573         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16574         assuming that a DECL necesarily has a constant size.
16576 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
16578         PR middle-end/85862
16579         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16581 2018-05-22  Richard Biener  <rguenther@suse.de>
16583         PR tree-optimization/85834
16584         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16585         non-constant and non-zero memset arguments.
16587 2018-05-22  Martin Liska  <mliska@suse.cz>
16589         PR ipa/85607
16590         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16592 2018-05-22  Richard Biener  <rguenther@suse.de>
16594         PR tree-optimization/85863
16595         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16596         comparisons when vectype is specified.
16597         (vectorizable_condition): Do not specify vectype for
16598         vect_is_simple_cond when SLP vectorizing.
16600 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
16602         PR target/85657
16603         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16604         define __ibm128 as long double.
16605         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16606         as a distinct type when IEEE 128-bit support is enabled.
16607         (init_float128_ieee): Fix up conversions between IFmode and IEEE
16608         128-bit types to use the correct functions.
16609         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16610         convert between 128-bit floating point types that have different
16611         modes but the same representation, instead of using gen_lowpart to
16612         makean alias.
16613         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16614         KFmode.
16615         (IFKF_reg): New attributes to give the register constraints for
16616         IFmode and KFmode.
16617         (extend<mode>tf2_internal): New insns to mark an explicit
16618         conversion between 128-bit floating point types that have a
16619         different mode but share the same representation.
16621 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
16623         PR tree-optimization/85814
16624         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16625         a null return from get_strinfo when unsharing the next
16626         strinfo in the chain.
16628 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
16630         PR gcc/84923
16631         * varasm.c (weak_finish): Clean up weak_decls.
16633 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16635         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16636         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16637         UNSPEC_UADALP values.
16638         * config/aarch64/iterators.md (ABAL): New int iterator.
16639         (ABDL2): Likewise.
16640         (ADALP): Likewise.
16641         (sur): Add mappings for the above.
16642         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16643         New define_insn.
16644         (aarch64_<sur>abal<mode>_4): Likewise.
16645         (aarch64_<sur>adalp<mode>_3): Likewise.
16646         (<sur>sadv16qi): New define_expand.
16648 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
16650         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16651         (*movdf_internal): Ditto.
16652         (*rcpsf2_sse): Ditto.
16653         (*rsqrtsf2_sse): Ditto.
16654         (*sqrt<mode>2_sse): Ditto.
16656 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
16658         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16659         eor3q<mode>4.
16660         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16661         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16662         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16663         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16664         vbcaxq_s64): New.
16665         * config/aarch64/arm_neon.h: Likewise.
16666         * config/aarch64/iterators.md (VQ_I): New.
16668 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
16670         * config.gcc: Add arc/t-multilib-linux to tmake_file for
16671         arc*-*-linux*.
16672         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16673         MULTILIB_DIRNAMES
16675 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16677         * config/nds32/constraints.md (S): New constraint.
16678         * config/nds32/nds32.md (call_internal): Use constraint S.
16679         (call_value_internal): Likewise.
16680         (sibcall_internal): Likewise.
16681         (sibcall_value_internal): Likewise.
16683 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16684             Chung-Ju Wu  <jasonwucj@gmail.com>
16686         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16687         into consideration.
16689 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16690             Chung-Ju Wu  <jasonwucj@gmail.com>
16692         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16693         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16694         (nds32_rtx_costs_impl): Simplify.
16695         (nds32_address_cost_impl): Simplify.
16696         (nds32_init_rtx_costs): New function.
16697         (nds32_rtx_costs_speed_prefer): Likewise.
16698         (nds32_rtx_costs_size_prefer): Likewise.
16699         (nds32_address_cost_speed_prefer): Likewise.
16700         (nds32_address_cost_speed_fwprop): Likewise.
16701         (nds32_address_cost_size_prefer): Likewise.
16702         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16703         * config/nds32/nds32.c (nds32_option_override): Use
16704         nds32_init_rtx_costs function.
16706 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16708         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16709         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16710         (TARGET_PIPELINE_N8): Likewise.
16711         (TARGET_PIPELINE_N10): Likewise.
16712         (TARGET_PIPELINE_N13): Likewise.
16713         (TARGET_PIPELINE_GRAYWOLF): Likewise.
16715 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
16717         * config/nds32/nds32-fpu.md: Update copyright year.
16719 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16721         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16723 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16725         * config/nds32/nds32.c
16726         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16727         * config/nds32/nds32.opt (minline-asm-r15): New option.
16729 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16731         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16732         MASK_HW_ABS.
16733         * config/nds32/nds32.md (abssi2): New pattern.
16735 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
16737         * config/i386/i386.md (rex64namesuffix): New mode attribute.
16738         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16739         Merge insn pattern from sse_cvtsi2ss<round_name> and
16740         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16741         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16742         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16743         using SWI48 mode iterator.
16744         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16745         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16746         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16747         pattern from sse_cvttss2si<round_saeonly_name>
16748         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16749         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16750         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16751         using SWI48 mode iterator.
16752         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16753         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16754         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16755         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16756         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16757         using SWI48 mode iterator.
16758         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16759         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16760         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16761         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16762         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16763         SWI48 mode iterator.
16764         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16765         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16766         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16767         pattern from sse_cvttsd2si<round_saeonly_name>
16768         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16770 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16772         * config/nds32/nds32-md-auxiliary.c
16773         (nds32_valid_smw_lwm_base_p): Refine.
16774         (nds32_output_smw_single_word): Refine.
16775         (nds32_output_smw_double_word): New.
16776         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16778 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16780         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16781         (nds32_output_stack_pop): Refine.
16782         (nds32_expand_unaligned_load): Refine.
16783         (nds32_expand_unaligned_store): Refine.
16785 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16786             Chung-Ju Wu  <jasonwucj@gmail.com>
16788         * config/nds32/constants.md: Add TP_REGNUM constant.
16789         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16790         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16791         UNSPEC_ADD32.
16792         * config/nds32/nds32-doubleword.md: Consider flag_pic.
16793         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16794         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16795         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16796         and PIC code generation.
16797         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16798         code generation.
16799         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16800         optimization.
16801         * config/nds32/nds32.md: Support TLS and PIC.
16802         * config/nds32/nds32.c: Support TLS and PIC.
16803         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16804         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16805         predicate.
16807 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16809         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16810         mode with E_ prefix.
16812 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16813             Chung-Ju Wu  <jasonwucj@gmail.com>
16815         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16816         * config/nds32/nds32-md-auxiliary.c
16817         (symbolic_reference_mentioned_p): New.
16818         (nds32_legitimize_ict_address): New.
16819         (nds32_expand_ict_move): New.
16820         (nds32_indirect_call_referenced_p): New.
16821         (nds32_symbol_binds_local_p): Delete.
16822         (nds32_long_call_p): Modify.
16823         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16824         * config/nds32/nds32-protos.h
16825         (symbolic_reference_mentioned_p): Declare.
16826         (nds32_legitimize_ict_address): Declare.
16827         (nds32_expand_ict_move): Declare.
16828         (nds32_indirect_call_referenced_p): Declare.
16829         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16830         (nds32_relax_group): Use nds32_ict_const_p as condition.
16831         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16832         (nds32_asm_file_start): Output ict_model directive in asm code.
16833         (nds32_legitimate_address_p): Consider indirect call.
16834         (nds32_print_operand): Consider indirect call.
16835         (nds32_print_operand_address): Consider indirect call.
16836         (nds32_insert_attributes): Handle "indirect_call" attribute.
16837         (TARGET_LEGITIMATE_ADDRESS_P): Define.
16838         (TARGET_LEGITIMATE_CONSTANT_P): Define.
16839         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16840         (TARGET_DELEGITIMIZE_ADDRESS): Define.
16841         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16842         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16843         (TARGET_ICT_MODEL_SMALL): Define.
16844         (TARGET_ICT_MODEL_LARGE): Define.
16845         * config/nds32/nds32.md (movsi): Consider ict model.
16846         (call, call_value): Consider ict model.
16847         (sibcall, sibcall_value): Consider ict model.
16848         * config/nds32/nds32.opt (mict-model): New option.
16849         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16850         model.
16852 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
16853             Monk Chiang  <sh.chiang04@gmail.com>
16854             Jim Wilson <jimw@sifive.com>
16856         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16857         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
16858         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16859         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16860         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
16861         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16862         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16863         compute save_libcall_adjustment properly.
16864         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16865         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16866         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16867         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16868         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16869         (ABI_SPEC): Handle mabi=ilp32e.
16870         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16871         (RVE): Add RVE mask.
16872         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16873         <-march>: Add rv32e as an example.
16875 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
16877         PR c++/82899
16878         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
16879         (intra_create_variable_infos): Handle C++ constructors.
16881 2018-05-18  Martin Liska  <mliska@suse.cz>
16883         * passes.def: Remove a redundant pass.
16885 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
16887         PR bootstrap/85838
16888         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
16890 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16892         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
16893         (ARMv4): Update.
16894         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16895         (ARMv6m): Update.
16896         (armv2, armv2a, armv3, armv3m): Delete architectures.
16897         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
16898         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
16899         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
16900         Delete cpus.
16901         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
16902         (*mulsidi3adddi): Likewise.
16903         (mulsidi3): Likewise.
16904         (*mulsidi3_nov6): Likewise.
16905         (umulsidi3): Likewise.
16906         (umulsidi3_nov6): Likewise.
16907         (umaddsidi4): Likewise.
16908         (*umulsidi3adddi): Likewise.
16909         (smulsi3_highpart): Likewise.
16910         (*smulsi3_highpart_nov6): Likewise.
16911         (umulsi3_highpart): Likewise.
16912         (*umulsi3_highpart_nov6): Likewise.
16913         * config/arm/arm.h (arm_arch3m): Delete.
16914         * config/arm/arm.c (arm_arch3m): Delete.
16915         (arm_option_override_internal): Update armv3-related comment.
16916         (arm_configure_build_target): Delete use of isa_bit_mode32.
16917         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
16918         (arm_rtx_costs_internal): Delete check of arm_arch3m.
16919         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
16920         (mulsa3): Likewise.
16921         (mulusa3): Likewise.
16922         * config/arm/arm-protos.h (arm_arch3m): Delete.
16923         * config/arm/arm-tables.opt: Regenerate.
16924         * config/arm/arm-tune.md: Likewise.
16925         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
16926         deleted architectures.
16928 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16930         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
16931         (armv5t, armv5te): New features.
16932         (ARMv5, ARMv5e): Delete fgroups.
16933         (ARMv5t, ARMv5te): Adjust for above changes.
16934         (ARMv6m): Likewise.
16935         (armv5, armv5e): Delete arches.
16936         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
16937         arm_arch5.
16938         (*call_reg_arm): Likewise.
16939         (*call_value_reg_armv5): Likewise.
16940         (*call_value_reg_arm): Likewise.
16941         (*call_symbol): Likewise.
16942         (*call_value_symbol): Likewise.
16943         (*sibcall_insn): Likewise.
16944         (*sibcall_value_insn): Likewise.
16945         (clzsi2): Likewise.
16946         (prefetch): Likewise.
16947         (define_split and define_peephole2 dependent on arm_arch5):
16948         Likewise.
16949         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
16950         arm_arch5e.
16951         (TARGET_ARM_QBIT): Likewise.
16952         (TARGET_DSP_MULTIPLY): Likewise.
16953         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
16954         (arm_arch5, arm_arch5e): Delete.
16955         (arm_arch5t, arm_arch5te): Declare.
16956         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
16957         (arm_arch5t): Declare.
16958         (arm_option_reconfigure_globals): Update for the above.
16959         (arm_options_perform_arch_sanity_checks): Update comment, replace
16960         use of arm_arch5 with arm_arch5t.
16961         (use_return_insn): Likewise.
16962         (arm_emit_call_insn): Likewise.
16963         (output_return_instruction): Likewise.
16964         (arm_final_prescan_insn): Likewise.
16965         (arm_coproc_builtin_available): Likewise.
16966         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
16967         arm_arch5e with arm_arch5t and arm_arch5te.
16968         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
16969         (arm_arch5t, arm_arch5te): Declare.
16970         * config/arm/arm-tables.opt: Regenerate.
16971         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
16972         * config/arm/t-multilib: Likewise.
16973         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
16974         instead of arm_arch5.
16975         (*call_reg_thumb1): Likewise.
16976         (*call_value_reg_thumb1_v5): Likewise.
16977         (*call_value_reg_thumb1): Likewise.
16978         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
16979         unreachable path.
16980         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
16982 2018-05-18  Martin Liska  <mliska@suse.cz>
16984         PR gcov-profile/84846
16985         * doc/gcov.texi: Document -t option of gcov tool.
16987 2018-05-18  Martin Liska  <mliska@suse.cz>
16989         PR gcov-profile/84846
16990         * gcov.c (print_usage): Add new -t option.
16991         (process_args): Handle the option.
16992         (generate_results): Use stdout as output when requested by
16993         the option.
16995 2018-05-18  Martin Liska  <mliska@suse.cz>
16997         PR gcov-profile/84846
16998         * coverage.c (coverage_init): Write PWD to .gcno file.
16999         * doc/gcov.texi: Document how working directory is printed.
17000         * gcov-dump.c (dump_gcov_file): Print PWD.
17001         * gcov.c (output_intermediate_file): Likewise.
17002         (read_graph_file): Read PWD string.
17003         (output_lines): Print PWD.
17005 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17007         PR middle-end/85817
17008         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
17009         for retval and return false if all args to phi are zero.
17011 2018-05-18  Richard Biener  <rguenther@suse.de>
17013         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
17014         method.
17015         (evrp_dom_walker::before_dom_children): Call it.
17017 2018-05-18  Richard Biener  <rguenther@suse.de>
17019         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
17020         results when processing array refs with variable index.
17022 2018-05-18  Toon Moene  <toon@moene.org>
17024         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
17025         directly after that of -floop-interchange. Indicate that both
17026         options are enabled by default when specifying -O3.
17028 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17030         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
17031         iterator.  Delete separate integer-mode vec_set<mode> expander.
17032         (aarch64_simd_vec_setv2di): Delete.
17033         (vec_setv2di): Delete.
17034         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
17035         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
17036         the "w, r" alternative.
17038 2018-05-18  Martin Liska  <mliska@suse.cz>
17040         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
17041         * tree-pass.h (make_pass_lower_switch_O0): New function.
17042         * tree-switch-conversion.c (node_has_low_bound): Remove.
17043         (node_has_high_bound): Likewise.
17044         (node_is_bounded): Likewise.
17045         (class pass_lower_switch): Make it a template type and create
17046         two instances.
17047         (pass_lower_switch::execute): Add template argument.
17048         (make_pass_lower_switch): New function.
17049         (make_pass_lower_switch_O0): New function.
17050         (do_jump_if_equal): Remove.
17051         (emit_case_nodes): Simplify to just handle all 3 cases and leave
17052         all the hard work to tree optimization passes.
17054 2018-05-18  Martin Liska  <mliska@suse.cz>
17056         * dbgcnt.c (limit_low): Renamed from limit.
17057         (limit_high): New variable.
17058         (dbg_cnt_is_enabled): Check for upper limit.
17059         (dbg_cnt): Adjust dumping.
17060         (dbg_cnt_set_limit_by_index): Add new argument for high
17061         value.
17062         (dbg_cnt_set_limit_by_name): Likewise.
17063         (dbg_cnt_process_single_pair): Parse new format.
17064         (dbg_cnt_process_opt): Use strtok.
17065         (dbg_cnt_list_all_counters): Remove 'value' and add
17066         'limit_high'.
17067         * doc/invoke.texi: Document changes.
17069 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
17071         * doc/sourcebuild.texi (scalar_all_fma): Document.
17072         * tree.def (FMA_EXPR): Delete.
17073         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
17074         * internal-fn.c (ternary_direct): New macro.
17075         (expand_ternary_optab_fn): Likewise.
17076         (direct_ternary_optab_supported_p): Likewise.
17077         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
17078         * builtins.c (fold_builtin_fma): Delete.
17079         (fold_builtin_3): Don't call it.
17080         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
17081         * expr.c (expand_expr_real_2): Likewise.
17082         * fold-const.c (operand_equal_p): Likewise.
17083         (fold_ternary_loc): Likewise.
17084         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
17085         * gimple.c (DEFTREECODE): Likewise.
17086         * gimplify.c (gimplify_expr): Likewise.
17087         * optabs-tree.c (optab_for_tree_code): Likewise.
17088         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
17089         * tree-eh.c (operation_could_trap_p): Likewise.
17090         (stmt_could_throw_1_p): Likewise.
17091         * tree-inline.c (estimate_operator_cost): Likewise.
17092         * tree-pretty-print.c (dump_generic_node): Likewise.
17093         (op_code_prio): Likewise.
17094         * tree-ssa-loop-im.c (stmt_cost): Likewise.
17095         * tree-ssa-operands.c (get_expr_operands): Likewise.
17096         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
17097         * fold-const-call.h (fold_fma): Delete.
17098         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
17099         CFN_FNMA and CFN_FNMS.
17100         (fold_fma): Delete.
17101         * genmatch.c (combined_fn): New enum.
17102         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
17103         (commutative_op): New function.
17104         (commutate): Use it.  Handle more than 2 operands.
17105         (dt_operand::gen_gimple_expr): Use commutative_op.
17106         (parser::parse_expr): Allow :c to be used with non-binary
17107         operators if the commutative operand is known.
17108         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
17109         CFN_FMS, CFN_FNMA and CFN_FNMS.
17110         (backprop::process_assign_use): Remove FMA_EXPR handling.
17111         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
17112         (gen_hsa_fma): New function.
17113         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
17114         IFN_FNMA and IFN_FNMS.
17115         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
17116         * gimple-fold.h (follow_all_ssa_edges): Declare.
17117         * gimple-fold.c (follow_all_ssa_edges): New function.
17118         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
17119         gimple_build interface and use follow_all_ssa_edges to fold the result.
17120         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
17121         instead of checking for optabs directly.
17122         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
17123         rather than FMA_EXPRs.
17124         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
17125         call to IFN_FMA instead of an FMA_EXPR.
17127 2018-05-17  Jim Wilson  <jimw@sifive.com>
17129         * expr.c (do_tablejump): When converting index to Pmode, if we have a
17130         sign extended promoted subreg, and the range does not have the sign bit
17131         set, then do a sign extend.
17133         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
17134         test, check for sign extended subreg and/or constant operands, and
17135         do a sign extend in that case.
17137 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
17139         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
17140         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
17141         Add untyped.
17142         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
17143         Change logics_shift_reg to logics_shift_imm.
17144         (thunderx2t99_fp_loadpair_basic): Delete.
17145         (thunderx2t99_fp_storepair_basic): Delete.
17146         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
17147         (thunderx2t99_asimd_polynomial): Delete.
17148         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
17149         and neon_fp_mul_d_scalar_q.
17150         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
17151         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
17152         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
17153         (thunderx2t99_asimd_lut): Add missing tbl types.
17154         (thunderx2t99_asimd_ext): Delete.
17155         (thunderx2t99_asimd_load1_1_mult): Delete.
17156         (thunderx2t99_asimd_load1_2_mult): Delete.
17157         (thunderx2t99_asimd_load1_ldp): New.
17158         (thunderx2t99_asimd_load1): New.
17159         (thunderx2t99_asimd_load2): Add missing *load2* types.
17160         (thunderx2t99_asimd_load3): New.
17161         (thunderx2t99_asimd_load4): New.
17162         (thunderx2t99_asimd_store1_1_mult): Delete.
17163         (thunderx2t99_asimd_store1_2_mult): Delete.
17164         (thunderx2t99_asimd_store2_mult): Delete.
17165         (thunderx2t99_asimd_store2_onelane): Delete.
17166         (thunderx2t99_asimd_store_stp): New.
17167         (thunderx2t99_asimd_store1): New.
17168         (thunderx2t99_asimd_store2): New.
17169         (thunderx2t99_asimd_store3): New.
17170         (thunderx2t99_asimd_store4): New.
17172 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
17174         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
17175         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
17177 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
17178             Segher Boessenkool  <segher@kernel.crashing.org>
17180         PR target/85698
17181         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
17182         operand.
17184 2018-05-17  Richard Biener  <rguenther@suse.de>
17186         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
17187         for pruning loop and prune defs feeding only already visited PHIs.
17189 2018-05-17  Richard Biener  <rguenther@suse.de>
17191         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
17193 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
17194             Richard Biener  <rguenther@suse.de>
17196         PR tree-optimization/85793
17197         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
17198         for VMAT_ELEMENTWISE.
17200 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17202         * internal-fn.h (lookup_internal_fn): Declare
17203         * internal-fn.c (lookup_internal_fn): New function.
17204         * gimple.c (gimple_build_call_from_tree): Handle calls to
17205         internal functions.
17206         * gimple-pretty-print.c (dump_gimple_call): Print "." before
17207         internal function names.
17208         * tree-pretty-print.c (dump_generic_node): Likewise.
17209         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
17211 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17213         * gimple-fold.h (gimple_build): Make the function forms take
17214         combined_fn rather than built_in_function.
17215         (gimple_simplify): Likewise.
17216         * gimple-match-head.c (gimple_simplify): Likewise.
17217         * gimple-fold.c (gimple_build): Likewise.
17218         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
17219         rather than gimple_build_call_internal.
17220         (get_initial_defs_for_reduction): Likewise.
17221         (vect_create_epilog_for_reduction): Likewise.
17222         (vectorizable_live_operation): Likewise.
17224 2018-05-17  Martin Liska  <mliska@suse.cz>
17226         * gimple-ssa-sprintf.c (format_directive): Do not use
17227         space in between 'G_' and '('.
17229 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
17231         PR target/85323
17232         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
17233         even if the mask is not all ones.
17235         PR target/85323
17236         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
17237         vector.
17238         (ix86_gimple_fold_builtin): Likewise.
17240         PR target/85323
17241         * config/i386/i386.c: Include tree-vector-builder.h.
17242         (ix86_vector_shift_count): New function.
17243         (ix86_fold_builtin): Fold shift builtins by scalar count.
17244         (ix86_gimple_fold_builtin): Likewise.
17246         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
17247         _mm512_setzero): New intrinsics.
17249 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
17250             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17252         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
17253         code generation for cases where splatting a value is not useful.
17254         * simplify-rtx.c (simplify_ternary_operation): Simplify
17255         vec_merge across a vec_duplicate and a paradoxical subreg forming
17256         a vector mode to a vec_concat.
17258 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
17260         * config.gcc: Support "goldmont-plus".
17261         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17262         "goldmont-plus".
17263         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17264         PROCESSOR_GOLDMONT_PLUS.
17265         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
17266         (processor_target_table): Add "goldmont-plus".
17267         (PTA_GOLDMONT_PLUS): Define.
17268         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
17269         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
17270         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
17271         (fold_builtin_cpu): Add "goldmont-plus".
17272         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
17273         (ix86_option_override_internal): Add "goldmont-plus".
17274         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
17275         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
17276         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
17277         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
17279 2018-05-17  Richard Biener  <rguenther@suse.de>
17281         PR tree-optimization/85757
17282         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
17283         remove defs that only feed that PHI from further processing.
17285 2018-05-16  Jim Wilson  <jimw@sifive.com>
17287         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
17288         asterisk to name.
17289         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
17291 2018-05-16  Mark Wielaard  <mark@klomp.org>
17293         * dwarf2out.c (count_index_strings): New function.
17294         (output_indirect_strings): Call count_index_strings and generate
17295         header for dwarf_version >= 5.
17297 2018-05-16  Mark Wielaard  <mark@klomp.org>
17299         * dwarf2out.c (dwarf_FORM): New function.
17300         (set_indirect_string): Use dwarf_FORM.
17301         (reset_indirect_string): Likewise.
17302         (size_of_die): Likewise.
17303         (value_format): Likewise.
17304         (output_die): Likewise.
17305         (add_skeleton_AT_string): Likewise.
17306         (output_macinfo_op): Likewise.
17307         (index_string): Likewise.
17308         (output_index_string_offset): Likewise.
17309         (output_index_string): Likewise.
17310         (count_index_strings): Likewise.
17312 2018-05-16  Carl Love  <cel@us.ibm.com>
17314         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17315         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
17317 2018-05-16  Martin Jambor  <mjambor@suse.cz>
17319         * ipa-prop.c (ipa_free_all_edge_args): Remove.
17320         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
17322 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
17324         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
17325         (fnma<mode>4): Likewise.
17326         (fms<mode>4): Likewise.
17327         (fnms<mode>4): Likewise.
17328         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
17329         (aarch64_fnma<mode>4): Likewise.
17330         (aarch64_fms<mode>4): Likewise.
17331         (aarch64_fnms<mode>4): Likewise.
17332         (aarch64_fnmadd<mode>4): Likewise.
17334 2018-05-16  Jason Merrill  <jason@redhat.com>
17336         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
17338 2018-05-16  Richard Biener  <rguenther@suse.de>
17340         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
17341         (dump_stmt_cost): Declare.
17342         (add_stmt_cost): Dump cost we add.
17343         (add_stmt_costs): New function.
17344         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
17345         No longer exported.
17346         (vect_analyze_stmt): Adjust prototype.
17347         (vectorizable_condition): Likewise.
17348         (vectorizable_live_operation): Likewise.
17349         (vectorizable_reduction): Likewise.
17350         (vectorizable_induction): Likewise.
17351         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
17352         cost vector to pass to vectorizable_ and record afterwards.
17353         (vect_model_reduction_cost): Take cost vector argument and adjust.
17354         (vect_model_induction_cost): Likewise.
17355         (vectorizable_reduction): Likewise.
17356         (vectorizable_induction): Likewise.
17357         (vectorizable_live_operation): Likewise.
17358         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
17359         SLP_TREE_NUMBER_OF_VEC_STMTS.
17360         (vect_analyze_slp_cost_1): Remove.
17361         (vect_analyze_slp_cost): Likewise.
17362         (vect_slp_analyze_node_operations): Take visited args and
17363         a target cost vector.  Avoid processing already visited stmt sets.
17364         (vect_slp_analyze_operations): Use a local cost vector to gather
17365         costs and register those of non-discarded instances.
17366         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
17367         (vect_schedule_slp_instance): Remove copying of
17368         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
17369         zero.
17370         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
17371         adding cost.  Record cost entry location.
17372         (vect_prologue_cost_for_slp_op): Function to compute cost of
17373         a constant or invariant generated for SLP vect in the prologue,
17374         split out from vect_analyze_slp_cost_1.
17375         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
17376         (vect_model_promotion_demotion_cost): Likewise.
17377         (vect_model_store_cost): Likewise, make static.
17378         (vect_model_load_cost): Likewise.
17379         (vectorizable_bswap): Add cost vector arg and adjust.
17380         (vectorizable_call): Likewise.
17381         (vectorizable_simd_clone_call): Likewise.
17382         (vectorizable_conversion): Likewise.
17383         (vectorizable_assignment): Likewise.
17384         (vectorizable_shift): Likewise.
17385         (vectorizable_operation): Likewise.
17386         (vectorizable_store): Likewise.
17387         (vectorizable_load): Likewise.
17388         (vectorizable_condition): Likewise.
17389         (vectorizable_comparison): Likewise.
17390         (can_vectorize_live_stmts): Likewise.
17391         (vect_analyze_stmt): Likewise.
17392         (vect_transform_stmt): Adjust calls to vectorizable_*.
17393         * tree-vectorizer.c: Include gimple-pretty-print.h.
17394         (dump_stmt_cost): New function.
17396 2018-05-16  Richard Biener  <rguenther@suse.de>
17398         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
17399         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
17400         * tree-ssa-dse.c: Include tree-ssa-loop.h.
17401         (check_name): New callback.
17402         (dse_classify_store): Track cycles via a visited bitmap of PHI
17403         defs and simplify handling of in-loop and across loop dead stores
17404         and properly fail for loop-variant refs.  Handle byte-tracking with
17405         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
17406         limiting the walk.
17408 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
17410         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
17411         (vect_get_mask_type_for_stmt): Likewise.
17412         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
17413         split out from...
17414         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
17415         to determine the statement's vector type and the vector type that
17416         should be used for calculating nunits.  Deal with cases in which
17417         the type has to be deferred.
17418         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
17419         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
17420         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
17421         (vect_determine_vf_for_stmt): New functions, split out from...
17422         (vect_determine_vectorization_factor): ...here.
17423         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
17424         (vect_get_mask_type_for_stmt): New functions, split out from
17425         vect_determine_vectorization_factor.
17427 2018-05-16  Richard Biener  <rguenther@suse.de>
17429         * tree-cfg.c (verify_gimple_assign_ternary): Properly
17430         verify the [VEC_]COND_EXPR embedded comparison.
17432 2018-05-15  Martin Sebor  <msebor@redhat.com>
17434         PR tree-optimization/85753
17435         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
17436         RECORD_TYPE in addition to ARRAY_TYPE.
17438 2018-05-15  Martin Sebor  <msebor@redhat.com>
17440         PR middle-end/85643
17441         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17443 2018-05-15  Richard Biener  <rguenther@suse.de>
17445         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
17446         add by_clobber_p one.  Change algorithm to collect all defs
17447         representing uses we need to walk and try reducing them to
17448         a single one before failing.
17449         (dse_dom_walker::dse_optimize_stmt): Adjust.
17451 2018-05-13  Mark Wielaard  <mark@klomp.org>
17453         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
17454         (size_of_loc_descr): Likewise.
17455         (output_loc_operands): Likewise.
17456         (output_loc_operands_raw): Likewise.
17457         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
17458         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17459         (hash_loc_operands): Likewise.
17460         (compare_loc_operands): Likewise.
17462 2018-05-14  Mark Wielaard  <mark@klomp.org>
17464         * dwarf2out.c (count_index_addrs): New function.
17465         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
17467 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17469         PR tree-optimization/83648
17470         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
17471         return value as malloc candidate.
17473 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17475         PR ipa/85734
17476         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
17477         param as true in call to suggest_attribute.
17479 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
17481         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
17482         -mreadonly-in-sdata.
17484 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17486         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
17487         New pattern.
17488         (aarch64_crypto_aesd_fused): Likewise.
17490 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
17492         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17493         (movsi_aarch64): Likewise.
17494         (load_pairsi): Likewise.
17495         (load_pairdi): Likewise.
17496         (store_pairsi): Likewise.
17497         (store_pairdi): Likewise.
17498         (load_pairsf): Likewise.
17499         (load_pairdf): Likewise.
17500         (store_pairsf): Likewise.
17501         (store_pairdf): Likewise.
17502         (zero_extend): Likewise.
17503         (trunc): Swap alternatives.
17504         (fcvt_target): Add '?' to prefer w over r.
17506 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
17508         PR target/85756
17509         * config/i386/i386.md: Disallow non-commutative arithmetics in
17510         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17511         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
17512         in the peephole2 before it.
17514 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
17516         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17517         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17518         (ix86_handle_option): Handle -mcldemote.
17519         * config.gcc: New header.
17520         * config/i386/cldemoteintrin.h: New file.
17521         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17522         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17523         -mcldemote.
17524         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17525         OPTION_MASK_ISA_CLDEMOTE.
17526         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17527         (ix86_valid_target_attribute_inner_p): Ditto.
17528         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17529         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17530         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17531         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17532         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17533         (cldemote): New.
17534         * config/i386/i386.opt: Add -mcldemote.
17535         * config/i386/x86intrin.h: New header.
17536         * doc/invoke.texi: Add -mcldemote.
17538 2018-05-14  Richard Biener  <rguenther@suse.de>
17540         * doc/match-and-simplify.texi: Adjust :s documentation.
17542 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
17544         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17545         intended memcpy size.
17546         (REORDER_45): Likewise.
17548 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
17550         * sort.cc: New file.
17551         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17552         * vec.c (qsort_chk): Use gcc_qsort.
17553         * Makefile.in (OBJS-libcommon): Add sort.o.
17554         (build/sort.o): New target.  Use it...
17555         (BUILD_RTL): ... here, and...
17556         (build/gencfn-macros): ... here, and...
17557         (build/genmatch): ... here.
17559 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17560             Chung-Ju Wu  <jasonwucj@gmail.com>
17562         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17563         * config/nds32/nds32-graywolf.md: New file.
17564         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17565         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17566         pipeline.
17567         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17568         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17569         * config/nds32/nds32.md (pipeline_model): Add graywolf.
17570         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17571         * config/nds32/pipelines.md: Include n15 settings.
17573 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17574             Chung-Ju Wu  <jasonwucj@gmail.com>
17576         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17577         * config/nds32/nds32-n13.md: New file.
17578         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17579         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17580         pipeline.
17581         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17582         * config/nds32/nds32.md (pipeline_model): Add n13.
17583         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17584         * config/nds32/pipelines.md: Include n13 settings.
17586 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17587             Chung-Ju Wu  <jasonwucj@gmail.com>
17589         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17590         * config/nds32/nds32-n10.md: New file.
17591         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17592         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17593         pipeline.
17594         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17595         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17596         * config/nds32/nds32.md (pipeline_model): Add n10.
17597         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17598         * config/nds32/pipelines.md: Include n10 settings.
17600 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
17601             Kito Cheng  <kito.cheng@gmail.com>
17602             Chung-Ju Wu  <jasonwucj@gmail.com>
17604         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17605         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17606         Add enum values for DSP extension instructions.
17607         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17608         New constraints.
17609         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17610         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17611         New code iterators.
17612         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17613         * config/nds32/nds32-dspext.md: New file for DSP implementation.
17614         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17615         * config/nds32/nds32-intrinsic.md: Likewise.
17616         * config/nds32/nds32_intrinsic.h: Likewise.
17617         * config/nds32/nds32-md-auxiliary.c: Likewise.
17618         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17619         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17620         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17621         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17622         * config/nds32/nds32-protos.h: New declarations for DSP extension.
17623         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17624         TYPE_DMAC in switch statement.
17625         * config/nds32/nds32.c: New checking and implementation for DSP
17626         extension instructions.
17627         * config/nds32/nds32.h: Likewise.
17628         * config/nds32/nds32.md: Likewise.
17629         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17630         * config/nds32/predicates.md: Implement new predicates for DSP
17631         extension.
17633 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
17635         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17636         Reformat alternatives and attributes so it is easier to identify
17637         which constraints/attributes go with which instruction.
17638         (mov<mode>_hardfloat32, FMOVE64): Likewise.
17639         (mov<mode>_softfloat32, FMOVE64): Likewise.
17640         (mov<mode>_hardfloat64, FMOVE64): Likewise.
17641         (mov<mode>_softfloat64, FMOVE64): Likewise.
17643 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17645         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17646         subsection.
17647         (Basic PowerPC Built-in Functions): The new name of the
17648         subsection previously known as "PowerPC Built-in Functions".
17649         (Basic PowerPC Built-in Functions Available on all Configurations):
17650         New subsubsection.
17651         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17652         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17653         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17654         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17656 2018-05-11  Martin Jambor  <mjambor@suse.cz>
17658         PR ipa/85655
17659         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17660         single const.
17662 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17664         PR target/85733
17665         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17667 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
17669         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17670         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17671         (ix86_handle_option): Handle -mwaitpkg.
17672         * config.gcc: New header.
17673         * config/i386/cpuid.h (bit_WAITPKG): New bit.
17674         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17675         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17676         function type.
17677         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17678         OPTION_MASK_ISA_WAITPKG.
17679         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17680         (ix86_option_override_internal): Add PTA_WAITPKG.
17681         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17682         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17683         IX86_BUILTIN_TPAUSE.
17684         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17685         __builtin_ia32_umwait and __builtin_ia32_tpause.
17686         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17687         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17688         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17689         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17690         UNSPECV_TPAUSE): New.
17691         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17692         * config/i386/i386.opt: Add -mwaitpkg.
17693         * config/i386/waitpkgintrin.h: New file.
17694         * config/i386/x86intrin.h: New header.
17695         * doc/invoke.texi: Add -mwaitpkg.
17697 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17699         PR target/85606
17700         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17701         equivalent.
17702         (cortex-m0): Use armv6s-m isa.
17703         (cortex-m0plus): Likewise.
17704         (cortex-m1): Likewise.
17705         (cortex-m0.small-multiply): Likewise.
17706         (cortex-m0plus.small-multiply): Likewise.
17707         (cortex-m1.small-multiply): Likewise.
17709 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
17710             Jakub Jelinek  <jakub@redhat.com>
17712         PR tree-optimization/85692
17713         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17714         source permute as well.
17716 2018-05-11  Martin Liska  <mliska@suse.cz>
17718         PR sanitizer/85556
17719         * doc/extend.texi: Document LLVM style format for no_sanitize
17720         attribute.
17722 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17724         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17725         mode_supports_vsx_dform_quad to mode_supports_dq_form.
17726         (mode_supports_vsx_dform_quad): Likewise.
17727         (mode_supports_vmx_dform): Move these functions to be next to the
17728         other mode_supports functions.
17729         (mode_supports_dq_form): Likewise.
17730         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17731         mode_supports_dq_form.
17732         (reg_offset_addressing_ok_p): Likewise.
17733         (offsettable_ok_by_alignment): Likewise.
17734         (rs6000_legitimate_offset_address_p): Likewise.
17735         (legitimate_lo_sum_address_p): Likewise.
17736         (rs6000_legitimize_address): Likewise.
17737         (rs6000_legitimize_reload_address): Likewise.
17738         (rs6000_secondary_reload_inner): Likewise.
17739         (rs6000_preferred_reload_class): Likewise.
17740         (rs6000_output_move_128bit): Likewise.
17742 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17744         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17745         Generate SImode target register for null target.
17746         <case IX86_BUILTIN_XGETBV>: Ditto.
17747         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17748         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17750 2018-05-10  Carl Love  <cel@us.ibm.com>
17752         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17753         dcbtt and dcbtstt if operands[2] is 0.
17755 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17757         PR target/85693
17758         * config/i386/sse.md (usadv64qi): New expander.
17760 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
17762         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17763         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17764         -maltivec=be support.
17765         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17766         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17767         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17768         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17769         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17770         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17771         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17772         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17773         altivec_vsumsws): Adjust.
17774         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17775         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17776         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17777         support.
17778         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17779         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17780         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17781         (altivec_lve<VI_char>x): Delete expand.
17782         (*altivec_lve<VI_char>x_internal): Rename to...
17783         (altivec_lve<VI_char>x): ... this.
17784         (altivec_lvxl_<mode>): Delete expand.
17785         (*altivec_lvxl_<mode>_internal): Rename to ...
17786         (altivec_lvxl_<mode>): ... this.
17787         (altivec_stvxl_<mode>): Delete expand.
17788         (*altivec_stvxl_<mode>_internal): Rename to ...
17789         (altivec_stvxl_<mode>): ... this.
17790         (altivec_stve<VI_char>x): Delete expand.
17791         (*altivec_stve<VI_char>x_internal): Rename to ...
17792         (altivec_stve<VI_char>x): ... this.
17793         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17794         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17795         reduc_plus_scal_<mode>): Adjust.
17796         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17797         comment.
17798         (rs6000_cpu_cpp_builtins): Adjust.
17799         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17800         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17801         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17802         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17803         -maltivec=be support.
17804         (rs6000_split_vec_extract_var): Adjust.
17805         (rs6000_split_v4si_init): Adjust.
17806         (swap_selector_for_mode): Delete.
17807         (altivec_expand_lvx_be, altivec_expand_stvx_be,
17808         altivec_expand_stvex_be): Delete.
17809         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17810         -maltivec=be support.
17811         (rs6000_gimple_fold_builtin): Ditto.
17812         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17813         Adjust.
17814         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17815         (TARGET_DIRECT_MOVE_64BIT): Adjust.
17816         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17817         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17818         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17819         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17820         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17821         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17822         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17823         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17824         anonymous split): Adjust.
17825         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17826         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17828 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
17830         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17831         when --with-gxx-include-dir is also specified.
17832         * configure: Regenerate.
17834 2018-05-09  Jim Wilson  <jimw@sifive.com>
17836         PR target/84797
17837         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17838         * config/riscv/t-withmultilib: New.
17839         * config/riscv/withmultilib.h: New.
17840         * doc/install.texi: Document RISC-V --with-multilib-list support.
17842 2018-05-09  Richard Biener  <rguenther@suse.de>
17844         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17845         vector.
17846         (vect_bb_vectorization_profitable_p): Adjust.  Compute
17847         actual scalar cost using the cost vector and the add_stmt_cost
17848         machinery.
17850 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17852         PR rtl-optimization/85645
17853         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17854         in the REG_CFA_REGISTER note for LR, don't leave it empty.
17856 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17858         PR rtl-optimization/85645
17859         * shrink-wrap.c (spread_components): Return a boolean saying if
17860         anything was changed.
17861         (try_shrink_wrapping_separate): Iterate spread_components until
17862         nothing changes anymore.
17864 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17866         PR rtl-optimization/85645
17867         * regrename.c (build_def_use): Also kill the chains that include the
17868         destination of a REG_CFA_REGISTER note.
17870 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17872         PR rtl-optimization/85645
17873         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17874         insn that has a REG_CFA_REGISTER note.
17876 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
17878         * cfgexpand.c (expand_clobber): New function.
17879         (expand_gimple_stmt_1): Use it.
17880         * tree-vect-stmts.c (vect_clobber_variable): New function,
17881         split out from...
17882         (vectorizable_simd_clone_call): ...here.
17883         (vectorizable_store): Emit a clobber either side of an
17884         IFN_STORE_LANES sequence.
17885         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
17887 2018-05-09  Tom de Vries  <tom@codesourcery.com>
17889         PR target/85626
17890         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
17891         (define_insn "trap_if_false"): Add exit after trap.
17893 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
17895         PR rtl-optimization/85638
17896         * bb-reorder.c: Include common/common-target.h.
17897         (create_forwarder_block): New function extracted from...
17898         (fix_up_crossing_landing_pad): ...here.  Rename into...
17899         (dw2_fix_up_crossing_landing_pad): ...this.
17900         (sjlj_fix_up_crossing_landing_pad): New function.
17901         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
17902         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
17903         from both partitions and exit the loop after one iteration.
17905 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17907         Revert:
17908         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17909         subsection.
17910         (Basic PowerPC Built-in Functions): The new name of the
17911         subsection previously known as "PowerPC Built-in Functions".
17912         (Basic PowerPC Built-in Functions Available on all Configurations):
17913         New subsubsection.
17914         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17915         subsubsection.
17916         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17917         subsubsection.
17918         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17919         subsubsection.
17920         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17921         subsubsection.
17923 2018-05-08  Jim Wilson  <jimw@sifive.com>
17925         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
17926         (LD_EMUL_SUFFIX): New.
17927         (LINK_SPEC): Use it.
17929 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17931         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17932         subsection.
17933         (Basic PowerPC Built-in Functions): The new name of the
17934         subsection previously known as "PowerPC Built-in Functions".
17935         (Basic PowerPC Built-in Functions Available on all Configurations):
17936         New subsubsection.
17937         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17938         subsubsection.
17939         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17940         subsubsection.
17941         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17942         subsubsection.
17943         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17944         subsubsection.
17946 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17948         PR target/85683
17949         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
17950         after cmpelim optimization.
17952 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
17954         * config.gcc: Support "goldmont".
17955         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
17956         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17957         PROCESSOR_GOLDMONT.
17958         * config/i386/i386.c (m_GOLDMONT): Define.
17959         (processor_target_table): Add "goldmont".
17960         (PTA_GOLDMONT): Define.
17961         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
17962         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
17963         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
17964         (fold_builtin_cpu): Add "goldmont".
17965         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
17966         (ix86_option_override_internal): Add "goldmont".
17967         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
17968         (processor_type): Add PROCESSOR_GOLDMONT.
17969         * config/i386/i386.md: Add CPU "glm".
17970         * config/i386/glm.md: New file.
17971         * config/i386/x86-tune.def: Add m_GOLDMONT.
17972         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
17974 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17976         PR target/85572
17977         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
17978         E_V4DImode.
17979         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
17980         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
17981         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
17983         PR target/85317
17984         * config/i386/i386.c (ix86_fold_builtin): Handle
17985         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
17987         PR target/85480
17988         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
17989         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
17991 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
17993         PR target/85658
17994         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
17995         (check_arch): Likewise.
17996         (check_fpu): Return the result rather than printing it.
17997         (end arch): Fix operator precedence.
17998         (end cpu): Likewise.
17999         (END): Print the result from check_fpu.
18001 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
18002             Alan Hayward  <alan.hayward@arm.com>
18003             David Sherwood  <david.sherwood@arm.com>
18005         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
18006         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
18007         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
18008         (*fcmuo<mode>_and): New patterns.
18010 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
18012         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
18013         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
18014         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
18015         (cmp_op, sve_imm_con): New code attributes.
18016         (SVE_COND_INT_CMP, imm_con): Delete.
18017         (cmp_op): Remove above unspecs from int attribute.
18018         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
18019         to...
18020         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
18021         comparison-specific unspecs.
18022         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
18023         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
18024         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
18025         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
18026         (*vec_fcm<cmp_op><mode>): Rename to...
18027         (*fcm<cmp_op><mode>): ...this and adjust likewise.
18028         (*vec_fcmuo<mode>): Rename to...
18029         (*fcmuo<mode>): ...this and adjust likewise.
18030         (*pred_fcm<cmp_op><mode>): New pattern.
18031         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
18032         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
18033         functions.
18034         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
18035         and UNORDERED.
18036         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
18037         (aarch64_emit_sve_predicated_cond): New function.
18038         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
18039         (aarch64_emit_unspec_cond_or): Replace with...
18040         (aarch64_emit_sve_or_conds): ...this new function.  Use
18041         aarch64_emit_sve_ptrue_op for the individual comparisons and
18042         aarch64_emit_binop to OR them together.
18043         (aarch64_emit_inverted_unspec_cond): Replace with...
18044         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
18045         aarch64_emit_sve_ptrue_op for the comparison and
18046         aarch64_emit_unop to invert the result.
18047         (aarch64_expand_sve_vec_cmp_float): Update after the above
18048         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
18050 2018-05-07  Nathan Sidwell  <nathan@acm.org>
18052         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
18053         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
18054         (Backwards Compatibility): Likewise.
18056 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18058         PR bootstrap/85681
18059         Revert:
18060         2018-05-07  Luis Machado  <luis.machado@linaro.org>
18062         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18063         <prefetch_dynamic_strides>: New const bool field.
18064         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18065         prefetch_dynamic_strides.
18066         (exynosm1_prefetch_tune): Likewise.
18067         (thunderxt88_prefetch_tune): Likewise.
18068         (thunderx_prefetch_tune): Likewise.
18069         (thunderx2t99_prefetch_tune): Likewise.
18070         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18071         to false.
18072         (aarch64_override_options_internal): Update to set
18073         PARAM_PREFETCH_DYNAMIC_STRIDES.
18074         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18075         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18076         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18077         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18078         prefetch-dynamic-strides setting.
18080         2018-05-07  Luis Machado  <luis.machado@linaro.org>
18082         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18083         <minimum_stride>: New const int field.
18084         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18085         minimum_stride field.
18086         (exynosm1_prefetch_tune): Likewise.
18087         (thunderxt88_prefetch_tune): Likewise.
18088         (thunderx_prefetch_tune): Likewise.
18089         (thunderx2t99_prefetch_tune): Likewise.
18090         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18091         (aarch64_override_options_internal): Update to set
18092         PARAM_PREFETCH_MINIMUM_STRIDE.
18093         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18094         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18095         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18096         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18097         stride is constant and is below the minimum stride threshold.
18099 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18101         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
18102         to 512.
18104 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18106         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18107         <prefetch_dynamic_strides>: New const bool field.
18108         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18109         prefetch_dynamic_strides.
18110         (exynosm1_prefetch_tune): Likewise.
18111         (thunderxt88_prefetch_tune): Likewise.
18112         (thunderx_prefetch_tune): Likewise.
18113         (thunderx2t99_prefetch_tune): Likewise.
18114         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18115         to false.
18116         (aarch64_override_options_internal): Update to set
18117         PARAM_PREFETCH_DYNAMIC_STRIDES.
18118         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18119         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18120         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18121         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18122         prefetch-dynamic-strides setting.
18124 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18126         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18127         <minimum_stride>: New const int field.
18128         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18129         minimum_stride field.
18130         (exynosm1_prefetch_tune): Likewise.
18131         (thunderxt88_prefetch_tune): Likewise.
18132         (thunderx_prefetch_tune): Likewise.
18133         (thunderx2t99_prefetch_tune): Likewise.
18134         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18135         (aarch64_override_options_internal): Update to set
18136         PARAM_PREFETCH_MINIMUM_STRIDE.
18137         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18138         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18139         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18140         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18141         stride is constant and is below the minimum stride threshold.
18143 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
18145         PR c++/85659
18146         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
18147         the type is addressable.  Don't force op into register if it has
18148         BLKmode.
18150 2018-05-05  Roland McGrath  <mcgrathr@google.com>
18152         PR other/77609
18153         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
18154         any section for which we don't know a specific type it should have,
18155         regardless of name.  Previously this was done only for the exact
18156         names ".init_array", ".fini_array", and ".preinit_array".
18157         (default_elf_asm_named_section): Add comment about
18158         relationship with default_section_type_flags and SECTION_NOTYPE.
18159         (get_section): Don't consider it a type conflict if one side has
18160         SECTION_NOTYPE and the other doesn't, as long as neither has the
18161         SECTION_BSS et al used in the default_section_type_flags logic.
18163 2018-05-05  Tom de Vries  <tom@codesourcery.com>
18165         PR target/85653
18166         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
18167         (workaround_barsyncs): New function.
18168         (nvptx_reorg): Use workaround_barsyncs.
18169         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18170         (define_expand "nvptx_membar_cta"): New define_expand.
18171         (define_insn "*nvptx_membar_cta"): New insn.
18173 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
18175         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
18176         To improve optimization opportunities.
18177         * builtin-types.def: The new needed builtin types for the above.
18179 2018-05-04  Richard Biener  <rguenther@suse.de>
18181         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
18182         * gimple-ssa-store-merging.c
18183         (imm_store_chain_info::output_merged_store): Remove redundant create,
18184         release split_store vector contents on failure.
18185         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
18186         scalar stmt vector on cache hit.
18188 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
18190         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
18191         Xilinx FP support.
18192         * config.gcc (powerpc-xilinx-eabi*): Remove.
18193         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
18194         support.
18195         (fusion_addis_mem_combo_load): Ditto.
18196         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
18197         FP support.
18198         (rs6000_cpu_cpp_builtins): Ditto.
18199         * config/rs6000/rs6000-linux.c
18200         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
18201         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
18202         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
18203         support.
18204         (rs6000_setup_reg_addr_masks): Ditto.
18205         (rs6000_init_hard_regno_mode_ok): Ditto.
18206         (rs6000_option_override_internal): Ditto.
18207         (legitimate_lo_sum_address_p): Ditto.
18208         (rs6000_legitimize_address): Ditto.
18209         (rs6000_legitimize_reload_address): Ditto.
18210         (rs6000_legitimate_address_p): Ditto.
18211         (abi_v4_pass_in_fpr): Ditto.
18212         (setup_incoming_varargs): Ditto.
18213         (rs6000_gimplify_va_arg): Ditto.
18214         (rs6000_split_multireg_move): Ditto.
18215         (rs6000_savres_strategy): Ditto.
18216         (rs6000_emit_prologue_components): Ditto.
18217         (rs6000_emit_epilogue_components): Ditto.
18218         (rs6000_emit_prologue): Ditto.
18219         (rs6000_emit_epilogue): Ditto.
18220         (rs6000_elf_file_end): Ditto.
18221         (rs6000_function_value): Ditto.
18222         (rs6000_libcall_value): Ditto.
18223         * config/rs6000/rs6000.h: Ditto.
18224         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
18225         (TARGET_MINMAX): ... this.  New.
18226         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
18227         * config/rs6000/rs6000.md: Remove Xilinx FP support.
18228         (*movsi_internal1_single): Delete.
18229         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
18230         mfpu=, mxilinx-fpu): Delete.
18231         * config/rs6000/singlefp.h: Delete.
18232         * config/rs6000/sysv4.h: Remove Xilinx FP support.
18233         * config/rs6000/t-rs6000: Ditto.
18234         * config/rs6000/t-xilinx: Delete.
18235         * config/rs6000/titan.md: Adjust for fp_type removal.
18236         * config/rs6000/vsx.md: Remove Xilinx FP support.
18237         (VStype_simple): Delete.
18238         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
18239         * config/rs6000/xfpu.h: Delete.
18240         * config/rs6000/xfpu.md: Delete.
18241         * config/rs6000/xilinx.h: Delete.
18242         * config/rs6000/xilinx.opt: Delete.
18243         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
18244         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
18246 2018-05-04  Tom de Vries  <tom@codesourcery.com>
18248         PR libgomp/85639
18249         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
18250         if ignore == 0.
18252 2018-05-04  Richard Biener  <rguenther@suse.de>
18254         PR middle-end/85627
18255         * tree-complex.c (update_complex_assignment): We are always in SSA form.
18256         (expand_complex_div_wide): Likewise.
18257         (expand_complex_operations_1): Likewise.
18258         (expand_complex_libcall): Preserve EH info of the original stmt.
18259         (tree_lower_complex): Handle removed blocks.
18260         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
18261         on complex multiplication and division libcall builtins.
18263 2018-05-04  Richard Biener  <rguenther@suse.de>
18265         PR middle-end/85574
18266         * fold-const.c (negate_expr_p): Restrict negation of operand
18267         zero of a division to when we know that can happen without
18268         overflow.
18269         (fold_negate_expr_1): Likewise.
18271 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
18273         PR libstdc++/85466
18274         * real.h (real_nextafter): Declare.
18275         * real.c (real_nextafter): New function.
18276         * fold-const-call.c (fold_const_nextafter): New function.
18277         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
18278         CASE_CFN_NEXTTOWARD.
18279         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
18280         even when arg1_mode is different from arg0_mode.
18282 2018-05-03  Nathan Sidwell  <nathan@acm.org>
18284         * doc/extend.texi (Deprecated Features): Remove
18285         -ffriend-injection.
18286         (Backwards Compatibility): Likewise.
18287         * doc/invoke.texi (C++ Language Options): Likewise.
18288         (C++ Dialect Options): Likewise.
18290 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
18292         PR target/85530
18293         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
18294         _mm512_mask_mullox_epi64): New intrinsics.
18296 2018-05-03  Tom de Vries  <tom@codesourcery.com>
18298         PR testsuite/85106
18299         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18300         dump files): Add offload-tree.
18302 2018-05-03  Richard Biener  <rguenther@suse.de>
18304         PR tree-optimization/85615
18305         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
18306         to loops not nested in BBs loop father to avoid creating multi-entry
18307         loops.
18309 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18311         PR tree-optimization/70291
18312         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
18313         arguments.  Change return type to tree.  Emit libcall as a new
18314         statement rather than replacing existing one when inplace_p is true.
18315         (expand_complex_multiplication_components): New function.
18316         (expand_complex_multiplication): Expand floating-point complex
18317         multiplication using the above.
18318         (expand_complex_division): Rename inner_type parameter to type.
18319         Update expand_complex_libcall call-site.
18320         (expand_complex_operations_1): Update expand_complex_multiplication
18321         and expand_complex_division call-sites.
18323 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18325         PR target/85582
18326         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18327         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18328         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
18329         the highest significant bit of the shift count mask is clear.  In
18330         check whether and[sq]i3 is needed verify that all significant bits
18331         of the shift count other than the highest are set.
18333 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18335         PR libgomp/82428
18336         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
18337         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
18338         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
18339         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
18340         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
18341         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
18342         __builtin_goacc_parlevel_size.
18344 2018-05-02  Richard Biener  <rguenther@suse.de>
18346         PR tree-optimization/85597
18347         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
18348         do not use split vect_get_vec_defs call but call vect_get_slp_defs
18349         directly.
18351 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18353         PR testsuite/85106
18354         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18355         dump files): Add ltrans-tree.
18357 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18359         PR testsuite/85106
18360         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18361         dump files): Add wpa-ipa.
18363 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
18365         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
18366         powerpc*-*-linux*paired* target.
18367         * config/rs6000/750cl.h: Delete.
18368         * config/rs6000/paired.h: Delete.
18369         * config/rs6000/paired.md: Delete.
18370         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
18371         float support.
18372         * config/rs6000/rs6000-builtin.def: Remove paired float support.
18373         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
18374         comment.  Remove paired float support.
18375         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
18376         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
18377         VECTOR_PAIRED.
18378         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
18379         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
18380         declarations.
18381         * config/rs6000/rs6000.c: Remove paired float support.
18382         (paired_expand_vector_init, paired_expand_vector_move,
18383         paired_emit_vector_compare, paired_emit_vector_cond_expr,
18384         (paired_expand_lv_builtin, paired_expand_stv_builtin,
18385         paired_expand_builtin, paired_expand_predicate_builtin,
18386         paired_init_builtins): Delete.
18387         * config/rs6000/rs6000.h: Remove paired float support.
18388         * config/rs6000/rs6000.md: Remove paired float support.
18389         (move_from_CR_ov_bit): Delete.
18390         * config/rs6000/rs6000.opt (mpaired): Delete.
18391         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
18392         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
18394 2018-05-02  Richard Biener  <rguenther@suse.de>
18396         PR middle-end/85567
18397         * gimplify.c (gimplify_save_expr): When in SSA form allow
18398         SAVE_EXPRs to compute to SSA vars.
18400 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18402         PR target/85582
18403         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18404         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18405         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
18406         clobber operands[2], instead use a new pseudo.  Formatting fixes.
18408 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
18410         PR tree-optimization/85586
18411         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
18412         exit early for statements in the same group if the accesses are
18413         not strided.
18415 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18417         PR lto/85451
18418         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
18419         error message.
18421 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
18423         PR tree-optimization/85143
18424         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
18426 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18428         PR lto/85451
18429         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
18430         not found" error message.
18432 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18434         PR other/83786
18435         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
18436         * vec.c (test_ordered_remove_if): New function.
18437         (vec_c_tests): Call test_ordered_remove_if.
18438         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
18439         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
18440         * tree-vect-patterns.c (vect_pattern_recog_1): Use
18441         VEC_ORDERED_REMOVE_IF.
18443 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18445         PR tree-optimization/82665
18446         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
18447         pointer subtraction where arguments come from a memchr call.
18449 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
18451         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
18452         --push-state --as-needed and --pop-state instead of --as-needed and
18453         --no-as-needed if ld supports it.
18454         * configure: Regenerated.
18456         PR web/85578
18457         * doc/install.texi2html: Replace _002d with - and _002a with * in
18458         generated html files using sed.
18460 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18462         PR c++/85523
18463         * gcc-rich-location.c (blank_line_before_p): New function.
18464         (use_new_line): New function.
18465         (gcc_rich_location::add_fixit_insert_formatted): New function.
18466         * gcc-rich-location.h
18467         (gcc_rich_location::add_fixit_insert_formatted): New function.
18469 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18471         * selftest.c (assert_streq): Rename "expected" and "actual" to
18472         "val1" and "val2".  Extend NULL-handling to cover both inputs
18473         symmetrically, while still requiring both to be non-NULL for a pass.
18474         * selftest.h (assert_streq): Rename "expected" and "actual" to
18475         "val1" and "val2".
18476         (ASSERT_EQ): Likewise.
18477         (ASSERT_EQ_AT): Likewise.
18478         (ASSERT_KNOWN_EQ): Likewise.
18479         (ASSERT_KNOWN_EQ_AT): Likewise.
18480         (ASSERT_NE): Likewise.
18481         (ASSERT_MAYBE_NE): Likewise.
18482         (ASSERT_MAYBE_NE_AT): Likewise.
18483         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
18484         the assertion to pass.
18485         (ASSERT_STREQ_AT): Likewise.
18487 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18489         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
18490         interaction with -pie.
18492 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18494         * selftest.h: Fix alphabetization of per-source-file selftest
18495         declarations.
18497 2018-04-30  Jason Merrill  <jason@redhat.com>
18499         PR c++/61982 - dead stores to destroyed objects.
18500         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18501         of clobber.
18503 2018-04-30  Jason Merrill  <jason@redhat.com>
18505         * tree.c (build_clobber): New.
18506         * tree.h: Declare it.
18507         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18509 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18511         * diagnostic-show-locus.c (layout::layout): Update for
18512         location_get_source_line returning a char_span.
18513         (struct char_span): Move to input.h.
18514         (struct correction): Update for fields in char_span becoming
18515         private.
18516         (struct source_line): Update for location_get_source_line
18517         returning a char_span.
18518         (layout::print_line): Likewise.
18519         * edit-context.c (edited_file::print_content): Likewise.
18520         (edited_file::print_diff_hunk): Likewise.
18521         (edited_file::print_run_of_changed_lines): Likewise.
18522         (edited_file::get_num_lines): Likewise.
18523         (edited_line::edited_line): Likewise.
18524         * final.c (asm_show_source): Likewise.
18525         * input.c (location_get_source_line): Convert return type
18526         from const char * to char_span, losing the final "line_len"
18527         param.
18528         (dump_location_info): Update for the above.
18529         (get_substring_ranges_for_loc): Likewise.  Use a char_span
18530         when handling the literal within the line.
18531         (test_reading_source_line): Update for location_get_source_line
18532         returning a char_span.
18533         * input.h (class char_span): Move here from
18534         diagnostic-show-locus.c, converting from a struct to a class.
18535         Make data members private.
18536         (char_span::operator bool): New.
18537         (char_span::length): New.
18538         (char_span::get_buffer): New.
18539         (char_span::operator[]): New.
18540         (char_span::subspan): Make const.
18541         (char_span::xstrdup): New.
18542         (location_get_source_line): Convert return type from const char *
18543         to char_span, losing the final "line_size" param.
18545 2018-04-30  Jan Hubicka  <jh@suse.cz>
18547         * lto-wrapper.c (ltrans_priorities): New static var.
18548         (cmp_priority): New.
18549         (run_gcc): Read priorities and if doing parallel build order
18550         the Makefile by them.
18552 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18554         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18556 2018-04-30  Richard Biener  <rguenther@suse.de>
18558         * tree-cfg.c (verify_address): Remove base argument, add
18559         flag whether to check TREE_ADDRESSABLE and do that.
18560         (verify_expr): Remove.
18561         (verify_types_in_gimple_reference): Add pieces from verify_expr.
18562         (verify_gimple_assign_single): Likewise.
18563         (verify_gimple_switch): Likewise.
18564         (verify_expr_location_1): Dereference tp once.  Add (disabled)
18565         piece from verify_expr.
18566         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18568 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18570         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18572 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18574         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18575         (small_data_pattern): Likewise.
18576         (arc_rewrite_small_data): Likewise.
18577         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18578         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18579         (get_symbol_alignment): New function.
18580         (legitimate_small_data_address_p): Likewise.
18581         (legitimate_scaled_address): Update, call
18582         legitimate_small_data_address_p.
18583         (output_sdata): New static variable.
18584         (arc_print_operand): Update how we handle small data operands.
18585         (arc_print_operand_address): Likewise.
18586         (arc_legitimate_address_p): Update, use
18587         legitimate_small_data_address_p.
18588         (arc_rewrite_small_data_p): Remove.
18589         (arc_rewrite_small_data_1): Likewise.
18590         (arc_rewrite_small_data): Likewise.
18591         (small_data_pattern): Likewise.
18592         (compact_sda_memory_operand): Update to use
18593         legitimate_small_data_address_p and get_symbol_alignment.
18594         (prepare_move_operands): Don't rewite sdata pattern.
18595         (prepare_extend_operands): Remove.
18596         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18597         pattern.
18598         (zero_extendqisi2): Likewise.
18599         (zero_extendhisi2): Likewise.
18600         (extendqihi2): Likewise.
18601         (extendqisi2): Likewise.
18602         (extendhisi2): Likewise.
18603         (addsi3): Likewise.
18604         (subsi3): Likewise.
18605         (andsi3): Likewise.
18606         * config/arc/constraints.md (Usd): Change it to memory constraint.
18608 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18610         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18611         as source of std instructions.
18612         * config/arc/arc.md (movsi_insn): Update pattern predicate to
18613         allow 6-bit constants as source for store instructions.
18614         (movdi_insn): Update instruction pattern to allow 6-bit constants
18615         as source for store instructions.
18617 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18619         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18621 2018-04-30  Nathan Sidwell  <nathan@acm.org>
18622             Sandra Loosemore <sandra@codesourcery.com>
18624         * dumpfile.c (dump_open): Allow '-' for stdout.
18625         * doc/invoke.texi (Developer Options): Document dump filename
18626         determination early.  Document stdin/stdout selection.
18628 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
18630         Microblaze Target: PIC data text relative
18632         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18633         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18634         Add declaration.
18635         * config/microblaze/microblaze.h (microblaze_constant_address_p):
18636         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18637         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18638         New addressing mode for data-text relative position indepenedent code.
18639         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18640         'ADDRESS_SYMBOLIC_TXT_REL'.
18641         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18642         (microblaze_legitimate_pic_operand): Exclude function calls from
18643         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18644         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18645         addresses cases.
18646         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18647         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18648         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18649         for 'address + offset'.
18650         (microblaze_expand_prologue): Add new function prologue call for
18651         'r20' assignation.
18652         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18653         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18654         table in case of TARGET_PIC_DATA_TEXT_REL.
18655         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18656         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18657         Add new macros 'UNSPEC_TEXT',
18658         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18659         + exclude function calls from 'UNSPEC_PLT' in case of data text
18660         relative mode.
18661         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18662         new target hook for generating address diff vector tables in case of
18663         flag_pic.
18664         * doc/tm.texi : Regenerate.
18665         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18666         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18667         of addr diff vector generation.
18668         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18669         target hook definition.
18670         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18671         Add default function for generate_pic_addr_diff_vec -> flag_pic.
18672         * doc/invoke.texi (Add new pic option): Add new microblaze pic
18673         option for data text relative.
18675 2018-04-30  Richard Biener  <rguenther@suse.de>
18677         * tree-chrec.h (evolution_function_is_constant_p): Remove
18678         redundant check.
18679         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18681 2018-04-30  Richard Biener  <rguenther@suse.de>
18683         PR bootstrap/85571
18684         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18686 2018-04-30  Richard Biener  <rguenther@suse.de>
18688         PR tree-optimization/28364
18689         PR tree-optimization/85275
18690         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18691         copying first exit test.
18693 2018-04-28  Mark Wielaard  <mark@klomp.org>
18695         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18696         dwarf_version >= 5.
18697         (dwarf_AT): Handle DW_AT_addr_base.
18698         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18700 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
18702         PR target/84431
18703         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18704         (*ashl<dwi>3_doubleword_mask_1): Ditto.
18705         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18706         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18708 2018-04-28  Richard Biener  <rguenther@suse.de>
18710         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18711         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18712         to reflect use.  Only add interesting stmts.
18714 2018-04-27  Martin Jambor  <mjambor@suse.cz>
18716         PR ipa/85549
18717         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18718         the jump function allows for passing through aggregate values.
18720 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
18722         * input.h (in_system_header_at): Convert from macro to inline
18723         function.
18724         (from_macro_expansion_at): Likewise.
18725         (from_macro_definition_at): Likewise.
18727 2018-04-27  Jeff Law  <law@redhat.com>
18729         * config.gcc: Mark tile* targets as deprecated/obsolete.
18731 2018-04-27  Richard Biener  <rguenther@suse.de>
18733         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18734         fix for ILP32.
18736 2018-04-27  Richard Biener  <rguenther@suse.de>
18738         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18740 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
18742         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18743         with Yd constraint. Set "preferred_for_speed" attribute from
18744         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18745         with Yd constraint.
18746         (*movdi_internal): Ditto.
18747         (movti_interunit splitters): Remove
18748         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18749         (movdi_interunit splitters): Ditto.
18750         * config/i386/constraints.md (Ye): Remove.
18751         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18753 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18755         PR target/85512
18756         * config/aarch64/constraints.md (Usg): Limit to 31.
18757         (Usj): Limit to 63.
18759 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
18761         PR tree-optimization/85529
18762         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18763         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18764         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18765         zero extension or masking of the MSB bit.
18766         (optimize_range_tests): Add FIRST_BB argument, pass it through
18767         to optimize_range_tests_var_bound.
18768         (maybe_optimize_range_tests, reassociate_bb): Adjust
18769         optimize_range_tests callers.
18771 2018-04-26  Richard Biener  <rguenther@suse.de>
18772             Jakub Jelinek  <jakub@redhat.com>
18774         * cgraph.h (symbol_table): Just declare debug method here.
18775         * symtab.c (symbol_table::debug): Define.
18777 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
18779         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18781 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
18783         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18784         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18785         (*movdi_internal): Substitute Yi and Yj constraint with x
18786         and Ym and Yn constraint with y constraint.  Update "isa"
18787         attribute and set "preferred_for_speed" attribute from
18788         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18789         (*movsi_internal): Ditto.
18790         (*movdf_internal): Ditto.
18791         (*movsf_internal): Ditto.
18792         (*zero_extendsidi2): Ditto.
18793         * config/i386/sse.md (vec_set<mode>_0): Ditto.
18794         (sse2_loadld): Ditto.
18795         (*vec_extract<ssevecmodelower>_0): Ditto.
18796         (*vec_extractv4si_0_zext_sse4): Ditto.
18797         (vec_concatv2di): Ditto.
18798         (*vec_dup<mode>): Ditto.
18799         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18800         * config/i386/constraints.md (Yi): Remove.
18801         (Yj): Remove.
18802         (Ym): Remove.
18803         (Yn): Remove.
18805 2018-04-26  Nathan Sidwell  <nathan@acm.org>
18807         * dumpfile.c (dump_open): New.
18808         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18809         (dump_finish): Detect stdio/stderr by value not name.
18811 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
18813         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18815 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18817         PR target/84952
18818         * config/nvptx/nvptx.c (verify_neutering_jumps)
18819         (verify_neutering_labels): New function
18820         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18822 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18824         PR target/84025
18825         * config/nvptx/nvptx.c (needs_neutering_p): New function.
18826         (nvptx_single): Use needs_neutering_p to skip over insns that do not
18827         need neutering.
18829 2018-04-26  Richard Biener <rguenther@suse.de>
18830             Tom de Vries  <tom@codesourcery.com>
18832         PR lto/85422
18833         * lto-streamer-out.c (output_function): Fixup loops if required to match
18834         discovery done in the reader.
18836 2018-04-26  Richard Biener  <rguenther@suse.de>
18838         PR tree-optimization/85116
18839         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18840         have a loop exit from the single latch predecessor.  Remove
18841         case of header with just condition.
18842         (ch_base::copy_headers): Exclude infinite loops from any
18843         processing.
18844         (pass_ch::execute): Record exits.
18846 2018-04-26  Richard Biener  <rguenther@suse.de>
18848         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18849         prologue cost vector and pass it to vect_get_load_cost.
18850         (vect_get_peeling_costs_all_drs): Likewise.
18851         (vect_peeling_hash_get_lowest_cost): Likewise.
18852         (vect_enhance_data_refs_alignment): Likewise.
18854 2018-04-26  Richard Biener  <rguenther@suse.de>
18856         PR middle-end/85450
18857         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18858         checking of integer<->pointer conversions.
18859         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18860         sign-/zero-extending pointer types.
18861         (expand_omp_for_static_chunk): Likewise.
18863 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
18864             Jean Lee  <xiaoyur347@gmail.com>
18866         * config/mips/mips.c (mips_asan_shadow_offset): New function.
18867         (TARGET_ASAN_SHADOW_OFFSET): Define.
18868         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18869         true for -fsanitize=address.
18871 2018-04-25  Mark Wielaard  <mark@klomp.org>
18873         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18874         shorter ones.
18876 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18878         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
18879         than "alu", remove explicit "memory" and "imm_disp" attributes.
18880         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
18882         PR middle-end/85414
18883         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
18884         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
18885         gen_lowpart_no_emit.
18887 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
18889         PR target/85473
18890         * config/i386/i386.c (ix86_expand_builtin): Change memory
18891         operand to XI, extend p0 to Pmode.
18892         * config/i386/i386.md: Change unspec volatile and operand
18893         1 mode to XI, change operand 0 mode to P.
18895 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18897         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
18898         GET_MODE_MASK before any checking.
18899         (nds32_can_use_bset_p): Likewise.
18900         (nds32_can_use_btgl_p): Likewise.
18902 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18904         * config/nds32/nds32-doubleword.md: New define_split pattern for
18905         illegal register number.
18907 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18909         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
18911 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18913         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
18915 2018-04-25  Richard Biener  <rguenther@suse.de>
18917         * lto-streamer.h (LTO_major_version): Bump to 8.
18919 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18921         * BASE-VER: Set to 9.0.0.
18923 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
18925         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
18926         in __abskf2 and __powikf2.
18928 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18930         PR target/85512
18931         * config/aarch64/constraints.md (Usg, Usj): New constraints.
18932         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
18933         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
18934         Use the above on operand 2.  Reindent.
18935         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
18937 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18939         PR target/85485
18940         * common/config/i386/i386-common.c (ix86_handle_option): Don't
18941         handle OPT_mcet.
18942         * config/i386/i386.opt (mcet): Removed.
18943         * doc/install.texi: Remove -mcet documentation.
18944         * doc/invoke.texi: Likewise.
18946 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18948         PR target/85485
18949         * doc/install.texi: Remove -mcet from bootstrap-cet.
18951 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18953         PR target/85511
18954         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
18955         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
18956         if TARGET_64BIT.
18958         PR target/85503
18959         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
18960         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
18961         containing a CONST_VECTOR.
18963 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
18965         * doc/install.texi: Update newlib dependency for nvptx.
18967 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18969         PR target/85508
18970         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
18971         instead of INTVAL when shifting x left.
18973 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
18975         PR tree-optimization/85478
18976         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
18977         vect_grouped_store_supported for single element vectors.
18979 2018-04-24  Richard Biener  <rguenther@suse.de>
18981         PR target/85491
18982         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
18983         load cost increase to the case of non-constant step.
18985 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18987         PR target/84828
18988         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
18989         destination if any_malformed_asm.
18991 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
18993         PR middle-end/85496
18994         * expr.c (store_field): In the bitfield case, if the value comes from
18995         a function call and is returned in registers by means of a PARALLEL,
18996         do not change the mode of the temporary unless BLKmode and VOIDmode.
18998 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
19000         PR rtl-optimization/85423
19001         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
19002         dependencies to debug insns when the previous insn is non-debug.
19004 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
19006         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
19007         enums into a single definition.
19008         (fls): Fix predicates and printing.
19009         (seti): Likewise.
19011 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
19013         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
19014         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
19015         and short u6 immediate.
19016         (check_if_valid_sleep_operand): Remove.
19017         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
19019 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19021         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
19022         flag_always_save_lp condition.
19023         * config/nds32/nds32.opt (malways-save-lp): New option.
19025 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19027         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
19028         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
19029         * config/nds32/nds32.h
19030         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
19031         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
19033 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19035         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
19036         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
19038 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19039             Chung-Ju Wu  <jasonwucj@gmail.com>
19041         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
19042         Declare.
19043         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
19044         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
19046 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19048         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
19050 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19052         * config/nds32/nds32-protos.h (nds32_data_alignment,
19053         nds32_local_alignment): Declare.
19054         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
19055         nds32_local_alignment): New functions.
19056         (TARGET_CONSTANT_ALIGNMENT): Define.
19057         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
19059 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19061         * config/nds32/nds32.c
19062         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
19063         (TARGET_MODES_TIEABLE_P): Likewise.
19065 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19067         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
19068         level Ofast and Og.
19070 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
19071             Chung-Ju Wu  <jasonwucj@gmail.com>
19073         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19074         for unaligned access.
19075         * config/nds32/nds32-intrinsic.c: Implementation of expanding
19076         unaligned access.
19077         * config/nds32/nds32-intrinsic.md: Likewise.
19078         * config/nds32/nds32_intrinsic.h: Likewise.
19079         * config/nds32/nds32.h (nds32_builtins): Likewise.
19080         * config/nds32/nds32.opt (munaligned-access): New option.
19081         * config/nds32/nds32.c (nds32_asm_file_start): Display
19082         flag_unaligned_access status.
19084 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
19086         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
19087         -mno-relax is present.
19088         * config/riscv/linux.h (LINK_SPEC): Ditto.
19090 2018-04-20  Martin Sebor  <msebor@redhat.com>
19092         PR c/85365
19093         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
19094         for null pointers.
19095         (gimple_fold_builtin_stxcpy_chk): Same.
19096         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
19098 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
19100         PR target/85456
19101         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
19102         __powikf2 when long double is IEEE 128-bit.
19104 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
19106         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
19107         step to make sure stack always aligned.
19109 2018-04-20  Carl Love  <cel@us.ibm.com>
19111         PR target/83402
19112         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
19113         size check for arg0.
19115 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
19116             Tom de Vries  <tom@codesourcery.com>
19118         PR target/85445
19119         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
19120         Emit insns for calls too.
19121         (nvptx_find_par): Always look for worker-level predecessor insn.
19122         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
19123         calls.
19124         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
19125         (nvptx_process_pars): Propagate frames for calls.
19127 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
19129         PR target/85469
19130         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
19131         Removed.
19132         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
19133         (ix86_handle_option): Don't handle OPT_mibt.
19134         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
19135         __SHSTK__.
19136         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
19137         has_ibt and ibt.
19138         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
19139         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
19140         (ix86_target_macros): Define __CET__ with flag_cf_protection
19141         for -fcf-protection.
19142         * config/i386/i386.c (isa2_opts): Remove -mibt.
19143         * config/i386/i386.h (TARGET_IBT): Removed.
19144         (TARGET_IBT_P): Likewise.
19145         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
19146         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
19147         * config/i386/i386.opt (mcet): Update help message.
19148         (mshstk): Likewise.
19149         (mibt): Removed.
19150         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
19151         -mcet as an alias for -mshstk.
19153 2018-04-20  Richard Biener <rguenther@suse.de>
19155         PR middle-end/85475
19156         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
19157         complexity by forcing a single use of the multiply operand.
19159 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19161         ipa/85449
19162         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
19163         recursion dependency to only apply to non-clones.
19165 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19167         ipa/85447
19168         * ipa-cp.c (create_specialized_node): Check that clones of
19169         self-recursive edges exist during IPA-CP.
19171 2018-04-19  Toon Moene  <toon@moene.org>
19173         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
19174         by -O3.
19176 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19178         PR tree-optimization/85467
19179         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
19180         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
19181         VECTOR_CST element to type.
19183 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19185         PR target/85397
19186         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
19187         * config/i386/i386.md (builtin_setjmp_setup): Removed.
19188         (builtin_longjmp): Likewise.
19189         (save_stack_nonlocal): New pattern.
19190         (restore_stack_nonlocal): Likewise.
19192 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19194         PR target/85404
19195         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19196         Replace ASM_OUTPUT_LABEL with fprintf.
19198 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19200         PR target/85417
19201         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19202         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
19203         * config/i386/i386-c.c (ix86_target_macros_internal): Also
19204         define __IBT__ and __SHSTK__ for -fcf-protection.
19205         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
19206         TARGET_IBT.
19207         (ix86_trampoline_init): Likewise.
19208         (x86_output_mi_thunk): Likewise.
19209         (ix86_notrack_prefixed_insn_p): Likewise.
19210         (ix86_option_override_internal): Don't disallow -fcf-protection.
19211         * config/i386/i386.md (rdssp<mode>): Also enable for
19212         -fcf-protection.
19213         (incssp<mode>): Likewise.
19214         (nop_endbr): Likewise.
19215         * config/i386/i386.opt (mcet): Change help message to built-in
19216         functions only.
19217         (mibt): Likewise.
19218         (mshstk): Likewise.
19219         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
19220         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
19221         enable CET built-in functions.
19223 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
19225         * common/config/i386/i386-common.c
19226         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
19227         OPTION_MASK_ISA_MOVDIRI_UNSET,
19228         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
19229         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
19230         * config.gcc (movdirintrin.h): New header.
19231         * config/i386/cpuid.h (bit_MOVDIRI,
19232         bit_MOVDIR64B): New bits.
19233         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
19234         and -mmvodir64b.
19235         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
19236         (VOID, PVOID, PCVOID)): New function types.
19237         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
19238         __builtin_ia32_directstoreu_u64,
19239         __builtin_ia32_movdir64b): New builtins.
19240         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
19241         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
19242         and -mmovdiri.
19243         (ix86_valid_target_attribute_inner_p): Ditto.
19244         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
19245         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
19246         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
19247         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
19248         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
19249         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
19250         (movdiri<mode>, movdir64b_<mode>): New.
19251         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
19252         * config/i386/immintrin.h: Include movdirintrin.h.
19253         * config/i386/movdirintrin.h: New file.
19254         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
19256 2018-04-19  Richard Biener  <rguenther@suse.de>
19258         PR middle-end/85455
19259         * cfg.c (clear_bb_flags): When loop state says we have
19260         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
19262 2018-04-19  Richard Biener  <rguenther@suse.de>
19264         PR tree-optimization/84737
19265         * tree-vect-data-refs.c (vect_copy_ref_info): New function
19266         copying restrict info.
19267         (vect_setup_realignment): Use it.
19268         * tree-vectorizer.h (vect_copy_ref_info): Declare.
19269         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
19270         the first DR to all generated stores.
19271         (vectorizable_load): Likewise for loads.
19273 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19275         PR tree-optimization/85446
19276         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
19277         the integral and pointer types to have the same precision.
19279         * doc/install.texi: Document --disable-cet being the default and
19280         --enable-cet=auto.
19282 2018-04-18  Martin Liska  <mliska@suse.cz>
19284         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
19285         style.
19287 2018-04-18  Martin Liska  <mliska@suse.cz>
19289         Revert
19290         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
19292         PR ipa/83983
19293         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19294         arguments if they are comparable.
19296 2018-04-18  Martin Liska  <mliska@suse.cz>
19298         Revert
19299         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19301         PR lto/84805
19302         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19303         incomplete types.
19305 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
19307         PR target/85388
19308         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
19309         ENDBR after calling __morestack.
19311 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
19313         PR jit/85384
19314         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
19315         by using gcc_base_ver to generate a gcc_driver_version, and use
19316         it when generating GCC_DRIVER_NAME.
19317         * configure: Regenerate.
19319 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
19321         PR target/81084
19322         * config.gcc: Obsolete powerpc*-*-*spe*.
19324 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19326         PR debug/84637
19327         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
19328         (stabstr_D): Change type of unum from unsigned int to
19329         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
19330         type.
19332 2018-04-17  Jim Wilson  <jimw@sifive.com>
19334         PR 84856
19335         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
19336         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
19337         Set arg_pointer_offset after using pretend_args_size.
19339 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19341         PR rtl-optimization/85431
19342         * dse.c (record_store): Ignore zero width stores.
19344         PR sanitizer/85230
19345         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
19346         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
19347         __builtin_stack_restore rather than after it.
19348         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
19349         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
19350         argument instead of virtual_dynamic_stack_rtx.
19352 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19354         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
19355         New prototype.
19356         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19357         Add note to error message to explain internal mapping of overloaded
19358         built-in function name to non-overloaded built-in function name.
19359         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
19360         function.
19362 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
19364         PR target/85424
19365         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
19366         where the inputs overlap with the output.
19368 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19370         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
19371         (=v, v) alternative and explicit "memory" attribute.
19372         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
19373         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19374         attributes.
19375         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
19376         "sselog1" type instead of "sselog".
19377         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
19378         "sselog".  Remove explicit "memory" attribute.
19379         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
19380         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19381         attributes.
19382         (vec_extract_hi_v32hi): Merge all alternatives into one, use
19383         "sselog1" type instead of "sselog".  Remove explicit "memory"
19384         attribute.
19385         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
19386         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19387         attribute.
19388         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
19389         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19390         attributes.
19391         (vec_extract_hi_v64qi): Merge all alternatives into one, use
19392         "sselog1" type instead of "sselog".  Remove explicit "memory"
19393         attribute.
19394         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
19395         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19396         attribute.
19398         PR target/85430
19399         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
19401         PR middle-end/85414
19402         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
19403         on a SUBREG.
19405 2018-04-17  Martin Jambor  <mjambor@suse.cz>
19407         PR ipa/85421
19408         * ipa-cp.c (create_specialized_node): Call
19409         expand_all_artificial_thunks if necessary.
19411 2018-04-17  Martin Liska  <mliska@suse.cz>
19413         PR lto/85405
19414         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
19415         in message, remote space in between '_G' and '('.
19417 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19419         PR target/85281
19420         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
19421         avx512f_vmcmp<mode>3<round_saeonly_name>,
19422         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
19423         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
19424         avx512f_rndscale<mode><round_saeonly_name>,
19425         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
19426         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
19427         Use %<iptr>2 instead of %2 for -masm=intel.
19428         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
19429         avx512f_vcvttss2usi<round_saeonly_name>,
19430         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
19431         -masm=intel.
19432         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
19433         avx512f_vcvttsd2usi<round_saeonly_name>,
19434         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
19435         Use %q1 instead of %1 for -masm=intel.
19436         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
19437         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
19438         of %3 for -masm=intel.
19439         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
19440         -masm=intel.
19441         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
19442         -masm=intel.
19443         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
19444         -masm=intel.
19445         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
19446         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
19447         %g1.
19448         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
19449         -masm=intel.
19450         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
19451         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
19452         %g1 and one with %0 and %1.
19453         (avx512er_vmrcp28<mode><round_saeonly_name>,
19454         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
19455         %1 for -masm=intel.
19456         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
19457         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
19458         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
19459         of %0 and %{%4%} for -masm=intel.
19460         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
19461         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
19462         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
19463         order of %0 and %{%5%}%{z%} for -masm=intel.
19465 2018-04-17  Jan Hubicka  <jh@suse.cz>
19467         PR lto/85405
19468         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
19470 2018-04-17  Martin Liska  <mliska@suse.cz>
19472         PR ipa/85329
19473         * multiple_target.c (create_dispatcher_calls): Set apostrophes
19474         for target_clone error message.  Make default implementation
19475         clone to be a local declaration.
19476         (separate_attrs): Add new argument and check for an empty
19477         string.
19478         (expand_target_clones): Handle it.
19479         (ipa_target_clone): Make redirection just for target_clones
19480         functions.
19482 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
19483             Tom de Vries  <tom@codesourcery.com>
19485         PR middle-end/84955
19486         * omp-expand.c (expand_oacc_for): Add dummy false branch for
19487         tiled basic blocks without omp continue statements.
19489 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
19491         PR target/83660
19492         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19493         vec_extract expression as having side effects to make sure it gets
19494         a cleanup point.
19496 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
19498         PR target/85403
19499         * config/i386/i386.c (get_builtin_code_for_version): Check
19500         error_mark_node.
19502 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
19504         PR target/84331
19505         * config.gcc: Support "skylake".
19506         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19507         PROCESSOR_SKYLAKE.
19508         * config/i386/i386.c (m_SKYLAKE): Define.
19509         (processor_target_table): Add "skylake".
19510         (ix86_option_override_internal): Add "skylake".
19511         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19512         PROCESSOR_CANNONLAKE.
19513         (get_builtin_code_for_version): Fix priority for
19514         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19515         PROCESSOR_SKYLAKE-AVX512.
19516         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19517         (processor_type): Add PROCESSOR_SKYLAKE.
19519 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
19520             Jason Merrill  <jason@redhat.com>
19522         PR c++/85112
19523         * convert.c (convert_to_integer_1): Use direct recursion for
19524         enumeral types and types with a precision less than the number
19525         of bits in their mode.
19527 2018-04-16  Julia Koval  <julia.koval@intel.com>
19529         PR target/84413
19530         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19531         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19533 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
19535         PR target/85293
19536         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19537         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19538         and -mno-direct-move.
19540 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
19542         PR target/83402
19543         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19544         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19545         Ensure negative shifts result in {0}.
19547 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
19549         PR rtl-optimization/79916
19550         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19551         regs (if any) to define how to gnerate SD moves when LRA is in
19552         progress.
19554 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
19556         PR rtl-optimization/85393
19557         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19558         * except.c (expand_dw2_landing_pad_for_region): Make static.
19559         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19560         a label and unconditional jump to old_bb, rather than
19561         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19562         basic block.
19564         PR rtl-optimization/85376
19565         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19566         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19567         instead of a specific value.
19569 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19570             Bin Cheng  <bin.cheng@arm.com>
19572         PR tree-optimization/82965
19573         PR tree-optimization/83991
19574         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19575         by_profile_only parameter.
19576         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19577         information if the loop was predicted to iterate too many times.
19578         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19580 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19582         PR lto/71991
19583         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19584         always inline.
19586 2018-04-13  Martin Liska  <mliska@suse.cz>
19587             Jakub Jelinek  <jakub@redhat.com>
19589         PR middle-end/81657
19590         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19591         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19592         * builtins.c (expand_builtin_memory_copy_args): Use
19593         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19594         handle dest_addr == pc_rtx.
19596 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
19598         PR target/85291
19599         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19600         asked to not generate direct moves.
19601         (fix_trunc<mode>si2_stfiwx): Similar.
19602         (fix_trunc<mode>si2_internal): Similar.
19604 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19606         PR debug/83157
19607         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19608         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19609         lookup if dest in some wider mode is known to be const0_rtx and
19610         if so, record permanent equivalence for it to be ZERO_EXTEND of
19611         the narrower mode destination.
19613 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19615         * lto-streamer-out.c (output_function): Revert 259346.
19616         * omp-expand.c (expand_oacc_for): Likewise.
19618 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
19620         PR rtl-optimization/85354
19621         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19622         * sel-sched.c (sel_global_init): ... here.
19624 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
19626         PR target/85238
19627         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19628         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19629         mode for PE-COFF targets.
19630         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19631         (i386_pe_asm_lto_end): Likewise.
19632         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19633         (TARGET_ASM_LTO_END): Likewise.
19634         * config/i386/winnt.c (saved_debug_info_level): New static variable.
19635         (i386_pe_asm_lto_start): New function.
19636         (i386_pe_asm_lto_end): Likewise.
19638 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19639             Richard Biener  <rguenther@suse.de>
19641         PR middle-end/84955
19642         * lto-streamer-out.c (output_function): Fix CFG loop state before
19643         streaming out.
19644         * omp-expand.c (expand_oacc_for): Handle calls to internal
19645         functions like regular functions.
19647 2018-04-12  Richard Biener  <rguenther@suse.de>
19649         PR lto/85371
19650         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19651         for the early LTO debug to properly generate references to it
19652         during DIE emission.  Do not re-use that for the skeleton for
19653         split-dwarf.
19654         (dwarf2out_early_finish): Likewise.
19656 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19658         PR target/85328
19659         * config/i386/sse.md
19660         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19661         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19662         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19663         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19664         and output is a reg, avoid creating invalid lowpart subreg, but
19665         instead split into a 512-bit move.  Don't split if not AVX512VL,
19666         input is xmm16+ reg and output is a mem.
19667         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19668         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19669         xmm16+ reg and output is a mem.
19671 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19673         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19674         also for flag_dwarf2_cfi_asm.
19676 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19678         PR rtl-optimization/85342
19679         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19680         a bool scalar var inside of the loop instead.  Don't try to update
19681         recog_data.operand after failed apply_change_group.
19683 2018-04-12  Tom de Vries  <tom@codesourcery.com>
19685         PR target/85296
19686         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19687         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
19688         array with flexible array member as array without given dimension.
19689         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19690         argument for undefined param to true.
19692 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
19694         PR target/85321
19695         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19696         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19697         from PowerPC section.
19698         * config/rs6000/sysv4.opt (mcall-): Improve help text.
19699         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19700         help text that is too long.
19701         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19702         help text that is too long.
19703         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19704         help text that is too long.
19706 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
19708         * config/alpha/alpha.md (stack_probe_internal): Rename
19709         from "probe_stack".  Update all callers.
19711 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19713         PR rtl-optimization/84566
19714         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19715         sched_macro_fuse_insns.
19717 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19719         PR target/84301
19720         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19721         (compute_block_dependences): ... from here.
19723 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19725         PR tree-optimization/85331
19726         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19727         from int to HOST_WIDE_INT.
19729 2018-04-11  Martin Jambor  <mjambor@suse.cz>
19731         PR ipa/84149
19732         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19733         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19734         not the same as the source val.
19735         (cgraph_edge_brings_value_p): New parameter.
19736         (gather_edges_for_value): Pass destination value to
19737         cgraph_edge_brings_value_p.
19738         (perhaps_add_new_callers): Likewise.
19739         (get_info_about_necessary_edges): Likewise and exclude values brought
19740         only by self-recursive edges.
19741         (create_specialized_node): Redirect only clones of self-calling edges.
19742         (+self_recursive_pass_through_p): New function.
19743         (find_more_scalar_values_for_callers_subset): Use it.
19744         (find_aggregate_values_for_callers_subset): Likewise.
19745         (known_aggs_to_agg_replacement_list): Removed.
19746         (decide_whether_version_node): Re-calculate known constants for all
19747         remaining context clones.
19749 2018-04-11  Richard Biener  <rguenther@suse.de>
19751         PR lto/85339
19752         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19753         from early DWARF output.
19754         (dwarf2out_early_finish): Output line info unconditionally into
19755         early DWARF and add reference to it.
19757 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19759         PR target/85281
19760         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19761         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19762         other than V2DFmode using iptr mode attribute.
19763         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19765 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19767         PR rtl-optimization/84659
19768         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19770 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19772         PR debug/85302
19773         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19774         SIZEP is NULL.
19775         (output_loc_list): Pass address of a dummy size variable even in the
19776         locview handling loop.
19777         (index_location_lists): Add comment on why skip_loc_list_entry can't
19778         call size_of_locs.
19780 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19782         PR target/85261
19783         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19784         into register.
19786 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
19788         PR target/85321
19789         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19790         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19791         and -mstring-compare-inline-limit.
19793 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19795         PR target/85287
19796         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19797         for stack clash protection in a register whenever we need it to be in
19798         a register.
19800 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19802         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19803         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19805 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19807         PR target/85321
19808         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19809         the help text.
19810         (mlong-double-): Ditto.
19811         * config/rs6000/sysv4.opt (msdata=): Ditto.
19812         (mtls-size=): Ditto.
19814 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19816         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19817         erroneous entries for
19818         "vector int vec_ldl (int, long int *)", and
19819         "vector unsigned int vec_ldl (int, unsigned long int *)".
19820         Add comments and entries for
19821         "vector bool char vec_ldl (int, bool char *)",
19822         "vector bool short vec_ldl (int, bool short *)",
19823         "vector bool int vec_ldl (int, bool int *)",
19824         "vector bool long long vec_ldl (int, bool long long *)",
19825         "vector pixel vec_ldl (int, pixel *)",
19826         "vector long long vec_ldl (int, long long *)",
19827         "vector unsigned long long vec_ldl (int, unsigned long long *)".
19828         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19829         type tree bool_long_long_type_node and correct definition of
19830         bool_V2DI_type_node to make reference to this new type tree.
19831         (rs6000_mangle_type): Replace erroneous reference to
19832         bool_long_type_node with bool_long_long_type_node.
19833         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19834         comments to emphasize sign distinctions for char and int types and
19835         replace RS6000_BTI_bool_long constant with
19836         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
19837         use of RS6000_BTI_pixel.
19838         (bool_long_type_node): Remove this macro definition.
19839         (bool_long_long_type_node): New macro definition
19841 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19843         PR rtl-optimization/85300
19844         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19845         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19846         simplify_unary_operation fails.
19848 2018-04-10  Martin Liska  <mliska@suse.cz>
19850         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19851         cgraph_edge and ipa_ref.
19853 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19855         PR target/85177
19856         PR target/85255
19857         * config/i386/sse.md
19858         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19859         computation of the VEC_MERGE selector from mask.
19860         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19861         Fix decoding of the VEC_MERGE selector into mask.
19863 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19865         PR tree-optimization/85286
19866         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19868 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19870         * final.c (final_1): Set insn_last_address as well as
19871         insn_current_address.
19873 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19875         PR target/85173
19876         * explow.c (emit_stack_probe): Call validize_mem on memory location
19877         before passing it to gen_probe_stack.  Create address operand and
19878         legitimize it for the probe_stack_address case.
19880 2018-04-09  Jan Hubicka  <jh@suse.cz>
19882         PR lto/85078
19883         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
19884         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
19885         * tree.c (free_lang_data_in_type): Fix handling of binfos;
19886         walk basetypes.
19887         (free_lang_data): Rebuild type inheritance graph.
19889 2018-04-09  Martin Sebor  <msebor@redhat.com>
19891         * invoke.texi (-finline-small-functions): Mention other optimization
19892         options.
19893         (-findirect-inlining, -fpartial-inlining): Same.
19894         (-finline-functions-called-once): Same.
19895         (-freorder-blocks-and-partition): Same.
19897 2018-04-09  Jan Hubicka  <jh@suse.cz>
19899         PR rtl/84058
19900         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
19901         jumps; choose last target that matches the criteria (i.e.
19902         no partition changes for non-crossing jumps).
19903         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
19904         support for redirecting crossing jumps to non-crossing.
19906 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
19908         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
19909         also for naked functions.
19911 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
19913         * config/arc/arc.md (add_shift): New pattern.
19914         (add_shift2): Likewise.
19915         (sub_shift): Likewise.
19916         (sub_shift_cmp0_noout): Likewise.
19917         (compare_si_ashiftsi): Likewise.
19918         (xbfu_cmp0_noout): New combine pattern.
19919         (xbfu_cmp0"): Likewise.
19920         (movsi_set_cc_insn): Place the predicable variant first.
19921         (commutative_binary_cmp0_noout): Remove clobber.
19922         (commutative_binary_cmp0): New pattern.
19923         (noncommutative_binary_cmp0): Likewise.
19924         (noncommutative_binary_cmp0_noout): Likewise.
19925         (noncommutative_binary_comparison_result_used): Removed.
19926         (rsub_cmp0): New pattern.
19927         (rsub_cmp0_noout): Likewise.
19928         (extzvsi): Changed, keep only meaningful variants.
19929         (SQH, SEZ): New iterators.
19930         (SQH_postfix): New mode attribute.
19931         (SEZ_prefix): New code attribute.
19932         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
19933         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
19934         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
19935         of numerical value.
19936         (noncommutative_operator): Check the availability of barrel
19937         shifter option.
19939 2018-04-09  Richard Biener  <rguenther@suse.de>
19941         PR tree-optimization/85284
19942         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
19943         Only use the niter constraining form of simple_iv when the exit
19944         is always executed.
19946 2018-04-09  Tom de Vries  <tom@codesourcery.com>
19948         PR target/84041
19949         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19950         (define_expand "*memory_barrier"): New define_expand.
19951         (define_insn "memory_barrier"): New insn.
19953 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19955         PR rtl-optimization/80463
19956         PR rtl-optimization/83972
19957         PR rtl-optimization/83480
19959         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
19960         correct producer for the insn.
19961         (tidy_control_flow): Fixup seqnos in case of debug insns.
19963 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19965         PR rtl-optimization/83913
19967         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
19968         different sched-times when merging exprs.
19970 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19972         PR rtl-optimization/83962
19974         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
19975         tidy_fallthru_edge and tidy_control_flow.
19977 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19979         PR rtl-optimization/83530
19981         * sel-sched.c (force_next_insn): New global variable.
19982         (remove_insn_for_debug): When force_next_insn is true, also leave only
19983         next insn in the ready list.
19984         (sel_sched_region): When the region wasn't scheduled, make another pass
19985         over it with force_next_insn set to 1.
19987 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19989         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
19990         into tm_file.
19991         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19992         for interrupt control.
19993         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19994         functions for interrupt control.
19995         * config/nds32/nds32-intrinsic.md: Likewise.
19996         * config/nds32/nds32_intrinsic.h: Likewise.
19997         * config/nds32/nds32.h (nds32_builtins): Likewise.
19999 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
20001         * config/nds32/nds32.c (nds32_init_machine_status,
20002         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
20003         strict_aligned_p field.
20004         (nds32_expand_to_rtl_hook): New function.
20005         (TARGET_EXPAND_TO_RTL_HOOK): Define.
20006         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
20008 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20009             Chung-Ju Wu  <jasonwucj@gmail.com>
20011         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
20012         * config/nds32/nds32-n7.md: New file.
20013         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
20014         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
20015         pipeline.
20016         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
20017         * config/nds32/nds32.md (pipeline_model): Add n7.
20018         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
20019         * config/nds32/pipelines.md: Include n7 settings.
20021 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20022             Chung-Ju Wu  <jasonwucj@gmail.com>
20024         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
20025         * config/nds32/nds32-e8.md: New file.
20026         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
20027         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
20028         pipeline.
20029         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
20030         * config/nds32/nds32.md (pipeline_model): Add e8.
20031         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
20032         * config/nds32/pipelines.md: Include e8 settings.
20034 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20035             Chung-Ju Wu  <jasonwucj@gmail.com>
20037         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
20038         * config/nds32/nds32-n8.md: New file.
20039         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
20040         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
20041         pipeline.
20042         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
20043         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
20044         * config/nds32/nds32.md (pipeline_model): Add n8.
20045         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
20046         * config/nds32/pipelines.md: Include n8 settings.
20048 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20049             Chung-Ju Wu  <jasonwucj@gmail.com>
20051         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
20052         * config/nds32/nds32-n9-2r1w.md: New file.
20053         * config/nds32/nds32-n9-3r2w.md: New file.
20054         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
20055         nds32_register_ports): New or modify for cpu n9.
20056         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
20057         pipeline.
20058         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
20059         * config/nds32/nds32-utils.c: New file.
20060         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
20061         TARGET_MUL_SLOW): Define.
20062         * config/nds32/nds32.md (pipeline_model): New attribute.
20063         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
20064         New options that support cpu n9.
20065         * config/nds32/pipelines.md: Include n9 settings.
20066         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
20068 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
20070         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
20071         information if necessary.
20072         (output_cond_branch_compare_zero): Likewise.
20073         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
20074         (nds32_target_alignment): Refine for alignment.
20075         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
20076         (FUNCTION_BOUNDARY): Modify.
20077         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
20078         align case.
20079         * config/nds32/nds32.opt (malways-align, malign-functions): New.
20081 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
20083         * config/nds32/constants.md (unspec_volatile_element): Add values for
20084         TLB operation and data prefetch.
20085         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
20086         functions for TLB operation and data prefetch.
20087         * config/nds32/nds32-intrinsic.md: Likewise.
20088         * config/nds32/nds32_intrinsic.h: Likewise.
20089         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
20090         (nds32_print_operand): Likewise.
20091         * config/nds32/nds32.h (nds32_builtins): Likewise.
20093 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
20094         Andrew Pinski <pinsika@gcc.gnu.org>
20096         PR middle-end/82976
20097         * match.pd: Use constant_boolean_node of correct type instead of
20098         boolean_true_node or boolean_false_node for simplifying
20099         pointer comparisons to zero.
20101 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20103         PR tree-optimization/80021
20104         * tree.c (verify_type_variant): Make error call in verify_variant_match
20105         translatable and remove final full stop.
20107 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20109         * config/nds32/constants.md (unspec_volatile_element): Add
20110         UNSPEC_VOLATILE_EH_RETURN.
20111         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
20112         nds32_output_stack_pop): Support dwarf exception handling process.
20113         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
20114         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
20115         exception handling process.
20116         (nds32_compute_stack_frame): Likewise.
20117         (nds32_return_addr_rtx): Likewise.
20118         (nds32_initial_elimination_offset): Likewise.
20119         (nds32_expand_prologue): Likewise.
20120         (nds32_expand_epilogue): Likewise.
20121         (nds32_dynamic_chain_address): New function.
20122         * config/nds32/nds32.h (machine_function): Add fields for dwarf
20123         exception handling.
20124         (DYNAMIC_CHAIN_ADDRESS): Define.
20125         (EH_RETURN_DATA_REGNO): Define.
20126         (EH_RETURN_STACKADJ_RTX): Define.
20127         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
20128         patterns for dwarf exception handling.
20130 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20132         * config/nds32/nds32.h: Clean up obsolete macros.
20134 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20136         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20137         Add enum values for particular instructions.
20138         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20139         particular intrinsic functions.
20140         * config/nds32/nds32-intrinsic.md: Likewise.
20141         * config/nds32/nds32_intrinsic.h: Likewise.
20142         * config/nds32/nds32.h (nds32_builtins): Likewise.
20143         * config/nds32/nds32.md (type): Add pbsad and pbsada.
20144         (btst, ave): New patterns for particular instructions.
20146 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20148         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20149         Add enum values for atomic load/store and memory sync.
20150         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
20151         and memory sync.
20152         * config/nds32/nds32-intrinsic.md: Likewise.
20153         * config/nds32/nds32_intrinsic.h: Likewise.
20154         * config/nds32/nds32.h (nds32_builtins): Likewise.
20156 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20158         PR tree-optimization/85257
20159         * fold-const.c (native_encode_vector): If not all elts could fit
20160         and off is -1, return 0 rather than offset.
20161         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
20162         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
20163         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
20164         adjust buffer in native_interpret_expr call.
20166 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20168         * config/nds32/constants.md (unspec_volatile_element): Add cache
20169         control enum values.
20170         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
20171         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
20172         * config/nds32/nds32.c (nds32_cctl_names): New.
20173         (nds32_print_operand): Handle cache control register names.
20174         * config/nds32/nds32.h (nds32_builtins): New enum values.
20175         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
20176         macros.
20177         * config/nds32/nds32.md (type): Add mmu.
20178         * config/nds32/pipelines.md (simple_insn): Add mmu.
20180 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20182         * config/nds32/nds32.md (type): Remove call.
20183         * config/nds32/pipelines.md (simple_insn): Likewise.
20185 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20187         * config/nds32/constants.md (unspec_volatile_element): Add
20188         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
20189         UNSPEC_VOLATILE_FMFCFG.
20190         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
20191         description for fmfcfg and fmfcsr.
20192         (bdesc_1arg): Add fmtcsr.
20193         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
20194         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
20195         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
20196         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
20197         unspec_fmfcfg): New patterns.
20198         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
20199         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
20200         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
20201         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
20202         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
20203         __nds32__fmfcfg): Define.
20205 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20207         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
20208         intrinsic register names.
20209         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
20210         intrinsic register enum values and macros.
20212 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20214         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
20215         for load/store addressing form.
20216         (nds32_print_operand_address): Likewise.
20218 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
20220         PR target/85196
20221         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
20222         based on LABEL_REF.  Remove useless assertion.
20223         (pic_address_needs_scratch): Fix formatting.
20224         (sparc_legitimize_pic_address): Minor tweaks.
20225         (sparc_delegitimize_address): Adjust assertion accordingly.
20226         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
20227         into symbolic_operand.
20228         (movsi_high_pic_label_ref): Likewise.
20229         (movsi_lo_sum_pic_label_ref): Likewise.
20230         (movdi_pic_label_ref): Likewise.
20231         (movdi_high_pic_label_ref): Likewise.
20232         (movdi_lo_sum_pic_label_ref): Likewise.
20234 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
20236         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
20237         custom LIB_SPEC setup.
20239 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
20240             Kito Cheng  <kito.cheng@gmail.com>
20242         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
20243         * config/riscv/freebsd.h: New.
20245 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20247         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
20248         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
20249         file.
20251 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20252             Kito Cheng  <kito.cheng@gmail.com>
20254         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
20255         nds32_output_call, nds32_symbol_binds_local_p): New functions.
20256         * config/nds32/nds32-protos.h (nds32_output_call,
20257         nds32_output_return): Declare.
20258         * config/nds32/nds32.md: Refine all the call and return patterns.
20260 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20262         PR debug/85252
20263         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
20264         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
20266         PR rtl-optimization/84872
20267         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
20268         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
20269         EDGE_CROSSING edge.
20271 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
20273         * expr.c (copy_blkmode_to_reg): Revert 254862.
20274         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
20276 2018-04-06  Richard Biener  <rguenther@suse.de>
20278         PR middle-end/85244
20279         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
20280         after seeing a component reference with an adjacent field.  Treat
20281         refs to arrays at struct end of external decls similar to
20282         refs to unconstrained commons.
20284 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20286         PR sanitizer/85213
20287         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
20288         look through SAVE_EXPRs with non-side-effects argument.  Adjust
20289         recursive calls.
20290         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
20291         save_p here.
20293 2018-04-06  Richard Biener  <rguenther@suse.de>
20295         PR middle-end/85180
20296         * alias.c (find_base_term): New wrapper around find_base_term
20297         unwinding CSELIB_VAL_PTR changes.
20298         (find_base_term): Do not restore CSELIB_VAL_PTR during the
20299         recursion.
20301 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20303         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
20304         instructions.
20305         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
20306         constant definitions.
20307         ("nop"): lr 0,0 -> nopr r0
20308         ("nop_lr0", "nop_lr1"): New insn definitions.
20310 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20312         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
20313         NDS32_V3PUSH_AVAILABLE_P macro.
20315 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
20316             Chung-Ju Wu  <jasonwucj@gmail.com>
20318         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
20319         (nds32*-*-*): Add float and fpu_config into supported_defaults.
20320         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
20321         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
20322         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
20323         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
20324         * config/nds32/constraints.md: New constraints and checking for hard
20325         float configuration.
20326         * config/nds32/iterators.md: New mode iterator and attribute for hard
20327         float configuration.
20328         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
20329         patterns.
20330         * config/nds32/nds32-fpu.md: New file.
20331         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
20332         deal with hard float code generation.
20333         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
20334         ARCH_V3S.
20335         (abi_type, float_reg_number): New enum type.
20336         * config/nds32/nds32-predicates.c: New predicates for hard float.
20337         * config/nds32/nds32-protos.h: Declare functions for hard float.
20338         * config/nds32/nds32.c: Implementation for hard float configuration.
20339         * config/nds32/nds32.h: Definitions for hard float configuration.
20340         * config/nds32/nds32.md: Include hard float machine description and
20341         modify patterns for hard float configuration.
20342         * config/nds32/nds32.opt: New options for hard float configuration.
20343         * config/nds32/predicates.md: New predicates for hard float
20344         configuration.
20346 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20348         * common/config/nds32/nds32-common.c
20349         (nds32_option_optimization_table): Enable -mreleax-hint by default.
20351 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
20353         PR middle-end/85195
20354         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
20355         CONSTRUCTOR_ELT (ctor, ...)->value.
20357 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
20359         PR target/85193
20360         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
20362 2018-04-05  Tom de Vries  <tom@codesourcery.com>
20364         PR target/85204
20365         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
20366         cond jump.
20368 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20369             Kito Cheng  <kito.cheng@gmail.com>
20371         * config/nds32/constraints.md (U33): Fine-tune checking condition.
20372         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
20373         * config/nds32/nds32.h (nds32_16bit_address_type): Add
20374         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
20376 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20377             Kito Cheng  <kito.cheng@gmail.com>
20379         * config/nds32/constraints.md (Ufe): New memory constraint.
20380         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
20381         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
20382         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
20383         operands.
20384         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
20385         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
20387 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20389         * config/nds32/nds32.md: Use optimize_size in the condition for
20390         alu-shift instructions.
20392 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20394         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
20396 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20398         * config/nds32/nds32.md (negsi2): Refine pattern.
20400 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20401             Chung-Ju Wu  <jasonwucj@gmail.com>
20403         * config/nds32/iterators.md (shift_rotate): New code iterator.
20404         (shift): New code attribute.
20405         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
20406         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
20407         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
20408         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
20409         bit-wise operations.
20410         (andsi3, *andsi3): Ditto.
20411         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
20412         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
20413         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
20414         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
20415         nds32_ior_operand, nds32_xor_operand): New predicates.
20417 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20419         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
20420         (addsi3, subsi3): ... this.
20422 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20424         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
20426 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20428         * config/nds32/nds32.md: Adjust indention.
20430 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20432         * config/nds32/nds32.md (feature): New attribute.
20434 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20436         * config/nds32/nds32.md (subtype): New attribute.
20438 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20440         PR target/85203
20441         * config/arm/arm-builtins.c (arm_expand_builtin): Change
20442         expansion to perform a bitwise AND of the argument followed by a
20443         boolean negation of the result.
20445 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
20447         PR rtl-optimization/84878
20448         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
20449         the basic block.  Assert the use reference is not artificial and that
20450         it has an associated insn.
20452 2018-04-04  Michael Matz  <matz@suse.de>
20454         * builtins.c (compute_objsize): Pass correct operand
20455         to array_at_struct_end_p.
20457 2018-04-04  Richard Biener  <rguenther@suse.de>
20459         PR lto/85176
20460         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
20461         from contexts for DINFO_LEVEL_TERSE and below.
20463 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20465         * config/nds32/nds32-doubleword.md (move_<mode>): Require
20466         resiter_operand condition.
20467         * config/nds32/nds32.md (*move<mode>): Ditto.
20469 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20470             Monk Chiang  <sh.chiang04@gmail.com>
20472         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
20474 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20476         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
20478 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20479             Kito Cheng  <kito.cheng@gmail.com>
20481         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
20482         nds32_cond_code_str, output_cond_branch,
20483         output_cond_branch_compare_zero, nds32_expand_cbranch,
20484         nds32_expand_cstore, nds32_expand_movcc,
20485         nds32_output_cbranchsi4_equality_zero,
20486         nds32_output_cbranchsi4_equality_reg,
20487         nds32_output_cbranchsi4_equality_reg_or_const_int,
20488         nds32_output_cbranchsi4_greater_less_zero: New functions.
20489         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
20490         nds32_expand_cstore, nds32_expand_movcc,
20491         nds32_output_cbranchsi4_equality_zero,
20492         nds32_output_cbranchsi4_equality_reg,
20493         nds32_output_cbranchsi4_equality_reg_or_const_int,
20494         nds32_output_cbranchsi4_greater_less_zero): Declare.
20495         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20496         nds32_rimm11s_operand): New predicates.
20497         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20498         * config/nds32/nds32.md: Rewrite all the branch and conditional move
20499         patterns.
20501 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20503         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20504         * config/nds32/nds32.md: Ditto.
20505         * config/nds32/pipelines.md: Ditto.
20507 2018-04-04  Richard Biener  <rguenther@suse.de>
20509         PR tree-optimization/85168
20510         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20511         propagating abnormals.
20513 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20515         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20517 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20518             Kito Cheng  <kito.cheng@gmail.com>
20520         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20521         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20522         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20523         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20524         * config/nds32/nds32.md (sibcall_internal): New.
20525         (sibcall_register): Remove.
20526         (sibcall_immediate): Remove.
20527         (sibcall_value_internal): New.
20528         (sibcall_value_register): Remove.
20529         (sibcall_value_immediate): Remove.
20530         * config/nds32/predicates.md (nds32_general_register_operand): New.
20531         (nds32_call_address_operand): New.
20533 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20535         PR rtl-optimization/85167
20536         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20537         bb_defs if *split_p, instead preinitialize it to NULL.
20539         PR tree-optimization/85156
20540         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20541         evaluating the argument multiple times.
20543 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
20545         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20546         than vector.
20547         (_mm_cvtpd_ps): Likewise.
20548         (_mm_cvttpd_epi32): Likewise.
20549         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20550         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20551         vector, pixel, and bool following altivec.h include.
20553 2018-04-03  Martin Sebor  <msebor@redhat.com>
20555         * doc/extend.texi (Common Function Attributes): Clarify.
20556         (const attribute): Likewise.
20557         (pure attribute): Likewise.
20559 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20561         PR target/85169
20562         * config/i386/i386.c (ix86_expand_vector_set): Use
20563         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
20565 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
20567         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20568         instructions when changing rounding bits to preserve precision bits
20569         in the x87 control word.
20571 2018-04-03  Martin Liska  <mliska@suse.cz>
20573         PR tree-optimization/82491
20574         * rtl.h (strip_offset_and_add): Replace += suboffset with
20575         poly_uint64 () + suboffset.
20577 2018-03-29  Martin Liska  <mliska@suse.cz>
20578             Martin Jambor  <mjambor@suse.cz>
20580         PR ipa/84947
20581         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20582         param_type is not an integral or pointer type.
20584 2018-04-03  Richard Biener  <rguenther@suse.de>
20586         * sese.h (recompute_all_dominators): Remove.
20588 2018-04-02  Martin Sebor  <msebor@redhat.com>
20590         * doc/invoke.texi (-Wrestrict): Fix typos.
20592 2018-04-02  Jim Wilson  <jimw@sifive.com>
20594         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20595         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20596         (<optab>di3, <optab>si3_extend): Likewise.
20597         (<optab>si3_mask, <optab>si3_mask_1): New.
20598         (<optab>di3_mask, <optab>di3_mask_1): New.
20599         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20600         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20601         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20603 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
20605         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20606         example.
20608 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20610         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20611         (nds32_canonicalize_comparison): New function.
20613 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20614             Kito Cheng  <kito.cheng@gmail.com>
20615             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20617         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20618         * config/nds32/constants.md (unspec_volatile_element): Add
20619         UNSPEC_VOLATILE_RELAX_GROUP.
20620         * config/nds32/nds32-relax-opt.c: New file.
20621         * config/nds32/nds32-predicates.c
20622         (nds32_symbol_load_store_p): New function.
20623         * config/nds32/nds32-protos.h
20624         (nds32_symbol_load_store_p): Declare function.
20625         (make_pass_nds32_relax_opt): Declare new rtl pass function.
20626         * config/nds32/nds32.c
20627         (nds32_register_pass): New function to register pass.
20628         (nds32_register_passes): New function to register passes.
20629         * config/nds32/nds32.md (relax_group): New pattern.
20630         * config/nds32/nds32.opt (mrelax-hint): New option.
20631         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20633 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
20635         * config/nds32/t-nds32: Modify files dependency.
20637 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20639         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20640         (PROFILE_HOOK): Define its implementation.
20642 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20644         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20645         type and 32-bit size.
20647 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
20649         PR middle-end/85090
20650         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20651         (V_128_256): New mode iterator.
20652         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20653         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20654         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20655         of V.
20656         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20657         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20659 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
20661         PR target/83315
20662         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20663         NaN inputs correctly.
20665 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
20667         PR target/80546
20668         * config/rs6000/vsx.md (??r): New mode attribute.
20669         (*vsx_mov<mode>_64bit): Use it.
20670         (*vsx_mov<mode>_32bit): Likewise.
20672 2018-03-30  Martin Sebor  <msebor@redhat.com>
20674         PR tree-optimization/84818
20675         * builtins.c (check_access): Use warning_n.
20677 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20679         PR target/83822
20680         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20681         condition.
20682         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20683         condition.
20685 2018-03-30  Julia Koval  <julia.koval@intel.com>
20687         PR target/84413
20688         * x86-tune.def (movx, partial_reg_dependency): Enable for
20689         m_SKYLAKE_AVX512.
20691 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
20693         PR inline-asm/84985
20694         * lra-constraints.c (process_alt_operands): Move setting
20695         this_alternative_matches below.
20697 2018-03-29  Martin Liska  <mliska@suse.cz>
20699         PR lto/84995.
20700         * doc/invoke.texi: Document how LTO works with debug info.
20701         Describe auto-load support of binutils.  Mention 'x86-64'
20702         as valid option value of -march option.
20704 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
20706         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20708         PR c/85094
20709         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20710         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20711         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20712         checking.
20714 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20716         PR target/84912
20717         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20718         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20719         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20720         for RS6000_BTM_POWERPC64.
20721         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20722         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20723         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20724         definition.
20725         (DIVDE): Use it.
20726         (DIVDEU): Likewise.
20728 2018-03-28  Carl Love  <cel@us.ibm.com>
20730         Revert
20731         2017-09-27  Carl Love  <cel@us.ibm.com>
20733         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20734         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20735         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20736         fctiw instruction.
20738 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20740         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20741         instead of __vector bool.
20742         (_mm_max_pu8): Likewise.
20743         (_mm_min_pi16): Likewise.
20745 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20747         PR target/84912
20748         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20749         (DIVWEUO): Likewise.
20750         (DIVDEO): Likewise.
20751         (DIVDEUO): Likewise.
20752         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20753         DIVWEUO and DIVDEUO.
20754         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20755         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20756         (div_extend): Likewise.
20757         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20758         builtin function.
20759         (__builtin_divweuo): Likewise.
20760         (__builtin_divdeo): Likewise.
20761         (__builtin_divdeuo): Likewise.
20763 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20765         PR target/85095
20766         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20767         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20769         PR tree-optimization/82004
20770         * gimple-match-head.c (optimize_pow_to_exp): New function.
20771         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20772         Don't fold to exp if optimize_pow_to_exp is false.
20774 2018-03-28  Martin Liska  <mliska@suse.cz>
20776         PR other/84819
20777         * calls.c (initialize_argument_information): Fix trailing space.
20778         * common.opt: Fix typo and provide better explanation for
20779         -fsanitize-coverage option.
20780         * config/i386/i386.opt: Fix typo.
20782 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20783             Martin Liska  <mliska@suse.cz>
20785         PR sanitizer/85081
20786         * gimplify.c (asan_poison_variable): Don't do the check for
20787         gimplify_omp_ctxp here.
20788         (gimplify_decl_expr): Do it here.
20789         (gimplify_target_expr): Likewise.
20791 2018-03-28  Martin Liska  <mliska@suse.cz>
20793         PR target/84988
20794         * config/i386/i386.c (ix86_function_arg_advance): Do not call
20795         chkp_type_bounds_count if MPX is not enabled.
20797 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20799         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20801 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
20803         PR target/84914
20804         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20805         function to create the function decl for complex long double
20806         multiply and divide for -mabi=ieeelongdouble.
20807         (init_float128_ieee): Call it.
20809 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
20811         PR target/85044
20812         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20813         -fcf-protection=branch -mibt.
20814         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20816 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20818         PR target/81863
20819         * config/arm/arm.c (arm_valid_symbolic_address): Handle
20820         arm_word_relocations.
20822 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
20824         PR target/85056
20825         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20826         extern array declarations.
20828 2018-03-27  Richard Biener  <rguenther@suse.de>
20830         PR middle-end/84067
20831         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20832         explicit single_use checks.
20834 2018-03-27  Richard Biener  <rguenther@suse.de>
20836         PR tree-optimization/85082
20837         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20838         Valueize the VUSE.
20840 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20842         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20843         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20844         Turn on fasynchronous-unwind-tables and funwind-tables.
20846 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
20848         PR target/85073
20849         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20850         (*bmi_blsr_<mode>_ccz): Ditto.
20852 2018-03-26  Tom de Vries  <tom@codesourcery.com>
20854         PR tree-optimization/85063
20855         * omp-general.c (offloading_function_p): New function.  Factor out
20856         of ...
20857         * omp-offload.c (pass_omp_target_link::gate): ... here.
20858         * omp-general.h (offloading_function_p): Declare.
20859         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20860         with attribute omp declare target for offloading functions.
20862 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
20864         PR tree-optimization/84005
20865         * tree-data-ref.h (get_base_for_alignment): Declare.
20866         * tree-data-ref.c (get_base_for_alignment_1): New function.
20867         (get_base_for_alignment): Likewise.
20868         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20869         get_base_for_alignment to find a suitable base object, instead
20870         of always using drb->base_address.
20872 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20874         PR inline-asm/85022
20875         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
20876         known size by default.
20878 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
20880         PR inline-asm/85030
20881         * lra-constraints.c (process_alt_operands): Don't match BLKmode
20882         and non BLKmode operands.
20884 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20886         PR target/85026
20887         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
20888         Clean up attributes.
20890 2018-03-23  Richard Biener  <rguenther@suse.de>
20892         PR debug/85020
20893         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
20894         we are going to emit early debug for LTO.
20896 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20898         PR inline-asm/85034
20899         * function.c (match_asm_constraints_1): Don't optimize if input
20900         doesn't satisfy general_operand predicate for output's mode.
20902         PR inline-asm/85022
20903         * alias.c (write_dependence_p): Don't require for x_canonicalized
20904         non-VOIDmode if x has VOIDmode.
20906         PR sanitizer/85029
20907         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
20908         just don't try to optimize it rather than assert it never happens.
20910 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20912         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
20913         macro expansions for definition of ST_INTERNAL_<mode> and
20914         LD_INTERNAL_<mode> builtins.
20915         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
20916         Remove prototype.
20917         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
20918         function.
20919         (altivec_expand_st_builtin): Likewise.
20920         (altivec_expand_builtin): Remove calls to deleted functions.
20921         (rs6000_address_for_altivec): Delete this function.
20922         * config/rs6000/vector.md: Remove expands for
20923         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
20925 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
20927         PR target/84826
20928         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
20929         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
20930         re-computing once computed.
20931         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
20932         (arm_init_machine_status): Initialize
20933         machine->static_chain_stack_bytes.
20935 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20937         PR target/84760
20938         * doc/extend.texi: Add four new prototypes for vec_ld.
20939         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
20940         definitions for more logical presentation.
20941         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
20942         entries for V1TI variants of __builtin_altivec_ld builtin.
20943         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
20944         handling of V1TI variant of LVX icode pattern.
20945         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
20946         (rs6000_gimple_fold_builtin): Likewise.
20947         (altivec_init_builtins): Add code to define
20948         __builtin_altivec_lvx_v1ti function.
20950 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20952         PR inline-asm/84941
20953         * function.c (match_asm_constraints_1): Don't do the optimization
20954         if input isn't a REG, SUBREG, MEM or constant.
20956 2018-03-22  Tom de Vries  <tom@codesourcery.com>
20958         PR tree-optimization/84956
20959         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
20960         bb_has_abnormal_pred.
20962 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20964         PR sanitizer/85018
20965         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
20966         DECL_INITIAL (decl) to decl at the end.
20967         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
20968         adjust the comment.
20970 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
20972         * doc/extend.texi (__builtin_tgmath): Document when complex
20973         integer types are treated as _Complex _Float64.
20975 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20977         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
20979 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20981         PR tree-optimization/84960
20982         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
20983         if it is ENTRY block, move them into single succ of ENTRY in that case.
20985 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
20987         PR tree-optimization/84811
20988         * poly-int.h (poly_span_traits): Remove the T3 parameter and
20989         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
20990         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
20991         (known_subrange_p): Update accordingly.  Cast each value involved
20992         in the size comparison, rather than casting the result of the
20993         subtraction.
20995 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20997         PR tree-optimization/84982
20998         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
20999         by flipping the least significant bit rather than all bits from
21000         bitpos to bitpos + bitsize - 1.
21002 2018-03-21  Nathan Sidwell  <nathan@acm.org>
21004         * doc/extend.texi (Deprecated Features): Remove mention of
21005         long-deleted deprecations.
21007 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21009         PR jit/84288
21010         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
21011         * configure: Regenerate.
21013 2018-03-21  Tom de Vries  <tom@codesourcery.com>
21015         PR tree-optimization/83126
21016         * tree-parloops.c (num_phis): New function.
21017         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
21019 2018-03-21  Nathan Sidwell  <nathan@acm.org>
21021         * doc/extend.texi (Deprecated Features): Update deprecated flags,
21022         mention anon-struct/union members and trailing attributes.
21024 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
21026         PR tree-optimization/84969
21027         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
21028         builtin memset partitions if they set different rhs values.
21030 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
21032         PR rtl-optimization/84989
21033         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
21034         VEC_DUPLICATE with scalar result mode.
21036 2018-03-21  Martin Liska  <mliska@suse.cz>
21038         PR ipa/84963
21039         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
21040         not intended return statement.
21042 2018-03-21  Martin Liska  <mliska@suse.cz>
21044         PR target/84988
21045         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
21046         (chkp_find_bound_slots_1): Limit number of iterations.
21048 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
21050         PR target/84838
21051         * Minor grammar fixes for x86 options.
21053 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21055         PR debug/84875
21056         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
21057         holding REG_CFA_RESTORE notes, instead turn them into a USE.
21059 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
21061         PR target/83789
21062         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
21063         (altivec_lvx_<mode>_1op): Likewise.
21064         (altivec_stvx_<mode>_2op): Likewise.
21065         (altivec_stvx_<mode>_1op): Likewise.
21066         (altivec_lvx_<VM2:mode>): New define_expand.
21067         (altivec_stvx_<VM2:mode>): Likewise.
21068         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
21069         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21070         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
21071         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21072         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
21073         (rs6000_gen_lvx): Likewise.
21074         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
21075         (altivec_expand_stv_builtin): Likewise.
21076         (altivec_expand_builtin): Likewise.
21077         * config/rs6000/vector.md: Likewise.
21079 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21081         PR target/82518
21082         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
21083         BYTES_BIG_ENDIAN.
21085 2018-03-20  Richard Biener  <rguenther@suse.de>
21087         PR target/84986
21088         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
21089         sign-conversions as zero, fall back to standard scalar_stmt
21090         cost for the rest.
21092 2018-03-20  Martin Liska  <mliska@suse.cz>
21094         PR ipa/84825
21095         * predict.c (rebuild_frequencies): Handle case when we have
21096         PROFILE_ABSENT, but flag_guess_branch_prob is false.
21098 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21100         PR target/84990
21101         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
21102         flag_section_anchors.
21103         * varasm.c (use_blocks_for_decl_p): Remove hack for
21104         dw2_force_const_mem.
21106         PR target/84845
21107         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
21108         to ...
21109         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
21110         be created, use lowpart_subreg of operands[0] rather than operands[0]
21111         itself.
21112         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
21113         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
21114         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
21115         and n constraint instead of aarch64_shift_imm_di and Usd.
21116         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
21117         (*aarch64_<optab>_reg_minus<mode>3): ... this.
21119 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
21121         PR target/82989
21122         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
21123         to favor GPR over NEON registers.
21124         (<shift>di3_neon): Likewise.
21126 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21128         PR target/84952
21129         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
21130         (nvptx_process_pars): Emit bar.sync asap and alap.
21132 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21134         PR target/84954
21135         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
21136         seen_label if seen_label is already set.
21138 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21140         PR target/84945
21141         * config/i386/i386.c (fold_builtin_cpu): For features above 31
21142         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
21143         Use 1U instead of 1.  Formatting fixes.
21145         PR c/84953
21146         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
21147         instead of TREE_TYPE (s1) for the return value.
21149 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
21151         PR tree-optimization/84946
21152         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
21153         bitsize + bitsize in poly_uint64 rather than poly_int64.
21155         PR sanitizer/78651
21156         * dwarf2asm.c: Include fold-const.c.
21157         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
21158         of decl rather than decl itself.
21160         PR rtl-optimization/84643
21161         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
21163 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
21165         PR sanitizer/78651
21166         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
21167         calling assemble_variable.
21169 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
21171         PR target/81647
21172         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
21173         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
21175 2018-03-19  Jim Wilson  <jimw@sifive.com>
21177         PR bootstrap/84856
21178         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
21179         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
21180         (riscv_first_stack_step): Likewise.
21181         (riscv_option_override): Use STACK_BOUNDARY instead of
21182         MIN_STACK_BOUNDARY.
21183         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
21184         MIN_STACK_BOUNDARY.
21185         (BIGGEST_ALIGNMENT): Set to 128.
21186         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
21187         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
21188         STACK_BOUNDARY.
21190 2018-03-19  Richard Biener  <rguenther@suse.de>
21192         PR tree-optimization/84933
21193         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
21194         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
21196 2018-03-19  Richard Biener  <rguenther@suse.de>
21198         PR tree-optimization/84859
21199         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
21200         (cond_if_else_store_replacement): Perform sinking operation on
21201         single-store BBs regardless of MAX_STORES_TO_SINK setting.
21202         Generalize what a BB with a single eligible store is.
21204 2018-03-19  Richard Biener  <rguenther@suse.de>
21206         PR tree-optimization/84929
21207         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
21208         chrec_is_positive against non-chrec arg.
21210 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
21212         PR target/84711
21213         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
21215 2018-03-18  Martin Liska  <mliska@suse.cz>
21217         PR rtl-optimization/84635
21218         * regrename.c (build_def_use): Use matches_mode only when
21219         matches >= 0.
21221 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
21223         PR tree-optimization/84913
21224         * tree-vect-loop.c (vectorizable_reduction): Don't try to
21225         vectorize chains of COND_EXPRs.
21227 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21229         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
21231 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21233         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
21235 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21237         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21239 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21240             Kito Cheng  <kito.cheng@gmail.com>
21242         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
21243         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
21244         (nds32_adjust_reg_alloc_order): New function.
21245         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
21247 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
21249         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
21250         nds32_print_operand, nds32_print_operand_address): Use
21251         HOST_WIDE_INT_PRINT_DEC instead.
21253 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21255         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
21257 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
21259         PR target/84902
21260         * config/i386/i386.c (initial_ix86_tune_features,
21261         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
21262         unsigned long long.
21263         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
21264         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
21265         rather than 1u << ix86_tune.  Formatting fix.
21266         (ix86_option_override_internal): Change ix86_arch_mask from
21267         unsigned int to unsigned HOST_WIDE_INT, initialize to
21268         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
21269         (ix86_function_specific_restore): Likewise.
21271 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21273         PR target/84899
21274         * postreload.c (reload_combine_recognize_pattern): Perform
21275         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
21276         truncate_int_for_mode the result for the destination's mode.
21278         PR c/84909
21279         * hsa-gen.c (mem_type_for_type): Fix comment typo.
21280         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
21281         Likewise.
21282         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21283         Likewise.
21285 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
21287         PR target/84876
21288         * lra-assigns.c (lra_split_hard_reg_for): Don't use
21289         regno_allocno_class_array and sorted_pseudos.
21290         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
21291         insns where regno is used.
21293 2018-03-16  Martin Liska  <mliska@suse.cz>
21295         PR ipa/84833
21296         * multiple_target.c (create_dispatcher_calls): Redirect
21297         reference in the symbol table.
21299 2018-03-16  Martin Liska  <mliska@suse.cz>
21301         PR ipa/84722
21302         * multiple_target.c (create_dispatcher_calls): Redirect also
21303         an alias.
21305 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21307         PR c++/79937
21308         PR c++/82410
21309         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
21310         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
21311         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
21313 2018-03-16  Julia Koval  <julia.koval@intel.com>
21315         * doc/invoke.texi (Skylake Server): Add CLWB.
21316         Cannonlake): Remove CLWB.
21318 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21320         PR tree-optimization/84841
21321         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
21322         1 << 3.
21323         (FLOAT_ONE_CONST_TYPE): Define.
21324         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
21325         (sort_by_operand_rank): Put entries with higher constant_type last
21326         rather than first to match comments.
21328 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
21330         * config/nios2/nios2.md (movsi_internal): Fix thinko in
21331         split predicate.
21333 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21335         PR c++/79085
21336         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
21337         check and use address of target always.
21339 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
21341         PR target/84574
21342         * config/i386/i386.c (indirect_thunk_needed): Update comments.
21343         (indirect_thunk_bnd_needed): Likewise.
21344         (indirect_thunks_used): Likewise.
21345         (indirect_thunks_bnd_used): Likewise.
21346         (indirect_return_needed): New.
21347         (indirect_return_bnd_needed): Likewise.
21348         (output_indirect_thunk_function): Add a bool argument for
21349         function return.
21350         (output_indirect_thunk_function): Don't generate alias for
21351         function return thunk.
21352         (ix86_code_end): Call output_indirect_thunk_function to generate
21353         function return thunks.
21354         (ix86_output_function_return): Set indirect_return_bnd_needed
21355         and indirect_return_needed instead of indirect_thunk_bnd_needed
21356         and indirect_thunk_needed.
21358 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
21360         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
21361         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
21362         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
21364 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
21365             Paul Hua <paul.hua.gm@gmail.com>
21367         PR c/84852
21368         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
21370 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
21372         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
21373         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
21374         resp. SFmode cases.
21376 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
21378         PR target/84711
21379         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
21380         instead of GET_MODE_SIZE when comparing Units.
21382 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
21384         PR target/68256
21385         * varasm.c (hash_section): Return an unchangeble hash value
21386         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
21387         Return !aarch64_can_use_per_function_literal_pools_p ().
21389 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21391         PR target/84860
21392         * optabs.c (emit_conditional_move): Pass address of cmode's copy
21393         rather than address of cmode as last argument to prepare_cmp_insn.
21395 2018-03-15  Julia Koval  <julia.koval@intel.com>
21397         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
21398         F_AVX512VNNI, F_AVX512BITALG): New.
21400 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
21402         PR target/83451
21403         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
21404         insn for floating-point loads and stores.
21406 2018-03-14  Carl Love  <cel@us.ibm.com>
21408         * config/rs6000/rs6000-c.c: Add macro definitions for
21409         ALTIVEC_BUILTIN_VEC_PERMXOR.
21410         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
21411         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
21412         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
21413         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
21414         UNSPEC_VPERMXOR.
21415         * config/doc/extend.texi: Add prototypes for vec_permxor.
21417 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
21419         PR c/84852
21420         * diagnostic-show-locus.c (class layout_point): Convert m_line
21421         from int to linenum_type.
21422         (line_span::comparator): Use linenum "compare" function when
21423         comparing line numbers.
21424         (test_line_span): New function.
21425         (layout_range::contains_point): Convert param "row" from int to
21426         linenum_type.
21427         (layout_range::intersects_line_p): Likewise.
21428         (layout::will_show_line_p): Likewise.
21429         (layout::print_source_line): Likewise.
21430         (layout::should_print_annotation_line_p): Likewise.
21431         (layout::print_annotation_line): Likewise.
21432         (layout::print_leading_fixits): Likewise.
21433         (layout::annotation_line_showed_range_p): Likewise.
21434         (struct line_corrections): Likewise for field m_row.
21435         (line_corrections::line_corrections): Likewise for param "row".
21436         (layout::print_trailing_fixits): Likewise.
21437         (layout::get_state_at_point): Likewise.
21438         (layout::get_x_bound_for_row): Likewise.
21439         (layout::print_line): Likewise.
21440         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
21441         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
21442         * input.c (selftest::test_linenum_comparisons): New function.
21443         (selftest::input_c_tests): Call it.
21444         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
21445         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
21446         * selftest.h (ASSERT_GT): New macro.
21447         (ASSERT_GT_AT): New macro.
21448         (ASSERT_LT): New macro.
21449         (ASSERT_LT_AT): New macro.
21451 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
21453         PR rtl-optimization/84780
21454         * combine.c (distribute_links): Don't make a link based on pc_rtx.
21456 2018-03-14  Martin Liska  <mliska@suse.cz>
21458         * tree.c (record_node_allocation_statistics): Use
21459         get_stats_node_kind.
21460         (get_stats_node_kind): New function extracted from
21461         record_node_allocation_statistics.
21462         (free_node): Use get_stats_node_kind.
21464 2018-03-14  Richard Biener  <rguenther@suse.de>
21466         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
21467         that the value-set of ANTIC_IN doesn't grow.
21469         Revert
21470         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21471         member.
21472         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21473         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21475 2018-03-14  Julia Koval  <julia.koval@intel.com>
21477         * config.gcc (icelake-client, icelake-server): New.
21478         (icelake): Remove.
21479         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
21480         (initial_ix86_arch_features): Ditto.
21481         (PTA_SKYLAKE): Add SGX.
21482         (PTA_ICELAKE): Remove.
21483         (PTA_ICELAKE_CLIENT): New.
21484         (PTA_ICELAKE_SERVER): New.
21485         (ix86_option_override_internal): Split up icelake on icelake client and
21486         icelake server.
21487         (get_builtin_code_for_version): Ditto.
21488         (fold_builtin_cpu): Ditto.
21489         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
21490         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
21491         * config/i386/i386.h (processor_type): Ditto.
21492         * doc/invoke.texi: Ditto.
21494 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
21496         PR sanitizer/83392
21497         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21498         INTEGER_CST offset, add it together with bitpos / 8 and
21499         sign extend based on POINTER_SIZE.
21501         PR target/84844
21502         Revert
21503         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
21505         PR target/78090
21506         * config/i386/constraints.md (Yc): New register constraint.
21507         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21508         Use Yc constraint for alternative 2 of operand 0.  Remove
21509         preferred_for_speed attribute.
21511 2018-03-14  Richard Biener  <rguenther@suse.de>
21513         PR tree-optimization/84830
21514         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21515         with the old one to avoid oscillations.
21517 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
21519         PR target/83712
21520         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21521         pseudos.
21522         (assign_by_spills): Return a flag of reload assignment failure.
21523         Do not process the reload assignment failures.  Do not spill other
21524         reload pseudos if they has the same reg class.  Update n if
21525         necessary.
21526         (lra_assign): Add a return arg.  Set up from the result of
21527         assign_by_spills call.
21528         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21529         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21530         usage_insns if it is not NULL.
21531         (spill_hard_reg_in_range): New function.
21532         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21533         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21534         function prototypes.
21535         (lra_assign): Change prototype.
21536         * lra.c (lra): Add code to deal with fails by splitting hard reg
21537         live ranges.
21539 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
21541         * config/riscv/riscv.opt (mrelax): New option.
21542         * config/riscv/riscv.c (riscv_file_start): Emit ".option
21543         "norelax" when riscv_mrelax is disabled.
21544         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21546 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21548         PR target/84743
21549         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21550         reassociation for int modes.
21552 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21554         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21555         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21556         for big-endian.
21557         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21558         * config/aarch64/aarch64-sve.md
21559         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21560         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21561         (*extend<mode><Vwide>2): Rename to...
21562         (aarch64_sve_extend<mode><Vwide>2): ...this.
21563         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21564         renaming the old pattern to...
21565         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
21566         unsigned packs.
21567         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21568         define_expand, renaming the old pattern to...
21569         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21570         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21571         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21572         account when deciding which SVE instruction the optab should use.
21573         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21575 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21577         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21578         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21579         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21580         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21581         (tlsdesc_small_<mode>): Turn a define_expand and use
21582         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
21583         (tlsdesc_small_advsimd_<mode>): ...this.
21584         (tlsdesc_small_sve_<mode>): New pattern.
21586 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21588         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21589         (UNSPEC_UMUL_HIGHPART): New constants.
21590         (MUL_HIGHPART): New int iteraor.
21591         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21592         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21593         define_expand.
21594         (*<su>mul<mode>3_highpart): New define_insn.
21596 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21598         PR lto/84805
21599         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21600         incomplete types.
21602 2018-03-13  Martin Liska  <mliska@suse.cz>
21604         PR ipa/84658.
21605         * (sem_item_optimizer::sem_item_optimizer): Initialize new
21606         vector.
21607         (sem_item_optimizer::~sem_item_optimizer): Release it.
21608         (sem_item_optimizer::merge_classes): Register variable aliases.
21609         (sem_item_optimizer::fixup_pt_set): New function.
21610         (sem_item_optimizer::fixup_points_to_sets): Likewise.
21611         * ipa-icf.h: Declare new variables and functions.
21613 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
21615         PR middle-end/84834
21616         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21617         integer_pow2p@2 and test integer_pow2p in condition.
21618         (A < 0 ? C : 0): Similarly for @1.
21620         PR middle-end/84831
21621         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21622         characters starting at p contain '\0' character, don't look beyond
21623         that.
21625         PR target/84827
21626         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21627         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21629         PR target/84828
21630         * reg-stack.c (change_stack): Change update_end var from int to
21631         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21632         also call set_block_for_insn on the newly added insns and rescan.
21634         PR target/84786
21635         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21636         on the last operand.
21638         PR c++/84704
21639         * tree.c (stabilize_reference_1): Return save_expr (e) for
21640         STATEMENT_LIST even if it doesn't have side-effects.
21642 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
21644         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21646 2018-03-12  Renlin Li  <renlin.li@arm.com>
21648         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21649         aarch64_output_scalar_simd_mov_immediate.
21651 2018-03-12  Martin Sebor  <msebor@redhat.com>
21653         PR tree-optimization/83456
21654         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21655         for perfectly overlapping calls to memcpy.
21656         (gimple_fold_builtin_memory_chk): Same.
21657         (gimple_fold_builtin_strcpy): Handle no-warning.
21658         (gimple_fold_builtin_stxcpy_chk): Same.
21659         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21661 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21663         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21664         parameter.  Use it for SFmode.
21665         (rs6000_function_arg_advance_1): Adjust.
21666         (rs6000_function_arg): Adjust.
21667         (rs6000_gimplify_va_arg): Pass false for that new parameter.
21669 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21671         PR rtl-optimization/84169
21672         PR rtl-optimization/84780
21673         * combine.c (can_combine_p): Check for a 2-insn combination whether
21674         the destination register is used between the two insns, too.
21676 2018-03-12  Richard Biener  <rguenther@suse.de>
21678         PR tree-optimization/84803
21679         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21680         for refs DR analysis didn't process.
21682 2018-03-12  Richard Biener  <rguenther@suse.de>
21684         PR tree-optimization/84777
21685         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21686         force-vectorize loops ignore whether we are optimizing for size.
21688 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
21690         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21691         (TARGET_MD_ASM_ADJUST): Define.
21693 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
21694             Kito Cheng  <kito.cheng@gmail.com>
21695             Chung-Ju Wu  <jasonwucj@gmail.com>
21697         * config/nds32/nds32.c (nds32_compute_stack_frame,
21698         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21699         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21700         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21701         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21702         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21703         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21704         * config/nds32/nds32.md (prologue, epilogue): Use macro
21705         NDS32_V3PUSH_AVAILABLE_P to do checking.
21707 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
21709         PR debug/58150
21710         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21711         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21712         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
21713         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
21714         addition of most attributes on !orig_type_die or the attribute not
21715         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21717 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21718             Chung-Ju Wu  <jasonwucj@gmail.com>
21720         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21721         __NDS32_VH__ macro.
21722         * config/nds32/nds32.opt (mvh): New option.
21724 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21725             Chung-Ju Wu  <jasonwucj@gmail.com>
21727         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21728         function.
21729         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21730         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21731         definition.
21733 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21734             Chung-Ju Wu  <jasonwucj@gmail.com>
21736         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21737         function.
21738         * config/nds32/nds32-multiple.md (strlensi): New pattern.
21739         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21741 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
21742             Kito Cheng  <kito.cheng@gmail.com>
21743             Chung-Ju Wu  <jasonwucj@gmail.com>
21745         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21746         UNSPEC_FFMISM and UNSPEC_FLMISM.
21747         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21748         for ffb, ffmism and flmism.
21749         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21750         (unspec_ffmism): Ditto.
21751         (unspec_flmism): Ditto.
21752         (nds32_expand_builtin_impl): Check if string extension is available.
21753         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21754         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21756 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
21758         Reverting patch:
21759         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21761         PR target/83712
21762         * lra-assigns.c (assign_by_spills): Return a flag of reload
21763         assignment failure.  Do not process the reload assignment
21764         failures.  Do not spill other reload pseudos if they has the same
21765         reg class.
21766         (lra_assign): Add a return arg.  Set up from the result of
21767         assign_by_spills call.
21768         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21769         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21770         usage_insns if it is not NULL.
21771         (spill_hard_reg_in_range): New function.
21772         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21773         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21774         function prototypes.
21775         (lra_assign): Change prototype.
21776         * lra.c (lra): Add code to deal with fails by splitting hard reg
21777         live ranges.
21779 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
21781         PR target/84807
21782         * config/i386/i386.opt: Replace Enforcment with Enforcement.
21784 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
21786         PR debug/84620
21787         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21788         (dw_val_node): Add val_symbolic_view.
21789         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21790         (symview_upper_bound): New.
21791         (new_line_info_table): Initialize symviews_since_reset.
21792         (dwarf2out_source_line): Count symviews_since_reset and set
21793         symview_upper_bound.
21794         (dw_val_equal_p): Handle symview.
21795         (add_AT_symview): New.
21796         (print_dw_val): Handle symview.
21797         (attr_checksum, attr_checksum_ordered): Likewise.
21798         (same_dw_val_p, size_of_die): Likewise.
21799         (value_format, output_die): Likewise.
21800         (add_high_low_attributes): Use add_AT_symview for entry_view.
21801         (dwarf2out_finish): Reset symview_upper_bound, clear
21802         zero_view_p.
21804 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
21806         PR target/83969
21807         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21808         Add strict argument and use it.
21809         (rs6000_split_multireg_move): Update for new strict argument.
21810         (mem_operand_gpr): Disallow all non-offsettable addresses.
21811         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21813 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
21815         PR target/84772
21816         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21817         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21818         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21820         PR c++/84767
21821         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21822         decl, use remap_type if we want to use the type.
21824 2018-03-09  Martin Sebor  <msebor@redhat.com>
21826         PR tree-optimization/84526
21827         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21828         Remove dead code.
21829         (builtin_access::generic_overlap): Be prepared to handle non-array
21830         base objects.
21832 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
21834         PR rtl-optimization/84682
21835         * lra-constraints.c (process_address_1): Check is_address flag
21836         for address constraints.
21837         (process_alt_operands): Likewise.
21838         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21839         preprocess_constraints.
21840         * recog.h (preprocess_constraints): Add oploc parameter.
21841         Adjust callers.
21842         * recog.c (preprocess_constraints): Test address_operand for
21843         CT_ADDRESS constraints.
21845 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21847         PR target/83712
21848         * lra-assigns.c (assign_by_spills): Return a flag of reload
21849         assignment failure.  Do not process the reload assignment
21850         failures.  Do not spill other reload pseudos if they has the same
21851         reg class.
21852         (lra_assign): Add a return arg.  Set up from the result of
21853         assign_by_spills call.
21854         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21855         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21856         usage_insns if it is not NULL.
21857         (spill_hard_reg_in_range): New function.
21858         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21859         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21860         function prototypes.
21861         (lra_assign): Change prototype.
21862         * lra.c (lra): Add code to deal with fails by splitting hard reg
21863         live ranges.
21865 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21867         PR target/83193
21868         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21869         Accept complain bool parameter.  Only emit errors if it is true.
21870         (arm_parse_cpu_option_name): Likewise.
21871         (arm_target_thumb_only): Adjust callers of the above.
21872         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21873         prototype to take a default true bool parameter.
21874         (arm_parse_arch_option_name): Likewise.
21876 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
21877             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
21879         PR jit/64089
21880         PR jit/84288
21881         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
21882         * configure: Regenerate.
21883         * configure.ac ("linker --version-script option"): New.
21884         ("linker soname option"): New.
21886 2018-03-09  Richard Biener  <rguenther@suse.de>
21888         PR tree-optimization/84775
21889         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
21890         immediate uses of predicate stmts and mark them modified.
21892         Revert
21893         PR tree-optimization/84178
21894         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21895         to caller.
21896         (version_loop_for_if_conversion): Delay update_ssa call.
21897         (tree_if_conversion): Delay update_ssa until after predicate
21898         insertion.
21900 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
21902         PR target/84763
21903         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
21904         when the function accesses prior frames.
21906 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21908         PR debug/84456
21909         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
21910         gen_llsym, otherwise call maybe_gen_llsym.
21912         PR inline-asm/84742
21913         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
21914         has ',' character inside of it.
21916 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21918         PR target/84748
21919         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21920         as clobbering CC_REGNUM.
21922 2018-03-08  Richard Biener  <rguenther@suse.de>
21924         PR middle-end/84552
21925         * tree-scalar-evolution.c: Include tree-into-ssa.h.
21926         (follow_copies_to_constant): Do not follow SSA names registered
21927         for update.
21929 2018-03-08  Richard Biener  <rguenther@suse.de>
21931         PR tree-optimization/84178
21932         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21933         to caller.
21934         (version_loop_for_if_conversion): Delay update_ssa call.
21935         (tree_if_conversion): Delay update_ssa until after predicate
21936         insertion.
21938 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
21940         PR tree-optimization/84178
21941         * tree-if-conv.c (release_bb_predicate): Remove the
21942         the assertion that the stmts have NULL use_ops.
21943         Discard the statements, asserting that they haven't
21944         yet been added to a BB.
21946 2018-03-08  Richard Biener  <rguenther@suse.de>
21948         PR tree-optimization/84746
21949         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
21950         (phi_translate): Pass in destination ANTIC_OUT set.
21951         (phi_translate_1): Likewise.  For a simplified result lookup
21952         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
21953         (phi_translate_set): Adjust.
21954         (do_pre_regular_insertion): Likewise.
21955         (do_pre_partial_partial_insertion): Likewise.
21957 2018-03-08  Martin Liska  <mliska@suse.cz>
21959         PR gcov-profile/84735
21960         * doc/gcov.texi: Document usage of profile files.
21961         * gcov-io.h: Document changes in the format.
21963 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
21965         PR debug/84404
21966         PR debug/84408
21967         * dwarf2out.c (struct dw_line_info_table): Update comments for
21968         view == -1.
21969         (FORCE_RESET_NEXT_VIEW): New.
21970         (FORCE_RESETTING_VIEW_P): New.
21971         (RESETTING_VIEW_P): Check for -1 too.
21972         (ZERO_VIEW_P): Likewise.
21973         (new_line_info_table): Force-reset next view.
21974         (dwarf2out_begin_function): Likewise.
21975         (dwarf2out_source_line): Simplify zero_view_p initialization.
21976         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
21977         view directly.  Omit view when omitting .loc at line 0.
21979 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21981         PR tree-optimization/84740
21982         * tree-switch-conversion.c (process_switch): Call build_constructors
21983         only if info.phi_count is non-zero.
21985         PR tree-optimization/84739
21986         * tree-tailcall.c (find_tail_calls): Check call arguments against
21987         DECL_ARGUMENTS (current_function_decl) rather than
21988         DECL_ARGUMENTS (func) when checking for tail recursion.
21990 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21992         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
21993         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
21994         Volker Reichelt's entry and add entries for people that perform
21995         GCC fuzzy testing and report numerous bugs.
21997 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
21999         PR target/82411
22000         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
22001         readonly data in sdata, if that is disabled.
22002         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
22003         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
22004         -mreadonly-in-sdata option.
22006 2018-03-07  Martin Sebor  <msebor@redhat.com>
22008         PR tree-optimization/84468
22009         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
22010         basic block when looking for nul assignment.
22012 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
22014         PR target/84277
22015         * except.h (output_function_exception_table): Adjust prototype.
22016         * except.c (output_function_exception_table): Remove FNNAME parameter
22017         and add SECTION parameter.  Ouput one part of the table at a time.
22018         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
22019         the first part of the exception table and emit unwind directives.
22020         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
22021         (i386_pe_seh_cold_init): Likewise.
22022         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
22023         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
22024         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
22025         (ix86_output_call_insn): Emit a nop in one more case for SEH.
22026         * config/i386/winnt.c: Include except.h.
22027         (struct seh_frame_state): Add reg_offset, after_prologue and
22028         in_cold_section fields.
22029         (i386_pe_seh_end_prologue): Set seh->after_prologue.
22030         (i386_pe_seh_cold_init): New function.
22031         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
22032         to seh->in_cold_section.
22033         (seh_emit_push): Record the offset of the push.
22034         (seh_emit_save): Record the offet of the save.
22035         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
22036         Test seh->after_prologue to disregard the epilogue.
22037         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
22038         (i386_pe_end_cold_function): New function.
22040 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
22042         PR fortran/84565
22043         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
22044         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
22046         PR c++/84704
22047         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
22048         on tmp_var.
22049         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
22050         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
22052         PR middle-end/84723
22053         * multiple_target.c: Include tree-inline.h and intl.h.
22054         (expand_target_clones): Diagnose and fail if node->definition and
22055         !tree_versionable_function_p (node->decl).
22057 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
22059         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
22060         sprint_ul.
22061         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
22062         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
22063         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
22065 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
22067         PR target/84710
22068         * combine.c (try_combine): Use reg_or_subregno instead of handling
22069         just paradoxical SUBREGs and REGs.
22071 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
22073         * config/arc/arc.c (arc_finalize_pic): Remove function.
22074         (arc_must_save_register): We use single base PIC register, remove
22075         checks to save/restore the PIC register.
22076         (arc_expand_prologue): Likewise.
22077         * config/arc/arc-protos.h (arc_set_default_type_attributes):
22078         Remove.
22079         (arc_verify_short): Likewise.
22080         (arc_attr_type): Likewise.
22081         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
22082         (walk_stores): Likewise.
22083         (arc_address_cost): Make it static.
22084         (arc_verify_short): Likewise.
22085         (branch_dest): Likewise.
22086         (arc_attr_type): Likewise.
22087         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
22088         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
22089         (arc_final_prescan_insn): Remove inserting the nops due to
22090         hardware hazards.  It is done in reorg step.
22091         (insn_length_variant_t): Remove.
22092         (insn_length_parameters_t): Likewise.
22093         (arc_insn_length_parameters): Likewise.
22094         (arc_get_insn_variants): Likewise.
22095         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
22097 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
22099         PR inline-asm/84683
22100         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
22101         assertion failure.
22103         PR tree-optimization/84687
22104         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
22105         on new_node->decl.
22106         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
22108 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22110         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
22111         Rename to ppc_speculation_barrier.
22112         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
22113         __builtin_ppc_speculation_barrier.
22115 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22117         PR target/84700
22118         * combine.c (combine_simplify_rtx): Don't try to simplify if
22119         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
22120         are equal to x.
22122 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
22124         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
22125         to 32 bytes when compiling for POWER9.
22127 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22129         PR target/84564
22130         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
22131         regparm >= 3 with no arg reg available also for calls with
22132         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
22134         PR target/84524
22135         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
22136         orig,vex.
22137         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
22139 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
22141         PR target/84264
22142         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
22144 2018-03-05  Richard Biener  <rguenther@suse.de>
22146         PR tree-optimization/84486
22147         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
22148         When inserting a __builtin_assume_aligned call set the LHS
22149         SSA name alignment info accordingly.
22151 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
22153         PR tree-optimization/84114
22154         * config/aarch64/aarch64.c (aarch64_reassociation_width)
22155         Avoid reassociation of FLOAT_MODE addition.
22157 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
22159         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
22160         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
22161         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
22162         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
22163         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
22164         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
22165         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
22166         and -mwbnoinvd.
22167         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
22168         __builtin_ia32_wbinvd): New builtins.
22169         (SPECIAL_ARGS2): New.
22170         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
22171         (SPECIAL_ARGS2): New.
22172         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
22173         (ix86_valid_target_attribute_inner_p): Ditto.
22174         (ix86_init_mmx_sse_builtins): Add special_args2.
22175         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
22176         TARGET_WBNOINVD_P): New.
22177         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
22178         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
22179         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
22180         * config/i386/immintrin.h (_wbinvd): New intrinsic.
22181         * config/i386/pconfigintrin.h: New file.
22182         * config/i386/wbnoinvdintrin.h: Ditto.
22183         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
22184         wbnoinvdintrin.h.
22185         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
22187 2018-03-05  Richard Biener  <rguenther@suse.de>
22189         PR tree-optimization/84670
22190         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22191         member.
22192         (BB_VISITED_WITH_VISITED_SUCCS): New define.
22193         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22194         (compute_antic_aux): Only assert the number of values in ANTIC_IN
22195         doesn't grow if all successors (recursively) were visited at least
22196         once.
22198 2018-03-05  Richard Biener  <rguenther@suse.de>
22200         PR tree-optimization/84650
22201         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
22202         if executed in the loop pipeline.
22204 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
22206         * doc/configfiles.texi (Configuration Files): Move info about
22207         conditionalizing $target-protos.h to...
22208         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
22209         differs from $target-protos.h.
22211 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
22212             Chung-Ju Wu  <jasonwucj@gmail.com>
22214         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
22215         * config/nds32/nds32-multiple.md (setmemsi): Define.
22216         * config/nds32/nds32-memory-manipulation.c
22217         (nds32_gen_dup_4_byte_to_word_value): New.
22218         (emit_setmem_word_loop): New.
22219         (emit_setmem_byte_loop): New.
22220         (nds32_expand_setmem_loop): New.
22221         (nds32_expand_setmem_loop_v3m): New.
22222         (nds32_expand_setmem_unroll): New.
22223         (nds32_expand_setmem): New.
22225 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22226             Chung-Ju Wu  <jasonwucj@gmail.com>
22228         * config/nds32/nds32-memory-manipulation.c
22229         (nds32_emit_load_store): New.
22230         (nds32_emit_post_inc_load_store): New.
22231         (nds32_emit_mem_move): New.
22232         (nds32_emit_mem_move_block): New.
22233         (nds32_expand_movmemsi_loop_unknown_size): New.
22234         (nds32_expand_movmemsi_loop_known_size): New.
22235         (nds32_expand_movmemsi_loop): New.
22236         (nds32_expand_movmemsi_unroll): New.
22237         (nds32_expand_movmemqi): Rename ...
22238         (nds32_expand_movmemsi): ... to this.
22239         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
22240         (movmemsi): ... to this.
22241         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
22242         (nds32_expand_movmemsi): ... to this.
22244 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22245             Monk Chiang  <sh.chiang04@gmail.com>
22246             Chung-Ju Wu  <jasonwucj@gmail.com>
22248         * config/nds32/nds32-protos.h
22249         (nds32_expand_load_multiple): New arguments.
22250         (nds32_expand_store_multiple): Ditto.
22251         (nds32_valid_multiple_load_store): Rename ...
22252         (nds32_valid_multiple_load_store_p): ... to this.
22253         * config/nds32/nds32-memory-manipulation.c
22254         (nds32_expand_load_multiple): Refine implementation.
22255         (nds32_expand_store_multiple): Ditto.
22256         * config/nds32/nds32-multiple.md
22257         (load_multiple): Update nds32_expand_load_multiple interface.
22258         (store_multiple): Update nds32_expand_store_multiple interface.
22259         * config/nds32/nds32-predicates.c
22260         (nds32_valid_multiple_load_store): Rename ...
22261         (nds32_valid_multiple_load_store_p): ... to this and refine
22262         implementation.
22263         * config/nds32/predicates.md
22264         (nds32_load_multiple_and_update_address_operation): New predicate.
22265         (nds32_store_multiple_and_update_address_operation): New predicate.
22267 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22268             Chung-Ju Wu  <jasonwucj@gmail.com>
22270         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
22271         (combo): New attribute.
22272         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
22274 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
22276         * config/nds32/nds32.opt: Change -mcmodel= default value.
22278 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
22279             Monk Chiang  <sh.chiang04@gmail.com>
22280             Chung-Ju Wu  <jasonwucj@gmail.com>
22282         * config/nds32/constants.md (unspec_element): New enum.
22283         * config/nds32/constraints.md (Umw): New constraint.
22284         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
22285         * config/nds32/nds32-intrinsic.md: Likewise.
22286         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
22287         (nds32_valid_smw_lwm_base_p): New.
22288         (nds32_output_smw_single_word): New.
22289         (nds32_output_lmw_single_word): New.
22290         (nds32_expand_unaligned_load): New.
22291         (nds32_expand_unaligned_store): New.
22292         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
22293         (nds32_output_smw_single_word): Declare.
22294         (nds32_output_lmw_single_word): Declare.
22295         (nds32_expand_unaligned_load): Declare.
22296         (nds32_expand_unaligned_store): Declare.
22297         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
22298         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
22299         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
22300         NDS32_BUILTIN_UASTORE_DW.
22301         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
22302         predicate.
22304 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
22305             Kito Cheng  <kito.cheng@gmail.com>
22306             Chung-Ju Wu  <jasonwucj@gmail.com>
22308         * config/nds32/nds32-intrinsic.c
22309         (nds32_expand_builtin_null_ftype_reg): Delete.
22310         (nds32_expand_builtin_reg_ftype_imm): Ditto.
22311         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
22312         (nds32_read_argument): New.
22313         (nds32_legitimize_target): Ditto.
22314         (nds32_legitimize_argument): Ditto.
22315         (nds32_check_constant_argument): Ditto.
22316         (nds32_expand_unop_builtin): Ditto.
22317         (nds32_expand_unopimm_builtin): Ditto.
22318         (nds32_expand_binop_builtin): Ditto.
22319         (nds32_builtin_decl_impl): Ditto.
22320         (builtin_description): Ditto.
22321         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
22322         (nds32_init_builtins_impl): Ditto.
22323         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
22324         (nds32_builtin_decl): New.
22325         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
22326         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
22328 2018-03-02  Jeff Law  <law@redhat.com>
22330         * reorg.c (stop_search_p): Handle DEBUG_INSN.
22331         (redundant_insn, fill_simple_delay_slots): Likewise.
22332         (fill_slots_from_thread): Likewise.
22333         * resource.c (mark_referenced_resources): Likewise.
22334         (mark_set_resources, find_dead_or_set_registers): Likewise.
22336 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22338         * substring-locations.h (format_warning_va): Formatting fix for
22339         ATTRIBUTE_GCC_DIAG.
22340         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
22341         argument.
22342         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
22343         * substring-locations.c: Include intl.h.
22344         (format_warning_va): Turned into small wrapper around
22345         format_warning_n_va, renamed to ...
22346         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
22347         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
22348         use ngettext.
22349         (format_warning_at_substring_n): New function.
22350         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
22351         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
22352         format_warning_at_substring with just a shorter name instead of
22353         const function pointer.
22354         (fmtwarn_n): New function.
22355         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
22356         appropriate, get rid of all the fmtstr temporaries, move conditionals
22357         with G_() wrapped string literals directly into fmtwarn arguments,
22358         cast dir.len to (int), formatting fixes.
22360 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
22362         * doc/invoke.texi: Remove "Cilk Plus" references.
22364 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22365             Richard Biener  <rguenther@suse.de>
22367         PR ipa/84628
22368         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
22369         for error or warning attributes if CALL_FROM_THUNK_P is set.
22370         Formatting fixes.
22372 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22374         PR target/56540
22375         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
22376         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
22378         PR target/56540
22379         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
22380         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
22382         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
22383         instead of -1U in last predictors element's probability member.
22385 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
22387         PR ipa/83983
22388         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22389         arguments if they are comparable.
22391 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22393         PR tree-optimization/84634
22394         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
22395         masks and masked_loop_p with a single loop_masks, making sure it's
22396         null for bb vectorization.
22398 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22400         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
22401         (vect_analyze_data_ref_access): Use loop->safe_len rather than
22402         loop->force_vectorize to check whether there is no alias.
22404 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22406         PR target/84614
22407         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
22408         prototypes.
22409         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
22410         comments.
22411         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
22412         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
22413         instead of a loop around prev_real_insn.
22414         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
22415         prev_real_insn.
22417         PR inline-asm/84625
22418         * config/i386/i386.c (ix86_print_operand): Use conditional
22419         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
22420         zero vector.
22422 2018-03-02  Richard Biener  <rguenther@suse.de>
22424         PR tree-optimization/84427
22425         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
22426         (bitmap_set_subtract_values): Rewrite to handle multiple
22427         exprs per value.
22428         (clean): Likewise.
22429         (prune_clobbered_mems): Likewise.
22430         (phi_translate): Take edge instead of pred/phiblock.
22431         (phi_translate_1): Likewise.
22432         (phi_translate_set): Likewise.  Insert all translated
22433         exprs for a value into the set, keeping possibly multiple
22434         expressions per value.
22435         (compute_antic_aux): Adjust for phi_translate changes.
22436         When intersecting union the expressions and prune those
22437         not in the final value set, keeping possibly multiple
22438         expressions per value.  Do not use value-insertion
22439         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
22440         all expressions.  Add verification that the value-sets
22441         only shrink during iteration.
22442         (compute_partial_antic_aux): Adjust for the phi_translate changes.
22443         (do_pre_regular_insertion): Likewise.
22444         (do_pre_partial_partial_insertion): Likewise.
22446 2018-03-02  Richard Biener  <rguenther@suse.de>
22448         PR target/82005
22449         * config/darwin.c (saved_debug_info_level): New static global.
22450         (darwin_asm_lto_start): Disable debug info generation for LTO out.
22451         (darwin_asm_lto_end): Restore debug info generation settings.
22453 2018-03-01  Martin Liska  <mliska@suse.cz>
22455         PR sanitizer/82484
22456         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
22457         volatile arguments.
22459 2018-03-01  Richard Biener  <rguenther@suse.de>
22461         PR debug/84645
22462         * dwarf2out.c (gen_variable_die): Properly handle late VLA
22463         type annotation with LTO when debug was disabled at compile-time.
22465 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
22467         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
22468         XINT with INTVAL.
22469         (mips_final_postscan_insn): Likewise.
22471 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
22473         PR rtl-optimization/84528
22474         * alias.c (init_alias_target): Add commentary.
22475         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
22476         a unique base value if the frame pointer is not eliminated
22477         to the stack pointer.
22479 2018-03-01  Tom de Vries  <tom@codesourcery.com>
22481         PR rtl-optimization/83327
22482         * lra-int.h (hard_regs_spilled_into): Declare.
22483         * lra.c (hard_regs_spilled_into): Define.
22484         (init_reg_info): Init hard_regs_spilled_into.
22485         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22486         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
22487         (process_bb_lives): Handle hard_regs_spilled_into.
22488         (lra_create_live_ranges_1): Before doing liveness propagation, clear
22489         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
22491 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
22493         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22494         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22495         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22496         * config/rs6000/aix72.h: New file.
22498 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
22500         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22501         instead of warning_at with conditional singular and plural messages
22502         where possible.
22504         PR target/52991
22505         * stor-layout.c (update_alignment_for_field): For
22506         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22507         && !DECL_PACKED (field), do the alignment update, just use
22508         only desired_align instead of MAX (type_align, desired_align)
22509         as the alignment.
22510         (place_field): Don't do known_align < desired_align handling
22511         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22512         is non-NULL, instead do it after rli->prev_field handling and
22513         only if not within a bitfield word.  For DECL_PACKED (field)
22514         use type_align of BITS_PER_UNIT.
22516 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
22518         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22519         superfluous parentheses and trailing spaces.
22521 2018-02-28  Richard Biener  <rguenther@suse.de>
22523         PR tree-optimization/84584
22524         * graphite-scop-detection.c (scop_detection::add_scop): Discard
22525         SCoPs with fake exit edge.
22527 2018-02-28  Martin Liska  <mliska@suse.cz>
22529         PR testsuite/84597
22530         * timevar.c (timer::print): Fix format to properly print 100%
22531         values.
22533 2018-02-28  Richard Biener  <rguenther@suse.de>
22535         PR middle-end/84607
22536         * genmatch.c (capture_info::walk_match): Do not mark
22537         captured expressions without operands as expr_p given
22538         they act more like predicates and should be subject to
22539         "lost tail" side-effect preserving.
22541 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
22543         PR rtl-optimization/81611
22544         * auto-inc-dec.c (attempt_change): Move dead note from
22545         mem_insn if it's the next use of regno
22546         (find_address): Take address use of reg holding
22547         non-incremented value.  Add parm to limit search to the named
22548         reg only.
22549         (merge_in_block): Attempt to use a mem insn that is the next
22550         use of the original regno.
22552 2018-02-27  Martin Sebor  <msebor@redhat.com>
22554         PR c++/83871
22555         * doc/invoke.texi (-Wmissing-attributes): New option.
22556         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22558 2018-02-27  Martin Sebor  <msebor@redhat.com>
22560         PR translation/84207
22561         * diagnostic-core.h (warning_n, error_n, inform_n): Change
22562         n argument to unsigned HOST_WIDE_INT.
22563         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22564         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
22565         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22566         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22568 2018-02-27  Richard Biener  <rguenther@suse.de>
22570         PR tree-optimization/84512
22571         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22572         Do not use the estimate returned from record_stmt_cost for
22573         the scalar iteration cost but sum properly using add_stmt_cost.
22575 2018-02-27  Richard Biener  <rguenther@suse.de>
22577         PR tree-optimization/84466
22578         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22579         Adjust last change to less strictly validate use operands.
22581 2018-02-27  Martin Liska  <mliska@suse.cz>
22583         PR gcov-profile/84548
22584         * gcov.c (process_file): Allow partial overlap and consider it
22585         also as group functions.
22586         (output_lines): Properly calculate range of lines for a group.
22588 2018-02-27  Martin Liska  <mliska@suse.cz>
22590         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22591         'ggc' suffixes.  Change first column width.
22592         (timer::print): Fix formatting of the column.
22594 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
22596         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22597         preserve inline entry blocks for the sake of debug inline
22598         entry point markers alone.
22599         (remove_unused_locals): Suggest in comments a better place to
22600         force the preservation of inline entry blocks that are
22601         otherwise unused, but do not preserve them.
22603 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22605         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22607 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22609         PR target/84039
22610         * config/i386/constraints.md (Bs): Replace
22611         ix86_indirect_branch_register with
22612         TARGET_INDIRECT_BRANCH_REGISTER.
22613         (Bw): Likewise.
22614         * config/i386/i386.md (indirect_jump): Likewise.
22615         (tablejump): Likewise.
22616         (*sibcall_memory): Likewise.
22617         (*sibcall_value_memory): Likewise.
22618         Peepholes of indirect call and jump via memory: Likewise.
22619         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22620         (*sibcall_value_GOT_32): Likewise.
22621         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22622         (GOT_memory_operand): Likewise.
22623         (call_insn_operand): Likewise.
22624         (sibcall_insn_operand): Likewise.
22625         (GOT32_symbol_operand): Likewise.
22626         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22628 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22630         PR rtl-optimization/83496
22631         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22632         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
22633         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22634         redundant insn, if any.
22635         (relax_delay_slots): Likewise.
22636         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22638 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
22640         PR tree-optimization/83965
22641         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22642         that grouped statements are part of a reduction chain.  Return
22643         true if the statement is not marked as a reduction itself but
22644         is part of a group.
22645         (vect_recog_dot_prod_pattern): Don't check whether the statement
22646         is part of a group here.
22647         (vect_recog_sad_pattern): Likewise.
22648         (vect_recog_widen_sum_pattern): Likewise.
22650 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22652         PR debug/84545
22653         * final.c (rest_of_clean_state): Also look for calls inside sequences.
22655 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22657         PR target/84530
22658         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22659         the bool argument.
22660         (ix86_output_indirect_function_return): New prototype.
22661         (ix86_split_simple_return_pop_internal): Likewise.
22662         * config/i386/i386.c (indirect_return_via_cx): New.
22663         (indirect_return_via_cx_bnd): Likewise.
22664         (indirect_thunk_name): Handle return va CX_REG.
22665         (output_indirect_thunk_function): Create alias for
22666         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22667         (ix86_output_indirect_jmp): Remove the bool argument.
22668         (ix86_output_indirect_function_return): New function.
22669         (ix86_split_simple_return_pop_internal): Likewise.
22670         * config/i386/i386.md (*indirect_jump): Don't pass false
22671         to ix86_output_indirect_jmp.
22672         (*tablejump_1): Likewise.
22673         (simple_return_pop_internal): Change it to define_insn_and_split.
22674         Call ix86_split_simple_return_pop_internal to split it for
22675         -mfunction-return=.
22676         (simple_return_indirect_internal): Call
22677         ix86_output_indirect_function_return instead of
22678         ix86_output_indirect_jmp.
22680 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
22682         PR bootstrap/84405
22683         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22684         memset and value initialization afterwards.
22686 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
22688         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22690 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22692         PR target/84521
22693         * common/config/aarch64/aarch64-common.c
22694         (aarch_option_optimization_table[]): Switch
22695         off fomit-frame-pointer
22697 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22698             Chung-Ju Wu  <jasonwucj@gmail.com>
22700         * config/nds32/nds32-multiple.md (load_multiple): Disallow
22701         volatile memory.
22702         (store_multiple): Ditto.
22704 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22706         * config.gcc: Add --with-cpu support for nds32 target.
22707         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22708         * config/nds32/nds32.opt: Add -mcpu= option.
22710 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
22712         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22713         isel=yes): Warn for these deprecated options.
22715 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
22717         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22718         ISA_2_5_MASKS_EMBEDDED.
22720 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
22722         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22723         p->max as pointers rather than using iterative_hash_expr.
22725 2018-02-23  Carl Love  <cel@us.ibm.com>
22727         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22728         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22729         BU_P8V_OVERLOAD_2.
22730         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22731         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22732         P8V_BUILTIN_VEC_VUNSIGNED2.
22734 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
22736         PR target/81572
22737         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22738         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22739         LRA_UNKNOWN_ALT.
22740         * lra-constraints.c (curr_insn_transform): Set up
22741         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
22742         LRA_UNKNOWN_ALT.
22743         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22744         * lra-eliminations.c (spill_pseudos): Ditto.
22745         (process_insn_for_elimination): Ditto.
22746         * lra-lives.c (reg_early_clobber_p): Use the new macros.
22747         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22748         LRA_NON_CLOBBERED_ALT.
22750 2018-02-22  Martin Sebor  <msebor@redhat.com>
22752         PR tree-optimization/84480
22753         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22754         to maybe_diag_stxncpy_trunc.  Call it.
22755         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22756         from gimple_fold_builtin_strcpy.  Print inlining stack.
22757         (handle_builtin_stxncpy): Print inlining stack.
22758         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22760 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
22762         PR target/84176
22763         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22764         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22765         and -fcheck-pointer-bounds are used together.
22766         (indirect_thunk_prefix): New enum.
22767         (indirect_thunk_need_prefix): New function.
22768         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
22769         "_nt" instead of "_bnd" for NOTRACK prefix.
22770         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22771         (output_indirect_thunk_function): Likewise.
22772         (): Likewise.
22773         (ix86_code_end): Update output_indirect_thunk_function calls.
22774         (ix86_output_indirect_branch_via_reg): Replace
22775         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22776         (ix86_output_indirect_branch_via_push): Likewise.
22777         (ix86_output_function_return): Likewise.
22778         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22779         incompatible with -fcf-protection=branch and
22780         -fcheck-pointer-bounds.
22782 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22784         PR target/83335
22785         * config/aarch64/aarch64.c (aarch64_print_address_internal):
22786         Change gcc_assert call to output_operand_lossage.
22788 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22790         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22792 2018-02-22  DJ Delorie  <dj@redhat.com>
22793             Sebastian Perta  <sebastian.perta@renesas.com>
22794             Oleg Endo  <olegendo@gcc.gnu.org>
22796         * config/rx/rx.c (rx_rtx_costs): New function.
22797         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22799 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22801         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22803 2018-02-22  Martin Liska  <mliska@suse.cz>
22805         PR driver/83193
22806         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22807         Add "native" as a possible value.
22809 2018-02-22  Martin Liska  <mliska@suse.cz>
22811         PR driver/83193
22812         * config/i386/i386.c (ix86_option_override_internal):
22813         Add "native" as a possible value for -march and -mtune.
22815 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
22817         PR target/84502
22818         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22819         to all type variants.
22821         PR tree-optimization/84503
22822         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22823         width as info->bitpos + info->bitsize - start.
22824         (merged_store_group::merge_overlapping): Simplify width computation.
22825         (check_no_overlap): New function.
22826         (imm_store_chain_info::try_coalesce_bswap): Compute expected
22827         start + width and last_order of the group, fail if check_no_overlap
22828         fails.
22829         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22830         to group if check_no_overlap fails.
22832 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22834         * config/rs6000/altivec.md: Delete contraint arguments to
22835         define_expand, define_split, and define_peephole2, and in
22836         define_insn_and_split if always unused.
22837         * config/rs6000/darwin.md: Ditto.
22838         * config/rs6000/dfp.md: Ditto.
22839         * config/rs6000/rs6000.md: Ditto.
22840         * config/rs6000/sync.md: Ditto.
22841         * config/rs6000/vector.md: Ditto.
22842         * config/rs6000/vsx.md: Ditto.
22844 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22846         * config/rs6000/altivec.md: Write output control strings as braced
22847         blocks instead of double-quoted strings.
22848         * config/rs6000/darwin.md: Ditto.
22849         * config/rs6000/rs6000.md: Ditto.
22850         * config/rs6000/vector.md: Ditto.
22851         * config/rs6000/vsx.md: Ditto.
22853 2018-02-21  Jason Merrill  <jason@redhat.com>
22855         PR c++/84314 - ICE with templates and fastcall attribute.
22856         * attribs.c (build_type_attribute_qual_variant): Remove assert.
22858 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22860         * ipa-cp.c (determine_versionability): Fix comment typos.
22862 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22864         PR c/84229
22865         * ipa-cp.c (determine_versionability): Do not version functions caling
22866         va_arg_pack.
22868 2018-02-21  Martin Liska  <mliska@suse.cz>
22870         PR driver/83193
22871         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22872         Add "native" as a possible value.
22873         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
22874         the macro when native cpu detection is available.
22876 2018-02-21  Martin Liska  <mliska@suse.cz>
22878         PR driver/83193
22879         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
22880         Add "native" as a possible value.
22881         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
22882         when native cpu detection is available.
22884 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
22885             Martin Sebor  <msebor@redhat.com>
22887         PR tree-optimization/84478
22888         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22889         false.
22890         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
22891         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
22892         support which is conservatively correct, for 2 only stay conservative
22893         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
22894         argument to the 2 argument get_range_strlen, adjust 6 arg
22895         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
22896         false.
22897         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
22898         (gimple_fold_builtin_strlen): Pass true as last argument to
22899         get_range_strlen.
22901 2018-02-20  Martin Sebor  <msebor@redhat.com>
22903         PR middle-end/84095
22904         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
22905         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
22906         (builtin_memref::builtin_memref): Factor out parts into
22907         set_base_and_offset and call it.
22909 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22911         PR middle-end/84406
22912         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
22913         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
22914         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
22915         search at the associated MODE_INT.
22917 2018-02-20  Jeff Law  <law@redhat.com>
22919         PR middle-end/82123
22920         PR tree-optimization/81592
22921         PR middle-end/79257
22922         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
22923         for range data rather than using global data.
22924         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
22925         range data rather than using global data.
22926         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
22927         pass it to children as needed.
22928         (struct directive::fmtresult): Similarly.
22929         (struct directive::set_width): Similarly.
22930         (struct directive::set_precision): Similarly.
22931         (format_integer, format_directive, parse_directive): Similarly.
22932         (format_none): Accept unnamed vr_values parameter.
22933         (format_percent, format_floating, format_character): Similarly.
22934         (format_string, format_plain): Similarly.
22935         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
22936         the EVRP range analyzer for range data rather than using global data.
22937         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
22938         gimple-ssa-evrp-analyze.h
22939         (class sprintf_dom_walker): Add after_dom_children member function.
22940         Add evrp_range_analyzer member.
22941         (sprintf_dom_walker::before_dom_children): Call into the EVRP
22942         range analyzer as needed.
22943         (sprintf_dom_walker::after_dom_children): New member function.
22944         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
22945         if not optimizing.
22946         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
22947         (evrp_range_analyzer::pop_to_marker): Likewise.
22949 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22951         PR tree-optimization/84419
22952         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
22953         with the required type if its current type is compatible but
22954         different.
22956 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
22958         PR middle-end/82004
22959         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
22960         after vectorization.
22962 2018-02-20  Martin Liska  <mliska@suse.cz>
22964         PR driver/83193
22965         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
22966         possible values if we don't have a hint.
22968 2018-02-20  Martin Liska  <mliska@suse.cz>
22970         PR c/84310
22971         PR target/79747
22972         * final.c (shorten_branches): Build align_tab array with one
22973         more element.
22974         * opts.c (finish_options): Add alignment option limit check.
22975         (MAX_CODE_ALIGN): Likewise.
22976         (MAX_CODE_ALIGN_VALUE): Likewise.
22977         * doc/invoke.texi: Document maximum allowed option value for
22978         all -falign-* options.
22980 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
22982         PR target/84146
22983         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
22984         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
22985         * var-tracking.c (emit_note_insn_var_location): Remove all references
22986         to NOTE_INSN_CALL_ARG_LOCATION.
22987         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
22988         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
22989         Use copy_rtx_if_shared.
22990         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
22991         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
22992         (dwarf2out_var_location): Remove handling of
22993         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
22994         on call_insn.
22995         * final.c (final_scan_insn): Remove all references to
22996         NOTE_INSN_CALL_ARG_LOCATION.
22997         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
22998         before dumping final insns.
22999         * except.c (emit_note_eh_region_end): Remove all references to
23000         NOTE_INSN_CALL_ARG_LOCATION.
23001         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
23002         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
23003         * config/arc/arc.c (hwloop_optimize): Likewise.
23004         * config/arm/arm.c (create_fix_barrier): Likewise.
23005         * config/s390/s390.c (s390_chunkify_start): Likewise.
23006         * config/sh/sh.c (find_barrier): Likewise.
23007         * config/i386/i386.c (rest_of_insert_endbranch,
23008         ix86_seh_fixup_eh_fallthru): Likewise.
23009         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
23010         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
23011         * config/frv/frv.c (frv_function_prologue): Likewise.
23012         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
23013         reg note.
23014         (note_outside_basic_block_p): Remove all references to
23015         NOTE_INSN_CALL_ARG_LOCATION.
23016         * gengtype.c (adjust_field_rtx_def): Likewise.
23017         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
23018         Likewise.
23019         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
23020         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
23022         PR c++/84444
23023         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
23024         is ADDR_EXPR.
23026         PR tree-optimization/84452
23027         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
23028         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
23029         is NULL.
23031 2018-02-19  Martin Liska  <mliska@suse.cz>
23033         PR sanitizer/82183
23034         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
23036 2018-02-19  Martin Liska  <mliska@suse.cz>
23037             Richard Sandiford  <richard.sandiford@linaro.org>
23039         PR tree-optimization/82491
23040         * gimple-fold.c (get_base_constructor): Make earlier bail out
23041         to prevent ubsan.
23043 2018-02-19  Carl Love  <cel@us.ibm.com>
23045         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
23046         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
23047         BU_P8V_OVERLOAD_1.
23048         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
23049         P8V_BUILTIN_VEC_NEG.
23051 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
23053         * config/rl78/rl78.md (movdf): New define expand.
23055 2018-02-19  Martin Liska  <mliska@suse.cz>
23057         PR other/80589
23058         * doc/invoke.texi: Fix typo.
23059         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
23061 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
23063         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
23064         handle rs6000_single_float and rs6000_double_float specially for
23065         e500 family CPUs.
23067 2018-02-16  Jeff Law  <law@redhat.com>
23069         * config/rx/rx.c (add_pop_cfi_notes): New function.;
23070         (pop_regs): Use it.
23072 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
23074         PR ipa/84425
23075         * ipa-inline.c (inline_small_functions): Fix a typo.
23077 2018-02-16  Nathan Sidwell  <nathan@acm.org>
23079         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
23081 2018-02-16  Carl Love  <cel@us.ibm.com>
23083         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
23084         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
23085         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
23086         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
23087         expansion to P8V_BUILTIN_VEC_FLOAT2.
23089 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
23091         PR rtl-optimization/70023
23092         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
23093         src_regno into account.
23095 2018-02-16  Carl Love  <cel@us.ibm.com>
23097         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
23098         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
23099         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
23100         * config/rs6000/rs6000.c: Remove case statements for
23101         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
23102         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
23103         and P9V_BUILTIN_VEC_VINSERT4B.
23104         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
23105         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
23106         * config/rs6000/vsx.md:
23107         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
23108         vec_insert4b.
23110 2018-02-16  Carl Love  <cel@us.ibm.com>
23112         * config/rs6000/altivec.h: Add builtin names vec_extract4b
23113         vec_insert4b.
23114         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
23115         definitions.
23116         * config/rs6000/rs6000-c.c: Add the definitions for
23117         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
23118         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
23119         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
23120         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
23121         definition for insert4b and define insn *insert3b_internal.
23122         * doc/extend.texi: Add documentation for vec_extract4b.
23124 2018-02-16  Nathan Sidwell  <nathan@acm.org>
23126         * doc/extend.texi (Backwards Compatibility): Mention friend
23127         injection.  Note for-scope is deprecated.
23128         * doc/invoke.texi (-ffriend-injection): Deprecate.
23130 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
23132         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
23133         that moved to I2, also allow destinations that are a paradoxical
23134         subreg (instead of a normal reg).
23136 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
23138         PR target/83831
23139         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
23140         to QImode.
23142 2018-02-16  Richard Biener  <rguenther@suse.de>
23144         PR tree-optimization/84037
23145         PR tree-optimization/84016
23146         PR target/82862
23147         * config/i386/i386.c (ix86_builtin_vectorization_cost):
23148         Adjust vec_construct for the fact we need additional higher latency
23149         128bit inserts for AVX256 and AVX512 vector builds.
23150         (ix86_add_stmt_cost): Scale vector construction cost for
23151         elementwise loads.
23153 2018-02-16  Richard Biener  <rguenther@suse.de>
23155         PR tree-optimization/84417
23156         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
23157         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
23158         (non_rewritable_lvalue_p): Likewise, use poly-ints.
23160 2018-02-16  Martin Liska  <mliska@suse.cz>
23162         PR sanitizer/84307
23163         * internal-fn.def (ASAN_CHECK): Set proper flags.
23164         (ASAN_MARK): Likewise.
23166 2018-02-16  Julia Koval  <julia.koval@intel.com>
23168         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
23169         from PTA_CANNONLAKE.
23171 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
23173         PR target/84272
23174         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
23175         Use ++iter rather than iter++ for std::list iterators.
23176         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
23177         defer deleting them until all nodes in the forest are processed.  Do
23178         free even leaf nodes.  Change to_process into auto_vec.
23180         PR bootstrap/84405
23181         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
23182         * vec.h (vec_default_construct): Use memset instead of placement new
23183         if BROKEN_VALUE_INITIALIZATION is defined.
23184         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
23185         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
23186         is defined.
23188         PR rtl-optimization/83723
23189         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
23190         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
23191         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
23192         recursive calls.
23193         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
23194         callers.
23195         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
23197 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
23199         PR rtl-optimization/81443
23200         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
23201         from inner REGs to paradoxical SUBREGs.
23203 2018-02-16  Richard Biener  <rguenther@suse.de>
23205         PR tree-optimization/84399
23206         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23207         For operands we can analyze at their definition make sure we can
23208         analyze them at each use as well.
23210 2018-02-16  Richard Biener  <rguenther@suse.de>
23212         PR tree-optimization/84190
23213         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
23214         volatile accesses if the decl isn't volatile.
23216 2018-02-15  Jason Merrill  <jason@redhat.com>
23218         PR c++/84314 - ICE with templates and fastcall attribute.
23219         * attribs.c (build_type_attribute_qual_variant): Don't clobber
23220         TYPE_CANONICAL on an existing type.
23222 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
23224         PR tree-optimization/84383
23225         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
23226         dstoff nor call operand_equal_p if dstbase is NULL.
23228         PR tree-optimization/84334
23229         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
23230         also a CONSTANT_CLASS_P, punt.
23232 2018-02-14  Jim Wilson  <jimw@sifive.com>
23234         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
23235         first SMALL_OPERAND check.  New local min_second_step.  Move assert
23236         to where locals are set.  Add TARGET_RVC support.
23237         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
23239 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
23241         * doc/invoke.texi: Correct -Wformat-overflow code sample.
23243 2018-02-14  Martin Sebor  <msebor@redhat.com>
23245         PR tree-optimization/83698
23246         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
23247         arrays constrain the offset range to their bounds.
23248         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
23249         (builtin_access::overlap): Avoid setting the size of overlap if it's
23250         already been set.
23251         (maybe_diag_overlap): Also consider arrays when deciding what values
23252         of offsets to include in diagnostics.
23254 2018-02-14  Martin Sebor  <msebor@redhat.com>
23256         PR c/84108
23257         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
23258         that correspond to the kind of a declaration.
23260 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
23262         PR target/83984
23263         * config/pa/pa.md: Load address of PIC label using the linkage table
23264         if the label is nonlocal.
23266 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23268         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
23269         warning message if user requests -maltivec=be.
23270         * doc/invoke.texi: Document deprecation of -maltivec=be.
23272 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
23274         PR target/84220
23275         * config/rs6000/rs6000-c.c: Update definitions for
23276         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
23277         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
23279 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23281         PR target/84239
23282         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
23283         add _get_ssp intrinsics. Remove argument from
23284         __builtin_ia32_rdssp[d|q].
23285         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
23286         * config/i386/i386-builtin.def: Remove argument from
23287         __builtin_ia32_rdssp[d|q].
23288         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
23289         ix86_expand_special_args_builtin for _rdssp[d|q].
23290         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
23291         Clear register before usage.
23292         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
23293         Add documentation for new _get_ssp and _inc_ssp intrinsics.
23295 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
23297         PR tree-optimization/84357
23298         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
23299         operand 1 of an ARRAY_REF too.
23301 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
23303         PR target/83831
23304         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
23305         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23306         declarations.
23307         (set_of_reg): New struct.
23308         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
23309         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
23310         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23311         functions.
23312         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
23313         Split into bitclr, bitset, bitinvert patterns if appropriate.
23314         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
23315         use rx_fuse_in_memory_bitop.
23316         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
23317         to named insn, correct maximum insn length.
23319 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
23321         PR target/79242
23322         * machmode.def: Define a complex mode for PARTIAL_INT.
23323         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
23324         MODE_PARTIAL_INT.
23325         * doc/rtl.texi: Document CSPImode.
23326         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
23327         handling.
23328         (msp430_hard_regno_nregs_with_padding): Likewise.
23330 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
23332         PR target/84279
23333         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
23335 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
23337         PR rtl-optimization/84169
23338         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
23339         we generated a parallel as new i3 and we split that to new i2 and i3
23340         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
23341         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
23342         those to i2, not i1.  Partially rewrite this scan code.
23344 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23346         PR c/82210
23347         * stor-layout.c (place_field): For variable length fields, adjust
23348         offset_align afterwards not just based on the field's alignment,
23349         but also on the size.
23351         PR middle-end/84309
23352         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
23353         of exps and logs in the use_exp2 case.
23355 2018-02-13  Jeff Law  <law@redhat.com>
23357         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
23358         entry for "vector".
23360         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
23361         ARGS as unused.
23363 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
23365         PR debug/84342
23366         PR debug/84319
23367         * common.opt (gas-loc-support, gas-locview-support): New.
23368         (ginline-points, ginternal-reset-location-views): New.
23369         * doc/invoke.texi: Document them.  Use @itemx where intended.
23370         (gvariable-location-views): Adjust.
23371         * target.def (reset_location_view): New.
23372         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
23373         (TARGET_RESET_LOCATION_VIEW): New.
23374         * doc/tm.texi: Rebuilt.
23375         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
23376         (dwarf2out_default_as_locview_support): New.
23377         (output_asm_line_debug_info): Use option variables.
23378         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
23379         (output_loc_list): Likewise.
23380         (add_high_low_attributes): Check option variables.
23381         Don't output entry view attribute in strict mode.
23382         (gen_inlined_subroutine_die): Check option variables.
23383         (dwarf2out_inline_entry): Likewise.
23384         (init_sections_and_labels): Likewise.
23385         (dwarf2out_early_finish): Likewise.
23386         (maybe_reset_location_view): New, from...
23387         (dwarf2out_var_location): ... here.  Call it.
23388         * debug.h (dwarf2out_default_as_loc_support): Declare.
23389         (dwarf2out_default_as_locview_support): Declare.
23390         * hooks.c (hook_int_rtx_insn_0): New.
23391         * hooks.h (hook_int_rtx_insn_0): Declare.
23392         * toplev.c (process_options): Take -gas-loc-support and
23393         -gas-locview-support from dwarf2out.  Enable
23394         -gvariable-location-views by default only with locview
23395         assembler support.  Enable -ginternal-reset-location-views by
23396         default only if the target defines the corresponding hook.
23397         Enable -ginline-points by default if location views are
23398         enabled; force it disabled if statement frontiers are
23399         disabled.
23400         * tree-inline.c (expand_call_inline): Check option variables.
23401         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
23403 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
23405         PR tree-optimization/84321
23406         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
23407         handling.  Also check whether the anti-range contains any values
23408         that satisfy the mask; switch to a VR_RANGE if not.
23410 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
23412         PR sanitizer/84340
23413         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
23415 2018-02-13  Martin Jambor  <mjambor@suse.cz>
23417         PR c++/83990
23418         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
23419         of call statements, also set location of a load to a temporary.
23421 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
23423         * config/rl78/rl78.c (add_vector_labels): New function.
23424         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
23425         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
23426         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
23427         which checks that no arguments are passed.
23428         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
23429         * doc/extend.texi: Documentation for the new attribute.
23431 2018-02-13  Andreas Schwab  <schwab@suse.de>
23433         * config/riscv/linux.h (CPP_SPEC): Define.
23435 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23437         PR target/84335
23438         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
23439         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
23440         OPTION_MASK_ISA_AES as first argument to def_builtin_const
23441         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
23442         instead of OPTION_MASK_ISA_PCLMUL as first argument to
23443         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
23444         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
23445         temporarily for AES and PCLMUL builtins.
23447         PR tree-optimization/84339
23448         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
23449         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
23450         Formatting fixes.
23452         PR middle-end/84309
23453         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
23454         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
23455         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
23456         inline function.
23457         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
23458         inline function.
23459         * omp-simd-clone.h: New file.
23460         * omp-simd-clone.c: Include omp-simd-clone.h.
23461         (expand_simd_clones): No longer static.
23462         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
23463         cgraph.h and omp-simd-clone.h.
23464         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
23465         (vect_recog_widen_shift_pattern): Formatting fix.
23466         (vect_pattern_recog_1): Don't check optab for calls.
23468         PR target/84336
23469         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
23470         operands[2] into a REG before using gen_lowpart on it.
23472 2018-02-12  Jeff Law  <law@redhat.com>
23474         PR target/83760
23475         * config/sh/sh.c (find_barrier): Consider a sibling call
23476         a barrier as well.
23478         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
23479         successfully back substituting a reg.
23481 2018-02-12  Richard Biener  <rguenther@suse.de>
23483         PR tree-optimization/84037
23484         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
23485         parameter, move visited init to caller.
23486         (vect_slp_analyze_operations): Separate cost from validity
23487         check, initialize visited once for all instances.
23488         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
23489         for all instances.
23490         * tree-vect-stmts.c (vect_model_simple_cost): Make early
23491         out an assert.
23492         (vect_model_promotion_demotion_cost): Likewise.
23493         (vectorizable_bswap): Guard cost modeling with !slp_node
23494         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23495         SLP stmts.
23496         (vectorizable_call): Likewise.
23497         (vectorizable_conversion): Likewise.
23498         (vectorizable_assignment): Likewise.
23499         (vectorizable_shift): Likewise.
23500         (vectorizable_operation): Likewise.
23501         (vectorizable_store): Likewise.
23502         (vectorizable_load): Likewise.
23503         (vectorizable_condition): Likewise.
23504         (vectorizable_comparison): Likewise.
23506 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
23508         PR sanitizer/84307
23509         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23510         (ASAN_MARK): Fix fnspec to account for return value, change pointer
23511         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23513 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23515         PR middle-end/83665
23516         * params.def (inline-min-speedup): Increase from 8 to 15.
23517         (max-inline-insns-auto): Decrease from 40 to 30.
23518         * ipa-split.c (consider_split): Add some buffer for function to
23519         be considered inlining candidate.
23520         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23521         default values.
23523 2018-02-12  Richard Biener  <rguenther@suse.de>
23525         PR tree-optimization/84037
23526         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23527         matched stmts if we cannot swap the non-matched ones.
23529 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23531         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23532         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23533         _mm_maskz_scalef_round_ss): New intrinsics.
23534         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23535         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23536         __builtin_ia32_scalefss_round): Remove.
23537         (__builtin_ia32_scalefsd_mask_round,
23538         __builtin_ia32_scalefss_mask_round): New intrinsics.
23539         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23540         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23541         ((match_operand:VF_128 2 "<round_nimm_predicate>"
23542         "<round_constraint>")): Changed to ...
23543         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23544         "<round_scalar_constraint>")): ... this.
23545         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23546         %0, %1, %2<round_op3>}"): Changed to ...
23547         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23548         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23549         %2<round_scalar_mask_op3>}"): ... this.
23550         * config/i386/subst.md (round_scalar_nimm_predicate): New.
23552 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23554         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23555         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23556         (_mm_maskz_sqrt_round_ss): New intrinsics.
23557         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23558         (__builtin_ia32_sqrtsd_mask_round)
23559         (__builtin_ia32_sqrtss_mask_round): New builtins.
23560         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23561         (__builtin_ia32_sqrtss_round): Remove.
23562         (__builtin_ia32_sqrtsd_mask_round)
23563         (__builtin_ia32_sqrtss_mask_round): New builtins.
23564         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23565         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23566         ((match_operand:VF_128 1 "vector_operand"
23567         "xBm,<round_constraint>")): Changed to ...
23568         ((match_operand:VF_128 1 "vector_operand"
23569         "xBm,<round_scalar_constraint>")): ... this.
23570         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23571         %0, %2, %<iptr>1<round_op3>}): Changed to ...
23572         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23573         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23574         %<iptr>1<round_scalar_mask_op3>}): ... this.
23575         ((set_attr "prefix" "<round_prefix>")): Changed to ...
23576         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23578 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
23580         PR target/84266
23581         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23582         Cast vec_cmpeq result to correct type.
23583         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23584         Cast vec_cmpgt result to correct type.
23586 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
23588         * final.c (final_scan_insn_1): Renamed from...
23589         (final_scan_insn): ... this.  New wrapper, to recover
23590         seen from the outermost call in recursive ones.
23591         * config/sparc/sparc.c (output_return): Drop seen from call.
23592         (output_sibcall): Likewise.
23593         * config/visium/visium.c (output_branch): Likewise.
23595 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
23597         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23598         function label.
23600 2018-02-10  Alan Modra  <amodra@gmail.com>
23602         PR target/84300
23603         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23604         Specify LR as an input.
23606 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
23608         PR sanitizer/83987
23609         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23610         remove_member_access_dummy_vars): New functions.
23611         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23612         lower_omp_1, execute_lower_omp): Use them.
23614         PR rtl-optimization/84308
23615         * shrink-wrap.c (spread_components): Release todo vector.
23617 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
23619         PR rtl-optimization/57193
23620         * ira-color.c (struct allocno_color_data): Add member
23621         conflict_allocno_hard_prefs.
23622         (update_conflict_allocno_hard_prefs): New.
23623         (bucket_allocno_compare_func): Add a preference based on
23624         conflict_allocno_hard_prefs.
23625         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23626         (color_allocnos): Remove a dead code.  Initiate
23627         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
23629 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23631         PR target/84226
23632         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23633         constraint from =wa to wa.  Avoid a subreg on the output operand,
23634         instead use a pseudo and subreg it in a move.
23635         (p9_xxbrd_<mode>): Changed to ...
23636         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23637         (p9_xxbrd_v2df): New expander.
23638         (p9_xxbrw_<mode>): Changed to ...
23639         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23640         (p9_xxbrw_v4sf): New expander.
23642 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23644         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23646 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
23648         PR target/83926
23649         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23650         multiply in 32-bit mode.
23651         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23652         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23653         mode.
23655 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23657         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23658         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23659         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23660         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23662 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
23664         PR lto/84213
23665         * dwarf2out.c (is_trivial_indirect_ref): New function.
23666         (dwarf2out_late_global_decl): Do not generate a location
23667         attribute for variables that have a non-trivial DECL_VALUE_EXPR
23668         and that are not defined in the current unit.
23670 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23672         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23673         instead of a libcall for UNORDERED.
23675 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
23677         PR target/82641
23678         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23679         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23681 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23683         PR target/PR84295
23684         * config/s390/s390.c (s390_set_current_function): Invoke
23685         s390_indirect_branch_settings also if fndecl didn't change.
23687 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23689         * config/rs6000/rs6000.md (blockage): Set length to zero.
23691 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23693         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23695 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23697         PR sanitizer/84285
23698         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23699         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23700         -static-lib*san.
23702         PR debug/84252
23703         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23704         PARALLEL incoming that failed vt_get_decl_and_offset check.
23706         PR middle-end/84237
23707         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23708         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23709         TREE_READONLY bit.
23710         (get_variable_section): For decls in named .bss* sections pass true as
23711         second argument to bss_initializer_p.
23713 2018-02-09  Marek Polacek  <polacek@redhat.com>
23714             Jakub Jelinek  <jakub@redhat.com>
23716         PR c++/83659
23717         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23718         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
23719         Sync some changes from cxx_fold_indirect_ref.
23721 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23723         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23724         markers.
23725         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23726         (BLOCK_INLINE_ENTRY_LABEL): New.
23727         (dwarf2out_var_location): Disregard inline entry markers.
23728         (inline_entry_data): New struct.
23729         (inline_entry_data_hasher): New hashtable type.
23730         (inline_entry_data_hasher::hash): New.
23731         (inline_entry_data_hasher::equal): New.
23732         (inline_entry_data_table): New variable.
23733         (add_high_low_attributes): Add DW_AT_entry_pc and
23734         DW_AT_GNU_entry_view attributes if a pending entry is found
23735         in inline_entry_data_table.  Add old entry_pc attribute only
23736         if debug nonbinding markers are disabled.
23737         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23738         markers are enabled.
23739         (block_within_block_p, dwarf2out_inline_entry): New.
23740         (dwarf2out_finish): Check that no entries remained in
23741         inline_entry_data_table.
23742         * final.c (reemit_insn_block_notes): Handle inline entry notes.
23743         (final_scan_insn, notice_source_line): Likewise.
23744         (rest_of_clean_state): Skip inline entry markers.
23745         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23746         markers.
23747         * gimple.c (gimple_build_debug_inline_entry): New.
23748         * gimple.h (enum gimple_debug_subcode): Add
23749         GIMPLE_DEBUG_INLINE_ENTRY.
23750         (gimple_build_debug_inline_entry): Declare.
23751         (gimple_debug_inline_entry_p): New.
23752         (gimple_debug_nonbind_marker_p): Adjust.
23753         * insn-notes.def (INLINE_ENTRY): New.
23754         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23755         inline entry marker notes.
23756         (print_insn): Likewise.
23757         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23758         (INSN_DEBUG_MARKER_KIND): Likewise.
23759         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23760         * tree-inline.c (expand_call_inline): Build and insert
23761         debug_inline_entry stmt.
23762         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23763         inline entry blocks early, if nonbind markers are enabled.
23764         (dump_scope_block): Dump fragment info.
23765         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23766         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23767         (gimple_build_debug_inline_entry): New.
23768         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23769         Enable/disable inline entry points too.
23770         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23771         (DEBUG_INSN): Describe inline entry markers.
23773         * common.opt (gvariable-location-views): New.
23774         (gvariable-location-views=incompat5): New.
23775         * config.in: Rebuilt.
23776         * configure: Rebuilt.
23777         * configure.ac: Test assembler for view support.
23778         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23779         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23780         * dwarf2out.c (var_loc_view): New typedef.
23781         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23782         (dwarf2out_locviews_in_attribute): New.
23783         (dwarf2out_locviews_in_loclist): New.
23784         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23785         (enum dw_line_info_opcode): Add LI_adv_address.
23786         (struct dw_line_info_table): Add view.
23787         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23788         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23789         (zero_view_p): New variable.
23790         (ZERO_VIEW_P): New macro.
23791         (output_asm_line_debug_info): New.
23792         (struct var_loc_node): Add view.
23793         (add_AT_view_list, AT_loc_list): New.
23794         (add_var_loc_to_decl): Add view param.  Test it against last.
23795         (new_loc_list): Add view params.  Record them.
23796         (AT_loc_list_ptr): Handle loc and view lists.
23797         (view_list_to_loc_list_val_node): New.
23798         (print_dw_val): Handle dw_val_class_view_list.
23799         (size_of_die): Likewise.
23800         (value_format): Likewise.
23801         (loc_list_has_views): New.
23802         (gen_llsym): Set vl_symbol too.
23803         (maybe_gen_llsym, skip_loc_list_entry): New.
23804         (dwarf2out_maybe_output_loclist_view_pair): New.
23805         (output_loc_list): Output view list or entries too.
23806         (output_view_list_offset): New.
23807         (output_die): Handle dw_val_class_view_list.
23808         (output_dwarf_version): New.
23809         (output_compilation_unit_header): Use it.
23810         (output_skeleton_debug_sections): Likewise.
23811         (output_rnglists, output_line_info): Likewise.
23812         (output_pubnames, output_aranges): Update version comments.
23813         (output_one_line_info_table): Output view numbers in asm comments.
23814         (dw_loc_list): Determine current endview, pass it to new_loc_list.
23815         Call maybe_gen_llsym.
23816         (loc_list_from_tree_1): Adjust.
23817         (add_AT_location_description): Create view list attribute if
23818         needed, check it's absent otherwise.
23819         (convert_cfa_to_fb_loc_list): Adjust.
23820         (maybe_emit_file): Call output_asm_line_debug_info for test.
23821         (dwarf2out_var_location): Reset views as needed.  Precompute
23822         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
23823         attribute.  Set view.
23824         (new_line_info_table): Reset next view.
23825         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23826         (dwarf2out_source_line): Likewise.  Output view resets and labels to
23827         the assembler, or select appropriate line info opcodes.
23828         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23829         (optimize_string_length): Catch it.  Adjust.
23830         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
23831         dw_val_class_view_list, and remove it if no longer needed.
23832         (hash_loc_list): Hash view numbers.
23833         (loc_list_hasher::equal): Compare them.
23834         (optimize_location_lists): Check whether a view list symbol is
23835         needed, and whether the locview attribute is present, and
23836         whether they match.  Remove the locview attribute if no longer
23837         needed.
23838         (index_location_lists): Call skip_loc_list_entry for test.
23839         (dwarf2out_finish): Call output_asm_line_debug_info for test.
23840         Use output_dwarf_version.
23841         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23842         (struct dw_val_node): Add val_view_list.
23843         * final.c (SEEN_NEXT_VIEW): New.
23844         (set_next_view_needed): New.
23845         (clear_next_view_needed): New.
23846         (maybe_output_next_view): New.
23847         (final_start_function): Rename to...
23848         (final_start_function_1): ... this.  Take pointer to FIRST,
23849         add SEEN parameter.  Emit param bindings in the initial view.
23850         (final_start_function): Reintroduce SEEN-less interface.
23851         (final): Rename to...
23852         (final_1): ... this.  Take SEEN parameter.  Output final pending
23853         next view at the end.
23854         (final): Reintroduce seen-less interface.
23855         (final_scan_insn): Output pending next view before switching
23856         sections or ending a block.  Mark the next view as needed when
23857         outputting variable locations.  Notify debug backend of section
23858         changes, and of location view changes.
23859         (rest_of_handle_final): Adjust.
23860         * toplev.c (process_options): Autodetect value for debug variable
23861         location views option.  Warn on incompat5 without -gdwarf-5.
23862         * doc/invoke.texi (gvariable-location-views): New.
23863         (gvariable-location-views=incompat5): New.
23864         (gno-variable-location-views): New.
23866 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
23868         PR tree-optimization/84136
23869         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23870         that the result of find_edge is non-NULL.
23872 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23874         PR target/83008
23875         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
23876         storing integer register in SImode.  Fix cost of 256 and 512
23877         byte aligned SSE register store.
23879 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23881         * config/i386/i386.c (ix86_multiplication_cost): Fix
23882         multiplication cost for TARGET_AVX512DQ.
23884 2018-02-08  Marek Polacek  <polacek@redhat.com>
23886         PR tree-optimization/84238
23887         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
23888         get_range_strlen.
23890 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23892         PR tree-optimization/84265
23893         * tree-vect-stmts.c (vectorizable_store): Don't treat
23894         VMAT_CONTIGUOUS accesses as grouped.
23895         (vectorizable_load): Likewise.
23897 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23899         PR tree-optimization/81635
23900         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
23901         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
23902         (test_round_for_mask): New functions.
23903         (wide_int_cc_tests): Call test_round_for_mask.
23904         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
23905         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
23906         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
23907         range returned by get_range_info.
23909 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23911         PR ipa/81360
23912         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
23913         * symtab.c: Include builtins.h
23914         (symtab_node::output_to_lto_symbol_table_p): Move here
23915         from lto-streamer-out.c:output_symbol_p.
23916         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
23917         (output_symbol_p): Move all logic to symtab.c
23918         (produce_symtab): Update.
23920 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23922         * config/s390/s390-opts.h (enum indirect_branch): Define.
23923         * config/s390/s390-protos.h (s390_return_addr_from_memory)
23924         (s390_indirect_branch_via_thunk)
23925         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
23926         (enum s390_indirect_branch_type): Define.
23927         * config/s390/s390.c (struct s390_frame_layout, struct
23928         machine_function): Remove.
23929         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
23930         (indirect_branch_table_label_no, indirect_branch_table_name):
23931         Define variables.
23932         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
23933         (enum s390_indirect_branch_option): Define.
23934         (s390_return_addr_from_memory): New function.
23935         (s390_handle_string_attribute): New function.
23936         (s390_attribute_table): Add new attribute handler.
23937         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
23938         (s390_indirect_branch_via_thunk): New function.
23939         (s390_indirect_branch_via_inline_thunk): New function.
23940         (s390_function_ok_for_sibcall): When jumping via thunk disallow
23941         sibling call optimization for non z10 compiles.
23942         (s390_emit_call): Force indirect branch target to be a single
23943         register.  Add r1 clobber for non-z10 compiles.
23944         (s390_emit_epilogue): Emit return jump via return_use expander.
23945         (s390_reorg): Handle JUMP_INSNs as execute targets.
23946         (s390_option_override_internal): Perform validity checks for the
23947         new command line options.
23948         (s390_indirect_branch_attrvalue): New function.
23949         (s390_indirect_branch_settings): New function.
23950         (s390_set_current_function): Invoke s390_indirect_branch_settings.
23951         (s390_output_indirect_thunk_function):  New function.
23952         (s390_code_end): Implement target hook.
23953         (s390_case_values_threshold): Implement target hook.
23954         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
23955         macros.
23956         * config/s390/s390.h (struct s390_frame_layout)
23957         (struct machine_function): Move here from s390.c.
23958         (TARGET_INDIRECT_BRANCH_NOBP_RET)
23959         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
23960         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
23961         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
23962         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
23963         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
23964         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
23965         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
23966         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
23967         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
23968         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
23969         (mnemonic attribute): Add values which aren't recognized
23970         automatically.
23971         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
23972         pattern for branch conversion.  Fix mnemonic attribute.
23973         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
23974         indirect branch via thunk if requested.
23975         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
23976         ("*indirect_jump"): Disable for branch conversion using out of
23977         line thunks.
23978         ("indirect_jump_via_thunk<mode>_z10")
23979         ("indirect_jump_via_thunk<mode>")
23980         ("indirect_jump_via_inlinethunk<mode>_z10")
23981         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
23982         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
23983         ("casesi_jump_via_inlinethunk<mode>_z10")
23984         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
23985         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
23986         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
23987         ("*indirect2_jump"): Disable for branch conversion.
23988         ("casesi_jump"): Turn into expander and expand patterns for branch
23989         conversion.
23990         ("return_use"): New expander.
23991         ("*return"): Emit return via thunk and rename it to ...
23992         ("*return<mode>"): ... this one.
23993         * config/s390/s390.opt: Add new options and and enum for the
23994         option values.
23996 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23998         * lra-constraints.c (match_reload): Unconditionally use
23999         gen_lowpart_SUBREG, rather than selecting between that
24000         and equivalent gen_rtx_SUBREG code.
24002 2018-02-08  Richard Biener  <rguenther@suse.de>
24004         PR tree-optimization/84233
24005         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
24006         changed flag instead of boguously re-using phi_inserted.
24008 2018-02-08  Martin Jambor  <mjambor@suse.cz>
24010         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
24011         static local variables.
24013 2018-02-08  Richard Biener  <rguenther@suse.de>
24015         PR tree-optimization/84278
24016         * tree-vect-stmts.c (vectorizable_store): When looking for
24017         smaller vector types to perform grouped strided loads/stores
24018         make sure the mode is supported by the target.
24019         (vectorizable_load): Likewise.
24021 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
24023         * config/aarch64/aarch64.c (aarch64_components_for_bb):
24024         Increase LDP/STP opportunities by adding adjacent callee-saves.
24026 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
24028         PR rtl-optimization/84068
24029         PR rtl-optimization/83459
24030         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
24032 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
24034         PR tree-optimization/84224
24035         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
24036         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
24037         non-zero arguments.
24039 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
24041         PR target/84113
24042         * config/rs6000/altivec.md (*restore_world): Remove LR use.
24043         * config/rs6000/predicates.md (restore_world_operation): Adjust op
24044         count, remove one USE.
24046 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24048         * doc/install.texi (Configuration): Document the
24049         --with-long-double-format={ibm,ieee} PowerPC configuration
24050         options.
24052         PR target/84154
24053         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
24054         Convert from define_expand to be define_insn_and_split.  Rework
24055         float/double/_Float128 conversions to QI/HI/SImode to work with
24056         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
24057         conversions to QI/HImode types did a store and then a load to
24058         truncate the value.  For conversions to VSX registers, don't split
24059         the insn, instead emit the code directly.  Use the code iterator
24060         any_fix to combine signed and unsigned conversions.
24061         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
24062         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
24063         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24064         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
24065         (fix_<mode>di2_hw): Likewise.
24066         (fixuns_<mode>di2_hw): Likewise.
24067         (fix_<mode>si2_hw): Likewise.
24068         (fixuns_<mode>si2_hw): Likewise.
24069         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
24070         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24071         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
24072         fix<uns>_trunc<SFDF:mode>si2_p8.
24073         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
24074         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
24075         (fix<uns>_<mode>_mem): Likewise.
24076         (fctiw<u>z_<mode>_mem): Likewise.
24077         (fix<uns>_<mode>_mem): Likewise.
24078         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
24079         the register allocator from doing a direct move to the GPRs to do
24080         a store, and instead use the ISA 3.0 store byte/half-word from
24081         vector register instruction.  For IEEE 128-bit floating point,
24082         also optimize stores of 32-bit ints.
24083         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
24085 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
24087         * genextract.c (push_pathstr_operand): New function to support
24088         [a-zA-Z].
24089         (walk_rtx): Call push_pathstr_operand.
24090         (print_path): Support [a-zA-Z].
24092 2018-02-07  Richard Biener  <rguenther@suse.de>
24094         PR tree-optimization/84037
24095         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
24096         (cse_and_gimplify_to_preheader): Declare.
24097         (vect_get_place_in_interleaving_chain): Likewise.
24098         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24099         ivexpr_map.
24100         (_loop_vec_info::~_loop_vec_info): Delete it.
24101         (cse_and_gimplify_to_preheader): New function.
24102         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
24103         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
24104         (vectorizable_load): Likewise.  For grouped stores always base
24105         the IV on the first element.
24106         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
24107         condition before gimplifying.
24109 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24111         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
24112         *DIV_EXPR and *MOD_EXPR.
24114 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
24116         PR target/84248
24117         * config/i386/i386.c (ix86_option_override_internal): Mask out
24118         the CF_SET bit when checking -fcf-protection.
24120 2018-02-07  Tom de Vries  <tom@codesourcery.com>
24122         PR libgomp/84217
24123         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
24124         enough.
24126 2018-02-07  Richard Biener  <rguenther@suse.de>
24128         PR tree-optimization/84204
24129         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
24130         this place.
24132         PR tree-optimization/84205
24133         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
24134         special-case isl_ast_op_zdiv_r.
24136         PR tree-optimization/84223
24137         * graphite-scop-detection.c (gather_bbs::before_dom_children):
24138         Only add conditions from within the region.
24139         (gather_bbs::after_dom_children): Adjust.
24141 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
24143         PR target/84209
24144         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
24145         * config/avr/avr.md: Only post-reload split REG-REG moves if
24146         either register is GENERAL_REG_P.
24148 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24150         PR tree-optimization/84235
24151         * tree-ssa-scopedtables.c
24152         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
24153         if the subtraction is performed in floating point type where NaNs are
24154         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
24155         build 1.  Formatting fix.
24157 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
24159         PR target/84146
24160         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
24161         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
24162         and skip it regardless of bb boundaries.  Use CALL_P macro,
24163         don't test INSN_P (insn) together with CALL_P or JUMP_P check
24164         unnecessarily, formatting fix.
24166 2018-02-06  Michael Collison  <michael.collison@arm.com>
24168         * config/arm/thumb2.md:
24169         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
24170         (*thumb_mov_notscc): Ditto.
24172 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
24174         PR target/84154
24175         * config/rs6000/rs6000.md (su code attribute): Use "u" for
24176         unsigned_fix, not "s".
24178 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24180         * configure.ac (gcc_fn_eh_frame_ro): New function.
24181         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
24182         correct .eh_frame permissions.
24183         * configure: Regenerate.
24185 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
24187         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
24188         irrelevant options.
24190 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24192         * config/rs6000/rs6000.c (rs6000_option_override_internal):
24193         Display warning message for -mno-speculate-indirect-jumps.
24195 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
24197         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
24198         Undocumented.
24199         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
24201 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
24203         PR tree-optimization/84225
24204         * tree-eh.c (find_trapping_overflow): Only call
24205         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
24207 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24209         PR target/84145
24210         * config/i386/i386.c: Reimplement the check of possible options
24211         -mibt/-mshstk conbination. Change error messages.
24212         * doc/invoke.texi: Fix a typo: remove extra '='.
24214 2018-02-06  Marek Polacek  <polacek@redhat.com>
24216         PR tree-optimization/84228
24217         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
24219 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
24221         PR target/82641
24222         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
24223         emitted arch directives.
24224         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
24225         __ARM_FEATURE_COPROC before changing architectures.
24227 2018-02-06  Richard Biener  <rguenther@suse.de>
24229         * config/i386/i386.c (print_reg): Fix typo.
24230         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
24232 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
24234         * configure: Regenerate.
24236 2018-02-05  Martin Sebor  <msebor@redhat.com>
24238         PR tree-optimization/83369
24239         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
24240         inlining context.
24242 2018-02-05  Martin Liska  <mliska@suse.cz>
24244         * doc/invoke.texi: Cherry-pick upstream r323995.
24246 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
24248         * ira.c (ira_init_register_move_cost): Adjust comment.
24250 2018-02-05  Martin Liska  <mliska@suse.cz>
24252         PR gcov-profile/84137
24253         * doc/gcov.texi: Fix typo in documentation.
24255 2018-02-05  Martin Liska  <mliska@suse.cz>
24257         PR gcov-profile/83879
24258         * doc/gcov.texi: Document necessity of --dynamic-list-data when
24259         using dlopen functionality.
24261 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
24263         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
24264         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
24265         _mm_maskz_range_ss, _mm_mask_range_round_ss,
24266         _mm_maskz_range_round_ss): New intrinsics.
24267         (__builtin_ia32_rangesd128_round)
24268         (__builtin_ia32_rangess128_round): Remove.
24269         (__builtin_ia32_rangesd128_mask_round,
24270         __builtin_ia32_rangess128_mask_round): New builtins.
24271         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
24272         __builtin_ia32_rangess128_round): Remove.
24273         (__builtin_ia32_rangesd128_mask_round,
24274         __builtin_ia32_rangess128_mask_round): New builtins.
24275         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
24276         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
24277         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
24278         "<round_saeonly_constraint>")): Changed to ...
24279         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
24280         "<round_saeonly_scalar_constraint>")): ... this.
24281         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
24282         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
24283         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
24284         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
24285         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
24287 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
24289         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
24290         options.
24291         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
24292         Remove all values except native, 8540 and 8548.
24294 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
24296         * config/i386/i386.c (ix86_output_function_return): Pass
24297         INVALID_REGNUM, instead of -1, as invalid register number to
24298         indirect_thunk_name and output_indirect_thunk.
24300 2018-02-02  Julia Koval  <julia.koval@intel.com>
24302         * config.gcc: Add -march=icelake.
24303         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
24304         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
24305         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
24306         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
24307         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
24308         (processor_target_table): Add icelake.
24309         (ix86_option_override_internal): Handle new PTAs.
24310         (get_builtin_code_for_version): Handle icelake.
24311         (M_INTEL_COREI7_ICELAKE): New.
24312         (fold_builtin_cpu): Handle icelake.
24313         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
24314         * doc/invoke.texi: Add -march=icelake.
24316 2018-02-02  Julia Koval  <julia.koval@intel.com>
24318         * config/i386/i386.c (ix86_option_override_internal): Change flags type
24319         to wide_int_bitmask.
24320         * wide-int-bitmask.h: New.
24322 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24324         PR target/84066
24325         * config/i386/i386.md: Replace Pmode with word_mode in
24326         builtin_setjmp_setup and builtin_longjmp to support x32.
24328 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
24330         PR target/56010
24331         PR target/83743
24332         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
24333         #include "opts.h".
24334         (rs6000_supported_cpu_names): New static variable.
24335         (linux_cpu_translation_table): Likewise.
24336         (elf_platform) <cpu>: Define new static variable and use it.
24337         Translate kernel AT_PLATFORM name to canonical name if needed.
24338         Error if platform name is unknown.
24340 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
24342         PR target/84089
24343         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
24345 2018-02-01  Jeff Law  <law@redhat.com>
24347         PR target/84128
24348         * config/i386/i386.c (release_scratch_register_on_entry): Add new
24349         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
24350         the scratch if RELEASE_VIA_POP is false.
24351         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
24352         If we have to save a temporary register, decrement SIZE appropriately.
24353         Pass new arguments to release_scratch_register_on_entry.
24354         (ix86_adjust_stack_and_probe): Likewise.
24355         (ix86_emit_probe_stack_range): Pass new arguments to
24356         release_scratch_register_on_entry.
24358 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
24360         PR rtl-optimization/84157
24361         * combine.c (change_zero_ext): Use REG_P predicate in
24362         front of HARD_REGISTER_P predicate.
24364 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
24366         * config/avr/avr.c (avr_option_override): Move disabling of
24367         -fdelete-null-pointer-checks to...
24368         * common/config/avr/avr-common.c (avr_option_optimization_table):
24369         ...here.
24371 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24373         PR tree-optimization/81635
24374         * tree-data-ref.c (split_constant_offset_1): For types that
24375         wrap on overflow, try to use range info to prove that wrapping
24376         cannot occur.
24378 2018-02-01  Renlin Li  <renlin.li@arm.com>
24380         PR target/83370
24381         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
24382         TAILCALL_ADDR_REGS.
24383         (aarch64_register_move_cost): Likewise.
24384         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
24385         TAILCALL_ADDR_REGS.
24386         (REG_CLASS_NAMES): Likewise.
24387         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
24388         TAILCALL_ADDR_REGS. Remove IP registers.
24389         * config/aarch64/aarch64.md (Ucs): Update register constraint.
24391 2018-02-01  Richard Biener  <rguenther@suse.de>
24393         * domwalk.h (dom_walker::dom_walker): Add additional constructor
24394         for specifying RPO order and allow NULL for that.
24395         * domwalk.c (dom_walker::dom_walker): Likewise.
24396         (dom_walker::walk): Handle NULL RPO order.
24397         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
24398         in RPO order.
24399         (rewrite_update_dom_walker): Likewise.
24400         (mark_def_dom_walker): Likewise.
24402 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24404         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
24405         (aarch64_maybe_expand_sve_subreg_move): Declare.
24406         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
24407         * config/aarch64/predicates.md (aarch64_any_register_operand): New
24408         predicate.
24409         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
24410         that are semantically a reverse operation.
24411         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
24412         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
24413         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
24414         functions.
24415         (aarch64_can_change_mode_class): For big-endian, forbid changes
24416         between two SVE modes if they have different element sizes.
24418 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24420         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
24421         the TImode handling for big-endian targets.
24423 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24425         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
24426         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
24427         not just bytes.
24428         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
24429         Remove BSWAP handing for big-endian targets and use the form of
24430         LD1RQ appropariate for the mode.
24432 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24434         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
24435         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
24436         duplicated element.
24438 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24440         PR tearget/83845
24441         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
24442         check for operands that need to go through aarch64_sve_reload_be.
24444 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
24446         PR tree-optimization/81661
24447         PR tree-optimization/84117
24448         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
24449         * tree-eh.c: Include gimplify.h.
24450         (find_trapping_overflow, replace_trapping_overflow,
24451         rewrite_to_non_trapping_overflow): New functions.
24452         * tree-vect-loop.c: Include tree-eh.h.
24453         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
24454         * tree-data-ref.c: Include tree-eh.h.
24455         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
24457 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
24459         PR rtl-optimization/84123
24460         * combine.c (change_zero_ext): Check if hard register satisfies
24461         can_change_dest_mode before calling gen_lowpart_SUBREG.
24463 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
24465         PR target/82444
24466         * ira.c (ira_init_register_move_cost): Remove assert.
24468 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24470         PR rtl-optimization/84071
24471         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
24472         * doc/tm.texi: Regenerate.
24474 2018-01-31  Richard Biener  <rguenther@suse.de>
24476         PR tree-optimization/84132
24477         * tree-data-ref.c (analyze_miv_subscript): Properly
24478         check whether evolution_function_is_affine_multivariate_p
24479         before calling gcd_of_steps_may_divide_p.
24481 2018-01-31  Julia Koval  <julia.koval@intel.com>
24483         PR target/83618
24484         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
24485         * config/i386/i386.md (rdpid_rex64) New.
24486         (rdpid): Make 32bit only.
24488 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
24490         PR lto/84105
24491         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
24492         an IDENTIFIER_NODE for FUNCTION_TYPE's.
24494 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24496         Revert
24497         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24499         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24501 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24503         PR rtl-optimization/84071
24504         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24505         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24507 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24509         * config/arc/arc.c (arc_handle_aux_attribute): New function.
24510         (arc_attribute_table): Add 'aux' attribute.
24511         (arc_in_small_data_p): Consider aux like variables.
24512         (arc_is_aux_reg_p): New function.
24513         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24514         (arc_get_aux_arg): New function.
24515         (prepare_move_operands): Handle aux-register access.
24516         (arc_handle_aux_attribute): New function.
24517         * doc/extend.texi (ARC Variable attributes): Add subsection.
24519 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24521         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24522         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24523         (arc_attribute_table): Add 'uncached' attribute.
24524         (arc_print_operand): Print '.di' flag for uncached memory
24525         accesses.
24526         (arc_in_small_data_p): Do not consider for small data the uncached
24527         types.
24528         (arc_is_uncached_mem_p): New function.
24529         * config/arc/predicates.md (compact_store_memory_operand): Check
24530         for uncached memory accesses.
24531         (nonvol_nonimm_operand): Likewise.
24532         * doc/extend.texi (ARC Type Attribute): New subsection.
24534 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
24536         PR c/84100
24537         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24538         falign-loops=): Add Optimization flag.
24540 2018-01-30  Jeff Law  <law@redhat.com>
24542         PR target/84064
24543         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24544         INT_REGISTERS_SAVED.  Check it prior to calling
24545         get_scratch_register_on_entry.
24546         (ix86_adjust_stack_and_probe): Similarly.
24547         (ix86_emit_probe_stack_range): Similarly.
24548         (ix86_expand_prologue): Corresponding changes.
24550 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24552         PR target/40411
24553         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24554         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24556 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
24558         PR target/84112
24559         * lra-constraints.c (curr_insn_transform): Process AND in the
24560         address.
24562 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
24564         PR rtl-optimization/83986
24565         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24566         dependence against last_pending_memory_flush in addition to
24567         pending_jump_insns.
24569 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
24571         PR tree-optimization/81611
24572         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24573         copies.
24575 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24577         PR target/83758
24578         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24579         a reg rtx.
24581 2018-01-30  Richard Biener  <rguenther@suse.de>
24582             Jakub Jelinek  <jakub@redhat.com>
24584         PR tree-optimization/84111
24585         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24586         inner loops added during recursion, as they don't have up-to-date
24587         SSA form.
24589 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24591         PR ipa/81360
24592         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24593         (can_inline_edge_by_limits_p): ... here.
24594         (can_early_inline_edge_p, check_callers,
24595         update_caller_keys, update_callee_keys, recursive_inlining,
24596         add_new_edges_to_heap, speculation_useful_p,
24597         inline_small_functions,
24598         inline_small_functions, flatten_function,
24599         inline_to_all_callers_1): Update.
24601 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24603         * profile-count.c (profile_count::combine_with_ipa_count): Handle
24604         zeros correctly.
24606 2018-01-30  Richard Biener  <rguenther@suse.de>
24608         PR tree-optimization/83008
24609         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24610         invariant and constant vector uses in stmts when they need
24611         more than one stmt.
24613 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24615         PR bootstrap/84017
24616         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24617         * configure: Regenerate.
24619 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24621         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24622         pattern.
24623         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24624         Use gen_rtx_REG rather than gen_lowpart.
24626 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24628         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24629         rather than 0 when creating partial subregs.
24631 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24633         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24634         of usage.
24636 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
24638         PR target/81550
24639         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24640         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24641         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24642         flags.  This restores the settings used before the 2017-07-24.
24643         Turning off pre increment/decrement/modify allows IVOPTS to
24644         optimize DF/SF loops where the index is an int.
24646 2018-01-29  Richard Biener  <rguenther@suse.de>
24647             Kelvin Nilsen  <kelvin@gcc.gnu.org>
24649         PR bootstrap/80867
24650         * tree-vect-stmts.c (vectorizable_call): Don't call
24651         targetm.vectorize_builtin_md_vectorized_function if callee is
24652         NULL.
24654 2018-01-22  Carl Love  <cel@us.ibm.com>
24656         * doc/extend.tex: Fix typo in second arg in
24657         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24659 2018-01-29  Richard Biener  <rguenther@suse.de>
24661         PR tree-optimization/84086
24662         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24663         (flush_ssaname_freelist): When SSA names were released reset
24664         the SCEV hash table.
24666 2018-01-29  Richard Biener  <rguenther@suse.de>
24668         PR tree-optimization/84057
24669         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24670         removed paths when removing edges.
24672 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
24674         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24675         -mfunction-return=@var{choice}.
24677 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
24679         PR diagnostic/84034
24680         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24681         Handle CR like TAB.
24682         (layout::print_source_line): Likewise.
24683         (test_get_line_width_without_trailing_whitespace): Add test cases.
24685 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
24687         PR middle-end/84040
24688         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24689         debug insns.
24691 2018-01-26  Jim Wilson  <jimw@sifive.com>
24693         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24695         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24696         specified.
24698 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24700         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24701         and CMP + SUB-immediate -> SUBS.
24703 2018-01-26  Martin Sebor  <msebor@redhat.com>
24705         PR tree-optimization/83896
24706         * tree-ssa-strlen.c (get_string_len): Rename...
24707         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
24708         Avoid assuming length is constant.
24709         (handle_char_store): Use HOST_WIDE_INT for string length.
24711 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
24713         PR target/81763
24714         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24715         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24717 2018-01-26  Richard Biener  <rguenther@suse.de>
24719         PR rtl-optimization/84003
24720         * dse.c (record_store): Only record redundant stores when
24721         the earlier store aliases at least all accesses the later one does.
24723 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
24725         PR rtl-optimization/83985
24726         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24727         REG_CFA_RESTORE insns.
24728         (delete_unmarked_insns): Don't ignore separate shrink wrapping
24729         REG_CFA_RESTORE insns here.
24731         PR c/83989
24732         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24733         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24735 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24737         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24738         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24739         (arc_init): Likewise.
24740         (arc_override_options): Likewise.
24741         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24742         value.
24743         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24744         support.
24745         * config/arc/arc.h (TARGET_DBNZ): Define.
24746         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24747         properly set the tune attribute.
24748         (dbnz): Use TARGET_DBNZ guard.
24749         * config/arc/arc.opt (mtune): Add core3 option.
24751 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24753         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24754         recognize new pic like addresses.
24755         (arc_delegitimize_address): Clean up.
24757 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24759         * config/arc/arc-arches.def: Option mrf16 valid for all
24760         architectures.
24761         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24762         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24763         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24764         * config/arc/arc-tables.opt: Regenerate.
24765         * config/arc/arc.c (arc_conditional_register_usage): Handle
24766         reduced register file case.
24767         (arc_file_start): Set must have build attributes.
24768         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24769         mrf16 option value.
24770         * config/arc/arc.opt (mrf16): Add new option.
24771         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24772         * config/arc/genmultilib.awk: Handle new mrf16 option.
24773         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24774         * config/arc/t-multilib: Regenerate.
24775         * doc/invoke.texi (ARC Options): Document mrf16 option.
24777 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24779         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24780         * config/arc/arc.c (arc_handle_secure_attribute): New function.
24781         (arc_attribute_table): Add 'secure_call' attribute.
24782         (arc_print_operand): Print secure call operand.
24783         (arc_function_ok_for_sibcall): Don't optimize tail calls when
24784         secure.
24785         (arc_is_secure_call_p): New function.  * config/arc/arc.md
24786         (call_i): Add support for sjli instruction.
24787         (call_value_i): Likewise.
24788         * config/arc/constraints.md (Csc): New constraint.
24790 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24791             John Eric Martin  <John.Martin@emmicro-us.com>
24793         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24794         * config/arc/arc.c (_arc_jli_section): New struct.
24795         (arc_jli_section): New type.
24796         (rc_jli_sections): New static variable.
24797         (arc_handle_jli_attribute): New function.
24798         (arc_attribute_table): Add jli_always and jli_fixed attribute.
24799         (arc_file_end): New function.
24800         (TARGET_ASM_FILE_END): Define.
24801         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24802         (arc_add_jli_section): New function.
24803         (jli_call_scan): Likewise.
24804         (arc_reorg): Call jli_call_scan.
24805         (arc_output_addsi): Remove 'S' from printing asm operand.
24806         (arc_is_jli_call_p): New function.
24807         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24808         operand.
24809         (movhi_insn): Likewise.
24810         (movsi_insn): Likewise.
24811         (movsi_set_cc_insn): Likewise.
24812         (loadqi_update): Likewise.
24813         (load_zeroextendqisi_update): Likewise.
24814         (load_signextendqisi_update): Likewise.
24815         (loadhi_update): Likewise.
24816         (load_zeroextendhisi_update): Likewise.
24817         (load_signextendhisi_update): Likewise.
24818         (loadsi_update): Likewise.
24819         (loadsf_update): Likewise.
24820         (movsicc_insn): Likewise.
24821         (bset_insn): Likewise.
24822         (bxor_insn): Likewise.
24823         (bclr_insn): Likewise.
24824         (bmsk_insn): Likewise.
24825         (bicsi3_insn): Likewise.
24826         (cmpsi_cc_c_insn): Likewise.
24827         (movsi_ne): Likewise.
24828         (movsi_cond_exec): Likewise.
24829         (clrsbsi2): Likewise.
24830         (norm_f): Likewise.
24831         (normw): Likewise.
24832         (swap): Likewise.
24833         (divaw): Likewise.
24834         (flag): Likewise.
24835         (sr): Likewise.
24836         (kflag): Likewise.
24837         (ffs): Likewise.
24838         (ffs_f): Likewise.
24839         (fls): Likewise.
24840         (call_i): Remove 'S' asm letter, add jli instruction.
24841         (call_value_i): Likewise.
24842         * config/arc/arc.op (mjli-always): New option.
24843         * config/arc/constraints.md (Cji): New constraint.
24844         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24845         operand.
24846         (subsf3_fpx): Likewise.
24847         (mulsf3_fpx): Likewise.
24848         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24849         asm operand.
24850         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24851         function attrbutes.
24852         * doc/invoke.texi (ARC): Document mjli-always option.
24854 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
24856         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24857         avoid addition with 0 and use incw and decw where possible.
24859 2018-01-26  Richard Biener  <rguenther@suse.de>
24861         PR tree-optimization/81082
24862         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24863         association if it requires casting to unsigned.
24864         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24865         from fold_plusminus_mult_expr to catch important cases late when
24866         range info is available.
24868 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24870         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24871         * configure.ac (hidden_linkonce): New test.
24872         * configure: Regenerate.
24873         * config.in: Regenerate.
24875 2018-01-26  Julia Koval  <julia.koval@intel.com>
24877         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
24878         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
24879         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
24880         _mm_mask_bitshuffle_epi64_mask): Fix type.
24881         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
24882         USI_FTYPE_V4DI_V4DI_USI): Remove.
24883         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
24884         __builtin_ia32_vpshufbitqmb256_mask,
24885         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
24886         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
24887         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
24889 2018-01-26  Alan Modra  <amodra@gmail.com>
24891         PR target/84033
24892         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
24893         UNSPEC_VBPERMQ.  Sort other unspecs.
24895 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
24897         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
24899 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
24901         PR middle-end/83055
24902         * predict.c (drop_profile): Do not push/pop cfun; update also
24903         node->count.
24904         (handle_missing_profiles): Fix logic looking for zero profiles.
24906 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
24908         PR middle-end/83977
24909         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
24910         on functions with #pragma omp declare simd or functions with simd
24911         attribute.
24912         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
24913         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
24914         Remove trailing \n from warning_at calls.
24916 2018-01-25  Tom de Vries  <tom@codesourcery.com>
24918         PR target/84028
24919         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24920         for neutered workers.
24922 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
24924         PR target/68467
24925         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
24926         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
24928 2018-01-24  Jeff Law  <law@redhat.com>
24930         PR target/83994
24931         * i386.c (get_probe_interval): Move to earlier point.
24932         (ix86_compute_frame_layout): If -fstack-clash-protection and
24933         the frame is larger than the probe interval, then use pushes
24934         to save registers rather than reg->mem moves.
24935         (ix86_expand_prologue): Remove conditional for int_registers_saved
24936         assertion.
24938 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
24940         PR target/84014
24941         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
24942         min/max for never referenced object.
24944 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
24946         PR middle-end/83977
24947         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
24948         here.
24949         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
24950         attributes from DECL_ATTRIBUTES (decl) without affecting
24951         DECL_ATTRIBUTES (current_function_decl).
24952         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
24953         functions with non-NULL DECL_ABSTRACT_ORIGIN.
24955 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
24957         PR tree-optimization/83979
24958         * fold-const.c (fold_comparison): Use constant_boolean_node
24959         instead of boolean_{true,false}_node.
24961 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
24963         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
24964         with zero counts.
24966 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24968         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24969         Simplify the clause that sets the length attribute.
24970         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24971         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
24972         clause that sets the length attribute.
24973         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24975 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24977         PR target/83589
24978         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
24979         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
24980         Add strict parameter.
24981         (prevent_branch_around_nothing): Insert dummy insn between branch to
24982         label and label with no ptx insn inbetween.
24983         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
24985 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24987         PR target/81352
24988         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24989         for neutered threads in warp.
24990         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
24992 2018-01-24  Richard Biener  <rguenther@suse.de>
24994         PR tree-optimization/83176
24995         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
24996         operands.
24998 2018-01-24  Richard Biener  <rguenther@suse.de>
25000         PR tree-optimization/82819
25001         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
25002         code generating pluses that are no-ops in the target precision.
25004 2018-01-24  Richard Biener  <rguenther@suse.de>
25006         PR middle-end/84000
25007         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
25009 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25011         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
25012         to merge probabilities.
25013         * predict.c (probably_never_executed): Also mark as cold functions
25014         with global 0 profile and guessed local profile.
25015         * profile-count.c (profile_probability::combine_with_count): New
25016         member function.
25017         * profile-count.h (profile_probability::operator*,
25018         profile_probability::operator*=, profile_probability::operator/,
25019         profile_probability::operator/=): Reduce precision to adjusted
25020         and set value to guessed on contradictory divisions.
25021         (profile_probability::combine_with_freq): Remove.
25022         (profile_probability::combine_wiht_count): Declare.
25023         (profile_count::force_nonzero):: Set to adjusted.
25024         (profile_count::probability_in):: Set quality to adjusted.
25025         * tree-ssa-tail-merge.c (replace_block_by): Use
25026         combine_with_count.
25028 2018-01-23  Andrew Waterman  <andrew@sifive.com>
25029             Jim Wilson  <jimw@sifive.com>
25031         * config/riscv/riscv.c (riscv_stack_boundary): New.
25032         (riscv_option_override): Set riscv_stack_boundary.  Handle
25033         riscv_preferred_stack_boundary_arg.
25034         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
25035         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
25036         (STACK_BOUNDARY): Set to riscv_stack_boundary.
25037         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
25038         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
25039         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
25041 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
25043         PR target/83905
25044         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
25045         of struct ix86_frame.
25046         (ix86_expand_epilogue): Likewise.  Add a local variable for
25047         the reg_save_offset field in struct ix86_frame.
25049 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
25051         PR tree-optimization/82604
25052         * tree-loop-distribution.c (enum partition_kind): New enum item
25053         PKIND_PARTIAL_MEMSET.
25054         (partition_builtin_p): Support above new enum item.
25055         (generate_code_for_partition): Ditto.
25056         (compute_access_range): Differentiate cases that equality can be
25057         proven at all loops, the innermost loops or no loops.
25058         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
25059         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
25060         (finalize_partitions, distribute_loop): Don't fuse partition of
25061         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
25062         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
25063         parloop is enabled.
25065 2018-01-23  Martin Liska  <mliska@suse.cz>
25067         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
25068         order to ignore the predictor.
25069         (PRED_POLYMORPHIC_CALL): Likewise.
25070         (PRED_RECURSIVE_CALL): Likewise.
25072 2018-01-23  Martin Liska  <mliska@suse.cz>
25074         * tree-profile.c (tree_profiling): Print function header to
25075         aware reader which function we are working on.
25076         * value-prof.c (gimple_find_values_to_profile): Do not print
25077         not interesting value histograms.
25079 2018-01-23  Martin Liska  <mliska@suse.cz>
25081         * profile-count.h (enum profile_quality): Add
25082         profile_uninitialized as the first value. Do not number values
25083         as they are zero based.
25084         (profile_count::verify): Update sanity check.
25085         (profile_probability::verify): Likewise.
25087 2018-01-23  Nathan Sidwell  <nathan@acm.org>
25089         * doc/invoke.texi (ffor-scope): Deprecate.
25091 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25093         PR tree-optimization/83510
25094         * domwalk.c (set_all_edges_as_executable): New function.
25095         (dom_walker::dom_walker): Convert bool param
25096         "skip_unreachable_blocks" to enum reachability.  Move setup of
25097         edge flags to set_all_edges_as_executable and only do it when
25098         reachability is REACHABLE_BLOCKS.
25099         * domwalk.h (enum dom_walker::reachability): New enum.
25100         (dom_walker::dom_walker): Convert bool param
25101         "skip_unreachable_blocks" to enum reachability.
25102         (set_all_edges_as_executable): New decl.
25103         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
25104         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
25105         "reachability".
25106         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
25107         but converting true to REACHABLE_BLOCKS.
25108         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
25109         * tree-vrp.c
25110         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
25111         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
25112         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
25113         REACHABLE_BLOCKS.
25114         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
25115         if check_all_array_refs will be called.
25117 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25119         * tree.c (selftest::test_location_wrappers): Add more test
25120         coverage.
25122 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25124         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
25125         (selftest::test_bit_in_range): Likewise.
25127 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25129         PR testsuite/83888
25130         * doc/sourcebuild.texi (vect_float): Say that the selector
25131         only describes the situation when -funsafe-math-optimizations is on.
25132         (vect_float_strict): Document.
25134 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25136         PR tree-optimization/83965
25137         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
25138         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
25139         instead of checking only for a reduction.
25140         (vect_recog_widen_sum_pattern): Likewise.
25142 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25144         * predict.c (probably_never_executed): Only use precise profile info.
25145         (compute_function_frequency): Skip after inlining hack since we now
25146         have quality checking.
25148 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25150         * profile-count.h (profile_probability::very_unlikely,
25151         profile_probability::unlikely, profile_probability::even): Set
25152         precision to guessed.
25154 2018-01-23  Richard Biener  <rguenther@suse.de>
25156         PR tree-optimization/83963
25157         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
25158         Properly terminate dominator walk when crossing the exit edge not
25159         when visiting its source block.
25161 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
25163         PR c++/83918
25164         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
25165         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
25167 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
25169         PR tree-optimization/83957
25170         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
25171         semicolon after for body surrounded by braces.
25173         PR tree-optimization/83081
25174         * profile-count.h (profile_probability::split): New method.
25175         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
25176         Use profile_probability::split.
25177         (do_compare_rtx_and_jump): Fix adjustment of probabilities
25178         when splitting a single conditional jump into 2.
25180 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
25182         PR tree-optimization/69452
25183         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
25184         decl.
25186 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25188         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
25189         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
25190         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
25192 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25194         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
25195         declaration.
25196         * config/rl78/rl78.md (movdi): New define_expand.
25197         * config/rl78/rl78.c (rl78_split_movdi): New function.
25199 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
25201         PR target/83862
25202         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
25203         no longer used.
25204         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
25205         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
25206         128-bit to produce an UNSPEC move to get the double word with the
25207         signbit and then a shift directly to do signbit.
25208         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
25209         implementation with a new version that just does either a direct
25210         move or a regular move.  Move memory interface to separate insns.
25211         Move insns so they are next to the expander.
25212         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
25213         with signbit move.  Split big and little endian case.
25214         (signbit<mode>2_dm_mem_le): Likewise.
25215         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
25216         (signbit<mode>2_dm2): Likewise.
25218 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25220         * config/rl78/rl78.md (anddi3): New define_expand.
25222 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25224         * config/rl78/rl78.md (umindi3): New define_expand.
25226 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25228         * config/rl78/rl78.md (smindi3): New define_expand.
25230 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25232         * config/rl78/rl78.md (smaxdi3): New define_expand.
25234 2018-01-22  Carl Love  <cel@us.ibm.com>
25236         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
25237         LVX_V1TI): Add macro expansion.
25238         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
25239         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
25240         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
25241         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
25242         Change check to determine if the instruction is a byte reversing
25243         entry.  Fix typo in comment.
25244         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
25245         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
25246         Add def_builtin calls for new builtins.
25247         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
25248         Add define_insn expansion.
25250 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25252         * config/rl78/rl78.md (umaxdi3): New define_expand.
25254 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25256         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
25257         for non-QImode registers.
25259 2018-01-22  Richard Biener  <rguenther@suse.de>
25261         PR tree-optimization/83963
25262         * graphite-scop-detection.c (scop_detection::get_sese): Delay
25263         including the loop exit block.
25264         (scop_detection::merge_sese): Likewise.
25265         (scop_detection::add_scop): Do it here instead.
25267 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25269         * doc/sourcebuild.texi (arm_softfloat): Document.
25271 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
25273         PR gcc/77734
25274         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
25275         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
25276         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
25278 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25279             David Edelsohn  <dje.gcc@gmail.com>
25281         PR target/83946
25282         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25283         Change "crset eq" to "crset 2".
25284         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25285         (*call_indirect_aix<mode>_nospec): Likewise.
25286         (*call_value_indirect_aix<mode>_nospec): Likewise.
25287         (*call_indirect_elfv2<mode>_nospec): Likewise.
25288         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
25289         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
25290         change assembly output from . to $.
25291         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25292         (indirect_jump<mode>_nospec): Change assembly output from . to $.
25293         (*tablejump<mode>_internal1_nospec): Likewise.
25295 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
25297         PR target/80870
25298         * config/sh/sh_optimize_sett_clrt.cc:
25299         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
25301 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
25303         PR tree-optimization/83940
25304         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
25305         offset_dt to vect_constant_def rather than vect_unknown_def_type.
25306         (vect_check_load_store_mask): Add a mask_dt_out parameter and
25307         use it to pass back the definition type.
25308         (vect_check_store_rhs): Likewise rhs_dt_out.
25309         (vect_build_gather_load_calls): Add a mask_dt argument and use
25310         it instead of a call to vect_is_simple_use.
25311         (vectorizable_store): Update calls to vect_check_load_store_mask
25312         and vect_check_store_rhs.  Use the dt returned by the latter instead
25313         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
25314         instead of calls to vect_is_simple_use.  Pass the scalar rather
25315         than the vector operand to vect_is_simple_use when handling
25316         second and subsequent copies of an rhs value.
25317         (vectorizable_load): Update calls to vect_check_load_store_mask
25318         and vect_build_gather_load_calls.  Use the cached mask_dt and
25319         gs_info.offset_dt instead of calls to vect_is_simple_use.
25321 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
25323         PR middle-end/83945
25324         * tree-emutls.c: Include gimplify.h.
25325         (lower_emutls_2): New function.
25326         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
25327         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
25328         it before further processing.
25330         PR target/83930
25331         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
25332         UINTVAL (trueop1) instead of INTVAL (op1).
25334 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
25336         PR debug/81570
25337         PR debug/83728
25338         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
25339         INCOMING_FRAME_SP_OFFSET if not defined.
25340         (scan_trace): Add ENTRY argument.  If true and
25341         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
25342         emit a note to adjust the CFA offset.
25343         (create_cfi_notes): Adjust scan_trace callers.
25344         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
25345         INCOMING_FRAME_SP_OFFSET in the CIE.
25346         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
25347         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
25348         Likewise.
25349         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
25350         * doc/tm.texi: Regenerated.
25352 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25354         PR rtl-optimization/83147
25355         * lra-constraints.c (remove_inheritance_pseudos): Use
25356         lra_substitute_pseudo_within_insn.
25358 2018-01-19  Tom de Vries  <tom@codesourcery.com>
25359             Cesar Philippidis  <cesar@codesourcery.com>
25361         PR target/83920
25362         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
25364 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
25366         PR target/83790
25367         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
25368         spaces for function labels.
25370 2018-01-19  Martin Liska  <mliska@suse.cz>
25372         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
25373         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
25374         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
25375         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
25376         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
25377         (PRED_CONST_RETURN): Change from 69 to 65.
25378         (PRED_NULL_RETURN): Change from 91 to 71.
25379         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
25380         (PRED_LOOP_GUARD): Change from 66 to 73.
25382 2018-01-19  Martin Liska  <mliska@suse.cz>
25384         * predict.c (predict_insn_def): Add new assert.
25385         (struct branch_predictor): Change type to signed integer.
25386         (test_prediction_value_range): Amend test to cover
25387         PROB_UNINITIALIZED.
25388         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
25389         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
25390         (PRED_LOOP_ITERATIONS_MAX): Likewise.
25391         (PRED_LOOP_IV_COMPARE): Likewise.
25392         * predict.h (PROB_UNINITIALIZED): Define new constant.
25394 2018-01-19  Martin Liska  <mliska@suse.cz>
25396         * predict.c (dump_prediction): Add new format for
25397         analyze_brprob.py script which is enabled with -details
25398         suboption.
25399         * profile-count.h (precise_p): New function.
25401 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25403         PR tree-optimization/83922
25404         * tree-vect-loop.c (vect_verify_full_masking): Return false if
25405         there are no statements that need masking.
25406         (vect_active_double_reduction_p): New function.
25407         (vect_analyze_loop_operations): Use it when handling phis that
25408         are not in the loop header.
25410 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25412         PR tree-optimization/83914
25413         * tree-vect-loop.c (vectorizable_induction): Don't convert
25414         init_expr or apply the peeling adjustment for inductions
25415         that are nested within the vectorized loop.
25417 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25419         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
25420         instead of NEG.
25422 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25424         PR sanitizer/81715
25425         PR testsuite/83882
25426         * function.h (gimplify_parameters): Add gimple_seq * argument.
25427         * function.c: Include gimple.h and options.h.
25428         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
25429         for the added local temporaries if needed.
25430         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
25431         if there are any parameter cleanups, wrap whole body into a
25432         try/finally with the cleanups.
25434 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
25436         PR target/82964
25437         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
25438         Use GET_MODE_CLASS for scalar floating point.
25440 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25442         PR ipa/82256
25443         patch by PaX Team
25444         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
25445         Fix call of call_cgraph_insertion_hooks.
25447 2018-01-18  Martin Sebor  <msebor@redhat.com>
25449         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
25451 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25453         PR ipa/83619
25454         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
25455         frequencies.
25457 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
25459         PR other/70268
25460         * common.opt: (-ffile-prefix-map): New option.
25461         * opts.c (common_handle_option): Defer it.
25462         * opts-global.c (handle_common_deferred_options): Handle it.
25463         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
25464         * file-prefix-map.h: New file.
25465         (remap_debug_filename, add_debug_prefix_map): ...here.
25466         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
25467         * final.c (debug_prefix_map, add_debug_prefix_map
25468         remap_debug_filename): Move to...
25469         * file-prefix-map.c: New file.
25470         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
25471         generalize, get rid of alloca(), use strrchr() instead of strchr().
25472         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
25473         Implement in terms of add_prefix_map().
25474         (remap_macro_filename, remap_debug_filename): Implement in term of
25475         remap_filename().
25476         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
25477         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
25478         * dbxout.c: Include file-prefix-map.h.
25479         * varasm.c: Likewise.
25480         * vmsdbgout.c: Likewise.
25481         * xcoffout.c: Likewise.
25482         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
25483         * doc/cppopts.texi (-fmacro-prefix-map): Document.
25484         * doc/invoke.texi (-ffile-prefix-map): Document.
25485         (-fdebug-prefix-map): Update description.
25487 2018-01-18  Martin Liska  <mliska@suse.cz>
25489         * config/i386/i386.c (indirect_thunk_name): Document that also
25490         lfence is emitted.
25491         (output_indirect_thunk): Document why both instructions
25492         (pause and lfence) are generated.
25494 2018-01-18  Richard Biener  <rguenther@suse.de>
25496         PR tree-optimization/83887
25497         * graphite-scop-detection.c
25498         (scop_detection::get_nearest_dom_with_single_entry): Remove.
25499         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25500         (scop_detection::merge_sese): Re-implement with a flood-fill
25501         algorithm that properly finds a SESE region if it exists.
25503 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25505         PR c/61240
25506         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25507         pointer_diff optimizations use view_convert instead of convert.
25509 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25511         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25512         Generate different code for -mno-speculate-indirect-jumps.
25513         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25514         (*call_indirect_aix<mode>): Disable for
25515         -mno-speculate-indirect-jumps.
25516         (*call_indirect_aix<mode>_nospec): New define_insn.
25517         (*call_value_indirect_aix<mode>): Disable for
25518         -mno-speculate-indirect-jumps.
25519         (*call_value_indirect_aix<mode>_nospec): New define_insn.
25520         (*sibcall_nonlocal_sysv<mode>): Generate different code for
25521         -mno-speculate-indirect-jumps.
25522         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25524 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
25526         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25527         long double type, set the flags for noting the default long double
25528         type, even if we don't pass or return a long double type.
25530 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
25532         PR ipa/83051
25533         * ipa-inline.c (flatten_function): Do not overwrite final inlining
25534         failure.
25536 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
25538         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25539         support for merge[hl].
25540         (fold_mergehl_helper): New helper function.
25541         (tree-vector-builder.h): New #include for tree_vector_builder usage.
25542         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25543         (altivec_vmrglw_direct): Add xxmrglw insn.
25545 2018-01-17  Andrew Waterman  <andrew@sifive.com>
25547         * config/riscv/riscv.c (riscv_conditional_register_usage): If
25548         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25550 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
25552         PR lto/83121
25553         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25554         call the lto_location_cache before reading the
25555         DECL_SOURCE_LOCATION of the types.
25557 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
25558             Richard Sandiford  <richard.sandiford@linaro.org>
25560         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25561         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25562         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25563         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25564         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25565         Add declaration.
25566         * config/aarch64/constraints.md (aarch64_movti_operand):
25567         Limit immediates.
25568         * config/aarch64/predicates.md (Uti): Add new constraint.
25570 2018-01-17  Carl Love  <cel@us.ibm.com>
25572         * config/rs6000/vsx.md (define_expand xl_len_r,
25573         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25574         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25575         lxvll.
25576         (define_expand, define_insn): Move the shift left from  the
25577         define_insn to the define_expand for lxvl and stxvl instructions.
25578         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25579         and XL_LEN_R definitions to PURE.
25581 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
25583         * config/i386/i386.c (indirect_thunk_name): Declare regno
25584         as unsigned int.  Compare regno with INVALID_REGNUM.
25585         (output_indirect_thunk): Ditto.
25586         (output_indirect_thunk_function): Ditto.
25587         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
25588         in the call to output_indirect_thunk_function.
25590 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
25592         PR middle-end/83884
25593         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25594         rather than the size of inner_type to determine the stack slot size
25595         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25597 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
25599         PR target/83546
25600         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25601         to PTA_SILVERMONT.
25603 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
25605         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25606         endian Linux systems to optionally enable multilibs for selecting
25607         the long double type if the user configured an explicit type.
25608         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25609         have no long double multilibs if not defined.
25610         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25611         warn if the user used -mabi={ieee,ibm}longdouble and we built
25612         multilibs for long double.
25613         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25614         appropriate multilib option.
25615         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25616         multilib options.
25617         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25618         for building long double multilibs.
25619         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25621 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
25623         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25624         copies.
25626         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25627         64 bits.
25628         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25629         128 bits.
25631         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25632         variables.
25634         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25635         return value.
25637 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25639         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25640         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25642 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25644         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25645         different rtl trees depending on TARGET_64BIT.
25646         (rs6000_gen_lvx): Likewise.
25648 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25650         * config/visium/visium.md (nop): Tweak comment.
25651         (hazard_nop): Likewise.
25653 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25655         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25656         -mspeculate-indirect-jumps.
25657         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25658         for -mno-speculate-indirect-jumps.
25659         (*call_indirect_elfv2<mode>_nospec): New define_insn.
25660         (*call_value_indirect_elfv2<mode>): Disable for
25661         -mno-speculate-indirect-jumps.
25662         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25663         (indirect_jump): Emit different RTL for
25664         -mno-speculate-indirect-jumps.
25665         (*indirect_jump<mode>): Disable for
25666         -mno-speculate-indirect-jumps.
25667         (*indirect_jump<mode>_nospec): New define_insn.
25668         (tablejump): Emit different RTL for
25669         -mno-speculate-indirect-jumps.
25670         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25671         (tablejumpsi_nospec): New define_expand.
25672         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25673         (tablejumpdi_nospec): New define_expand.
25674         (*tablejump<mode>_internal1): Disable for
25675         -mno-speculate-indirect-jumps.
25676         (*tablejump<mode>_internal1_nospec): New define_insn.
25677         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25678         option.
25680 2018-01-16  Artyom Skrobov tyomitch@gmail.com
25682         * caller-save.c (insert_save): Drop unnecessary parameter.  All
25683         callers updated.
25685 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25686             Richard Biener  <rguenth@suse.de>
25688         PR libgomp/83590
25689         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25690         return early, inline manually is_gimple_sizepos.  Make sure if we
25691         call gimplify_expr we don't end up with a gimple constant.
25692         * tree.c (variably_modified_type_p): Don't return true for
25693         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
25694         * gimplify.h (is_gimple_sizepos): Remove.
25696 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25698         PR tree-optimization/83857
25699         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25700         vectorizable_live_operation for pure SLP statements.
25701         (vectorizable_live_operation): Handle PHIs.
25703 2018-01-16  Richard Biener  <rguenther@suse.de>
25705         PR tree-optimization/83867
25706         * tree-vect-stmts.c (vect_transform_stmt): Precompute
25707         nested_in_vect_loop_p since the scalar stmt may get invalidated.
25709 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25711         PR c/83844
25712         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25713         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25714         If off is not INTEGER_CST, issue a may not be aligned warning
25715         rather than isn't aligned.  Use isn%'t rather than isn't.
25716         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25717         into MULT_EXPR.
25718         <case MULT_EXPR>: Improve the case when bottom and one of the
25719         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25720         operand, in that case check if the other operand is multiple of
25721         bottom divided by the INTEGER_CST operand.
25723 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25725         PR target/83858
25726         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25727         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25728         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25729         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25730         * config/pa/pa.c (pa_function_arg_advance): Likewise.
25731         (pa_function_arg, pa_arg_partial_bytes): Likewise.
25732         (pa_function_arg_size): New function.
25734 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25736         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25737         in a separate statement.
25739 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25741         PR tree-optimization/83847
25742         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25743         group gathers and scatters.
25745 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25747         PR rtl-optimization/86620
25748         * params.def (max-sched-ready-insns): Bump minimum value to 1.
25750         PR rtl-optimization/83213
25751         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25752         to last if both are JUMP_INSNs.
25754         PR tree-optimization/83843
25755         * gimple-ssa-store-merging.c
25756         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25757         store_immediate_info for bswap/nop orig_stores.
25759 2018-01-15  Andrew Waterman  <andrew@sifive.com>
25761         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25762         !TARGET_MUL.
25763         <UDIV>: Increase cost if !TARGET_DIV.
25765 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
25767         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25768         (define_attr "cr_logical_3op"): New.
25769         (cceq_ior_compare): Adjust.
25770         (cceq_ior_compare_complement): Adjust.
25771         (*cceq_rev_compare): Adjust.
25772         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25773         (is_cracked_insn): Adjust.
25774         (insn_must_be_first_in_group): Adjust.
25775         * config/rs6000/40x.md: Adjust.
25776         * config/rs6000/440.md: Adjust.
25777         * config/rs6000/476.md: Adjust.
25778         * config/rs6000/601.md: Adjust.
25779         * config/rs6000/603.md: Adjust.
25780         * config/rs6000/6xx.md: Adjust.
25781         * config/rs6000/7450.md: Adjust.
25782         * config/rs6000/7xx.md: Adjust.
25783         * config/rs6000/8540.md: Adjust.
25784         * config/rs6000/cell.md: Adjust.
25785         * config/rs6000/e300c2c3.md: Adjust.
25786         * config/rs6000/e500mc.md: Adjust.
25787         * config/rs6000/e500mc64.md: Adjust.
25788         * config/rs6000/e5500.md: Adjust.
25789         * config/rs6000/e6500.md: Adjust.
25790         * config/rs6000/mpc.md: Adjust.
25791         * config/rs6000/power4.md: Adjust.
25792         * config/rs6000/power5.md: Adjust.
25793         * config/rs6000/power6.md: Adjust.
25794         * config/rs6000/power7.md: Adjust.
25795         * config/rs6000/power8.md: Adjust.
25796         * config/rs6000/power9.md: Adjust.
25797         * config/rs6000/rs64.md: Adjust.
25798         * config/rs6000/titan.md: Adjust.
25800 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25802         * config/i386/predicates.md (indirect_branch_operand): Rewrite
25803         ix86_indirect_branch_register logic.
25805 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25807         * config/i386/constraints.md (Bs): Update
25808         ix86_indirect_branch_register check.  Don't check
25809         ix86_indirect_branch_register with GOT_memory_operand.
25810         (Bw): Likewise.
25811         * config/i386/predicates.md (GOT_memory_operand): Don't check
25812         ix86_indirect_branch_register here.
25813         (GOT32_symbol_operand): Likewise.
25815 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25817         * config/i386/predicates.md (constant_call_address_operand):
25818         Rewrite ix86_indirect_branch_register logic.
25819         (sibcall_insn_operand): Likewise.
25821 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25823         * config/i386/constraints.md (Bs): Replace
25824         ix86_indirect_branch_thunk_register with
25825         ix86_indirect_branch_register.
25826         (Bw): Likewise.
25827         * config/i386/i386.md (indirect_jump): Likewise.
25828         (tablejump): Likewise.
25829         (*sibcall_memory): Likewise.
25830         (*sibcall_value_memory): Likewise.
25831         Peepholes of indirect call and jump via memory: Likewise.
25832         * config/i386/i386.opt: Likewise.
25833         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25834         (GOT_memory_operand): Likewise.
25835         (call_insn_operand): Likewise.
25836         (sibcall_insn_operand): Likewise.
25837         (GOT32_symbol_operand): Likewise.
25839 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25841         PR middle-end/83837
25842         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25843         type rather than type addr's type points to.
25844         (expand_omp_atomic_mutex): Likewise.
25845         (expand_omp_atomic): Likewise.
25847 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25849         PR target/83839
25850         * config/i386/i386.c (output_indirect_thunk_function): Use
25851         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25852         for  __x86_return_thunk.
25854 2018-01-15  Richard Biener  <rguenther@suse.de>
25856         PR middle-end/83850
25857         * expmed.c (extract_bit_field_1): Fix typo.
25859 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25861         PR target/83687
25862         * config/arm/iterators.md (VF): New mode iterator.
25863         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25864         Remove integer-related logic from pattern.
25865         (neon_vabd<mode>_3): Likewise.
25867 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25869         PR middle-end/82694
25870         * common.opt (fstrict-overflow): No longer an alias.
25871         (fwrapv-pointer): New option.
25872         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25873         also for pointer types based on flag_wrapv_pointer.
25874         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25875         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
25876         opts->x_flag_wrapv got set.
25877         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
25878         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
25879         POINTER_TYPE_OVERFLOW_UNDEFINED.
25880         * match.pd: Likewise in address comparison pattern.
25881         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
25883 2018-01-15  Richard Biener  <rguenther@suse.de>
25885         PR lto/83804
25886         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
25887         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
25888         Reset type names to their identifier if their TYPE_DECL doesn't
25889         have linkage (and thus is used for ODR and devirt).
25890         (save_debug_info_for_decl): Remove.
25891         (save_debug_info_for_type): Likewise.
25892         (add_tree_to_fld_list): Adjust.
25893         * tree-pretty-print.c (dump_generic_node): Make dumping of
25894         type names more robust.
25896 2018-01-15  Richard Biener  <rguenther@suse.de>
25898         * BASE-VER: Bump to 8.0.1.
25900 2018-01-14  Martin Sebor  <msebor@redhat.com>
25902         PR other/83508
25903         * builtins.c (check_access): Avoid warning when the no-warning bit
25904         is set.
25906 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
25908         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
25909         * ira-color (allocno_hard_regs_compare): Likewise.
25911 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
25913         PR target/83013
25914         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
25915         Use .pushsection/.popsection.
25917 2018-01-14  Martin Sebor  <msebor@redhat.com>
25919         PR c++/81327
25920         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
25922 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
25924         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
25925         entry from extra_headers.
25926         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
25927         extra_headers, make the list bitwise identical to the i?86-*-* one.
25929 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25931         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
25932         -mcmodel=large with -mindirect-branch=thunk,
25933         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
25934         -mfunction-return=thunk-extern.
25935         * doc/invoke.texi: Document -mcmodel=large is incompatible with
25936         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
25937         -mfunction-return=thunk and -mfunction-return=thunk-extern.
25939 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25941         * config/i386/i386.c (print_reg): Print the name of the full
25942         integer register without '%'.
25943         (ix86_print_operand): Handle 'V'.
25944         * doc/extend.texi: Document 'V' modifier.
25946 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25948         * config/i386/constraints.md (Bs): Disallow memory operand for
25949         -mindirect-branch-register.
25950         (Bw): Likewise.
25951         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25952         (GOT_memory_operand): Likewise.
25953         (call_insn_operand): Likewise.
25954         (sibcall_insn_operand): Likewise.
25955         (GOT32_symbol_operand): Likewise.
25956         * config/i386/i386.md (indirect_jump): Call convert_memory_address
25957         for -mindirect-branch-register.
25958         (tablejump): Likewise.
25959         (*sibcall_memory): Likewise.
25960         (*sibcall_value_memory): Likewise.
25961         Disallow peepholes of indirect call and jump via memory for
25962         -mindirect-branch-register.
25963         (*call_pop): Replace m with Bw.
25964         (*call_value_pop): Likewise.
25965         (*sibcall_pop_memory): Replace m with Bs.
25966         * config/i386/i386.opt (mindirect-branch-register): New option.
25967         * doc/invoke.texi: Document -mindirect-branch-register option.
25969 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25971         * config/i386/i386-protos.h (ix86_output_function_return): New.
25972         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
25973         set function_return_type.
25974         (indirect_thunk_name): Add ret_p to indicate thunk for function
25975         return.
25976         (output_indirect_thunk_function): Pass false to
25977         indirect_thunk_name.
25978         (ix86_output_indirect_branch_via_reg): Likewise.
25979         (ix86_output_indirect_branch_via_push): Likewise.
25980         (output_indirect_thunk_function): Create alias for function
25981         return thunk if regno < 0.
25982         (ix86_output_function_return): New function.
25983         (ix86_handle_fndecl_attribute): Handle function_return.
25984         (ix86_attribute_table): Add function_return.
25985         * config/i386/i386.h (machine_function): Add
25986         function_return_type.
25987         * config/i386/i386.md (simple_return_internal): Use
25988         ix86_output_function_return.
25989         (simple_return_internal_long): Likewise.
25990         * config/i386/i386.opt (mfunction-return=): New option.
25991         (indirect_branch): Mention -mfunction-return=.
25992         * doc/extend.texi: Document function_return function attribute.
25993         * doc/invoke.texi: Document -mfunction-return= option.
25995 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25997         * config/i386/i386-opts.h (indirect_branch): New.
25998         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
25999         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
26000         with local indirect jump when converting indirect call and jump.
26001         (ix86_set_indirect_branch_type): New.
26002         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
26003         (indirectlabelno): New.
26004         (indirect_thunk_needed): Likewise.
26005         (indirect_thunk_bnd_needed): Likewise.
26006         (indirect_thunks_used): Likewise.
26007         (indirect_thunks_bnd_used): Likewise.
26008         (INDIRECT_LABEL): Likewise.
26009         (indirect_thunk_name): Likewise.
26010         (output_indirect_thunk): Likewise.
26011         (output_indirect_thunk_function): Likewise.
26012         (ix86_output_indirect_branch_via_reg): Likewise.
26013         (ix86_output_indirect_branch_via_push): Likewise.
26014         (ix86_output_indirect_branch): Likewise.
26015         (ix86_output_indirect_jmp): Likewise.
26016         (ix86_code_end): Call output_indirect_thunk_function if needed.
26017         (ix86_output_call_insn): Call ix86_output_indirect_branch if
26018         needed.
26019         (ix86_handle_fndecl_attribute): Handle indirect_branch.
26020         (ix86_attribute_table): Add indirect_branch.
26021         * config/i386/i386.h (machine_function): Add indirect_branch_type
26022         and has_local_indirect_jump.
26023         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
26024         to true.
26025         (tablejump): Likewise.
26026         (*indirect_jump): Use ix86_output_indirect_jmp.
26027         (*tablejump_1): Likewise.
26028         (simple_return_indirect_internal): Likewise.
26029         * config/i386/i386.opt (mindirect-branch=): New option.
26030         (indirect_branch): New.
26031         (keep): Likewise.
26032         (thunk): Likewise.
26033         (thunk-inline): Likewise.
26034         (thunk-extern): Likewise.
26035         * doc/extend.texi: Document indirect_branch function attribute.
26036         * doc/invoke.texi: Document -mindirect-branch= option.
26038 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
26040         PR ipa/83051
26041         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
26043 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
26045         * ipa-inline.c (want_inline_small_function_p): Return false if
26046         inlining has already failed with CIF_FINAL_ERROR.
26047         (update_caller_keys): Call want_inline_small_function_p before
26048         can_inline_edge_p.
26049         (update_callee_keys): Likewise.
26051 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26053         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
26054         New function.
26055         (rs6000_quadword_masked_address_p): Likewise.
26056         (quad_aligned_load_p): Likewise.
26057         (quad_aligned_store_p): Likewise.
26058         (const_load_sequence_p): Add comment to describe the outer-most loop.
26059         (mimic_memory_attributes_and_flags): New function.
26060         (rs6000_gen_stvx): Likewise.
26061         (replace_swapped_aligned_store): Likewise.
26062         (rs6000_gen_lvx): Likewise.
26063         (replace_swapped_aligned_load): Likewise.
26064         (replace_swapped_load_constant): Capitalize argument name in
26065         comment describing this function.
26066         (rs6000_analyze_swaps): Add a third pass to search for vector loads
26067         and stores that access quad-word aligned addresses and replace
26068         with stvx or lvx instructions when appropriate.
26069         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
26070         New function prototype.
26071         (rs6000_quadword_masked_address_p): Likewise.
26072         (rs6000_gen_lvx): Likewise.
26073         (rs6000_gen_stvx): Likewise.
26074         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
26075         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
26076         when memory address is aligned.
26077         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
26078         this split to select lvx instruction when memory address is aligned.
26079         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
26080         instruction when memory address is aligned.
26081         (*vsx_le_perm_load_v16qi): Likewise.
26082         (four unnamed splitters): Modify to select the stvx instruction
26083         when memory is aligned.
26085 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
26087         * predict.c (determine_unlikely_bbs): Handle correctly BBs
26088         which appears in the queue multiple times.
26090 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26091             Alan Hayward  <alan.hayward@arm.com>
26092             David Sherwood  <david.sherwood@arm.com>
26094         * tree-vectorizer.h (vec_lower_bound): New structure.
26095         (_loop_vec_info): Add check_nonzero and lower_bounds.
26096         (LOOP_VINFO_CHECK_NONZERO): New macro.
26097         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
26098         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
26099         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
26100         fields.  Make seg_len the distance travelled, not including the
26101         access size.
26102         (dr_direction_indicator): Declare.
26103         (dr_zero_step_indicator): Likewise.
26104         (dr_known_forward_stride_p): Likewise.
26105         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
26106         tree-ssanames.h.
26107         (runtime_alias_check_p): Allow runtime alias checks with
26108         variable strides.
26109         (operator ==): Compare access_size and align.
26110         (prune_runtime_alias_test_list): Rework for new distinction between
26111         the access_size and seg_len.
26112         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
26113         segment lengths.
26114         (get_segment_min_max): New function.
26115         (create_intersect_range_checks): Use it.
26116         (dr_step_indicator): New function.
26117         (dr_direction_indicator): Likewise.
26118         (dr_zero_step_indicator): Likewise.
26119         (dr_known_forward_stride_p): Likewise.
26120         * tree-loop-distribution.c (data_ref_segment_size): Return
26121         DR_STEP * (niters - 1).
26122         (compute_alias_check_pairs): Update call to the dr_with_seg_len
26123         constructor.
26124         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
26125         (vect_preserves_scalar_order_p): New function, split out from...
26126         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
26127         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
26128         (vect_vfa_access_size): New function.
26129         (vect_vfa_align): Likewise.
26130         (vect_compile_time_alias): Take access_size_a and access_b arguments.
26131         (dump_lower_bound): New function.
26132         (vect_check_lower_bound): Likewise.
26133         (vect_small_gap_p): Likewise.
26134         (vectorizable_with_step_bound_p): Likewise.
26135         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
26136         depencies if the vectorization factor is 1.  Convert the checks
26137         for nonzero steps into checks on the bounds of DR_STEP.  Try using
26138         a bunds check for variable steps if the minimum required step is
26139         relatively small. Update calls to the dr_with_seg_len
26140         constructor and to vect_compile_time_alias.
26141         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
26142         function.
26143         (vect_loop_versioning): Call it.
26144         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
26145         when retrying.
26146         (vect_estimate_min_profitable_iters): Account for any bounds checks.
26148 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26149             Alan Hayward  <alan.hayward@arm.com>
26150             David Sherwood  <david.sherwood@arm.com>
26152         * doc/sourcebuild.texi (vect_scatter_store): Document.
26153         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
26154         optabs.
26155         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
26156         Document.
26157         * genopinit.c (main): Add supports_vec_scatter_store and
26158         supports_vec_scatter_store_cached to target_optabs.
26159         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
26160         IFN_MASK_SCATTER_STORE.
26161         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
26162         functions.
26163         * internal-fn.h (internal_store_fn_p): Declare.
26164         (internal_fn_stored_value_index): Likewise.
26165         * internal-fn.c (scatter_store_direct): New macro.
26166         (expand_scatter_store_optab_fn): New function.
26167         (direct_scatter_store_optab_supported_p): New macro.
26168         (internal_store_fn_p): New function.
26169         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
26170         IFN_MASK_SCATTER_STORE.
26171         (internal_fn_mask_index): Likewise.
26172         (internal_fn_stored_value_index): New function.
26173         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
26174         for scatter stores.
26175         * optabs-query.h (supports_vec_scatter_store_p): Declare.
26176         * optabs-query.c (supports_vec_scatter_store_p): New function.
26177         * tree-vectorizer.h (vect_get_store_rhs): Declare.
26178         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
26179         true for scatter stores.
26180         (vect_gather_scatter_fn_p): Handle scatter stores too.
26181         (vect_check_gather_scatter): Consider using scatter stores if
26182         supports_vec_scatter_store_p.
26183         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
26184         scatter stores too.
26185         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26186         internal_fn_stored_value_index.
26187         (check_load_store_masking): Handle scatter stores too.
26188         (vect_get_store_rhs): Make public.
26189         (vectorizable_call): Use internal_store_fn_p.
26190         (vectorizable_store): Handle scatter store internal functions.
26191         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
26192         when deciding whether the end of the group has been reached.
26193         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
26194         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
26195         (mask_scatter_store<mode>): New insns.
26197 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26198             Alan Hayward  <alan.hayward@arm.com>
26199             David Sherwood  <david.sherwood@arm.com>
26201         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
26202         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
26203         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
26204         function.
26205         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
26206         Use vect_truncate_gather_scatter_offset if we can't treat the
26207         operation as a normal gather load or scatter store.
26208         (get_group_load_store_type): Take the gather_scatter_info
26209         as argument.  Try using a gather load or scatter store for
26210         single-element groups.
26211         (get_load_store_type): Update calls to get_group_load_store_type
26212         and vect_use_strided_gather_scatters_p.
26214 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26215             Alan Hayward  <alan.hayward@arm.com>
26216             David Sherwood  <david.sherwood@arm.com>
26218         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
26219         optional tree argument.
26220         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
26221         null target hooks.
26222         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
26223         but continue to use the current value as a fallback.
26224         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
26225         to compare the updates.
26226         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
26227         (get_load_store_type): Use it when handling a strided access.
26228         (vect_get_strided_load_store_ops): New function.
26229         (vect_get_data_ptr_increment): Likewise.
26230         (vectorizable_load): Handle strided gather loads.  Always pass
26231         a step to vect_create_data_ref_ptr and bump_vector_ptr.
26233 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26234             Alan Hayward  <alan.hayward@arm.com>
26235             David Sherwood  <david.sherwood@arm.com>
26237         * doc/md.texi (gather_load@var{m}): Document.
26238         (mask_gather_load@var{m}): Likewise.
26239         * genopinit.c (main): Add supports_vec_gather_load and
26240         supports_vec_gather_load_cached to target_optabs.
26241         * optabs-tree.c (init_tree_optimization_optabs): Use
26242         ggc_cleared_alloc to allocate target_optabs.
26243         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
26244         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
26245         functions.
26246         * internal-fn.h (internal_load_fn_p): Declare.
26247         (internal_gather_scatter_fn_p): Likewise.
26248         (internal_fn_mask_index): Likewise.
26249         (internal_gather_scatter_fn_supported_p): Likewise.
26250         * internal-fn.c (gather_load_direct): New macro.
26251         (expand_gather_load_optab_fn): New function.
26252         (direct_gather_load_optab_supported_p): New macro.
26253         (direct_internal_fn_optab): New function.
26254         (internal_load_fn_p): Likewise.
26255         (internal_gather_scatter_fn_p): Likewise.
26256         (internal_fn_mask_index): Likewise.
26257         (internal_gather_scatter_fn_supported_p): Likewise.
26258         * optabs-query.c (supports_at_least_one_mode_p): New function.
26259         (supports_vec_gather_load_p): Likewise.
26260         * optabs-query.h (supports_vec_gather_load_p): Declare.
26261         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
26262         and memory_type field.
26263         (NUM_PATTERNS): Bump to 15.
26264         * tree-vect-data-refs.c: Include internal-fn.h.
26265         (vect_gather_scatter_fn_p): New function.
26266         (vect_describe_gather_scatter_call): Likewise.
26267         (vect_check_gather_scatter): Try using internal functions for
26268         gather loads.  Recognize existing calls to a gather load function.
26269         (vect_analyze_data_refs): Consider using gather loads if
26270         supports_vec_gather_load_p.
26271         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
26272         (vect_get_gather_scatter_offset_type): Likewise.
26273         (vect_convert_mask_for_vectype): Likewise.
26274         (vect_add_conversion_to_patterm): Likewise.
26275         (vect_try_gather_scatter_pattern): Likewise.
26276         (vect_recog_gather_scatter_pattern): New pattern recognizer.
26277         (vect_vect_recog_func_ptrs): Add it.
26278         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26279         internal_fn_mask_index and internal_gather_scatter_fn_p.
26280         (check_load_store_masking): Take the gather_scatter_info as an
26281         argument and handle gather loads.
26282         (vect_get_gather_scatter_ops): New function.
26283         (vectorizable_call): Check internal_load_fn_p.
26284         (vectorizable_load): Likewise.  Handle gather load internal
26285         functions.
26286         (vectorizable_store): Update call to check_load_store_masking.
26287         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
26288         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
26289         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
26290         (aarch64_gather_scale_operand_d): New predicates.
26291         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
26292         (mask_gather_load<mode>): New insns.
26294 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26295             Alan Hayward  <alan.hayward@arm.com>
26296             David Sherwood  <david.sherwood@arm.com>
26298         * optabs.def (fold_left_plus_optab): New optab.
26299         * doc/md.texi (fold_left_plus_@var{m}): Document.
26300         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
26301         * internal-fn.c (fold_left_direct): Define.
26302         (expand_fold_left_optab_fn): Likewise.
26303         (direct_fold_left_optab_supported_p): Likewise.
26304         * fold-const-call.c (fold_const_fold_left): New function.
26305         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
26306         * tree-parloops.c (valid_reduction_p): New function.
26307         (gather_scalar_reductions): Use it.
26308         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
26309         (vect_finish_replace_stmt): Declare.
26310         * tree-vect-loop.c (fold_left_reduction_fn): New function.
26311         (needs_fold_left_reduction_p): New function, split out from...
26312         (vect_is_simple_reduction): ...here.  Accept reductions that
26313         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
26314         (vect_force_simple_reduction): Also store the reduction type in
26315         the assignment's STMT_VINFO_REDUC_TYPE.
26316         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
26317         (merge_with_identity): New function.
26318         (vect_expand_fold_left): Likewise.
26319         (vectorize_fold_left_reduction): Likewise.
26320         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
26321         scalar phi in place for it.  Check for target support and reject
26322         cases that would reassociate the operation.  Defer the transform
26323         phase to vectorize_fold_left_reduction.
26324         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
26325         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
26326         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
26328 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26330         * tree-if-conv.c (predicate_mem_writes): Remove redundant
26331         call to ifc_temp_var.
26333 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26334             Alan Hayward  <alan.hayward@arm.com>
26335             David Sherwood  <david.sherwood@arm.com>
26337         * target.def (legitimize_address_displacement): Take the original
26338         offset as a poly_int.
26339         * targhooks.h (default_legitimize_address_displacement): Update
26340         accordingly.
26341         * targhooks.c (default_legitimize_address_displacement): Likewise.
26342         * doc/tm.texi: Regenerate.
26343         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
26344         as an argument, moving assert of ad->disp == ad->disp_term to...
26345         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
26346         Try calling targetm.legitimize_address_displacement before expanding
26347         the address rather than afterwards, and adjust for the new interface.
26348         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
26349         Match the new hook interface.  Handle SVE addresses.
26350         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
26351         new hook interface.
26353 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26355         * Makefile.in (OBJS): Add early-remat.o.
26356         * target.def (select_early_remat_modes): New hook.
26357         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
26358         * doc/tm.texi: Regenerate.
26359         * targhooks.h (default_select_early_remat_modes): Declare.
26360         * targhooks.c (default_select_early_remat_modes): New function.
26361         * timevar.def (TV_EARLY_REMAT): New timevar.
26362         * passes.def (pass_early_remat): New pass.
26363         * tree-pass.h (make_pass_early_remat): Declare.
26364         * early-remat.c: New file.
26365         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
26366         function.
26367         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
26369 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26370             Alan Hayward  <alan.hayward@arm.com>
26371             David Sherwood  <david.sherwood@arm.com>
26373         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
26374         vfm1 with a bound_epilog parameter.
26375         (vect_do_peeling): Update calls accordingly, and move the prologue
26376         call earlier in the function.  Treat the base bound_epilog as 0 for
26377         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
26378         this base when peeling for gaps.
26379         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
26380         with fully-masked loops.
26381         (vect_estimate_min_profitable_iters): Handle the single peeled
26382         iteration in that case.
26384 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26385             Alan Hayward  <alan.hayward@arm.com>
26386             David Sherwood  <david.sherwood@arm.com>
26388         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
26389         single-element interleaving even if the size is not a power of 2.
26390         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
26391         accesses for single-element interleaving if the group size is
26392         not a power of 2.
26394 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26395             Alan Hayward  <alan.hayward@arm.com>
26396             David Sherwood  <david.sherwood@arm.com>
26398         * doc/md.texi (fold_extract_last_@var{m}): Document.
26399         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
26400         * optabs.def (fold_extract_last_optab): New optab.
26401         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
26402         * internal-fn.c (fold_extract_direct): New macro.
26403         (expand_fold_extract_optab_fn): Likewise.
26404         (direct_fold_extract_optab_supported_p): Likewise.
26405         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
26406         * tree-vect-loop.c (vect_model_reduction_cost): Handle
26407         EXTRACT_LAST_REDUCTION.
26408         (get_initial_def_for_reduction): Do not create an initial vector
26409         for EXTRACT_LAST_REDUCTION reductions.
26410         (vectorizable_reduction): Leave the scalar phi in place for
26411         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
26412         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
26413         epilogue code for EXTRACT_LAST_REDUCTION and defer the
26414         transform phase to vectorizable_condition.
26415         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
26416         split out from...
26417         (vect_finish_stmt_generation): ...here.
26418         (vect_finish_replace_stmt): New function.
26419         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
26420         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
26421         pattern.
26422         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
26424 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26425             Alan Hayward  <alan.hayward@arm.com>
26426             David Sherwood  <david.sherwood@arm.com>
26428         * doc/md.texi (extract_last_@var{m}): Document.
26429         * optabs.def (extract_last_optab): New optab.
26430         * internal-fn.def (EXTRACT_LAST): New internal function.
26431         * internal-fn.c (cond_unary_direct): New macro.
26432         (expand_cond_unary_optab_fn): Likewise.
26433         (direct_cond_unary_optab_supported_p): Likewise.
26434         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
26435         loops using EXTRACT_LAST.
26436         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
26437         (extract_last_<mode>): ...this optab.
26438         (vec_extract<mode><Vel>): Update accordingly.
26440 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26441             Alan Hayward  <alan.hayward@arm.com>
26442             David Sherwood  <david.sherwood@arm.com>
26444         * target.def (empty_mask_is_expensive): New hook.
26445         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
26446         * doc/tm.texi: Regenerate.
26447         * targhooks.h (default_empty_mask_is_expensive): Declare.
26448         * targhooks.c (default_empty_mask_is_expensive): New function.
26449         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
26450         if the target says that empty masks are expensive.
26451         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
26452         New function.
26453         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
26455 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26456             Alan Hayward  <alan.hayward@arm.com>
26457             David Sherwood  <david.sherwood@arm.com>
26459         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
26460         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
26461         (vect_use_loop_mask_for_alignment_p): New function.
26462         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
26463         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
26464         niters_skip argument.  Make sure that the first niters_skip elements
26465         of the first iteration are inactive.
26466         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
26467         Update call to vect_set_loop_masks_directly.
26468         (get_misalign_in_elems): New function, split out from...
26469         (vect_gen_prolog_loop_niters): ...here.
26470         (vect_update_init_of_dr): Take a code argument that specifies whether
26471         the adjustment should be added or subtracted.
26472         (vect_update_init_of_drs): Likewise.
26473         (vect_prepare_for_masked_peels): New function.
26474         (vect_do_peeling): Skip prologue peeling if we're using a mask
26475         instead.  Update call to vect_update_inits_of_drs.
26476         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26477         mask_skip_niters.
26478         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
26479         alignment.  Do not include the number of peeled iterations in
26480         the minimum threshold in that case.
26481         (vectorizable_induction): Adjust the start value down by
26482         LOOP_VINFO_MASK_SKIP_NITERS iterations.
26483         (vect_transform_loop): Call vect_prepare_for_masked_peels.
26484         Take the number of skipped iterations into account when calculating
26485         the loop bounds.
26486         * tree-vect-stmts.c (vect_gen_while_not): New function.
26488 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26489             Alan Hayward  <alan.hayward@arm.com>
26490             David Sherwood  <david.sherwood@arm.com>
26492         * doc/sourcebuild.texi (vect_fully_masked): Document.
26493         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26494         default value to 0.
26495         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26496         split out from...
26497         (vect_analyze_loop_2): ...here. Don't check the vectorization
26498         factor against the number of loop iterations if the loop is
26499         fully-masked.
26501 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26502             Alan Hayward  <alan.hayward@arm.com>
26503             David Sherwood  <david.sherwood@arm.com>
26505         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26506         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26507         (dump_groups): Update accordingly.
26508         (iv_use::mem_type): New member variable.
26509         (address_p): New function.
26510         (record_use): Add a mem_type argument and initialize the new
26511         mem_type field.
26512         (record_group_use): Add a mem_type argument.  Use address_p.
26513         Remove obsolete null checks of base_object.  Update call to record_use.
26514         (find_interesting_uses_op): Update call to record_group_use.
26515         (find_interesting_uses_cond): Likewise.
26516         (find_interesting_uses_address): Likewise.
26517         (get_mem_type_for_internal_fn): New function.
26518         (find_address_like_use): Likewise.
26519         (find_interesting_uses_stmt): Try find_address_like_use before
26520         calling find_interesting_uses_op.
26521         (addr_offset_valid_p): Use the iv mem_type field as the type
26522         of the addressed memory.
26523         (add_autoinc_candidates): Likewise.
26524         (get_address_cost): Likewise.
26525         (split_small_address_groups_p): Use address_p.
26526         (split_address_groups): Likewise.
26527         (add_iv_candidate_for_use): Likewise.
26528         (autoinc_possible_for_pair): Likewise.
26529         (rewrite_groups): Likewise.
26530         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26531         (determine_group_iv_cost): Update after split of USE_ADDRESS.
26532         (get_alias_ptr_type_for_ptr_address): New function.
26533         (rewrite_use_address): Rewrite address uses in calls that were
26534         identified by find_address_like_use.
26536 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26537             Alan Hayward  <alan.hayward@arm.com>
26538             David Sherwood  <david.sherwood@arm.com>
26540         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26541         TARGET_MEM_REFs.
26542         * gimple-expr.h (is_gimple_addressable: Likewise.
26543         * gimple-expr.c (is_gimple_address): Likewise.
26544         * internal-fn.c (expand_call_mem_ref): New function.
26545         (expand_mask_load_optab_fn): Use it.
26546         (expand_mask_store_optab_fn): Likewise.
26548 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26549             Alan Hayward  <alan.hayward@arm.com>
26550             David Sherwood  <david.sherwood@arm.com>
26552         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26553         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26554         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26555         (cond_umax@var{mode}): Document.
26556         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26557         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26558         (cond_umin_optab, cond_umax_optab): New optabs.
26559         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26560         (COND_IOR, COND_XOR): New internal functions.
26561         * internal-fn.h (get_conditional_internal_fn): Declare.
26562         * internal-fn.c (cond_binary_direct): New macro.
26563         (expand_cond_binary_optab_fn): Likewise.
26564         (direct_cond_binary_optab_supported_p): Likewise.
26565         (get_conditional_internal_fn): New function.
26566         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26567         Cope with reduction statements that are vectorized as calls rather
26568         than assignments.
26569         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26570         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26571         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26572         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26573         (UNSPEC_COND_EOR): New unspecs.
26574         (optab): Add mappings for them.
26575         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26576         (sve_int_op, sve_fp_op): New int attributes.
26578 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26579             Alan Hayward  <alan.hayward@arm.com>
26580             David Sherwood  <david.sherwood@arm.com>
26582         * optabs.def (while_ult_optab): New optab.
26583         * doc/md.texi (while_ult@var{m}@var{n}): Document.
26584         * internal-fn.def (WHILE_ULT): New internal function.
26585         * internal-fn.h (direct_internal_fn_supported_p): New override
26586         that takes two types as argument.
26587         * internal-fn.c (while_direct): New macro.
26588         (expand_while_optab_fn): New function.
26589         (convert_optab_supported_p): Likewise.
26590         (direct_while_optab_supported_p): New macro.
26591         * wide-int.h (wi::udiv_ceil): New function.
26592         * tree-vectorizer.h (rgroup_masks): New structure.
26593         (vec_loop_masks): New typedef.
26594         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26595         and fully_masked_p.
26596         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26597         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26598         (vect_max_vf): New function.
26599         (slpeel_make_loop_iterate_ntimes): Delete.
26600         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26601         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26602         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26603         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26604         internal-fn.h, stor-layout.h and optabs-query.h.
26605         (vect_set_loop_mask): New function.
26606         (add_preheader_seq): Likewise.
26607         (add_header_seq): Likewise.
26608         (interleave_supported_p): Likewise.
26609         (vect_maybe_permute_loop_masks): Likewise.
26610         (vect_set_loop_masks_directly): Likewise.
26611         (vect_set_loop_condition_masked): Likewise.
26612         (vect_set_loop_condition_unmasked): New function, split out from
26613         slpeel_make_loop_iterate_ntimes.
26614         (slpeel_make_loop_iterate_ntimes): Rename to..
26615         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
26616         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26617         (vect_do_peeling): Update call accordingly.
26618         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26619         loops.
26620         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26621         mask_compare_type, can_fully_mask_p and fully_masked_p.
26622         (release_vec_loop_masks): New function.
26623         (_loop_vec_info): Use it to free the loop masks.
26624         (can_produce_all_loop_masks_p): New function.
26625         (vect_get_max_nscalars_per_iter): Likewise.
26626         (vect_verify_full_masking): Likewise.
26627         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26628         retries, and free the mask rgroups before retrying.  Check loop-wide
26629         reasons for disallowing fully-masked loops.  Make the final decision
26630         about whether use a fully-masked loop or not.
26631         (vect_estimate_min_profitable_iters): Do not assume that peeling
26632         for the number of iterations will be needed for fully-masked loops.
26633         (vectorizable_reduction): Disable fully-masked loops.
26634         (vectorizable_live_operation): Likewise.
26635         (vect_halve_mask_nunits): New function.
26636         (vect_double_mask_nunits): Likewise.
26637         (vect_record_loop_mask): Likewise.
26638         (vect_get_loop_mask): Likewise.
26639         (vect_transform_loop): Handle the case in which the final loop
26640         iteration might handle a partial vector.  Call vect_set_loop_condition
26641         instead of slpeel_make_loop_iterate_ntimes.
26642         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26643         (check_load_store_masking): New function.
26644         (prepare_load_store_mask): Likewise.
26645         (vectorizable_store): Handle fully-masked loops.
26646         (vectorizable_load): Likewise.
26647         (supportable_widening_operation): Use vect_halve_mask_nunits for
26648         booleans.
26649         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26650         (vect_gen_while): New function.
26651         * config/aarch64/aarch64.md (umax<mode>3): New expander.
26652         (aarch64_uqdec<mode>): New insn.
26654 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26655             Alan Hayward  <alan.hayward@arm.com>
26656             David Sherwood  <david.sherwood@arm.com>
26658         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26659         (reduc_xor_scal_optab): New optabs.
26660         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26661         (reduc_xor_scal_@var{m}): Document.
26662         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26663         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26664         internal functions.
26665         * fold-const-call.c (fold_const_call): Handle them.
26666         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26667         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26668         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26669         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26670         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26671         (UNSPEC_XORV): New unspecs.
26672         (optab): Add entries for them.
26673         (BITWISEV): New int iterator.
26674         (bit_reduc_op): New int attributes.
26676 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26677             Alan Hayward  <alan.hayward@arm.com>
26678             David Sherwood  <david.sherwood@arm.com>
26680         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26681         * internal-fn.def (VEC_SHL_INSERT): New internal function.
26682         * optabs.def (vec_shl_insert_optab): New optab.
26683         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26684         (duplicate_and_interleave): Likewise.
26685         * tree-vect-loop.c: Include internal-fn.h.
26686         (neutral_op_for_slp_reduction): New function, split out from
26687         get_initial_defs_for_reduction.
26688         (get_initial_def_for_reduction): Handle option 2 for variable-length
26689         vectors by loading the neutral value into a vector and then shifting
26690         the initial value into element 0.
26691         (get_initial_defs_for_reduction): Replace the code argument with
26692         the neutral value calculated by neutral_op_for_slp_reduction.
26693         Use gimple_build_vector for constant-length vectors.
26694         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26695         but the first group_size elements have a neutral value.
26696         Use duplicate_and_interleave otherwise.
26697         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26698         Update call to get_initial_defs_for_reduction.  Handle SLP
26699         reductions for variable-length vectors by creating one vector
26700         result for each scalar result, with the elements associated
26701         with other scalar results stubbed out with the neutral value.
26702         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26703         Require IFN_VEC_SHL_INSERT for double reductions on
26704         variable-length vectors, or SLP reductions that have
26705         a neutral value.  Require can_duplicate_and_interleave_p
26706         support for variable-length unchained SLP reductions if there
26707         is no neutral value, such as for MIN/MAX reductions.  Also require
26708         the number of vector elements to be a multiple of the number of
26709         SLP statements when doing variable-length unchained SLP reductions.
26710         Update call to vect_create_epilog_for_reduction.
26711         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26712         and remove initial values.
26713         (duplicate_and_interleave): Make public.
26714         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26715         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26717 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26718             Alan Hayward  <alan.hayward@arm.com>
26719             David Sherwood  <david.sherwood@arm.com>
26721         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26722         (can_duplicate_and_interleave_p): New function.
26723         (vect_get_and_check_slp_defs): Take the vector of statements
26724         rather than just the current one.  Remove excess parentheses.
26725         Restriction rejectinon of vect_constant_def and vect_external_def
26726         for variable-length vectors to boolean types, or types for which
26727         can_duplicate_and_interleave_p is false.
26728         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26729         (duplicate_and_interleave): New function.
26730         (vect_get_constant_vectors): Use gimple_build_vector for
26731         constant-length vectors and suitable variable-length constant
26732         vectors.  Use duplicate_and_interleave for other variable-length
26733         vectors.  Don't defer the update when inserting new statements.
26735 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26736             Alan Hayward  <alan.hayward@arm.com>
26737             David Sherwood  <david.sherwood@arm.com>
26739         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26740         min_profitable_iters doesn't go negative.
26742 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26743             Alan Hayward  <alan.hayward@arm.com>
26744             David Sherwood  <david.sherwood@arm.com>
26746         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26747         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26748         * optabs.def (vec_mask_load_lanes_optab): New optab.
26749         (vec_mask_store_lanes_optab): Likewise.
26750         * internal-fn.def (MASK_LOAD_LANES): New internal function.
26751         (MASK_STORE_LANES): Likewise.
26752         * internal-fn.c (mask_load_lanes_direct): New macro.
26753         (mask_store_lanes_direct): Likewise.
26754         (expand_mask_load_optab_fn): Handle masked operations.
26755         (expand_mask_load_lanes_optab_fn): New macro.
26756         (expand_mask_store_optab_fn): Handle masked operations.
26757         (expand_mask_store_lanes_optab_fn): New macro.
26758         (direct_mask_load_lanes_optab_supported_p): Likewise.
26759         (direct_mask_store_lanes_optab_supported_p): Likewise.
26760         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26761         parameter.
26762         (vect_load_lanes_supported): Likewise.
26763         * tree-vect-data-refs.c (strip_conversion): New function.
26764         (can_group_stmts_p): Likewise.
26765         (vect_analyze_data_ref_accesses): Use it instead of checking
26766         for a pair of assignments.
26767         (vect_store_lanes_supported): Take a masked_p parameter.
26768         (vect_load_lanes_supported): Likewise.
26769         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26770         vect_store_lanes_supported and vect_load_lanes_supported.
26771         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26772         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26773         parameter.  Don't allow gaps for masked accesses.
26774         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
26775         and vect_load_lanes_supported.
26776         (get_load_store_type): Take a masked_p parameter and update
26777         call to get_group_load_store_type.
26778         (vectorizable_store): Update call to get_load_store_type.
26779         Handle IFN_MASK_STORE_LANES.
26780         (vectorizable_load): Update call to get_load_store_type.
26781         Handle IFN_MASK_LOAD_LANES.
26783 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26784             Alan Hayward  <alan.hayward@arm.com>
26785             David Sherwood  <david.sherwood@arm.com>
26787         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26788         modes for SVE.
26789         * config/aarch64/aarch64-protos.h
26790         (aarch64_sve_struct_memory_operand_p): Declare.
26791         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26792         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26793         (VPRED, vpred): Handle SVE structure modes.
26794         * config/aarch64/constraints.md (Utx): New constraint.
26795         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26796         (aarch64_sve_struct_nonimmediate_operand): New predicates.
26797         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26798         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26799         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26800         structure modes.  Split into pieces after RA.
26801         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26802         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26803         New patterns.
26804         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26805         SVE structure modes.
26806         (aarch64_classify_address): Likewise.
26807         (sizetochar): Move earlier in file.
26808         (aarch64_print_operand): Handle SVE register lists.
26809         (aarch64_array_mode): New function.
26810         (aarch64_sve_struct_memory_operand_p): Likewise.
26811         (TARGET_ARRAY_MODE): Redefine.
26813 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26814             Alan Hayward  <alan.hayward@arm.com>
26815             David Sherwood  <david.sherwood@arm.com>
26817         * target.def (array_mode): New target hook.
26818         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26819         * doc/tm.texi: Regenerate.
26820         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26821         * hooks.c (hook_optmode_mode_uhwi_none): New function.
26822         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26823         targetm.array_mode.
26824         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
26825         type sizes.
26827 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26828             Alan Hayward  <alan.hayward@arm.com>
26829             David Sherwood  <david.sherwood@arm.com>
26831         * fold-const.c (fold_binary_loc): Check the argument types
26832         rather than the result type when testing for a vector operation.
26834 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26836         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26837         * doc/tm.texi: Regenerate.
26839 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26840             Alan Hayward  <alan.hayward@arm.com>
26841             David Sherwood  <david.sherwood@arm.com>
26843         * doc/invoke.texi (-msve-vector-bits=): Document new option.
26844         (sve): Document new AArch64 extension.
26845         * doc/md.texi (w): Extend the description of the AArch64
26846         constraint to include SVE vectors.
26847         (Upl, Upa): Document new AArch64 predicate constraints.
26848         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26849         enum.
26850         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26851         (msve-vector-bits=): New option.
26852         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26853         SVE when these are disabled.
26854         (sve): New extension.
26855         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26856         modes.  Adjust their number of units based on aarch64_sve_vg.
26857         (MAX_BITSIZE_MODE_ANY_MODE): Define.
26858         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26859         aarch64_addr_query_type.
26860         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26861         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26862         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26863         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26864         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26865         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26866         (aarch64_simd_imm_zero_p): Delete.
26867         (aarch64_check_zero_based_sve_index_immediate): Declare.
26868         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26869         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26870         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26871         (aarch64_sve_float_mul_immediate_p): Likewise.
26872         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26873         rather than an rtx.
26874         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26875         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
26876         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
26877         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
26878         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
26879         (aarch64_regmode_natural_size): Likewise.
26880         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
26881         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
26882         left one place.
26883         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
26884         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
26885         for VG and the SVE predicate registers.
26886         (V_ALIASES): Add a "z"-prefixed alias.
26887         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
26888         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
26889         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
26890         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
26891         (REG_CLASS_NAMES): Add entries for them.
26892         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
26893         and the predicate registers.
26894         (aarch64_sve_vg): Declare.
26895         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
26896         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
26897         (REGMODE_NATURAL_SIZE): Define.
26898         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
26899         SVE macros.
26900         * config/aarch64/aarch64.c: Include cfgrtl.h.
26901         (simd_immediate_info): Add a constructor for series vectors,
26902         and an associated step field.
26903         (aarch64_sve_vg): New variable.
26904         (aarch64_dbx_register_number): Handle VG and the predicate registers.
26905         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
26906         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
26907         (VEC_ANY_DATA, VEC_STRUCT): New constants.
26908         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
26909         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
26910         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
26911         (aarch64_get_mask_mode): New functions.
26912         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
26913         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
26914         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
26915         predicate modes and predicate registers.  Explicitly restrict
26916         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
26917         to store a vector mode if it is recognized by
26918         aarch64_classify_vector_mode.
26919         (aarch64_regmode_natural_size): New function.
26920         (aarch64_hard_regno_caller_save_mode): Return the original mode
26921         for predicates.
26922         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
26923         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
26924         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
26925         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
26926         functions.
26927         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
26928         does not overlap dest if the function is frame-related.  Handle
26929         SVE constants.
26930         (aarch64_split_add_offset): New function.
26931         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
26932         them aarch64_add_offset.
26933         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
26934         and update call to aarch64_sub_sp.
26935         (aarch64_add_cfa_expression): New function.
26936         (aarch64_expand_prologue): Pass extra temporary registers to the
26937         functions above.  Handle the case in which we need to emit new
26938         DW_CFA_expressions for registers that were originally saved
26939         relative to the stack pointer, but now have to be expressed
26940         relative to the frame pointer.
26941         (aarch64_output_mi_thunk): Pass extra temporary registers to the
26942         functions above.
26943         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
26944         IP0 and IP1 values for SVE frames.
26945         (aarch64_expand_vec_series): New function.
26946         (aarch64_expand_sve_widened_duplicate): Likewise.
26947         (aarch64_expand_sve_const_vector): Likewise.
26948         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
26949         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
26950         into the register, rather than emitting a SET directly.
26951         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
26952         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
26953         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
26954         (offset_9bit_signed_scaled_p): New functions.
26955         (aarch64_replicate_bitmask_imm): New function.
26956         (aarch64_bitmask_imm): Use it.
26957         (aarch64_cannot_force_const_mem): Reject expressions involving
26958         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
26959         (aarch64_classify_index): Handle SVE indices, by requiring
26960         a plain register index with a scale that matches the element size.
26961         (aarch64_classify_address): Handle SVE addresses.  Assert that
26962         the mode of the address is VOIDmode or an integer mode.
26963         Update call to aarch64_classify_symbol.
26964         (aarch64_classify_symbolic_expression): Update call to
26965         aarch64_classify_symbol.
26966         (aarch64_const_vec_all_in_range_p): New function.
26967         (aarch64_print_vector_float_operand): Likewise.
26968         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
26969         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
26970         and the FP immediates 1.0 and 0.5.
26971         (aarch64_print_address_internal): Handle SVE addresses.
26972         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
26973         (aarch64_regno_regclass): Handle predicate registers.
26974         (aarch64_secondary_reload): Handle big-endian reloads of SVE
26975         data modes.
26976         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
26977         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
26978         (aarch64_convert_sve_vector_bits): New function.
26979         (aarch64_override_options): Use it to handle -msve-vector-bits=.
26980         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26981         rather than an rtx.
26982         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
26983         Handle SVE vector and predicate modes.  Accept VL-based constants
26984         that need only one temporary register, and VL offsets that require
26985         no temporary registers.
26986         (aarch64_conditional_register_usage): Mark the predicate registers
26987         as fixed if SVE isn't available.
26988         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
26989         Return true for SVE vector and predicate modes.
26990         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
26991         rather than an unsigned int.  Handle SVE modes.
26992         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
26993         SVE modes.
26994         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
26995         if SVE is enabled.
26996         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26997         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26998         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26999         (aarch64_sve_float_mul_immediate_p): New functions.
27000         (aarch64_sve_valid_immediate): New function.
27001         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
27002         Explicitly reject structure modes.  Check for INDEX constants.
27003         Handle PTRUE and PFALSE constants.
27004         (aarch64_check_zero_based_sve_index_immediate): New function.
27005         (aarch64_simd_imm_zero_p): Delete.
27006         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
27007         vector modes.  Accept constants in the range of CNT[BHWD].
27008         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
27009         ask for an Advanced SIMD mode.
27010         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
27011         (aarch64_simd_vector_alignment): Handle SVE predicates.
27012         (aarch64_vectorize_preferred_vector_alignment): New function.
27013         (aarch64_simd_vector_alignment_reachable): Use it instead of
27014         the vector size.
27015         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
27016         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
27017         functions.
27018         (MAX_VECT_LEN): Delete.
27019         (expand_vec_perm_d): Add a vec_flags field.
27020         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
27021         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
27022         (aarch64_evpc_ext): Don't apply a big-endian lane correction
27023         for SVE modes.
27024         (aarch64_evpc_rev): Rename to...
27025         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
27026         (aarch64_evpc_rev_global): New function.
27027         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
27028         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
27029         MAX_VECT_LEN.
27030         (aarch64_evpc_sve_tbl): New function.
27031         (aarch64_expand_vec_perm_const_1): Update after rename of
27032         aarch64_evpc_rev.  Handle SVE permutes too, trying
27033         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
27034         than aarch64_evpc_tbl.
27035         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
27036         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
27037         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
27038         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
27039         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
27040         (aarch64_expand_sve_vcond): New functions.
27041         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
27042         of aarch64_vector_mode_p.
27043         (aarch64_dwarf_poly_indeterminate_value): New function.
27044         (aarch64_compute_pressure_classes): Likewise.
27045         (aarch64_can_change_mode_class): Likewise.
27046         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
27047         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
27048         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
27049         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
27050         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
27051         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
27052         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
27053         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
27054         constraints.
27055         (Dn, Dl, Dr): Accept const as well as const_vector.
27056         (Dz): Likewise.  Compare against CONST0_RTX.
27057         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
27058         of "vector" where appropriate.
27059         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
27060         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
27061         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
27062         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
27063         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
27064         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
27065         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
27066         (v_int_equiv): Extend to SVE modes.
27067         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
27068         mode attributes.
27069         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
27070         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
27071         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
27072         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
27073         (SVE_COND_FP_CMP): New int iterators.
27074         (perm_hilo): Handle the new unpack unspecs.
27075         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
27076         attributes.
27077         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
27078         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
27079         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
27080         (aarch64_equality_operator, aarch64_constant_vector_operand)
27081         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
27082         (aarch64_sve_nonimmediate_operand): Likewise.
27083         (aarch64_sve_general_operand): Likewise.
27084         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
27085         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
27086         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
27087         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
27088         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
27089         (aarch64_sve_float_arith_immediate): Likewise.
27090         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
27091         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
27092         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
27093         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
27094         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
27095         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
27096         (aarch64_sve_float_arith_operand): Likewise.
27097         (aarch64_sve_float_arith_with_sub_operand): Likewise.
27098         (aarch64_sve_float_mul_operand): Likewise.
27099         (aarch64_sve_vec_perm_operand): Likewise.
27100         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
27101         (aarch64_mov_operand): Accept const_poly_int and const_vector.
27102         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
27103         as well as const_vector.
27104         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
27105         in file.  Use CONST0_RTX and CONSTM1_RTX.
27106         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
27107         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
27108         Use aarch64_simd_imm_zero.
27109         * config/aarch64/aarch64-sve.md: New file.
27110         * config/aarch64/aarch64.md: Include it.
27111         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
27112         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
27113         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
27114         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
27115         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
27116         (sve): New attribute.
27117         (enabled): Disable instructions with the sve attribute unless
27118         TARGET_SVE.
27119         (movqi, movhi): Pass CONST_POLY_INT operaneds through
27120         aarch64_expand_mov_immediate.
27121         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
27122         CNT[BHSD] immediates.
27123         (movti): Split CONST_POLY_INT moves into two halves.
27124         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
27125         Split additions that need a temporary here if the destination
27126         is the stack pointer.
27127         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
27128         (*add<mode>3_poly_1): New instruction.
27129         (set_clobber_cc): New expander.
27131 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27133         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
27134         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
27135         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
27136         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
27137         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
27138         Change innermode from fixed_mode_size to machine_mode.
27139         (simplify_subreg): Update call accordingly.  Handle a constant-sized
27140         subreg of a variable-length CONST_VECTOR.
27142 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27143             Alan Hayward  <alan.hayward@arm.com>
27144             David Sherwood  <david.sherwood@arm.com>
27146         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
27147         (add_offset_to_base): New function, split out from...
27148         (create_mem_ref): ...here.  When handling a scale other than 1,
27149         check first whether the address is valid without the offset.
27150         Add it into the base if so, leaving the index and scale as-is.
27152 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27154         PR c++/83778
27155         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
27156         fold_for_warn before checking if arg2 is INTEGER_CST.
27158 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
27160         * config/rs6000/predicates.md (load_multiple_operation): Delete.
27161         (store_multiple_operation): Delete.
27162         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
27163         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
27164         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
27165         guarded by TARGET_STRING.
27166         (rs6000_output_load_multiple): Delete.
27167         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
27168         OPTION_MASK_STRING / TARGET_STRING handling.
27169         (print_operand) <'N', 'O'>: Add comment that these are unused now.
27170         (const rs6000_opt_masks) <"string">: Change mask to 0.
27171         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
27172         (MASK_STRING): Delete.
27173         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
27174         parts.  Simplify.
27175         (load_multiple): Delete.
27176         (*ldmsi8): Delete.
27177         (*ldmsi7): Delete.
27178         (*ldmsi6): Delete.
27179         (*ldmsi5): Delete.
27180         (*ldmsi4): Delete.
27181         (*ldmsi3): Delete.
27182         (store_multiple): Delete.
27183         (*stmsi8): Delete.
27184         (*stmsi7): Delete.
27185         (*stmsi6): Delete.
27186         (*stmsi5): Delete.
27187         (*stmsi4): Delete.
27188         (*stmsi3): Delete.
27189         (movmemsi_8reg): Delete.
27190         (corresponding unnamed define_insn): Delete.
27191         (movmemsi_6reg): Delete.
27192         (corresponding unnamed define_insn): Delete.
27193         (movmemsi_4reg): Delete.
27194         (corresponding unnamed define_insn): Delete.
27195         (movmemsi_2reg): Delete.
27196         (corresponding unnamed define_insn): Delete.
27197         (movmemsi_1reg): Delete.
27198         (corresponding unnamed define_insn): Delete.
27199         * config/rs6000/rs6000.opt (mno-string): New.
27200         (mstring): Replace by deprecation warning stub.
27201         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
27203 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27205         * regrename.c (regrename_do_replace): If replacing the same
27206         reg multiple times, try to reuse last created gen_raw_REG.
27208         PR debug/81155
27209         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
27210         main to workaround a bug in GDB.
27212 2018-01-12  Tom de Vries  <tom@codesourcery.com>
27214         PR target/83737
27215         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
27217 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
27219         PR rtl-optimization/80481
27220         * ira-color.c (get_cap_member): New function.
27221         (allocnos_conflict_by_live_ranges_p): Use it.
27222         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
27223         (setup_slot_coalesced_allocno_live_ranges): Ditto.
27225 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
27227         PR target/83628
27228         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
27229         (*saddl_se_1): Ditto.
27230         (*ssubsi_1): Ditto.
27231         (*ssubl_se_1): Ditto.
27233 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27235         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
27236         rather than wi::to_widest for DR_INITs.
27237         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
27238         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
27239         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
27240         INTEGER_CSTs.
27241         (vect_analyze_group_access_1): Note that here.
27243 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27245         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
27246         polynomial type sizes.
27248 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27250         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
27251         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
27252         (gimple_add_tmp_var): Likewise.
27254 2018-01-12  Martin Liska  <mliska@suse.cz>
27256         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
27257         (gimple_alloc_sizes): Likewise.
27258         (dump_gimple_statistics): Use PRIu64 in printf format.
27259         * gimple.h: Change uint64_t to int.
27261 2018-01-12  Martin Liska  <mliska@suse.cz>
27263         * tree-core.h: Use uint64_t instead of int.
27264         * tree.c (tree_node_counts): Likewise.
27265         (tree_node_sizes): Likewise.
27266         (dump_tree_statistics): Use PRIu64 in printf format.
27268 2018-01-12  Martin Liska  <mliska@suse.cz>
27270         * Makefile.in: As qsort_chk is implemented in vec.c, add
27271         vec.o to linkage of gencfn-macros.
27272         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
27273         passing the info to record_node_allocation_statistics.
27274         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
27275         and pass the info.
27276         * ggc-common.c (struct ggc_usage): Add operator== and use
27277         it in operator< and compare function.
27278         * mem-stats.h (struct mem_usage): Likewise.
27279         * vec.c (struct vec_usage): Remove operator< and compare
27280         function. Can be simply inherited.
27282 2018-01-12  Martin Jambor  <mjambor@suse.cz>
27284         PR target/81616
27285         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
27286         * tree-ssa-math-opts.c: Include domwalk.h.
27287         (convert_mult_to_fma_1): New function.
27288         (fma_transformation_info): New type.
27289         (fma_deferring_state): Likewise.
27290         (cancel_fma_deferring): New function.
27291         (result_of_phi): Likewise.
27292         (last_fma_candidate_feeds_initial_phi): Likewise.
27293         (convert_mult_to_fma): Added deferring logic, split actual
27294         transformation to convert_mult_to_fma_1.
27295         (math_opts_dom_walker): New type.
27296         (math_opts_dom_walker::after_dom_children): New method, body moved
27297         here from pass_optimize_widening_mul::execute, added deferring logic
27298         bits.
27299         (pass_optimize_widening_mul::execute): Moved most of code to
27300         math_opts_dom_walker::after_dom_children.
27301         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
27302         * config/i386/i386.c (ix86_option_override_internal): Added
27303         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
27305 2018-01-12  Richard Biener  <rguenther@suse.de>
27307         PR debug/83157
27308         * dwarf2out.c (gen_variable_die): Do not reset old_die for
27309         inline instance vars.
27311 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
27313         PR target/81819
27314         * config/rx/rx.c (rx_is_restricted_memory_address):
27315         Handle SUBREG case.
27317 2018-01-12  Richard Biener  <rguenther@suse.de>
27319         PR tree-optimization/80846
27320         * target.def (split_reduction): New target hook.
27321         * targhooks.c (default_split_reduction): New function.
27322         * targhooks.h (default_split_reduction): Declare.
27323         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
27324         target requests first reduce vectors by combining low and high
27325         parts.
27326         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
27327         (get_vectype_for_scalar_type_and_size): Export.
27328         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
27329         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
27330         * doc/tm.texi: Regenerate.
27331         * config/i386/i386.c (ix86_split_reduction): Implement
27332         TARGET_VECTORIZE_SPLIT_REDUCTION.
27334 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27336         PR target/83368
27337         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
27338         in PIC mode except for TARGET_VXWORKS_RTP.
27339         * config/sparc/sparc.c: Include cfgrtl.h.
27340         (TARGET_INIT_PIC_REG): Define.
27341         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
27342         (sparc_pic_register_p): New predicate.
27343         (sparc_legitimate_address_p): Use it.
27344         (sparc_legitimize_pic_address): Likewise.
27345         (sparc_delegitimize_address): Likewise.
27346         (sparc_mode_dependent_address_p): Likewise.
27347         (gen_load_pcrel_sym): Remove 4th parameter.
27348         (load_got_register): Adjust call to above.  Remove obsolete stuff.
27349         (sparc_expand_prologue): Do not call load_got_register here.
27350         (sparc_flat_expand_prologue): Likewise.
27351         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
27352         (sparc_use_pseudo_pic_reg): New function.
27353         (sparc_init_pic_reg): Likewise.
27354         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27355         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
27357 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
27359         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
27360         Add item for branch_cost.
27362 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27364         PR rtl-optimization/83565
27365         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
27366         not extend the result to a larger mode for rotate operations.
27367         (num_sign_bit_copies1): Likewise.
27369 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27371         PR target/40411
27372         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
27373         -symbolic.
27374         Use values-Xc.o for -pedantic.
27375         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
27377 2018-01-12  Martin Liska  <mliska@suse.cz>
27379         PR ipa/83054
27380         * ipa-devirt.c (final_warning_record::grow_type_warnings):
27381         New function.
27382         (possible_polymorphic_call_targets): Use it.
27383         (ipa_devirt): Likewise.
27385 2018-01-12  Martin Liska  <mliska@suse.cz>
27387         * profile-count.h (enum profile_quality): Use 0 as invalid
27388         enum value of profile_quality.
27390 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
27392         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
27393         -mext-string options.
27395 2018-01-12  Richard Biener  <rguenther@suse.de>
27397         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
27398         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
27399         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
27400         Likewise.
27401         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
27403 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
27405         * configure.ac (--with-long-double-format): Add support for the
27406         configuration option to change the default long double format on
27407         PowerPC systems.
27408         * config.gcc (powerpc*-linux*-*): Likewise.
27409         * configure: Regenerate.
27410         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
27411         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
27412         used without modification.
27414 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27416         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
27417         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
27418         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
27419         MISC_BUILTIN_SPEC_BARRIER.
27420         (rs6000_init_builtins): Likewise.
27421         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
27422         enum value.
27423         (speculation_barrier): New define_insn.
27424         * doc/extend.texi: Document __builtin_speculation_barrier.
27426 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27428         PR target/83203
27429         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
27430         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
27431         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
27432         iterators.
27433         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
27434         integral modes instead of "ss" and "sd".
27435         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
27436         vectors with 32-bit and 64-bit elements.
27437         (vecdupssescalarmodesuffix): New mode attribute.
27438         (vec_dup<mode>): Use it.
27440 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
27442         PR target/83330
27443         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
27444         frame if argument is passed on stack.
27446 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27448         PR target/82682
27449         * ree.c (combine_reaching_defs): Optimize also
27450         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
27451         reg2=any_extend(exp); reg1=reg2;, formatting fix.
27453 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27455         PR middle-end/83189
27456         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
27458 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27460         PR middle-end/83718
27461         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
27462         after they are computed.
27464 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
27466         PR tree-optimization/83695
27467         * gimple-loop-linterchange.cc
27468         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
27469         reset cached scev information after interchange.
27470         (pass_linterchange::execute): Remove call to scev_reset_htab.
27472 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27474         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
27475         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
27476         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
27477         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
27478         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
27479         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
27480         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
27481         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
27482         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
27483         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
27484         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
27485         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
27486         (V_lane_reg): Likewise.
27487         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
27488         New define_expand.
27489         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
27490         (vfmal_lane_low<mode>_intrinsic,
27491         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
27492         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27493         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27494         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27495         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27496         vfmsl_lane_high<mode>_intrinsic): New define_insns.
27498 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27500         * config/arm/arm-cpus.in (fp16fml): New feature.
27501         (ALL_SIMD): Add fp16fml.
27502         (armv8.2-a): Add fp16fml as an option.
27503         (armv8.3-a): Likewise.
27504         (armv8.4-a): Add fp16fml as part of fp16.
27505         * config/arm/arm.h (TARGET_FP16FML): Define.
27506         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27507         when appropriate.
27508         * config/arm/arm-modes.def (V2HF): Define.
27509         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27510         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27511         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27512         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27513         vfmsl_low, vfmsl_high): New set of builtins.
27514         * config/arm/iterators.md (PLUSMINUS): New code iterator.
27515         (vfml_op): New code attribute.
27516         (VFMLHALVES): New int iterator.
27517         (VFML, VFMLSEL): New mode attributes.
27518         (V_reg): Define mapping for V2HF.
27519         (V_hi, V_lo): New mode attributes.
27520         (VF_constraint): Likewise.
27521         (vfml_half, vfml_half_selector): New int attributes.
27522         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27523         define_expand.
27524         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27525         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27526         New define_insn.
27527         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27528         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27529         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27530         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
27531         documentation.
27532         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27533         Document new effective target and option set.
27535 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27537         * config/arm/arm-cpus.in (armv8_4): New feature.
27538         (ARMv8_4a): New fgroup.
27539         (armv8.4-a): New arch.
27540         * config/arm/arm-tables.opt: Regenerate.
27541         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27542         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27543         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27544         Add matching rules for -march=armv8.4-a and extensions.
27545         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27547 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
27549         PR target/81821
27550         * config/rx/rx.md (BW): New mode attribute.
27551         (sync_lock_test_and_setsi): Add mode suffix to insn output.
27553 2018-01-11  Richard Biener  <rguenther@suse.de>
27555         PR tree-optimization/83435
27556         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27557         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27558         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27560 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27561             Alan Hayward  <alan.hayward@arm.com>
27562             David Sherwood  <david.sherwood@arm.com>
27564         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27565         field.
27566         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
27567         (aarch64_print_address_internal): Use it to check for a zero offset.
27569 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27570             Alan Hayward  <alan.hayward@arm.com>
27571             David Sherwood  <david.sherwood@arm.com>
27573         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27574         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27575         Return a poly_int64 rather than a HOST_WIDE_INT.
27576         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27577         rather than a HOST_WIDE_INT.
27578         * config/aarch64/aarch64.h (aarch64_frame): Protect with
27579         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
27580         hard_fp_offset, frame_size, initial_adjust, callee_offset and
27581         final_offset from HOST_WIDE_INT to poly_int64.
27582         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27583         to_constant when getting the number of units in an Advanced SIMD
27584         mode.
27585         (aarch64_builtin_vectorized_function): Check for a constant number
27586         of units.
27587         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27588         GET_MODE_SIZE.
27589         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27590         attribute instead of GET_MODE_NUNITS.
27591         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27592         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27593         GET_MODE_SIZE for fixed-size registers.
27594         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27595         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27596         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27597         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27598         (aarch64_print_operand, aarch64_print_address_internal)
27599         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27600         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27601         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27602         Handle polynomial GET_MODE_SIZE.
27603         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
27604         wider than SImode without modification.
27605         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27606         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27607         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27608         passing and returning SVE modes.
27609         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27610         rather than GEN_INT.
27611         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27612         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27613         (aarch64_allocate_and_probe_stack_space): Likewise.
27614         (aarch64_layout_frame): Cope with polynomial offsets.
27615         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27616         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
27617         polynomial offsets.
27618         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27619         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27620         poly_int64 rather than a HOST_WIDE_INT.
27621         (aarch64_get_separate_components, aarch64_process_components)
27622         (aarch64_expand_prologue, aarch64_expand_epilogue)
27623         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27624         (aarch64_anchor_offset): New function, split out from...
27625         (aarch64_legitimize_address): ...here.
27626         (aarch64_builtin_vectorization_cost): Handle polynomial
27627         TYPE_VECTOR_SUBPARTS.
27628         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27629         GET_MODE_NUNITS.
27630         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27631         number of elements from the PARALLEL rather than the mode.
27632         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27633         rather than GET_MODE_BITSIZE.
27634         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27635         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27636         (aarch64_expand_vec_perm_const_1): Handle polynomial
27637         d->perm.length () and d->perm elements.
27638         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
27639         Apply to_constant to d->perm elements.
27640         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27641         polynomial CONST_VECTOR_NUNITS.
27642         (aarch64_move_pointer): Take amount as a poly_int64 rather
27643         than an int.
27644         (aarch64_progress_pointer): Avoid temporary variable.
27645         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27646         the mode attribute instead of GET_MODE.
27648 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27649             Alan Hayward  <alan.hayward@arm.com>
27650             David Sherwood  <david.sherwood@arm.com>
27652         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27653         x exists before using it.
27654         (aarch64_add_constant_internal): Rename to...
27655         (aarch64_add_offset_1): ...this.  Replace regnum with separate
27656         src and dest rtxes.  Handle the case in which they're different,
27657         including when the offset is zero.  Replace scratchreg with an rtx.
27658         Use 2 additions if there is no spare register into which we can
27659         move a 16-bit constant.
27660         (aarch64_add_constant): Delete.
27661         (aarch64_add_offset): Replace reg with separate src and dest
27662         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
27663         Use aarch64_add_offset_1.
27664         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27665         an rtx rather than an int.  Take the delta as a poly_int64
27666         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
27667         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27668         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27669         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27670         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27671         and aarch64_add_sp.
27672         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27673         aarch64_add_constant.
27675 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27677         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27678         Use scalar_float_mode.
27680 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27682         * config/aarch64/aarch64-simd.md
27683         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27684         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27685         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27686         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27687         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27688         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27689         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27690         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27691         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27692         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27694 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27696         PR target/83514
27697         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27698         targ_options->x_arm_arch_string is non NULL.
27700 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
27702         * config/aarch64/aarch64.h
27703         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
27705 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
27707         PR target/82096
27708         * expmed.c (emit_store_flag_force): Swap if const op0
27709         and change VOIDmode to mode of op0.
27711 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27713         PR rtl-optimization/83761
27714         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27715         than bytes to mode_for_size.
27717 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27719         PR middle-end/83189
27720         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27721         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27722         profile.
27724 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27726         PR middle-end/83575
27727         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27728         when in layout mode.
27729         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27730         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27731         partition fixup.
27733 2018-01-10  Michael Collison  <michael.collison@arm.com>
27735         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27736         * config/aarch64/aarch64-option-extension.def: Add
27737         AARCH64_OPT_EXTENSION of 'fp16fml'.
27738         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27739         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27740         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27741         * config/aarch64/constraints.md (Ui7): New constraint.
27742         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27743         (VFMLA_SEL_W): Ditto.
27744         (f16quad): Ditto.
27745         (f16mac1): Ditto.
27746         (VFMLA16_LOW): New int iterator.
27747         (VFMLA16_HIGH): Ditto.
27748         (UNSPEC_FMLAL): New unspec.
27749         (UNSPEC_FMLSL): Ditto.
27750         (UNSPEC_FMLAL2): Ditto.
27751         (UNSPEC_FMLSL2): Ditto.
27752         (f16mac): New code attribute.
27753         * config/aarch64/aarch64-simd-builtins.def
27754         (aarch64_fmlal_lowv2sf): Ditto.
27755         (aarch64_fmlsl_lowv2sf): Ditto.
27756         (aarch64_fmlalq_lowv4sf): Ditto.
27757         (aarch64_fmlslq_lowv4sf): Ditto.
27758         (aarch64_fmlal_highv2sf): Ditto.
27759         (aarch64_fmlsl_highv2sf): Ditto.
27760         (aarch64_fmlalq_highv4sf): Ditto.
27761         (aarch64_fmlslq_highv4sf): Ditto.
27762         (aarch64_fmlal_lane_lowv2sf): Ditto.
27763         (aarch64_fmlsl_lane_lowv2sf): Ditto.
27764         (aarch64_fmlal_laneq_lowv2sf): Ditto.
27765         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27766         (aarch64_fmlalq_lane_lowv4sf): Ditto.
27767         (aarch64_fmlsl_lane_lowv4sf): Ditto.
27768         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27769         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27770         (aarch64_fmlal_lane_highv2sf): Ditto.
27771         (aarch64_fmlsl_lane_highv2sf): Ditto.
27772         (aarch64_fmlal_laneq_highv2sf): Ditto.
27773         (aarch64_fmlsl_laneq_highv2sf): Ditto.
27774         (aarch64_fmlalq_lane_highv4sf): Ditto.
27775         (aarch64_fmlsl_lane_highv4sf): Ditto.
27776         (aarch64_fmlalq_laneq_highv4sf): Ditto.
27777         (aarch64_fmlsl_laneq_highv4sf): Ditto.
27778         * config/aarch64/aarch64-simd.md:
27779         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27780         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27781         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27782         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27783         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27784         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27785         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27786         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27787         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27788         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27789         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27790         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27791         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27792         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27793         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27794         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27795         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27796         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27797         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27798         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27799         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27800         (vfmlsl_low_u32): Ditto.
27801         (vfmlalq_low_u32): Ditto.
27802         (vfmlslq_low_u32): Ditto.
27803         (vfmlal_high_u32): Ditto.
27804         (vfmlsl_high_u32): Ditto.
27805         (vfmlalq_high_u32): Ditto.
27806         (vfmlslq_high_u32): Ditto.
27807         (vfmlal_lane_low_u32): Ditto.
27808         (vfmlsl_lane_low_u32): Ditto.
27809         (vfmlal_laneq_low_u32): Ditto.
27810         (vfmlsl_laneq_low_u32): Ditto.
27811         (vfmlalq_lane_low_u32): Ditto.
27812         (vfmlslq_lane_low_u32): Ditto.
27813         (vfmlalq_laneq_low_u32): Ditto.
27814         (vfmlslq_laneq_low_u32): Ditto.
27815         (vfmlal_lane_high_u32): Ditto.
27816         (vfmlsl_lane_high_u32): Ditto.
27817         (vfmlal_laneq_high_u32): Ditto.
27818         (vfmlsl_laneq_high_u32): Ditto.
27819         (vfmlalq_lane_high_u32): Ditto.
27820         (vfmlslq_lane_high_u32): Ditto.
27821         (vfmlalq_laneq_high_u32): Ditto.
27822         (vfmlslq_laneq_high_u32): Ditto.
27823         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27824         (AARCH64_FL_FOR_ARCH8_4): New.
27825         (AARCH64_ISA_F16FML): New ISA flag.
27826         (TARGET_F16FML): New feature flag for fp16fml.
27827         (doc/invoke.texi): Document new fp16fml option.
27829 2018-01-10  Michael Collison  <michael.collison@arm.com>
27831         * config/aarch64/aarch64-builtins.c:
27832         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27833         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27834         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27835         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27836         (AARCH64_ISA_SHA3): New ISA flag.
27837         (TARGET_SHA3): New feature flag for sha3.
27838         * config/aarch64/iterators.md (sha512_op): New int attribute.
27839         (CRYPTO_SHA512): New int iterator.
27840         (UNSPEC_SHA512H): New unspec.
27841         (UNSPEC_SHA512H2): Ditto.
27842         (UNSPEC_SHA512SU0): Ditto.
27843         (UNSPEC_SHA512SU1): Ditto.
27844         * config/aarch64/aarch64-simd-builtins.def
27845         (aarch64_crypto_sha512hqv2di): New builtin.
27846         (aarch64_crypto_sha512h2qv2di): Ditto.
27847         (aarch64_crypto_sha512su0qv2di): Ditto.
27848         (aarch64_crypto_sha512su1qv2di): Ditto.
27849         (aarch64_eor3qv8hi): Ditto.
27850         (aarch64_rax1qv2di): Ditto.
27851         (aarch64_xarqv2di): Ditto.
27852         (aarch64_bcaxqv8hi): Ditto.
27853         * config/aarch64/aarch64-simd.md:
27854         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27855         (aarch64_crypto_sha512su0qv2di): Ditto.
27856         (aarch64_crypto_sha512su1qv2di): Ditto.
27857         (aarch64_eor3qv8hi): Ditto.
27858         (aarch64_rax1qv2di): Ditto.
27859         (aarch64_xarqv2di): Ditto.
27860         (aarch64_bcaxqv8hi): Ditto.
27861         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27862         (vsha512h2q_u64): Ditto.
27863         (vsha512su0q_u64): Ditto.
27864         (vsha512su1q_u64): Ditto.
27865         (veor3q_u16): Ditto.
27866         (vrax1q_u64): Ditto.
27867         (vxarq_u64): Ditto.
27868         (vbcaxq_u16): Ditto.
27869         * config/arm/types.md (crypto_sha512): New type attribute.
27870         (crypto_sha3): Ditto.
27871         (doc/invoke.texi): Document new sha3 option.
27873 2018-01-10  Michael Collison  <michael.collison@arm.com>
27875         * config/aarch64/aarch64-builtins.c:
27876         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
27877         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27878         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
27879         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
27880         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
27881         (AARCH64_ISA_SM4): New ISA flag.
27882         (TARGET_SM4): New feature flag for sm4.
27883         * config/aarch64/aarch64-simd-builtins.def
27884         (aarch64_sm3ss1qv4si): Ditto.
27885         (aarch64_sm3tt1aq4si): Ditto.
27886         (aarch64_sm3tt1bq4si): Ditto.
27887         (aarch64_sm3tt2aq4si): Ditto.
27888         (aarch64_sm3tt2bq4si): Ditto.
27889         (aarch64_sm3partw1qv4si): Ditto.
27890         (aarch64_sm3partw2qv4si): Ditto.
27891         (aarch64_sm4eqv4si): Ditto.
27892         (aarch64_sm4ekeyqv4si): Ditto.
27893         * config/aarch64/aarch64-simd.md:
27894         (aarch64_sm3ss1qv4si): Ditto.
27895         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
27896         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
27897         (aarch64_sm4eqv4si): Ditto.
27898         (aarch64_sm4ekeyqv4si): Ditto.
27899         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
27900         (sm3part_op): Ditto.
27901         (CRYPTO_SM3TT): Ditto.
27902         (CRYPTO_SM3PART): Ditto.
27903         (UNSPEC_SM3SS1): New unspec.
27904         (UNSPEC_SM3TT1A): Ditto.
27905         (UNSPEC_SM3TT1B): Ditto.
27906         (UNSPEC_SM3TT2A): Ditto.
27907         (UNSPEC_SM3TT2B): Ditto.
27908         (UNSPEC_SM3PARTW1): Ditto.
27909         (UNSPEC_SM3PARTW2): Ditto.
27910         (UNSPEC_SM4E): Ditto.
27911         (UNSPEC_SM4EKEY): Ditto.
27912         * config/aarch64/constraints.md (Ui2): New constraint.
27913         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
27914         * config/arm/types.md (crypto_sm3): New type attribute.
27915         (crypto_sm4): Ditto.
27916         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
27917         (vsm3tt1aq_u32): Ditto.
27918         (vsm3tt1bq_u32): Ditto.
27919         (vsm3tt2aq_u32): Ditto.
27920         (vsm3tt2bq_u32): Ditto.
27921         (vsm3partw1q_u32): Ditto.
27922         (vsm3partw2q_u32): Ditto.
27923         (vsm4eq_u32): Ditto.
27924         (vsm4ekeyq_u32): Ditto.
27925         (doc/invoke.texi): Document new sm4 option.
27927 2018-01-10  Michael Collison  <michael.collison@arm.com>
27929         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
27930         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
27931         (AARCH64_FL_FOR_ARCH8_4): New.
27932         (AARCH64_FL_V8_4): New flag.
27933         (doc/invoke.texi): Document new armv8.4-a option.
27935 2018-01-10  Michael Collison  <michael.collison@arm.com>
27937         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27938         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
27939         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
27940         * config/aarch64/aarch64-option-extension.def: Add
27941         AARCH64_OPT_EXTENSION of 'sha2'.
27942         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
27943         (crypto): Disable sha2 and aes if crypto disabled.
27944         (crypto): Enable aes and sha2 if enabled.
27945         (simd): Disable sha2 and aes if simd disabled.
27946         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
27947         New flags.
27948         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
27949         (TARGET_SHA2): New feature flag for sha2.
27950         (TARGET_AES): New feature flag for aes.
27951         * config/aarch64/aarch64-simd.md:
27952         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
27953         conditional on TARGET_AES.
27954         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
27955         (aarch64_crypto_sha1hsi): Make pattern conditional
27956         on TARGET_SHA2.
27957         (aarch64_crypto_sha1hv4si): Ditto.
27958         (aarch64_be_crypto_sha1hv4si): Ditto.
27959         (aarch64_crypto_sha1su1v4si): Ditto.
27960         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
27961         (aarch64_crypto_sha1su0v4si): Ditto.
27962         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
27963         (aarch64_crypto_sha256su0v4si): Ditto.
27964         (aarch64_crypto_sha256su1v4si): Ditto.
27965         (doc/invoke.texi): Document new aes and sha2 options.
27967 2018-01-10  Martin Sebor  <msebor@redhat.com>
27969         PR tree-optimization/83781
27970         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
27971         as string arrays.
27973 2018-01-11  Martin Sebor  <msebor@gmail.com>
27974             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27976         PR tree-optimization/83501
27977         PR tree-optimization/81703
27979         * tree-ssa-strlen.c (get_string_cst): Rename...
27980         (get_string_len): ...to this.  Handle global constants.
27981         (handle_char_store): Adjust.
27983 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
27984             Jim Wilson  <jimw@sifive.com>
27986         * config/riscv/riscv-protos.h (riscv_output_return): New.
27987         * config/riscv/riscv.c (struct machine_function): New naked_p field.
27988         (riscv_attribute_table, riscv_output_return),
27989         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
27990         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
27991         (riscv_compute_frame_info): Only compute frame->mask if not a naked
27992         function.
27993         (riscv_expand_prologue): Add early return for naked function.
27994         (riscv_expand_epilogue): Likewise.
27995         (riscv_function_ok_for_sibcall): Return false for naked function.
27996         (riscv_set_current_function): New.
27997         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
27998         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
27999         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
28000         * doc/extend.texi (RISC-V Function Attributes): New.
28002 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
28004         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
28005         check for 128-bit long double before checking TCmode.
28006         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
28007         128-bit long doubles before checking TFmode or TCmode.
28008         (FLOAT128_IBM_P): Likewise.
28010 2018-01-10  Martin Sebor  <msebor@redhat.com>
28012         PR tree-optimization/83671
28013         * builtins.c (c_strlen): Unconditionally return zero for the empty
28014         string.
28015         Use -Warray-bounds for warnings.
28016         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
28017         for non-constant array indices with COMPONENT_REF, arrays of
28018         arrays, and pointers to arrays.
28019         (gimple_fold_builtin_strlen): Determine and set length range for
28020         non-constant character arrays.
28022 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
28024         PR middle-end/81897
28025         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
28026         empty blocks.
28028 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
28030         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
28032 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
28034         PR target/83399
28035         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
28036         VECTOR_MEM_ALTIVEC_OR_VSX_P.
28037         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
28038         indexed_or_indirect_operand predicate.
28039         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
28040         (*vsx_le_perm_load_v8hi): Likewise.
28041         (*vsx_le_perm_load_v16qi): Likewise.
28042         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
28043         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
28044         (*vsx_le_perm_store_v8hi): Likewise.
28045         (*vsx_le_perm_store_v16qi): Likewise.
28046         (eight unnamed splitters): Likewise.
28048 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
28050         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
28051         * config/rs6000/emmintrin.h: Likewise.
28052         * config/rs6000/mmintrin.h: Likewise.
28053         * config/rs6000/xmmintrin.h: Likewise.
28055 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
28057         PR c++/43486
28058         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
28059         "public_flag".
28060         * tree.c (tree_nop_conversion): Return true for location wrapper
28061         nodes.
28062         (maybe_wrap_with_location): New function.
28063         (selftest::check_strip_nops): New function.
28064         (selftest::test_location_wrappers): New function.
28065         (selftest::tree_c_tests): Call it.
28066         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
28067         (maybe_wrap_with_location): New decl.
28068         (EXPR_LOCATION_WRAPPER_P): New macro.
28069         (location_wrapper_p): New inline function.
28070         (tree_strip_any_location_wrapper): New inline function.
28072 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
28074         PR target/83735
28075         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
28076         stack_realign_offset for the largest alignment of stack slot
28077         actually used.
28078         (ix86_find_max_used_stack_alignment): New function.
28079         (ix86_finalize_stack_frame_flags): Use it.  Set
28080         max_used_stack_alignment if we don't realign stack.
28081         * config/i386/i386.h (machine_function): Add
28082         max_used_stack_alignment.
28084 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
28086         * config/arm/arm.opt (-mbranch-cost): New option.
28087         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
28088         account.
28090 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
28092         PR target/83629
28093         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
28094         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
28096 2018-01-10  Richard Biener  <rguenther@suse.de>
28098         PR debug/83765
28099         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
28100         early out so it also covers the case where we have a non-NULL
28101         origin.
28103 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28105         PR tree-optimization/83753
28106         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
28107         for non-strided grouped accesses if the number of elements is 1.
28109 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28111         PR target/81616
28112         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
28113         * i386.h (TARGET_USE_GATHER): Define.
28114         * x86-tune.def (X86_TUNE_USE_GATHER): New.
28116 2018-01-10  Martin Liska  <mliska@suse.cz>
28118         PR bootstrap/82831
28119         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
28120         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
28121         partitioning.
28122         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
28123         CLEANUP_NO_PARTITIONING is not set.
28125 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28127         * doc/rtl.texi: Remove documentation of (const ...) wrappers
28128         for vectors, as a partial revert of r254296.
28129         * rtl.h (const_vec_p): Delete.
28130         (const_vec_duplicate_p): Don't test for vector CONSTs.
28131         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
28132         * expmed.c (make_tree): Likewise.
28134         Revert:
28135         * common.md (E, F): Use CONSTANT_P instead of checking for
28136         CONST_VECTOR.
28137         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
28138         checking for CONST_VECTOR.
28140 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28142         PR middle-end/83575
28143         * predict.c (force_edge_cold): Handle in more sane way edges
28144         with no prediction.
28146 2018-01-09  Carl Love  <cel@us.ibm.com>
28148         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
28149         V4SI, V4SF types.
28150         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
28151         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
28152         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
28153         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
28154         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
28155         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
28156         * config/rs6000/rs6000-protos.h: Add extern defition for
28157         rs6000_generate_float2_double_code.
28158         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
28159         function.
28160         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
28161         (float2_v2df): Add define_expand.
28163 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
28165         PR target/83628
28166         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
28167         op_mode in the force_to_mode call.
28169 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28171         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
28172         instead of checking each element individually.
28173         (aarch64_evpc_uzp): Likewise.
28174         (aarch64_evpc_zip): Likewise.
28175         (aarch64_evpc_ext): Likewise.
28176         (aarch64_evpc_rev): Likewise.
28177         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
28178         instead of checking each element individually.  Return true without
28179         generating rtl if
28180         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
28181         whether all selected elements come from the same input, instead of
28182         checking each element individually.  Remove calls to gen_rtx_REG,
28183         start_sequence and end_sequence and instead assert that no rtl is
28184         generated.
28186 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28188         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
28189         order of HIGH and CONST checks.
28191 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28193         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
28194         if the destination isn't an SSA_NAME.
28196 2018-01-09  Richard Biener  <rguenther@suse.de>
28198         PR tree-optimization/83668
28199         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
28200         move prologue...
28201         (canonicalize_loop_form): ... here, renamed from ...
28202         (canonicalize_loop_closed_ssa_form): ... this and amended to
28203         swap successor edges for loop exit blocks to make us use
28204         the RPO order we need for initial schedule generation.
28206 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
28208         PR tree-optimization/64811
28209         * match.pd: When optimizing comparisons with Inf, avoid
28210         introducing or losing exceptions from comparisons with NaN.
28212 2018-01-09  Martin Liska  <mliska@suse.cz>
28214         PR sanitizer/82517
28215         * asan.c (shadow_mem_size): Add gcc_assert.
28217 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
28219         Don't save registers in main().
28221         PR target/83738
28222         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
28223         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
28224         * config/avr/avr.c (avr_set_current_function): Don't error if
28225         naked, OS_task or OS_main are specified at the same time.
28226         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
28227         OS_main.
28228         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
28229         attribute.
28230         * common/config/avr/avr-common.c (avr_option_optimization_table):
28231         Switch on -mmain-is-OS_task for optimizing compilations.
28233 2018-01-09  Richard Biener  <rguenther@suse.de>
28235         PR tree-optimization/83572
28236         * graphite.c: Include cfganal.h.
28237         (graphite_transform_loops): Connect infinite loops to exit
28238         and remove fake edges at the end.
28240 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28242         * ipa-inline.c (edge_badness): Revert accidental checkin.
28244 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28246         PR ipa/80763
28247         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
28248         symbols; not inline clones.
28250 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
28252         PR target/83507
28253         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
28254         hard registers.  Formatting fixes.
28256         PR preprocessor/83722
28257         * gcc.c (try_generate_repro): Pass
28258         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
28259         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
28260         do_report_bug.
28262 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
28263             Kito Cheng  <kito.cheng@gmail.com>
28265         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
28266         (riscv_leaf_function_p): Delete.
28267         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
28269 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28271         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
28272         function.
28273         (do_ifelse): New function.
28274         (do_isel): New function.
28275         (do_sub3): New function.
28276         (do_add3): New function.
28277         (do_load_mask_compare): New function.
28278         (do_overlap_load_compare): New function.
28279         (expand_compare_loop): New function.
28280         (expand_block_compare): Call expand_compare_loop() when appropriate.
28281         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
28282         option description.
28283         (-mblock-compare-inline-loop-limit): New option.
28285 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28287         PR target/83677
28288         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
28289         Reverse order of second and third operands in first alternative.
28290         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
28291         of first and second elements in UNSPEC_VPERMR vector.
28292         (altivec_expand_vec_perm_le): Likewise.
28294 2018-01-08  Jeff Law  <law@redhat.com>
28296         PR rtl-optimizatin/81308
28297         * tree-switch-conversion.c (cfg_altered): New file scoped static.
28298         (process_switch): If group_case_labels makes a change, then set
28299         cfg_altered.
28300         (pass_convert_switch::execute): If a switch is converted, then
28301         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
28303         PR rtl-optimization/81308
28304         * recog.c (split_all_insns): Conditionally cleanup the CFG after
28305         splitting insns.
28307 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
28309         PR target/83663 - Revert r255946
28310         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
28311         generation for cases where splatting a value is not useful.
28312         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
28313         across a vec_duplicate and a paradoxical subreg forming a vector
28314         mode to a vec_concat.
28316 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28318         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
28319         -march=armv8.3-a variants.
28320         * config/arm/t-multilib: Likewise.
28321         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
28323 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28325         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
28326         to generate rtl.
28327         (cceq_ior_compare_complement): Give it a name so I can use it, and
28328         change boolean_or_operator predicate to boolean_operator so it can
28329         be used to generate a crand.
28330         (eqne): New code iterator.
28331         (bd/bd_neg): New code_attrs.
28332         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
28333         a single define_insn.
28334         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
28335         decrement (bdnzt/bdnzf/bdzt/bdzf).
28336         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
28337         with the new names of the branch decrement patterns, and added the
28338         names of the branch decrement conditional patterns.
28340 2018-01-08  Richard Biener  <rguenther@suse.de>
28342         PR tree-optimization/83563
28343         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
28344         cache.
28346 2018-01-08  Richard Biener  <rguenther@suse.de>
28348         PR middle-end/83713
28349         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
28351 2018-01-08  Richard Biener  <rguenther@suse.de>
28353         PR tree-optimization/83685
28354         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
28355         references to abnormals.
28357 2018-01-08  Richard Biener  <rguenther@suse.de>
28359         PR lto/83719
28360         * dwarf2out.c (output_indirect_strings): Handle empty
28361         skeleton_debug_str_hash.
28362         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
28364 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28366         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
28367         (emit_store_direct): Likewise.
28368         (arc_trampoline_adjust_address): Likewise.
28369         (arc_asm_trampoline_template): New function.
28370         (arc_initialize_trampoline): Use asm_trampoline_template.
28371         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
28372         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
28373         * config/arc/arc.md (flush_icache): Delete pattern.
28375 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28377         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
28378         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
28379         munaligned-access.
28381 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28383         PR target/83681
28384         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
28385         by not USED_FOR_TARGET.
28386         (make_pass_resolve_sw_modes): Likewise.
28388 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28390         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
28391         USED_FOR_TARGET.
28393 2018-01-08  Richard Biener  <rguenther@suse.de>
28395         PR middle-end/83580
28396         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
28398 2018-01-08  Richard Biener  <rguenther@suse.de>
28400         PR middle-end/83517
28401         * match.pd ((t * 2) / 2) -> t): Add missing :c.
28403 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
28405         PR middle-end/81897
28406         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
28407         basic blocks with a small number of successors.
28408         (convert_control_dep_chain_into_preds): Improve handling of
28409         forwarder blocks.
28410         (dump_predicates): Split apart into...
28411         (dump_pred_chain): ...here...
28412         (dump_pred_info): ...and here.
28413         (can_one_predicate_be_invalidated_p): Add debugging printfs.
28414         (can_chain_union_be_invalidated_p): Improve check for invalidation
28415         of paths.
28416         (uninit_uses_cannot_happen): Avoid unnecessary if
28417         convert_control_dep_chain_into_preds yielded nothing.
28419 2018-01-06  Martin Sebor  <msebor@redhat.com>
28421         PR tree-optimization/83640
28422         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
28423         subtracting negative offset from size.
28424         (builtin_access::overlap): Adjust offset bounds of the access to fall
28425         within the size of the object if possible.
28427 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
28429         PR rtl-optimization/83699
28430         * expmed.c (extract_bit_field_1): Restrict the vector usage of
28431         extract_bit_field_as_subreg to cases in which the extracted
28432         value is also a vector.
28434         * lra-constraints.c (process_alt_operands): Test for the equivalence
28435         substitutions when detecting a possible reload cycle.
28437 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
28439         PR debug/83480
28440         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
28441         by default if flag_selective_schedling{,2}.  Formatting fixes.
28443         PR rtl-optimization/83682
28444         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
28445         if it has non-VECTOR_MODE element mode.
28446         (vec_duplicate_p): Likewise.
28448         PR middle-end/83694
28449         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
28450         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
28452 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28454         PR target/83604
28455         * config/i386/i386-builtin.def
28456         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
28457         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
28458         Require also OPTION_MASK_ISA_AVX512F in addition to
28459         OPTION_MASK_ISA_GFNI.
28460         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
28461         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
28462         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
28463         to OPTION_MASK_ISA_GFNI.
28464         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
28465         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
28466         OPTION_MASK_ISA_AVX512BW.
28467         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
28468         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
28469         addition to OPTION_MASK_ISA_GFNI.
28470         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
28471         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
28472         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
28473         to OPTION_MASK_ISA_GFNI.
28474         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
28475         a requirement for all ISAs rather than any of them with a few
28476         exceptions.
28477         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
28478         processing.
28479         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
28480         bitmasks to be enabled with 3 exceptions, instead of requiring any
28481         enabled ISA with lots of exceptions.
28482         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
28483         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
28484         Change avx512bw in isa attribute to avx512f.
28485         * config/i386/sgxintrin.h: Add license boilerplate.
28486         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
28487         to __AVX512F__ and __AVX512VL to __AVX512VL__.
28488         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
28489         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
28490         defined.
28491         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
28492         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28493         temporarily sse2 rather than sse if not enabled already.
28495         PR target/83604
28496         * config/i386/sse.md (VI248_VLBW): Rename to ...
28497         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
28498         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28499         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28500         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28501         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28502         mode iterator instead of VI248_VLBW.
28504 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
28506         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28507         (record_modified): Skip clobbers; add debug output.
28508         (param_change_prob): Use sreal frequencies.
28510 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28512         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28513         punt for user-aligned variables.
28515 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28517         * tree-chrec.c (chrec_contains_symbols): Return true for
28518         POLY_INT_CST.
28520 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
28522         PR target/82439
28523         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28524         of (x|y) == x for BICS pattern.
28526 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28528         PR tree-optimization/83605
28529         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28530         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28531         can throw.
28533 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28535         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28536         * config/epiphany/rtems.h: New file.
28538 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28539             Uros Bizjak  <ubizjak@gmail.com>
28541         PR target/83554
28542         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28543         QIreg_operand instead of register_operand predicate.
28544         * config/i386/i386.c (ix86_rop_should_change_byte_p,
28545         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28546         comments instead of -fmitigate[-_]rop.
28548 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28550         PR bootstrap/81926
28551         * cgraphunit.c (symbol_table::compile): Switch to text_section
28552         before calling assembly_start debug hook.
28553         * run-rtl-passes.c (run_rtl_passes): Likewise.
28554         Include output.h.
28556 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28558         * tree-vrp.c (extract_range_from_binary_expr_1): Check
28559         range_int_cst_p rather than !symbolic_range_p before calling
28560         extract_range_from_multiplicative_op_1.
28562 2018-01-04  Jeff Law  <law@redhat.com>
28564         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28565         redundant test in assertion.
28567 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28569         * doc/rtl.texi: Document machine_mode wrapper classes.
28571 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28573         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28574         using tree_to_uhwi.
28576 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28578         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28579         the VEC_PERM_EXPR fold to fail.
28581 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28583         PR debug/83585
28584         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28585         to switched_sections.
28587 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28589         PR target/83680
28590         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28591         test for d.testing.
28593 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
28595         PR target/83387
28596         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28597         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28599 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28601         PR debug/83666
28602         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28603         is BLKmode and bitpos not zero or mode change is needed.
28605 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28607         PR target/83675
28608         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28609         TARGET_VIS2.
28611 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
28613         PR target/83628
28614         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28615         instead of MULT rtx.  Update all corresponding splitters.
28616         (*saddl_se): Ditto.
28617         (*ssub<modesuffix>): Ditto.
28618         (*ssubl_se): Ditto.
28619         (*cmp_sadd_di): Update split patterns.
28620         (*cmp_sadd_si): Ditto.
28621         (*cmp_sadd_sidi): Ditto.
28622         (*cmp_ssub_di): Ditto.
28623         (*cmp_ssub_si): Ditto.
28624         (*cmp_ssub_sidi): Ditto.
28625         * config/alpha/predicates.md (const23_operand): New predicate.
28626         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28627         Look for ASHIFT, not MULT inner operand.
28628         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28630 2018-01-04  Martin Liska  <mliska@suse.cz>
28632         PR gcov-profile/83669
28633         * gcov.c (output_intermediate_file): Add version to intermediate
28634         gcov file.
28635         * doc/gcov.texi: Document new field 'version' in intermediate
28636         file format. Fix location of '-k' option of gcov command.
28638 2018-01-04  Martin Liska  <mliska@suse.cz>
28640         PR ipa/82352
28641         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28643 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28645         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28647 2018-01-03  Martin Sebor  <msebor@redhat.com>
28649         PR tree-optimization/83655
28650         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28651         checking calls with invalid arguments.
28653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28655         * tree-vect-stmts.c (vect_get_store_rhs): New function.
28656         (vectorizable_mask_load_store): Delete.
28657         (vectorizable_call): Return false for masked loads and stores.
28658         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
28659         instead of gimple_assign_rhs1.
28660         (vectorizable_load): Handle IFN_MASK_LOAD.
28661         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28663 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28665         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28666         split out from..,
28667         (vectorizable_mask_load_store): ...here.
28668         (vectorizable_load): ...and here.
28670 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28672         * tree-vect-stmts.c (vect_build_all_ones_mask)
28673         (vect_build_zero_merge_argument): New functions, split out from...
28674         (vectorizable_load): ...here.
28676 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28678         * tree-vect-stmts.c (vect_check_store_rhs): New function,
28679         split out from...
28680         (vectorizable_mask_load_store): ...here.
28681         (vectorizable_store): ...and here.
28683 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28685         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28686         split out from...
28687         (vectorizable_mask_load_store): ...here.
28689 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28691         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28692         (vect_model_store_cost): Take a vec_load_store_type instead of a
28693         vect_def_type.
28694         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28695         (vect_model_store_cost): Take a vec_load_store_type instead of a
28696         vect_def_type.
28697         (vectorizable_mask_load_store): Update accordingly.
28698         (vectorizable_store): Likewise.
28699         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28701 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28703         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28704         IFN_MASK_LOAD calls here rather than...
28705         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28707 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28708             Alan Hayward  <alan.hayward@arm.com>
28709             David Sherwood  <david.sherwood@arm.com>
28711         * expmed.c (extract_bit_field_1): For vector extracts,
28712         fall back to extract_bit_field_as_subreg if vec_extract
28713         isn't available.
28715 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28716             Alan Hayward  <alan.hayward@arm.com>
28717             David Sherwood  <david.sherwood@arm.com>
28719         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28720         they are variable or constant sized.
28721         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28722         slots for constant-sized data.
28724 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28725             Alan Hayward  <alan.hayward@arm.com>
28726             David Sherwood  <david.sherwood@arm.com>
28728         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28729         handling COND_EXPRs with boolean comparisons, try to find a better
28730         basis for the mask type than the boolean itself.
28732 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28734         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28735         is calculated and how it can be overridden.
28736         * genmodes.c (max_bitsize_mode_any_mode): New variable.
28737         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28738         if defined.
28739         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28740         if nonzero.
28742 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28743             Alan Hayward  <alan.hayward@arm.com>
28744             David Sherwood  <david.sherwood@arm.com>
28746         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28747         Remove the mode argument.
28748         (aarch64_simd_valid_immediate): Remove the mode and inverse
28749         arguments.
28750         * config/aarch64/iterators.md (bitsize): New iterator.
28751         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28752         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28753         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28754         aarch64_simd_valid_immediate.
28755         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28756         (aarch64_reg_or_bic_imm): Likewise.
28757         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28758         with an insn_type enum and msl with a modifier_type enum.
28759         Replace element_width with a scalar_mode.  Change the shift
28760         to unsigned int.  Add constructors for scalar_float_mode and
28761         scalar_int_mode elements.
28762         (aarch64_vect_float_const_representable_p): Delete.
28763         (aarch64_can_const_movi_rtx_p)
28764         (aarch64_simd_scalar_immediate_valid_for_move)
28765         (aarch64_simd_make_constant): Update call to
28766         aarch64_simd_valid_immediate.
28767         (aarch64_advsimd_valid_immediate_hs): New function.
28768         (aarch64_advsimd_valid_immediate): Likewise.
28769         (aarch64_simd_valid_immediate): Remove mode and inverse
28770         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
28771         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28772         and aarch64_float_const_representable_p on the result.
28773         (aarch64_output_simd_mov_immediate): Remove mode argument.
28774         Update call to aarch64_simd_valid_immediate and use of
28775         simd_immediate_info.
28776         (aarch64_output_scalar_simd_mov_immediate): Update call
28777         accordingly.
28779 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28780             Alan Hayward  <alan.hayward@arm.com>
28781             David Sherwood  <david.sherwood@arm.com>
28783         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28784         (mode_nunits): Likewise CONST_MODE_NUNITS.
28785         * machmode.def (ADJUST_NUNITS): Document.
28786         * genmodes.c (mode_data::need_nunits_adj): New field.
28787         (blank_mode): Update accordingly.
28788         (adj_nunits): New variable.
28789         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28790         parameter.
28791         (emit_mode_size_inline): Set need_bytesize_adj for all modes
28792         listed in adj_nunits.
28793         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28794         listed in adj_nunits.  Don't emit case statements for such modes.
28795         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28796         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
28797         nothing if adj_nunits is nonnull.
28798         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28799         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28800         (emit_mode_fbit): Update use of print_maybe_const_decl.
28801         (emit_move_size): Likewise.  Treat the array as non-const
28802         if adj_nunits.
28803         (emit_mode_adjustments): Handle adj_nunits.
28805 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28807         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28808         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28809         (VECTOR_MODES): Use it.
28810         (make_vector_modes): Take the prefix as an argument.
28812 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28813             Alan Hayward  <alan.hayward@arm.com>
28814             David Sherwood  <david.sherwood@arm.com>
28816         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28817         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28818         for MODE_VECTOR_BOOL.
28819         * machmode.def (VECTOR_BOOL_MODE): Document.
28820         * genmodes.c (VECTOR_BOOL_MODE): New macro.
28821         (make_vector_bool_mode): New function.
28822         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28823         MODE_VECTOR_BOOL.
28824         * lto-streamer-in.c (lto_input_mode_table): Likewise.
28825         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28826         Likewise.
28827         * stor-layout.c (int_mode_for_mode): Likewise.
28828         * tree.c (build_vector_type_for_mode): Likewise.
28829         * varasm.c (output_constant_pool_2): Likewise.
28830         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28831         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
28832         for MODE_VECTOR_BOOL.
28833         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28834         of mode class checks.
28835         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28836         instead of a list of mode class checks.
28837         (expand_vector_scalar_condition): Likewise.
28838         (type_for_widest_vector_mode): Handle BImode as an inner mode.
28840 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28841             Alan Hayward  <alan.hayward@arm.com>
28842             David Sherwood  <david.sherwood@arm.com>
28844         * machmode.h (mode_size): Change from unsigned short to
28845         poly_uint16_pod.
28846         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28847         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28848         or if measurement_type is not polynomial.
28849         (fixed_size_mode::includes_p): Check for constant-sized modes.
28850         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28851         return a poly_uint16 rather than an unsigned short.
28852         (emit_mode_size): Change the type of mode_size from unsigned short
28853         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
28854         (emit_mode_adjustments): Cope with polynomial vector sizes.
28855         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28856         for GET_MODE_SIZE.
28857         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28858         for GET_MODE_SIZE.
28859         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28860         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28861         * caller-save.c (setup_save_areas): Likewise.
28862         (replace_reg_with_saved_mem): Likewise.
28863         * calls.c (emit_library_call_value_1): Likewise.
28864         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28865         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28866         (gen_lowpart_for_combine): Likewise.
28867         * convert.c (convert_to_integer_1): Likewise.
28868         * cse.c (equiv_constant, cse_insn): Likewise.
28869         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28870         (cselib_subst_to_values): Likewise.
28871         * dce.c (word_dce_process_block): Likewise.
28872         * df-problems.c (df_word_lr_mark_ref): Likewise.
28873         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28874         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28875         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
28876         (rtl_for_decl_location): Likewise.
28877         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
28878         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
28879         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
28880         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
28881         (expand_expr_real_1): Likewise.
28882         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
28883         (pad_below): Likewise.
28884         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28885         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
28886         * ira.c (get_subreg_tracking_sizes): Likewise.
28887         * ira-build.c (ira_create_allocno_objects): Likewise.
28888         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
28889         (ira_sort_regnos_for_alter_reg): Likewise.
28890         * ira-costs.c (record_operand_costs): Likewise.
28891         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
28892         (resolve_simple_move): Likewise.
28893         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
28894         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
28895         (lra_constraints): Likewise.
28896         (CONST_POOL_OK_P): Reject variable-sized modes.
28897         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
28898         (add_pseudo_to_slot, lra_spill): Likewise.
28899         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28900         * optabs-query.c (get_best_extraction_insn): Likewise.
28901         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28902         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
28903         (expand_mult_highpart, valid_multiword_target_p): Likewise.
28904         * recog.c (offsettable_address_addr_space_p): Likewise.
28905         * regcprop.c (maybe_mode_change): Likewise.
28906         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
28907         * regrename.c (build_def_use): Likewise.
28908         * regstat.c (dump_reg_info): Likewise.
28909         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
28910         (find_reloads, find_reloads_subreg_address): Likewise.
28911         * reload1.c (eliminate_regs_1): Likewise.
28912         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
28913         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
28914         (simplify_binary_operation_1, simplify_subreg): Likewise.
28915         * targhooks.c (default_function_arg_padding): Likewise.
28916         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
28917         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
28918         (verify_gimple_assign_ternary): Likewise.
28919         * tree-inline.c (estimate_move_cost): Likewise.
28920         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28921         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
28922         (get_address_cost_ainc): Likewise.
28923         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
28924         (vect_supportable_dr_alignment): Likewise.
28925         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28926         (vectorizable_reduction): Likewise.
28927         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
28928         (vectorizable_operation, vectorizable_load): Likewise.
28929         * tree.c (build_same_sized_truth_vector_type): Likewise.
28930         * valtrack.c (cleanup_auto_inc_dec): Likewise.
28931         * var-tracking.c (emit_note_insn_var_location): Likewise.
28932         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
28933         (ADDR_VEC_ALIGN): Likewise.
28935 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28936             Alan Hayward  <alan.hayward@arm.com>
28937             David Sherwood  <david.sherwood@arm.com>
28939         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
28940         unsigned short.
28941         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28942         or if measurement_type is polynomial.
28943         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
28944         * combine.c (make_extraction): Likewise.
28945         * dse.c (find_shift_sequence): Likewise.
28946         * dwarf2out.c (mem_loc_descriptor): Likewise.
28947         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
28948         (extract_bit_field, extract_low_bits): Likewise.
28949         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
28950         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
28951         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
28952         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
28953         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28954         * reload.c (find_reloads): Likewise.
28955         * reload1.c (alter_reg): Likewise.
28956         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
28957         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
28958         * tree-if-conv.c (predicate_mem_writes): Likewise.
28959         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
28960         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
28961         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
28962         * valtrack.c (dead_debug_insert_temp): Likewise.
28963         * varasm.c (mergeable_constant_section): Likewise.
28964         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
28966 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28967             Alan Hayward  <alan.hayward@arm.com>
28968             David Sherwood  <david.sherwood@arm.com>
28970         * expr.c (expand_assignment): Cope with polynomial mode sizes
28971         when assigning to a CONCAT.
28973 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28974             Alan Hayward  <alan.hayward@arm.com>
28975             David Sherwood  <david.sherwood@arm.com>
28977         * machmode.h (mode_precision): Change from unsigned short to
28978         poly_uint16_pod.
28979         (mode_to_precision): Return a poly_uint16 rather than an unsigned
28980         short.
28981         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
28982         or if measurement_type is not polynomial.
28983         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
28984         in which the mode is already known to be a scalar_int_mode.
28985         * genmodes.c (emit_mode_precision): Change the type of mode_precision
28986         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
28987         initializer.
28988         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28989         for GET_MODE_PRECISION.
28990         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28991         for GET_MODE_PRECISION.
28992         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
28993         as polynomial.
28994         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
28995         (expand_field_assignment, make_extraction): Likewise.
28996         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
28997         (get_last_value): Likewise.
28998         * convert.c (convert_to_integer_1): Likewise.
28999         * cse.c (cse_insn): Likewise.
29000         * expr.c (expand_expr_real_1): Likewise.
29001         * lra-constraints.c (simplify_operand_subreg): Likewise.
29002         * optabs-query.c (can_atomic_load_p): Likewise.
29003         * optabs.c (expand_atomic_load): Likewise.
29004         (expand_atomic_store): Likewise.
29005         * ree.c (combine_reaching_defs): Likewise.
29006         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
29007         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
29008         * tree.h (type_has_mode_precision_p): Likewise.
29009         * ubsan.c (instrument_si_overflow): Likewise.
29011 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29012             Alan Hayward  <alan.hayward@arm.com>
29013             David Sherwood  <david.sherwood@arm.com>
29015         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
29016         polynomial numbers of units.
29017         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
29018         (valid_vector_subparts_p): New function.
29019         (build_vector_type): Remove temporary shim and take the number
29020         of units as a poly_uint64 rather than an int.
29021         (build_opaque_vector_type): Take the number of units as a
29022         poly_uint64 rather than an int.
29023         * tree.c (build_vector_from_ctor): Handle polynomial
29024         TYPE_VECTOR_SUBPARTS.
29025         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
29026         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
29027         (build_vector_from_val): If the number of units is variable,
29028         use build_vec_duplicate_cst for constant operands and
29029         VEC_DUPLICATE_EXPR otherwise.
29030         (make_vector_type): Remove temporary is_constant ().
29031         (build_vector_type, build_opaque_vector_type): Take the number of
29032         units as a poly_uint64 rather than an int.
29033         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
29034         VECTOR_CST_NELTS.
29035         * cfgexpand.c (expand_debug_expr): Likewise.
29036         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
29037         (store_constructor, expand_expr_real_1): Likewise.
29038         (const_scalar_mask_from_tree): Likewise.
29039         * fold-const-call.c (fold_const_reduction): Likewise.
29040         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
29041         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
29042         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
29043         (fold_relational_const): Likewise.
29044         (native_interpret_vector): Likewise.  Change the size from an
29045         int to an unsigned int.
29046         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
29047         TYPE_VECTOR_SUBPARTS.
29048         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
29049         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
29050         duplicating a non-constant operand into a variable-length vector.
29051         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
29052         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
29053         * ipa-icf.c (sem_variable::equals): Likewise.
29054         * match.pd: Likewise.
29055         * omp-simd-clone.c (simd_clone_subparts): Likewise.
29056         * print-tree.c (print_node): Likewise.
29057         * stor-layout.c (layout_type): Likewise.
29058         * targhooks.c (default_builtin_vectorization_cost): Likewise.
29059         * tree-cfg.c (verify_gimple_comparison): Likewise.
29060         (verify_gimple_assign_binary): Likewise.
29061         (verify_gimple_assign_ternary): Likewise.
29062         (verify_gimple_assign_single): Likewise.
29063         * tree-pretty-print.c (dump_generic_node): Likewise.
29064         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29065         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
29066         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
29067         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
29068         (vect_shift_permute_load_chain): Likewise.
29069         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
29070         (expand_vector_condition, optimize_vector_constructor): Likewise.
29071         (lower_vec_perm, get_compute_type): Likewise.
29072         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29073         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
29074         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
29075         (vect_recog_mask_conversion_pattern): Likewise.
29076         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
29077         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
29078         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29079         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
29080         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
29081         (vectorizable_shift, vectorizable_operation, vectorizable_store)
29082         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
29083         (supportable_widening_operation): Likewise.
29084         (supportable_narrowing_operation): Likewise.
29085         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
29086         Likewise.
29087         * varasm.c (output_constant): Likewise.
29089 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29090             Alan Hayward  <alan.hayward@arm.com>
29091             David Sherwood  <david.sherwood@arm.com>
29093         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
29094         so that both the length == 3 and length != 3 cases set up their
29095         own permute vectors.  Add comments explaining why we know the
29096         number of elements is constant.
29097         (vect_permute_load_chain): Likewise.
29099 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29100             Alan Hayward  <alan.hayward@arm.com>
29101             David Sherwood  <david.sherwood@arm.com>
29103         * machmode.h (mode_nunits): Change from unsigned char to
29104         poly_uint16_pod.
29105         (ONLY_FIXED_SIZE_MODES): New macro.
29106         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
29107         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
29108         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
29109         New typedefs.
29110         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
29111         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
29112         or if measurement_type is not polynomial.
29113         * genmodes.c (ZERO_COEFFS): New macro.
29114         (emit_mode_nunits_inline): Make mode_nunits_inline return a
29115         poly_uint16.
29116         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
29117         Use ZERO_COEFFS when emitting initializers.
29118         * data-streamer.h (bp_pack_poly_value): New function.
29119         (bp_unpack_poly_value): Likewise.
29120         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29121         for GET_MODE_NUNITS.
29122         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29123         for GET_MODE_NUNITS.
29124         * tree.c (make_vector_type): Remove temporary shim and make
29125         the real function take the number of units as a poly_uint64
29126         rather than an int.
29127         (build_vector_type_for_mode): Handle polynomial nunits.
29128         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
29129         * emit-rtl.c (const_vec_series_p_1): Likewise.
29130         (gen_rtx_CONST_VECTOR): Likewise.
29131         * fold-const.c (test_vec_duplicate_folding): Likewise.
29132         * genrecog.c (validate_pattern): Likewise.
29133         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
29134         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29135         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
29136         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
29137         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
29138         * rtlanal.c (subreg_get_info): Likewise.
29139         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29140         (vect_grouped_load_supported): Likewise.
29141         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
29142         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
29143         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
29144         (simplify_const_unary_operation, simplify_binary_operation_1)
29145         (simplify_const_binary_operation, simplify_ternary_operation)
29146         (test_vector_ops_duplicate, test_vector_ops): Likewise.
29147         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
29148         instead of CONST_VECTOR_NUNITS.
29149         * varasm.c (output_constant_pool_2): Likewise.
29150         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
29151         explicit-encoded elements in the XVEC for variable-length vectors.
29153 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29155         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
29157 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29158             Alan Hayward  <alan.hayward@arm.com>
29159             David Sherwood  <david.sherwood@arm.com>
29161         * coretypes.h (fixed_size_mode): Declare.
29162         (fixed_size_mode_pod): New typedef.
29163         * builtins.h (target_builtins::x_apply_args_mode)
29164         (target_builtins::x_apply_result_mode): Change type to
29165         fixed_size_mode_pod.
29166         * builtins.c (apply_args_size, apply_result_size, result_vector)
29167         (expand_builtin_apply_args_1, expand_builtin_apply)
29168         (expand_builtin_return): Update accordingly.
29170 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29172         * cse.c (hash_rtx_cb): Hash only the encoded elements.
29173         * cselib.c (cselib_hash_rtx): Likewise.
29174         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
29175         CONST_VECTOR encoding.
29177 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29178             Jeff Law  <law@redhat.com>
29180         PR target/83641
29181         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
29182         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
29183         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
29184         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
29186         PR target/83641
29187         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
29188         explicitly probe *sp in a noreturn function if there were any callee
29189         register saves or frame pointer is needed.
29191 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29193         PR debug/83621
29194         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
29195         BLKmode for ternary, binary or unary expressions.
29197         PR debug/83645
29198         * var-tracking.c (delete_vta_debug_insn): New inline function.
29199         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
29200         insns from get_insns () to NULL instead of each bb separately.
29201         Use delete_vta_debug_insn.  No longer static.
29202         (vt_debug_insns_local, variable_tracking_main_1): Adjust
29203         delete_vta_debug_insns callers.
29204         * rtl.h (delete_vta_debug_insns): Declare.
29205         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
29206         instead of variable_tracking_main.
29208 2018-01-03  Martin Sebor  <msebor@redhat.com>
29210         PR tree-optimization/83603
29211         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
29212         arguments past the endof the argument list in functions declared
29213         without a prototype.
29214         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
29215         Avoid checking when arguments are null.
29217 2018-01-03  Martin Sebor  <msebor@redhat.com>
29219         PR c/83559
29220         * doc/extend.texi (attribute const): Fix a typo.
29221         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
29222         issuing -Wsuggest-attribute for void functions.
29224 2018-01-03  Martin Sebor  <msebor@redhat.com>
29226         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
29227         offset_int::from instead of wide_int::to_shwi.
29228         (maybe_diag_overlap): Remove assertion.
29229         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
29230         * gimple-ssa-sprintf.c (format_directive): Same.
29231         (parse_directive): Same.
29232         (sprintf_dom_walker::compute_format_length): Same.
29233         (try_substitute_return_value): Same.
29235 2018-01-03  Jeff Law  <law@redhat.com>
29237         PR middle-end/83654
29238         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
29239         non-constant residual for zero at runtime and avoid probing in
29240         that case.  Reorganize code for trailing problem to mirror handling
29241         of the residual.
29243 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29245         PR tree-optimization/83501
29246         * tree-ssa-strlen.c (get_string_cst): New.
29247         (handle_char_store): Call get_string_cst.
29249 2018-01-03  Martin Liska  <mliska@suse.cz>
29251         PR tree-optimization/83593
29252         * tree-ssa-strlen.c: Include tree-cfg.h.
29253         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
29254         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
29255         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
29256         to false.
29257         (strlen_dom_walker::before_dom_children): Call
29258         gimple_purge_dead_eh_edges. Dump tranformation with details
29259         dump flags.
29260         (strlen_dom_walker::before_dom_children): Update call by adding
29261         new argument cleanup_eh.
29262         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
29264 2018-01-03  Martin Liska  <mliska@suse.cz>
29266         PR ipa/83549
29267         * cif-code.def (VARIADIC_THUNK): New enum value.
29268         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
29269         thunks.
29271 2018-01-03  Jan Beulich  <jbeulich@suse.com>
29273         * sse.md (mov<mode>_internal): Tighten condition for when to use
29274         vmovdqu<ssescalarsize> for TI and OI modes.
29276 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29278         Update copyright years.
29280 2018-01-03  Martin Liska  <mliska@suse.cz>
29282         PR ipa/83594
29283         * ipa-visibility.c (function_and_variable_visibility): Skip
29284         functions with noipa attribure.
29286 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29288         * gcc.c (process_command): Update copyright notice dates.
29289         * gcov-dump.c (print_version): Ditto.
29290         * gcov.c (print_version): Ditto.
29291         * gcov-tool.c (print_version): Ditto.
29292         * gengtype.c (create_file): Ditto.
29293         * doc/cpp.texi: Bump @copying's copyright year.
29294         * doc/cppinternals.texi: Ditto.
29295         * doc/gcc.texi: Ditto.
29296         * doc/gccint.texi: Ditto.
29297         * doc/gcov.texi: Ditto.
29298         * doc/install.texi: Ditto.
29299         * doc/invoke.texi: Ditto.
29301 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29303         * vector-builder.h (vector_builder::m_full_nelts): Change from
29304         unsigned int to poly_uint64.
29305         (vector_builder::full_nelts): Update prototype accordingly.
29306         (vector_builder::new_vector): Likewise.
29307         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
29308         (vector_builder::operator ==): Likewise.
29309         (vector_builder::finalize): Likewise.
29310         * int-vector-builder.h (int_vector_builder::int_vector_builder):
29311         Take the number of elements as a poly_uint64 rather than an
29312         unsigned int.
29313         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
29314         from unsigned int to poly_uint64.
29315         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
29316         (vec_perm_indices::new_vector): Likewise.
29317         (vec_perm_indices::length): Likewise.
29318         (vec_perm_indices::nelts_per_input): Likewise.
29319         (vec_perm_indices::input_nelts): Likewise.
29320         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
29321         number of elements per input as a poly_uint64 rather than an
29322         unsigned int.  Use the original encoding for variable-length
29323         vectors, rather than clamping each individual element.
29324         For the second and subsequent elements in each pattern,
29325         clamp the step and base before clamping their sum.
29326         (vec_perm_indices::series_p): Handle polynomial element counts.
29327         (vec_perm_indices::all_in_range_p): Likewise.
29328         (vec_perm_indices_to_tree): Likewise.
29329         (vec_perm_indices_to_rtx): Likewise.
29330         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
29331         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
29332         (tree_vector_builder::new_binary_operation): Handle polynomial
29333         element counts.  Return false if we need to know the number
29334         of elements at compile time.
29335         * fold-const.c (fold_vec_perm): Punt if the number of elements
29336         isn't known at compile time.
29338 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29340         * vec-perm-indices.h (vec_perm_builder): Change element type
29341         from HOST_WIDE_INT to poly_int64.
29342         (vec_perm_indices::element_type): Update accordingly.
29343         (vec_perm_indices::clamp): Handle polynomial element_types.
29344         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29345         (vec_perm_indices::all_in_range_p): Likewise.
29346         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
29347         than shwi trees.
29348         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
29349         polynomial vec_perm_indices element types.
29350         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
29351         * fold-const.c (fold_vec_perm): Likewise.
29352         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
29353         * tree-vect-generic.c (lower_vec_perm): Likewise.
29354         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29355         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
29356         element type to HOST_WIDE_INT.
29358 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29359             Alan Hayward  <alan.hayward@arm.com>
29360             David Sherwood  <david.sherwood@arm.com>
29362         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
29363         rather than an int.  Use plus_constant.
29364         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
29365         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
29367 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29368             Alan Hayward  <alan.hayward@arm.com>
29369             David Sherwood  <david.sherwood@arm.com>
29371         * calls.c (emit_call_1, expand_call): Change struct_value_size from
29372         a HOST_WIDE_INT to a poly_int64.
29374 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29375             Alan Hayward  <alan.hayward@arm.com>
29376             David Sherwood  <david.sherwood@arm.com>
29378         * calls.c (load_register_parameters): Cope with polynomial
29379         mode sizes.  Require a constant size for BLKmode parameters
29380         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
29381         forces a parameter to be padded at the lsb end in order to
29382         fill a complete number of words, require the parameter size
29383         to be ordered wrt UNITS_PER_WORD.
29385 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29386             Alan Hayward  <alan.hayward@arm.com>
29387             David Sherwood  <david.sherwood@arm.com>
29389         * reload1.c (spill_stack_slot_width): Change element type
29390         from unsigned int to poly_uint64_pod.
29391         (alter_reg): Treat mode sizes as polynomial.
29393 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29394             Alan Hayward  <alan.hayward@arm.com>
29395             David Sherwood  <david.sherwood@arm.com>
29397         * reload.c (complex_word_subreg_p): New function.
29398         (reload_inner_reg_of_subreg, push_reload): Use it.
29400 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29401             Alan Hayward  <alan.hayward@arm.com>
29402             David Sherwood  <david.sherwood@arm.com>
29404         * lra-constraints.c (process_alt_operands): Reject matched
29405         operands whose sizes aren't ordered.
29406         (match_reload): Refer to this check here.
29408 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29409             Alan Hayward  <alan.hayward@arm.com>
29410             David Sherwood  <david.sherwood@arm.com>
29412         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
29413         that the mode size is in the set {1, 2, 4, 8, 16}.
29415 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29416             Alan Hayward  <alan.hayward@arm.com>
29417             David Sherwood  <david.sherwood@arm.com>
29419         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
29420         Use plus_constant instead of gen_rtx_PLUS.
29422 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29423             Alan Hayward  <alan.hayward@arm.com>
29424             David Sherwood  <david.sherwood@arm.com>
29426         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
29427         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
29428         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
29429         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
29430         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
29431         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
29432         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
29433         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
29434         * config/i386/i386.c (ix86_push_rounding): ...this new function.
29435         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29436         a poly_int64.
29437         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
29438         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
29439         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
29440         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
29441         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
29442         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
29443         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
29444         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
29445         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
29446         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
29447         function.
29448         * expr.c (emit_move_resolve_push): Treat the input and result
29449         of PUSH_ROUNDING as a poly_int64.
29450         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
29451         (emit_push_insn): Likewise.
29452         * lra-eliminations.c (mark_not_eliminable): Likewise.
29453         * recog.c (push_operand): Likewise.
29454         * reload1.c (elimination_effects): Likewise.
29455         * rtlanal.c (nonzero_bits1): Likewise.
29456         * calls.c (store_one_arg): Likewise.  Require the padding to be
29457         known at compile time.
29459 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29460             Alan Hayward  <alan.hayward@arm.com>
29461             David Sherwood  <david.sherwood@arm.com>
29463         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
29464         Use plus_constant instead of gen_rtx_PLUS.
29466 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29467             Alan Hayward  <alan.hayward@arm.com>
29468             David Sherwood  <david.sherwood@arm.com>
29470         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
29471         rather than an int.
29473 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29474             Alan Hayward  <alan.hayward@arm.com>
29475             David Sherwood  <david.sherwood@arm.com>
29477         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
29478         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
29479         via stack temporaries.  Treat the mode size as polynomial too.
29481 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29482             Alan Hayward  <alan.hayward@arm.com>
29483             David Sherwood  <david.sherwood@arm.com>
29485         * expr.c (expand_expr_real_2): When handling conversions involving
29486         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
29487         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
29488         as a poly_uint64 too.
29490 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29491             Alan Hayward  <alan.hayward@arm.com>
29492             David Sherwood  <david.sherwood@arm.com>
29494         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
29496 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29497             Alan Hayward  <alan.hayward@arm.com>
29498             David Sherwood  <david.sherwood@arm.com>
29500         * combine.c (can_change_dest_mode): Handle polynomial
29501         REGMODE_NATURAL_SIZE.
29502         * expmed.c (store_bit_field_1): Likewise.
29503         * expr.c (store_constructor): Likewise.
29504         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29505         and polynomial REGMODE_NATURAL_SIZE.
29506         (gen_lowpart_common): Likewise.
29507         * reginfo.c (record_subregs_of_mode): Likewise.
29508         * rtlanal.c (read_modify_subreg_p): Likewise.
29510 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29511             Alan Hayward  <alan.hayward@arm.com>
29512             David Sherwood  <david.sherwood@arm.com>
29514         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29515         numbers of elements.
29517 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29518             Alan Hayward  <alan.hayward@arm.com>
29519             David Sherwood  <david.sherwood@arm.com>
29521         * match.pd: Cope with polynomial numbers of vector elements.
29523 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29524             Alan Hayward  <alan.hayward@arm.com>
29525             David Sherwood  <david.sherwood@arm.com>
29527         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29528         in a POINTER_PLUS_EXPR.
29530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29531             Alan Hayward  <alan.hayward@arm.com>
29532             David Sherwood  <david.sherwood@arm.com>
29534         * omp-simd-clone.c (simd_clone_subparts): New function.
29535         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29536         (ipa_simd_modify_function_body): Likewise.
29538 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29539             Alan Hayward  <alan.hayward@arm.com>
29540             David Sherwood  <david.sherwood@arm.com>
29542         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29543         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29544         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29545         (expand_vector_condition, vector_element): Likewise.
29546         (subparts_gt): New function.
29547         (get_compute_type): Use subparts_gt.
29548         (count_type_subparts): Delete.
29549         (expand_vector_operations_1): Use subparts_gt instead of
29550         count_type_subparts.
29552 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29553             Alan Hayward  <alan.hayward@arm.com>
29554             David Sherwood  <david.sherwood@arm.com>
29556         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29557         (vect_compile_time_alias): ...this new function.  Do the calculation
29558         on poly_ints rather than trees.
29559         (vect_prune_runtime_alias_test_list): Update call accordingly.
29561 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29562             Alan Hayward  <alan.hayward@arm.com>
29563             David Sherwood  <david.sherwood@arm.com>
29565         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29566         numbers of units.
29567         (vect_schedule_slp_instance): Likewise.
29569 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29570             Alan Hayward  <alan.hayward@arm.com>
29571             David Sherwood  <david.sherwood@arm.com>
29573         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29574         constant and extern definitions for variable-length vectors.
29575         (vect_get_constant_vectors): Note that the number of units
29576         is known to be constant.
29578 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29579             Alan Hayward  <alan.hayward@arm.com>
29580             David Sherwood  <david.sherwood@arm.com>
29582         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29583         of units as polynomial.  Choose between WIDE and NARROW based
29584         on multiple_p.
29586 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29587             Alan Hayward  <alan.hayward@arm.com>
29588             David Sherwood  <david.sherwood@arm.com>
29590         * tree-vect-stmts.c (simd_clone_subparts): New function.
29591         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
29593 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29594             Alan Hayward  <alan.hayward@arm.com>
29595             David Sherwood  <david.sherwood@arm.com>
29597         * tree-vect-stmts.c (vectorizable_call): Treat the number of
29598         vectors as polynomial.  Use build_index_vector for
29599         IFN_GOMP_SIMD_LANE.
29601 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29602             Alan Hayward  <alan.hayward@arm.com>
29603             David Sherwood  <david.sherwood@arm.com>
29605         * tree-vect-stmts.c (get_load_store_type): Treat the number of
29606         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29607         for variable-length vectors.
29608         (vectorizable_mask_load_store): Treat the number of units as
29609         polynomial, asserting that it is constant if the condition has
29610         already been enforced.
29611         (vectorizable_store, vectorizable_load): Likewise.
29613 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29614             Alan Hayward  <alan.hayward@arm.com>
29615             David Sherwood  <david.sherwood@arm.com>
29617         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29618         of units as polynomial.  Punt if we can't tell at compile time
29619         which vector contains the final result.
29621 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29622             Alan Hayward  <alan.hayward@arm.com>
29623             David Sherwood  <david.sherwood@arm.com>
29625         * tree-vect-loop.c (vectorizable_induction): Treat the number
29626         of units as polynomial.  Punt on SLP inductions.  Use an integer
29627         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
29628         cast of such a series for variable-length floating-point
29629         reductions.
29631 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29632             Alan Hayward  <alan.hayward@arm.com>
29633             David Sherwood  <david.sherwood@arm.com>
29635         * tree.h (build_index_vector): Declare.
29636         * tree.c (build_index_vector): New function.
29637         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29638         of units as polynomial, forcibly converting it to a constant if
29639         vectorizable_reduction has already enforced the condition.
29640         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
29641         to create a {1,2,3,...} vector.
29642         (vectorizable_reduction): Treat the number of units as polynomial.
29643         Choose vectype_in based on the largest scalar element size rather
29644         than the smallest number of units.  Enforce the restrictions
29645         relied on above.
29647 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29648             Alan Hayward  <alan.hayward@arm.com>
29649             David Sherwood  <david.sherwood@arm.com>
29651         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29652         number of units as polynomial.
29654 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29655             Alan Hayward  <alan.hayward@arm.com>
29656             David Sherwood  <david.sherwood@arm.com>
29658         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29659         * target.def (autovectorize_vector_sizes): Return the vector sizes
29660         by pointer, using vector_sizes rather than a bitmask.
29661         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29662         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29663         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29664         Likewise.
29665         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29666         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29667         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29668         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29669         * omp-general.c (omp_max_vf): Likewise.
29670         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29671         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29672         * tree-vect-loop.c (vect_analyze_loop): Likewise.
29673         * tree-vect-slp.c (vect_slp_bb): Likewise.
29674         * doc/tm.texi: Regenerate.
29675         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29676         to a poly_uint64.
29677         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29678         the vector size as a poly_uint64 rather than an unsigned int.
29679         (current_vector_size): Change from an unsigned int to a poly_uint64.
29680         (get_vectype_for_scalar_type): Update accordingly.
29681         * tree.h (build_truth_vector_type): Take the size and number of
29682         units as a poly_uint64 rather than an unsigned int.
29683         (build_vector_type): Add a temporary overload that takes
29684         the number of units as a poly_uint64 rather than an unsigned int.
29685         * tree.c (make_vector_type): Likewise.
29686         (build_truth_vector_type): Take the number of units as a poly_uint64
29687         rather than an unsigned int.
29689 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29690             Alan Hayward  <alan.hayward@arm.com>
29691             David Sherwood  <david.sherwood@arm.com>
29693         * target.def (get_mask_mode): Take the number of units and length
29694         as poly_uint64s rather than unsigned ints.
29695         * targhooks.h (default_get_mask_mode): Update accordingly.
29696         * targhooks.c (default_get_mask_mode): Likewise.
29697         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29698         * doc/tm.texi: Regenerate.
29700 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29701             Alan Hayward  <alan.hayward@arm.com>
29702             David Sherwood  <david.sherwood@arm.com>
29704         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29705         * omp-general.c (omp_max_vf): Likewise.
29706         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29707         (expand_omp_simd): Handle polynomial safelen.
29708         * omp-low.c (omplow_simd_context): Add a default constructor.
29709         (omplow_simd_context::max_vf): Change from int to poly_uint64.
29710         (lower_rec_simd_input_clauses): Update accordingly.
29711         (lower_rec_input_clauses): Likewise.
29713 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29714             Alan Hayward  <alan.hayward@arm.com>
29715             David Sherwood  <david.sherwood@arm.com>
29717         * tree-vectorizer.h (vect_nunits_for_cost): New function.
29718         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29719         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29720         (vect_analyze_slp_cost): Likewise.
29721         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29722         (vect_model_load_cost): Likewise.
29724 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29725             Alan Hayward  <alan.hayward@arm.com>
29726             David Sherwood  <david.sherwood@arm.com>
29728         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29729         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29730         from an unsigned int * to a poly_uint64_pod *.
29731         (calculate_unrolling_factor): New function.
29732         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
29734 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29735             Alan Hayward  <alan.hayward@arm.com>
29736             David Sherwood  <david.sherwood@arm.com>
29738         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29739         from an unsigned int to a poly_uint64.
29740         (_loop_vec_info::slp_unrolling_factor): Likewise.
29741         (_loop_vec_info::vectorization_factor): Change from an int
29742         to a poly_uint64.
29743         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29744         (vect_get_num_vectors): New function.
29745         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29746         (vect_get_num_copies): Use vect_get_num_vectors.
29747         (vect_analyze_data_ref_dependences): Change max_vf from an int *
29748         to an unsigned int *.
29749         (vect_analyze_data_refs): Change min_vf from an int * to a
29750         poly_uint64 *.
29751         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29752         than an unsigned HOST_WIDE_INT.
29753         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29754         (vect_analyze_data_ref_dependence): Change max_vf from an int *
29755         to an unsigned int *.
29756         (vect_analyze_data_ref_dependences): Likewise.
29757         (vect_compute_data_ref_alignment): Handle polynomial vf.
29758         (vect_enhance_data_refs_alignment): Likewise.
29759         (vect_prune_runtime_alias_test_list): Likewise.
29760         (vect_shift_permute_load_chain): Likewise.
29761         (vect_supportable_dr_alignment): Likewise.
29762         (dependence_distance_ge_vf): Take the vectorization factor as a
29763         poly_uint64 rather than an unsigned HOST_WIDE_INT.
29764         (vect_analyze_data_refs): Change min_vf from an int * to a
29765         poly_uint64 *.
29766         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29767         vfm1 as a poly_uint64 rather than an int.  Make the same change
29768         for the returned bound_scalar.
29769         (vect_gen_vector_loop_niters): Handle polynomial vf.
29770         (vect_do_peeling): Likewise.  Update call to
29771         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29772         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29773         be constant.
29774         * tree-vect-loop.c (vect_determine_vectorization_factor)
29775         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29776         (vect_get_known_peeling_cost): Likewise.
29777         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29778         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29779         (vect_transform_loop): Likewise.  Use the lowest possible VF when
29780         updating the upper bounds of the loop.
29781         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
29782         rather than an int.
29783         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29784         polynomial unroll factors.
29785         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29786         (vect_make_slp_decision): Likewise.
29787         (vect_supported_load_permutation_p): Likewise, and polynomial
29788         vf too.
29789         (vect_analyze_slp_cost): Handle polynomial vf.
29790         (vect_slp_analyze_node_operations): Likewise.
29791         (vect_slp_analyze_bb_1): Likewise.
29792         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29793         than an unsigned HOST_WIDE_INT.
29794         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29795         (vectorizable_load): Handle polynomial vf.
29796         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29797         a poly_uint64.
29798         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29800 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29801             Alan Hayward  <alan.hayward@arm.com>
29802             David Sherwood  <david.sherwood@arm.com>
29804         * match.pd: Handle bit operations involving three constants
29805         and try to fold one pair.
29807 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29809         * tree-vect-loop-manip.c: Include gimple-fold.h.
29810         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29811         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
29812         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29813         Add a path that uses a step of VF instead of 1, but disable it
29814         for now.
29815         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29816         and niters_no_overflow parameters.  Update calls to
29817         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29818         Create a new SSA name if the latter choses to use a ste other
29819         than zero, and return it via niters_vector_mult_vf_var.
29820         * tree-vect-loop.c (vect_transform_loop): Update calls to
29821         vect_do_peeling, vect_gen_vector_loop_niters and
29822         slpeel_make_loop_iterate_ntimes.
29823         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29824         (vect_gen_vector_loop_niters): Update declarations after above changes.
29826 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
29828         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29829         128-bit round to integer instructions.
29830         (ceil<mode>2): Likewise.
29831         (btrunc<mode>2): Likewise.
29832         (round<mode>2): Likewise.
29834 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29836         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29837         unaligned VSX load/store on P8/P9.
29838         (expand_block_clear): Allow the use of unaligned VSX
29839         load/store on P8/P9.
29841 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29843         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29844         New function.
29845         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29846         swap associated with both a load and a store.
29848 2018-01-02  Andrew Waterman  <andrew@sifive.com>
29850         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29851         * config/riscv/riscv.md (clear_cache): Use it.
29853 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
29855         * web.c: Remove out-of-date comment.
29857 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29859         * expr.c (fixup_args_size_notes): Check that any existing
29860         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29861         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29862         (emit_single_push_insn): ...here.
29864 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29866         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29867         (const_vector_encoded_nelts): New function.
29868         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29869         (const_vector_int_elt, const_vector_elt): Declare.
29870         * emit-rtl.c (const_vector_int_elt_1): New function.
29871         (const_vector_elt): Likewise.
29872         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29873         of CONST_VECTOR_ELT.
29875 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29877         * expr.c: Include rtx-vector-builder.h.
29878         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
29879         directly on the tree encoding.
29880         (const_vector_from_tree): Likewise.
29881         * optabs.c: Include rtx-vector-builder.h.
29882         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
29883         sequence of "u" values.
29884         * vec-perm-indices.c: Include rtx-vector-builder.h.
29885         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
29886         directly on the vec_perm_indices encoding.
29888 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29890         * doc/rtl.texi (const_vector): Describe new encoding scheme.
29891         * Makefile.in (OBJS): Add rtx-vector-builder.o.
29892         * rtx-vector-builder.h: New file.
29893         * rtx-vector-builder.c: Likewise.
29894         * rtl.h (rtx_def::u2): Add a const_vector field.
29895         (CONST_VECTOR_NPATTERNS): New macro.
29896         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
29897         (CONST_VECTOR_DUPLICATE_P): Likewise.
29898         (CONST_VECTOR_STEPPED_P): Likewise.
29899         (CONST_VECTOR_ENCODED_ELT): Likewise.
29900         (const_vec_duplicate_p): Check for a duplicated vector encoding.
29901         (unwrap_const_vec_duplicate): Likewise.
29902         (const_vec_series_p): Check for a non-duplicated vector encoding.
29903         Say that the function only returns true for integer vectors.
29904         * emit-rtl.c: Include rtx-vector-builder.h.
29905         (gen_const_vec_duplicate_1): Delete.
29906         (gen_const_vector): Call gen_const_vec_duplicate instead of
29907         gen_const_vec_duplicate_1.
29908         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
29909         (gen_const_vec_duplicate): Use rtx_vector_builder.
29910         (gen_const_vec_series): Likewise.
29911         (gen_rtx_CONST_VECTOR): Likewise.
29912         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
29913         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29914         Build a new vector rather than modifying a CONST_VECTOR in-place.
29915         (handle_special_swappables): Update call accordingly.
29916         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
29917         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29918         Build a new vector rather than modifying a CONST_VECTOR in-place.
29919         (handle_special_swappables): Update call accordingly.
29921 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29923         * simplify-rtx.c (simplify_const_binary_operation): Use
29924         CONST_VECTOR_ELT instead of XVECEXP.
29926 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29928         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
29929         the selector elements to be different from the data elements
29930         if the selector is a VECTOR_CST.
29931         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
29932         ssizetype for the selector.
29934 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29936         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
29937         before testing each element individually.
29938         * tree-vect-generic.c (lower_vec_perm): Likewise.
29940 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29942         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
29943         * selftest-run-tests.c (selftest::run_tests): Call it.
29944         * vector-builder.h (vector_builder::operator ==): New function.
29945         (vector_builder::operator !=): Likewise.
29946         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
29947         (vec_perm_indices::all_from_input_p): New function.
29948         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29949         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
29950         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
29951         instead of reading the VECTOR_CST directly.  Detect whether both
29952         vector inputs are the same before constructing the vec_perm_indices,
29953         and update the number of inputs argument accordingly.  Use the
29954         utility functions added above.  Only construct sel2 if we need to.
29956 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29958         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
29959         the broadcast of the low byte.
29960         (expand_mult_highpart): Use an explicit encoding for the permutes.
29961         * optabs-query.c (can_mult_highpart_p): Likewise.
29962         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
29963         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29964         (vectorizable_bswap): Likewise.
29965         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
29966         explicit encoding for the power-of-2 permutes.
29967         (vect_permute_store_chain): Likewise.
29968         (vect_grouped_load_supported): Likewise.
29969         (vect_permute_load_chain): Likewise.
29971 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29973         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
29974         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
29975         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
29976         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29977         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
29978         (vect_gen_perm_mask_any): Likewise.
29980 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29982         * int-vector-builder.h: New file.
29983         * vec-perm-indices.h: Include int-vector-builder.h.
29984         (vec_perm_indices): Redefine as an int_vector_builder.
29985         (auto_vec_perm_indices): Delete.
29986         (vec_perm_builder): Redefine as a stand-alone class.
29987         (vec_perm_indices::vec_perm_indices): New function.
29988         (vec_perm_indices::clamp): Likewise.
29989         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
29990         (vec_perm_indices::new_vector): New function.
29991         (vec_perm_indices::new_expanded_vector): Update for new
29992         vec_perm_indices class.
29993         (vec_perm_indices::rotate_inputs): New function.
29994         (vec_perm_indices::all_in_range_p): Operate directly on the
29995         encoded form, without computing elided elements.
29996         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
29997         encoding.  Update for new vec_perm_indices class.
29998         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
29999         the given vec_perm_builder.
30000         (expand_vec_perm_var): Update vec_perm_builder constructor.
30001         (expand_mult_highpart): Use vec_perm_builder instead of
30002         auto_vec_perm_indices.
30003         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
30004         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
30005         or double series encoding as appropriate.
30006         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
30007         vec_perm_indices instead of auto_vec_perm_indices.
30008         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30009         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30010         (vect_permute_store_chain): Likewise.
30011         (vect_grouped_load_supported): Likewise.
30012         (vect_permute_load_chain): Likewise.
30013         (vect_shift_permute_load_chain): Likewise.
30014         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30015         (vect_transform_slp_perm_load): Likewise.
30016         (vect_schedule_slp_instance): Likewise.
30017         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30018         (vectorizable_mask_load_store): Likewise.
30019         (vectorizable_bswap): Likewise.
30020         (vectorizable_store): Likewise.
30021         (vectorizable_load): Likewise.
30022         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
30023         vec_perm_indices instead of auto_vec_perm_indices.  Use
30024         tree_to_vec_perm_builder to read the vector from a tree.
30025         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
30026         vec_perm_builder instead of a vec_perm_indices.
30027         (have_whole_vector_shift): Use vec_perm_builder and
30028         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
30029         truncation to calc_vec_perm_mask_for_shift.
30030         (vect_create_epilog_for_reduction): Likewise.
30031         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
30032         from auto_vec_perm_indices to vec_perm_indices.
30033         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30034         instead of changing individual elements.
30035         (aarch64_vectorize_vec_perm_const): Use new_vector to install
30036         the vector in d.perm.
30037         * config/arm/arm.c (expand_vec_perm_d::perm): Change
30038         from auto_vec_perm_indices to vec_perm_indices.
30039         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30040         instead of changing individual elements.
30041         (arm_vectorize_vec_perm_const): Use new_vector to install
30042         the vector in d.perm.
30043         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
30044         Update vec_perm_builder constructor.
30045         (rs6000_expand_interleave): Likewise.
30046         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
30047         (rs6000_expand_interleave): Likewise.
30049 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30051         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
30052         to qimode could truncate the indices.
30053         * optabs.c (expand_vec_perm_var): Likewise.
30055 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30057         * Makefile.in (OBJS): Add vec-perm-indices.o.
30058         * vec-perm-indices.h: New file.
30059         * vec-perm-indices.c: Likewise.
30060         * target.h (vec_perm_indices): Replace with a forward class
30061         declaration.
30062         (auto_vec_perm_indices): Move to vec-perm-indices.h.
30063         * optabs.h: Include vec-perm-indices.h.
30064         (expand_vec_perm): Delete.
30065         (selector_fits_mode_p, expand_vec_perm_var): Declare.
30066         (expand_vec_perm_const): Declare.
30067         * target.def (vec_perm_const_ok): Replace with...
30068         (vec_perm_const): ...this new hook.
30069         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
30070         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
30071         * doc/tm.texi: Regenerate.
30072         * optabs.def (vec_perm_const): Delete.
30073         * doc/md.texi (vec_perm_const): Likewise.
30074         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
30075         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
30076         expand_vec_perm for constant permutation vectors.  Assert that
30077         the mode of variable permutation vectors is the integer equivalent
30078         of the mode that is being permuted.
30079         * optabs-query.h (selector_fits_mode_p): Declare.
30080         * optabs-query.c: Include vec-perm-indices.h.
30081         (selector_fits_mode_p): New function.
30082         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
30083         is defined, instead of checking whether the vec_perm_const_optab
30084         exists.  Use targetm.vectorize.vec_perm_const instead of
30085         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
30086         fit in the vector mode before using a variable permute.
30087         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
30088         vec_perm_indices instead of an rtx.
30089         (expand_vec_perm): Replace with...
30090         (expand_vec_perm_const): ...this new function.  Take the selector
30091         as a vec_perm_indices rather than an rtx.  Also take the mode of
30092         the selector.  Update call to shift_amt_for_vec_perm_mask.
30093         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
30094         Use vec_perm_indices::new_expanded_vector to expand the original
30095         selector into bytes.  Check whether the indices fit in the vector
30096         mode before using a variable permute.
30097         (expand_vec_perm_var): Make global.
30098         (expand_mult_highpart): Use expand_vec_perm_const.
30099         * fold-const.c: Includes vec-perm-indices.h.
30100         * tree-ssa-forwprop.c: Likewise.
30101         * tree-vect-data-refs.c: Likewise.
30102         * tree-vect-generic.c: Likewise.
30103         * tree-vect-loop.c: Likewise.
30104         * tree-vect-slp.c: Likewise.
30105         * tree-vect-stmts.c: Likewise.
30106         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
30107         Delete.
30108         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
30109         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
30110         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
30111         (aarch64_vectorize_vec_perm_const): ...this new function.
30112         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30113         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30114         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
30115         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
30116         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30117         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30118         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
30119         into...
30120         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
30121         check for NEON modes.
30122         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
30123         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
30124         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
30125         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
30126         into...
30127         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
30128         the old VEC_PERM_CONST conditions.
30129         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
30130         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
30131         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
30132         (ia64_vectorize_vec_perm_const_ok): Merge into...
30133         (ia64_vectorize_vec_perm_const): ...this new function.
30134         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
30135         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
30136         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
30137         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
30138         * config/mips/mips.c (mips_expand_vec_perm_const)
30139         (mips_vectorize_vec_perm_const_ok): Merge into...
30140         (mips_vectorize_vec_perm_const): ...this new function.
30141         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
30142         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
30143         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
30144         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
30145         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
30146         (rs6000_expand_vec_perm_const): Delete.
30147         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
30148         Delete.
30149         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30150         (altivec_expand_vec_perm_const_le): Take each operand individually.
30151         Operate on constant selectors rather than rtxes.
30152         (altivec_expand_vec_perm_const): Likewise.  Update call to
30153         altivec_expand_vec_perm_const_le.
30154         (rs6000_expand_vec_perm_const): Delete.
30155         (rs6000_vectorize_vec_perm_const_ok): Delete.
30156         (rs6000_vectorize_vec_perm_const): New function.
30157         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30158         an element count and rtx array.
30159         (rs6000_expand_extract_even): Update call accordingly.
30160         (rs6000_expand_interleave): Likewise.
30161         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
30162         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
30163         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
30164         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
30165         (rs6000_expand_vec_perm_const): Delete.
30166         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30167         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30168         (altivec_expand_vec_perm_const_le): Take each operand individually.
30169         Operate on constant selectors rather than rtxes.
30170         (altivec_expand_vec_perm_const): Likewise.  Update call to
30171         altivec_expand_vec_perm_const_le.
30172         (rs6000_expand_vec_perm_const): Delete.
30173         (rs6000_vectorize_vec_perm_const_ok): Delete.
30174         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
30175         reference to the SPE evmerge intructions.
30176         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30177         an element count and rtx array.
30178         (rs6000_expand_extract_even): Update call accordingly.
30179         (rs6000_expand_interleave): Likewise.
30180         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
30181         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
30182         new function.
30183         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30185 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30187         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
30188         vector mode and that that mode matches the mode of the data
30189         being permuted.
30190         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
30191         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
30192         directly using expand_vec_perm_1 when forcing selectors into
30193         registers.
30194         (expand_vec_perm_var): New function, split out from expand_vec_perm.
30196 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30198         * optabs-query.h (can_vec_perm_p): Delete.
30199         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
30200         * optabs-query.c (can_vec_perm_p): Split into...
30201         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
30202         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
30203         particular selector is valid.
30204         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30205         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30206         (vect_grouped_load_supported): Likewise.
30207         (vect_shift_permute_load_chain): Likewise.
30208         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30209         (vect_transform_slp_perm_load): Likewise.
30210         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30211         (vectorizable_bswap): Likewise.
30212         (vect_gen_perm_mask_checked): Likewise.
30213         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
30214         implementations of variable permutation vectors into account
30215         when deciding which selector to use.
30216         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
30217         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
30218         with a false third argument.
30219         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
30220         to test whether the constant selector is valid and can_vec_perm_var_p
30221         to test whether a variable selector is valid.
30223 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30225         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
30226         * optabs-query.c (can_vec_perm_p): Likewise.
30227         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
30228         instead of vec_perm_indices.
30229         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
30230         (vect_gen_perm_mask_checked): Likewise,
30231         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
30232         (vect_gen_perm_mask_checked): Likewise,
30234 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30236         * optabs-query.h (qimode_for_vec_perm): Declare.
30237         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
30238         (qimode_for_vec_perm): ...this new function.
30239         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
30241 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30243         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
30244         does not have a conditional at the top.
30246 2018-01-02  Richard Biener  <rguenther@suse.de>
30248         * ipa-inline.c (big_speedup_p): Fix expression.
30250 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30252         PR target/81616
30253         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
30254         for generic 4->6.
30256 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30258         PR target/81616
30259         Generic tuning.
30260         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
30261         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
30262         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
30263         cond_taken_branch_cost 3->4.
30265 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
30267         PR tree-optimization/83581
30268         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
30269         TODO_cleanup_cfg if any changes have been made.
30271         PR middle-end/83608
30272         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
30273         convert_modes if target mode has the right side, but different mode
30274         class.
30276         PR middle-end/83609
30277         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
30278         last argument when extracting from CONCAT.  If either from_real or
30279         from_imag is NULL, use expansion through memory.  If result is not
30280         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
30281         the parts directly to inner mode, if even that fails, use expansion
30282         through memory.
30284         PR middle-end/83623
30285         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
30286         check for bswap in mode rather than HImode and use that in expand_unop
30287         too.
30289 Copyright (C) 2018 Free Software Foundation, Inc.
30291 Copying and distribution of this file, with or without modification,
30292 are permitted in any medium without royalty provided the copyright
30293 notice and this notice are preserved.