S/390: Add loc patterns for QImode and HImode
[official-gcc.git] / gcc / ChangeLog
bloba0f35e63327b1d1d8c4d90be268e50d923d7b3a2
1 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
3         * config/s390/s390.md: QImode and HImode for load on condition.
5 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
7         * config/s390/predicates.md: Fix typo.
8         * config/s390/s390.md: Allow immediates for load on condition.
10 2018-11-05  Martin Liska  <mliska@suse.cz>
12         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
13         * bitmap.h (struct bitmap_usage): Likewise.
14         * ggc-common.c (SCALE): Remove.
15         (LABEL): Likewise.
16         (struct ggc_usage): Use SIZE_AMOUNT. And update
17         compare method.
18         * ggc-page.c (SCALE): Remove.
19         (STAT_LABEL): Remove.
20         (ggc_print_statistics): Use SIZE_AMOUNT.
21         * gimple.h (SCALE): Remove.
22         (LABEL): Likewise.
23         * input.c (ONE_K): Remove.
24         (ONE_M): Likewise.
25         (SCALE): Likewise.
26         (STAT_LABEL): Likewise.
27         (FORMAT_AMOUNT): Likewise.
28         (dump_line_table_statistics): Use SIZE_AMOUNT.
29         * mem-stats.h (struct mem_usage): Likewise.
30         * rtl.c (dump_rtx_statistics): Likewise.
31         (rtx_alloc_counts): Change type to size_t.
32         (rtx_alloc_sizes): Likewise.
33         (rtx_count_cmp): New.
34         (dump_rtx_statistics): Sort first based on counts.
35         * tree.c (tree_nodes_cmp): New.
36         (tree_codes_cmp): New.
37         (dump_tree_statistics): Sort first based on counts.
38         * system.h (ONE_K): New.
39         (ONE_M): Likewise.
40         (SIZE_SCALE): Likewise.
41         (SIZE_LABEL): Likewise.
42         (SIZE_AMOUNT): Likewise.
43         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
44         * tree-dfa.c (dump_dfa_stats): Likewise.
45         * tree-phinodes.c (phinodes_print_statistics): Likewise.
46         * tree-ssanames.c (ssanames_print_statistics): Likewise.
47         * tree.c (dump_tree_statistics): Likewise.
48         * vec.c (struct vec_usage): Likewise.
49         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
50         trigger a -Werror=format-overflow with
51         --enable-gather-detailed-stats.
53 2018-11-05  Martin Liska  <mliska@suse.cz>
55         * mem-stats.h (mem_alloc_description::release_instance_overhead):
56         Return T *.
57         * vec.c (struct vec_usage): Register m_element_size.
58         (vec_prefix::register_overhead): New arguments: elements and
59         element_size.
60         (vec_prefix::release_overhead): Subtract elements.
61         * vec.h (struct vec_prefix): Change signature.
62         (va_heap::reserve): Pass proper arguments.
63         (va_heap::release): Likewise.
65 2018-11-05  Martin Liska  <mliska@suse.cz>
67         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
68         style.
69         * vec.c: Likewise.
71 2018-11-05  Richard Biener  <rguenther@suse.de>
73         * tree-scalar-evolution.h (final_value_replacement_loop): Update
74         prototype.
75         * tree-scalar-evolution.c (final_value_replacement_loop): Return
76         whether anything was done.
77         (scev_const_prop): Remove constant propagation part, fold
78         remains into ...
79         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
80         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
81         conditionally.
83 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
85         PR tree-optimization/87859
86         * gimple-ssa-store-merging.c (struct merged_store_group): Add
87         only_constants and first_nonmergeable_order members.
88         (merged_store_group::merged_store_group): Initialize them.
89         (merged_store_group::do_merge): Clear only_constants member if
90         adding something other than INTEGER_CST store.
91         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
92         stores with order >= first_nonmergeable_order.  Use
93         merged_store->only_constants instead of always recomputing it.
94         Set merged_store->first_nonmergeable_order if we've skipped any
95         stores.  Attempt to merge overlapping INTEGER_CST stores that
96         we would otherwise skip.
98         PR sanitizer/87837
99         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
101 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
103         PR target/87853
104         * config/i386/emmintrin.h (__v16qs): New to cope with option
105         -funsigned-char.
106         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
107         (_mm_cmplt_epi8): Likewise.
108         (_mm_cmpgt_epi8): Likewise.
110 2018-11-05  Richard Biener  <rguenther@suse.de>
112         PR rtl-optimization/87852
113         * fwprop.c (use_killed_between): Only consider single-defs of the
114         use whose definition statement dominates the use.
116 2018-11-05  Martin Liska  <mliska@suse.cz>
118         PR web/87829
119         * doc/invoke.texi: Remove options that are
120         not disabled with -Os.
122 2018-11-05  Martin Liska  <mliska@suse.cz>
124         PR c/87811
125         * doc/extend.texi: Update constrain about the last argument
126         of __builtin_expect_with_probability.
128 2018-11-05  Martin Liska  <mliska@suse.cz>
130         PR c/87811
131         * predict.c (expr_expected_value_1): Verify
132         that last argument is a real constants and emit
133         error.
135 2018-11-05  Martin Liska  <mliska@suse.cz>
137         PR gcov-profile/77698
138         * ipa-profile.c (ipa_profile): Adjust hotness threshold
139         only in LTO mode.
141 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
143         PR tree-optimization/86572
144         * builtins.c (c_strlen): Handle negative offsets in a safe way.
146 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
148         PR tree-optimization/87672
149         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
150         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
152 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
154         PR middle-end/58372
155         * cfgexpand.c (pass_expand::execute): Move the call to
156         finish_eh_generation in front of the call to expand_stack_alignment.
158 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
160         * common/config/i386/i386-common.c (processor_alias_table): Add
161         znver2 entry.
162         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
163         (case ${target}): Add znver2.
164         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
165         -march=native recognize znver2 processors.
166         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
167         * config/i386/i386.c (m_znver2): New definition.
168         (m_ZNVER): New definition.
169         (m_AMD_MULTIPLE): Includes m_znver2.
170         (processor_cost_table): Add znver2 entry.
171         (processor_target_table): Add znver2 entry.
172         (get_builtin_code_for_version): Set priority for
173         PROCESSOR_ZNVER2.
174         (processor_model): Add M_AMDFAM17H_ZNVER2.
175         (arch_names_table): Ditto.
176         (ix86_reassociation_width): Include znver2.
177         * config/i386/i386.h (TARGET_znver2): New definition.
178         (struct ix86_size_cost): Add TARGET_ZNVER2.
179         (enum processor_type): Add PROCESSOR_ZNVER2.
180         * config/i386/i386.md (define_attr "cpu"): Add znver2.
181         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
182         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
183         (ix86_adjust_cost): Add znver2.
184         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
185         * gcc/doc/extend.texi: Add details about znver2.
186         * gcc/doc/invoke.texi: Add details about znver2.
188 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
190         PR target/87079
192         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
193         pattern.
195 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
197         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
198         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
200 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
202         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
203         shrink-wrapping when -mtrack-speculation.
205 2018-11-02  Richard Biener  <rguenther@suse.de>
207         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
208         (pop_cost_one_pair): Do not free pair.
209         (pop_best_coalesce): Likewise.
210         (create_coalesce_list): Initialize obstack.
211         (delete_coalesce_list): Free obstack.
212         (find_coalesce_pair): Obstack-allocate coalesce pairs.
213         (add_cost_one_coalesce): Likewise.
214         (struct live_track): Remove bitmap pointer indirections.
215         (new_live_track): Adjust.
216         (delete_live_track): Likewise.
217         (live_track_remove_partition): Likewise.
218         (live_track_add_partition): Likewise.
219         (live_track_live_p): Likewise.
220         (live_track_process_def): Likewise.
221         (live_track_clear_base_vars): Likewise.
223 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
225         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
226         * configure: Regenerate.
227         * config.in: Regenerate.
228         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
229         HAVE_AS_REGISTER_PSEUDO_OP guard.
230         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
231         !HAVE_AS_REGISTER_PSEUDO_OP support.
232         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
233         guard.
235 2018-11-02  Richard Biener  <rguenther@suse.de>
237         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
238         with is_type_die.
240 2018-11-02  Richard Biener  <rguenther@suse.de>
242         PR tree-optimization/87776
243         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
244         executable when iterating but running into rpo-vn-max-loop-depth
245         and not eliding the iteration.
247 2018-11-30  Jan Hubicka  <jh@suse.cz>
249         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
250         TYPE_DECL.
252 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
254         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
255         prototype.
256         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
257         Combine with rs6000_address_for_fpconvert.
258         (rs6000_address_for_fpconvert): Combine with
259         rs6000_force_indexed_or_indirect_mem.
260         (rs6000_expand_vector_init): Change function call from
261         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
262         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
263         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
264         (floatsi<mode>2_lfiwax_mem): Ditto.
265         (floatunssi<mode>2_lfiwzx): Ditto.
266         (floatunssi<mode>2_lfiwzx_mem): Ditto.
267         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
268         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
269         (fix_trunc<mode>si2_stfiwx): Ditto.
270         (fixuns_trunc<mode>si2_stfiwx): Ditto.
271         (float_<mode>si2_hw): Ditto.
272         (floatuns_<mode>si2_hw): Ditto.
273         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
274         (vsx_splat_<mode>): Ditto.
276 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
278         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
279         at end of assembler input text.
280         * configure: Regenerate.
282 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
284         PR tree-optimization/87826
285         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
286         negative or larger or equal to type's precision.
288 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
290         * opts.c (default_options_table): Do not enable
291         OPT_fdelayed_branch at -Og.
292         * doc/invoke.texi (-fdelayed-branch): Document it.
294 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
296         * optabs-libfuncs.c (build_libfunc_function_visibility):
297         New, split out from...
298         (build_libfunc_function): ... here.
299         (init_one_libfunc_visibility): New, split out from ...
300         (init_one_libfunc): ... here.
302         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
303         scratch register need not be early-clobber.  Document the reason
304         why we cannot use ST<OP>.
306 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
308         PR bootstrap/82856
309         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
310         line for second argument of AC_DEFINE_UNQUOTED.
311         * doc/install.texi (Tools/packages necessary for modifying GCC):
312         Update to autoconf 2.69 and automake 1.15.1.
313         * aclocal.m4, config.in, configure: Regenerate.
315 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
317         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
318         initialization.
320 2018-10-31  Martin Liska  <mliska@suse.cz>
322         PR driver/83193
323         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
324         Add new argument invalid_extension.
325         (aarch64_get_all_extension_candidates): New function.
326         (aarch64_rewrite_selected_cpu): Add NULL to function call.
327         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
328         new argument.
329         (aarch64_get_all_extension_candidates): New function.
330         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
331         argument invalid_extension.
332         (aarch64_parse_cpu): Likewise.
333         (aarch64_print_hint_for_extensions): New function.
334         (aarch64_validate_mcpu): Provide hint about invalid extension.
335         (aarch64_validate_march): Likewise.
336         (aarch64_handle_attr_arch): Pass new argument.
337         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
338         (aarch64_handle_attr_isa_flags): Likewise.
340 2018-10-31  Richard Biener  <rguenther@suse.de>
342         PR middle-end/70359
343         PR middle-end/86270
344         * tree-outof-ssa.c (insert_backedge_copies): Restrict
345         copy generation to useful cases.  Place the copy before
346         the definition of the backedge value when possible.
348 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
350         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
351         * config/arc/arc.c (arc_active_insn): New function.
352         (check_store_cacheline_hazard): Likewise.
353         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
354         (arc_override_options): Disable delay slot scheduler for older
355         A7.
356         (arc_store_addr_hazard_p): New implementation, old one renamed to
357         ...
358         (arc_store_addr_hazard_internal_p): Renamed.
359         (arc_reorg): Don't combine into brcc instructions which are part
360         of hardware hazard solution.
361         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
362         (tune_arc700): Likewise.
363         * config/arc/arc.opt (arc7xx): New tune value.
364         * config/arc/arc700.md: Improve A7 scheduler.
366 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
368         * config/arc/arc.c (arc_override_options): Remove
369         TARGET_COMPACT_CASESI.
370         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
371         (CASE_VECTOR_MODE): Likewise.
372         (CASE_VECTOR_PC_RELATIVE): Likewise.
373         (CASE_VECTOR_SHORTEN_MODE): Likewise.
374         (CASE_VECTOR_SHORTEN_MODE1): Delete.
375         (ADDR_VEC_ALIGN): Update.
376         (ASM_OUTPUT_CASE_LABEL): Undefine.
377         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
378         (TARGET_BI_BIH): Define.
379         (DEFAULT_BRANCH_INDEX): Likewise.
380         * config/arc/arc.md (casesi): Rework to accept BI/BIH
381         instructions, remove compact_casesi use case.
382         (casesi_compact_jump): Remove.
383         (casesi_dispatch): New pattern.
384         * config/arc/arc.opt: Add mbranch-index option. Deprecate
385         compact_casesi option.
386         * doc/invoke.texi: Document mbranch-index option.
388 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
390         * config/arc/arc.c (arc_get_tp): Remove function.
391         (arc_emit_call_tls_get_addr): Likewise.
392         (arc_call_tls_get_addr): New function.
393         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
394         * config/arc/arc.md (tls_load_tp_soft): Remove.
395         (tls_gd_get_addr): Likewise.
397 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
399         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
400         (mulsi3_600_lib): Remove pattern.
401         (umulsi3_highpart_600_lib_le): Likewise.
402         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
403         (umulsidi3): Remove call to umulsidi3_600_lib.
404         (umulsidi3_600_lib): Remove pattern.
405         (peephole2): Remove peephole using the above deprecated patterns.
407 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
409         PR target/87374
410         * config/arm/arm.c (arm_option_check_internal): Disable the combined
411         use of -mslow-flash-data and -mword-relocations.
412         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
413         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
414         flag_pic.
415         * doc/invoke.texi (-mword-relocations): Mention conflict with
416         -mslow-flash-data.
417         (-mslow-flash-data): Reciprocally.
419 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
421         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
422         16-byte modes held in GP registers to use an even regno.
424         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
425         (aarch64_atomic_ldop_supported_p): Remove.
426         (aarch64_gen_atomic_ldop): Remove.
427         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
428         Fully expand LSE operations here.
429         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
430         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
431         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
432         and use ATOMIC_LDOP instead; use register_operand for the input;
433         drop the split and emit insns directly.
434         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
435         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
436         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
438         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
439         (aarch64_gen_atomic_ldop): Don't call it.
440         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
441         Use aarch64_reg_or_zero.
442         (aarch64_atomic_exchange<ALLI>): Likewise.
443         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
444         operand 0; use aarch64_reg_or_zero for input; merge ...
445         (@aarch64_atomic_swp<ALLI>): ... this and remove.
447         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
448         (aarch64_split_compare_and_swap): Use it.
449         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
450         test oldval against the proper predicate.
451         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
452         Use nonmemory_operand for expected.
453         (cas_short_expected_pred): New.
454         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
455         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
456         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
457         (aarch64_plushi_operand): New.
459         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
460         Force oldval into the rval register for TARGET_LSE; emit the compare
461         during initial expansion so that it may be deleted if unused.
462         (aarch64_gen_atomic_cas): Remove.
463         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
464         Change =&r to +r for operand 0; use match_dup for operand 2;
465         remove is_weak and mod_f operands as unused.  Drop the split
466         and merge with...
467         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
468         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
469         (@aarch64_atomic_cas<GPI>): Similarly.
471 2018-10-31  Richard Biener  <rguenther@suse.de>
473         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
474         using ABSU_EXPR.
476 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
478         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
479         * config/aarch64/aarch64.md: Include saphira.md
480         * config/aarch64/saphira.md: New file for pipeline description.
482 2018-10-30  Martin Sebor  <msebor@redhat.com>
484         PR middle-end/87041
485         * gimple-ssa-sprintf.c (format_directive): Use %G to include
486         inlining context.
487         (sprintf_dom_walker::compute_format_length):
488         Avoid setting POSUNDER4K here.
489         (get_destination_size): Handle null argument values.
490         (get_user_idx_format): New function.
491         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
492         functions, including user-defined with attribute format printf.
493         Use %G to include inlining context.
494         Set POSUNDER4K here.
496 2018-10-30  Jan Hubicka  <jh@suse.cz>
498         * params.def (lto-partitions): Bump from 32 to 128.
500 2018-10-30  Jan Hubicka  <jh@suse.cz>
502         * tree.c
503         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
504         head in file.
505         (free_lang_data_in_type): Forward declare.
506         (fld_type_variant_equal_p): New function.
507         (fld_type_variant): New function
508         (fld_incomplete_types): New hash.
509         (fld_incomplete_type_of): New function
510         (fld_simplfied-type): New function.
511         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
512         (free_lang_data): Allocate and free fld_incomplete_type; update call
513         of free_lang_data_in_decl.
515 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
517         * gcov.c (output_lines): Remove duplicate line.
519 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
521         * config/rs6000/rs6000.md (bswapdi2): Force address into register
522         if not in indexed or indirect form.
523         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
524         (bswapdi2_store): Ditto.
525         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
526         helper function.
527         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
528         Prototype for helper function.
530 2018-10-30  Martin Sebor  <msebor@redhat.com>
532         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
533         (target, pragma GCC optimize, pragma GCC target): Ditto.
535 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
537         * doc/extend.texi: Fix prototype and description of
538         __builtin_expect_with_probability.
540 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
542         * cgraph.h (clone_function_name_1): Replaced by new
543           clone_function_name_numbered that takes name as string; for
544           privatize_symbol_name_1 use only.
545           (clone_function_name): Renamed to
546           clone_function_name_numbered to be explicit about numbering.
547           (clone_function_name): New two-argument function that does
548           not number its output.
549           (clone_function_name): New three-argument function that
550           takes a number to append to its output.
551         * cgraphclones.c (duplicate_thunk_for_node):
552           (clone_function_name_1): Renamed.
553           (clone_function_name_numbered): Two new functions.
554           (clone_function_name): Improved documentation.
555           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
556         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
557         * final.c (final_scan_insn_1): Use the new clone_function_name
558           without numbering.
559         * multiple_target.c (create_dispatcher_calls): Ditto.
560           (create_target_clone): Ditto.
561         * omp-expand.c (grid_expand_target_grid_body): Ditto.
562         * omp-low.c (create_omp_child_function_name): Ditto.
563         * omp-simd-clone.c (simd_clone_create): Ditto.
564         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
565           new clone_function_name without numbering.
567 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
569         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
570         Assert that the allocation size is not zero.
572 2018-10-30  Richard Biener  <rguenther@suse.de>
574         PR tree-optimization/87800
575         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
576         non-induction or reduction PHIs.
578 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
580         * config/aarch64/falkor-tag-collision-avoidance.c
581         (execute_tag_collision_avoidance): Call df_note_add_problem.
583 2018-10-30  Martin Liska  <mliska@suse.cz>
585         * doc/extend.texi: Fix typo in documentation
586         of __builtin_expect_with_probability.
588 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
590         PR c++/87721
591         * input.c (get_substring_ranges_for_loc): Detect if
592         linemap_resolve_location gives us a NULL map, and reject
593         this case.
595 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
597         * config.gcc (xstormy16-*-elf): Set tm_d_file.
599 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
600             Martin Sebor  <msebor@redhat.com>
601             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
603         * doc/ux.texi (Quoting): New subsection, adapted from material at
604         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
605         MartinSebor and ManuelLopezIbanez.
606         (Fix-it hints): Note that fix-it hints shouldn't be marked for
607         translation.
609 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
611         PR middle-end/87469
612         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
613         max value.
615 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
617         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
619 2018-10-29  Paul Koning  <ni1d@arrl.net>
621         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
623 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
625         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
626         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
627         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
628         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
629         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
630         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
631         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
632         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
633         Change 'vector' to '__vector'.
634         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
635         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
636         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
637         _mm_avg_pu16): Likewise.  And, whitespace corrections.
639 2018-10-29  Richard Biener  <rguenther@suse.de>
641         PR tree-optimization/87785
642         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
643         internal defs.
645 2018-10-29  Olivier Hainque  <hainque@adacore.com>
647         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
649 2018-10-29  Olivier Hainque  <hainque@adacore.com>
651         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
652         ports configuration macro, defaults to "gnu".
653         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
654         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
656 2018-10-29  Olivier Hainque  <hainque@adacore.com>
658         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
660 2018-10-29  Richard Biener  <rguenther@suse.de>
662         PR tree-optimization/87790
663         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
664         (vect_make_slp_decision): Adjust.
665         (vect_slp_analyze_bb_1): Likewise.
666         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
667         edges.
669 2018-10-29  Richard Biener  <rguenther@suse.de>
671         PR tree-optimization/87785
672         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
673         and processing.
674         (vect_build_slp_tree): Likewise.
675         (vect_gather_slp_loads): New function.
676         (vect_analyze_slp_instance): Gather loads separately from the
677         SLP tree build.
679 2018-10-29  Martin Liska  <mliska@suse.cz>
681         * Makefile.in: Make dependency to json.o.
682         * doc/gcov.texi: Document new JSON format, remove
683         old intermediate format documentation.
684         * gcov.c (struct function_info): Come up with m_name and
685         m_demangled_name.
686         (function_info::function_info): Initialize it.
687         (function_info::~function_info): Release it.
688         (main): Rename flag_intermediate_format to flag_json_format.
689         (print_usage): Describe --json-format.
690         (process_args): Set flag_json_format.
691         (output_intermediate_line): Remove.
692         (output_intermediate_json_line): Likewise.
693         (get_gcov_intermediate_filename): Return new extension
694         ".gcov.json.gz".
695         (output_intermediate_file): Implement JSON emission.
696         (output_json_intermediate_file): Implement JSON emission.
697         (generate_results): Use ::get_name for function name.
698         Handle JSON output file.
699         (read_graph_file): Use ::get_name instead of cplus_demangle.
700         (read_count_file): Likewise.
701         (solve_flow_graph): Likewise.
702         (add_line_counts): Likewise.
703         (accumulate_line_counts): Use new flag_json_format.
704         (output_function_details): Use ::get_name instead of cplus_demangle.
705         (output_lines): Likewise.
706         * json.cc (test_writing_literals): Add new tests.
707         * json.h (class literal): Add new boolean constructor.
709 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
711         PR rtl-optimization/87701
712         PR rtl-optimization/87780
713         * combine.c (make_more_copies): Rewrite.
715 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
717         * doc/generic.texi (ABSU_EXPR): Document.
718         * match.pd (absu(x)*absu(x) -> x*x): Handle.
719         (absu(absu(X)) -> absu(X)): Likewise.
720         (absu(-X) -> absu(X)): Likewise.
721         (absu(X)  where X is nonnegative -> X): Likewise.
723 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
725         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
726         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
727         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
728         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
729         (s-tm-texi): Also check timestamp on d-target.def.
730         (generated_files): Add TM_D_H and d-target-hooks-def.h.
731         (build/genhooks.o): Also depend on D_TARGET_DEF.
732         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
733         variables.
734         * config/aarch64/aarch64-d.c: New file.
735         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
736         Define.
737         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
738         prototype.
739         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
740         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
741         * config/arm/arm-d.c: New file.
742         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
743         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
744         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
745         * config/arm/t-arm (arm-d.o): New rule.
746         * config/default-d.c: New file.
747         * config/glibc-d.c: New file.
748         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
749         * config/i386/i386-d.c: New file.
750         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
751         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
752         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
753         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
754         * config/i386/t-i386 (i386-d.o): New rule.
755         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
756         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
757         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
758         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
759         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
760         * config/mips/mips-d.c: New file.
761         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
762         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
763         * config/mips/t-mips (mips-d.o): New rule.
764         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
765         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
766         * config/powerpcspe/powerpcspe-d.c: New file.
767         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
768         New prototype.
769         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
770         Support GNU D by using 0 as the language type.
771         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
772         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
773         * config/riscv/riscv-d.c: New file.
774         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
775         prototype.
776         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
777         * config/riscv/t-riscv (riscv-d.o): New rule.
778         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
779         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
780         * config/rs6000/rs6000-d.c: New file.
781         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
782         prototype.
783         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
784         Support GNU D by using 0 as the language type.
785         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
786         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
787         * config/s390/s390-d.c: New file.
788         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
789         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
790         * config/s390/t-s390 (s390-d.o): New rule.
791         * config/sparc/sparc-d.c: New file.
792         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
793         prototype.
794         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
795         * config/sparc/t-sparc (sparc-d.o): New rule.
796         * config/t-glibc (glibc-d.o): New rule.
797         * configure: Regenerated.
798         * configure.ac (tm_d_file): New variable.
799         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
800         * doc/contrib.texi (Contributors): Add self for the D frontend.
801         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
802         * doc/install.texi (Configuration): Mention libphobos as an option for
803         --enable-shared.  Mention d as an option for --enable-languages.
804         (Testing): Mention check-d as a target.
805         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
806         name suffixes.  Mention d as a -x option.
807         * doc/sourcebuild.texi (Top Level): Mention libphobos.
808         * doc/standards.texi (Standards): Add section on D language.
809         * doc/tm.texi: Regenerated.
810         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
811         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
812         * dwarf2out.c (is_dlang): New function.
813         (gen_compile_unit_die): Use DW_LANG_D for D.
814         (declare_in_namespace): Return module die for D, instead of adding
815         extra declarations into the namespace.
816         (gen_namespace_die): Generate DW_TAG_module for D.
817         (gen_decl_die): Handle CONST_DECLSs for D.
818         (dwarf2out_decl): Likewise.
819         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
820         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
821         kinds of aggregates.
822         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
823         * genhooks.c: Include d/d-target.def.
825 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
827         PR target/85669
828         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
829         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
830         (STACK_DYNAMIC_OFFSET): Likewise.
832 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
834         PR target/80024
835         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
836         error message.
838 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
840         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
841         return type and other typos.
843 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
845         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
846         a shorter sequence with fewer branches.
847         (emit_final_str_compare_gpr): Ditto.
849 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
851         * config/rs6000/tmmintrin.h: New file.
852         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
854 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
856         * config/rs6000/mmintrin.h: Enable 32bit compilation.
857         * config/rs6000/xmmintrin.h: Likewise.
859 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
861         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
863 2018-10-26  Richard Biener  <rguenther@suse.de>
865         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
866         and wrapper.
867         (vect_mark_slp_stmts_relevant): Likewise.
868         (vect_detect_hybrid_slp_stmts): Likewise.
869         (vect_bb_slp_scalar_cost): Likewise.
870         (vect_remove_slp_scalar_calls): Likewise.
872 2018-10-26  Jan Hubicka  <jh@suse.cz>
874         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
875         (warn_types_mismatch): Fix walk of DECL_NAME.
876         (odr_types_equivalent_p): Fix overactive assert.
878 2018-10-26  Richard Biener  <rguenther@suse.de>
880         PR tree-optimization/87105
881         * tree-vectorizer.h (_slp_tree::refcnt): New member.
882         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
883         refcnt.
884         (vect_create_new_slp_node): Initialize refcnt to one.
885         (bst_traits): Move.
886         (scalar_stmts_set_t, bst_fail): Remove.
887         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
888         (vect_build_slp_tree): Add bst_map argument and lookup
889         already created SLP nodes.
890         (vect_print_slp_tree): Handle a SLP graph, print SLP node
891         addresses.
892         (vect_slp_rearrange_stmts): Handle a SLP graph.
893         (vect_analyze_slp_instance): Adjust and free SLP nodes from
894         the CSE map.  Fix indenting.
895         (vect_schedule_slp_instance): Add short-cut.
897 2018-10-26  Martin Liska  <mliska@suse.cz>
899         PR testsuite/86158
900         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
901         addr_expr and not with pointers.
903 2018-10-26  Jan Hubicka  <jh@suse.cz>
905         * tree.c (free_lang_data_in_type): Only check main variants.
906         * ipa-devirt.c (warn_odr): Make static.
907         (types_same_for_odr): Drop strict variant.
908         (types_odr_comparable): Likewise.
909         (odr_or_derived_type_p): Look for main variants.
910         (odr_name_hasher::equal): Cleanup comment.
911         (odr_subtypes_equivalent): Add warn and warned arguments; check main
912         variants.
913         (type_variants_equivalent_p): break out from ...
914         (odr_types_equivalent): ... here; go for main variants where needed.
915         (warn_odr): ... here; turn static.
916         (warn_types_mismatch): Compare mangled names of main variants.
917         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
918         (type_with_linkage_p): Sanity check that we look at main variant.
919         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
920         * tree.h (types_same_for_odr): Drop strict argument.
922 2018-10-26  Richard Biener  <rguenther@suse.de>
924         PR tree-optimization/87746
925         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
926         Simplify and fix WRT strided store groups with size not
927         equal to step in element count.
928         (vect_analyze_group_access_1): Dump the whole group.
930 2018-10-25  Carl Love  <cel@us.ibm.com>
932         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
933         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
934         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
935         precicion entry for each overloaded builtin.
936         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
937         VSCEDPUO): Rename overloaded name.
938         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
939         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
940         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
941         define_expand for xscmpexqp instruction.
942         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
944 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
945             Jinsong Ji <jji@us.ibm.com>
947         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
948         function with vec_sl.
949         (_mm_slli_epi32): Likewise.
950         (_mm_slli_epi64): Likewise.
951         (_mm_srai_epi16): Replace deprecated function with vec_sra.
952         (_mm_srai_epi32): Likewise.
953         (_mm_srli_epi16): Replace deprecated function with vec_sr.
954         (_mm_srli_epi32): Likewise.
955         (_mm_srli_epi64): Likewise.
956         (_mm_sll_epi16): Replace deprecated function with vec_sl.
957         (_mm_sll_epi32): Likewise.
958         (_mm_sll_epi64): Likewise.
959         (_mm_sra_epi16): Replace deprecated function with vec_sra.
960         (_mm_sra_epi32): Likewise.
961         (_mm_srl_epi16): Replace deprecated function with vec_sr.
962         (_mm_srl_epi32): Likewise.
963         (_mm_srl_epi64): Likewise.
965 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
966             Jinsong Ji <jji@us.ibm.com>
968         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
969         comparison operators with vec_cmp* for compatibility due to
970         unfortunate history; clean up formatting and use types more
971         appropriately.
972         (_mm_sll_epi32): Likewise.
973         (_mm_sll_epi64): Likewise.
974         (_mm_srl_epi16): Likewise.
975         (_mm_srl_epi32): Likewise.
976         (_mm_srl_epi64): Likewise.
978 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
979             Jinsong Ji <jji@us.ibm.com>
981         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
982         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
983         __vector __bool int.  Use vec_cmpgt in preference to deprecated
984         function vec_vcmpgtfp.
985         (_mm_max_ps): Likewise.
987 2018-10-25  Jeff Law  <law@redhat.com>
989         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
990         if returning false.
992 2018-10-25  Martin Sebor  <msebor@redhat.com>
994         * doc/extend.texi (aligned): Expand attribute description.
995         (Alignment): Rename section.  Discuss function arguments.
997 2018-10-25  Jan Hubicka  <jh@suse.cz>
999         * ipa-devirt.c (main_odr_variant): Remove.
1000         (hash_odr_name, types_same_for_odr, types_odr_comparable,
1001         odr_name_hasher::equal, odr_subtypes_equivalent_p):
1002         Drop use of main_odr_variant.
1003         (add_type_duplicate): Silence confused warnings on integer types.
1004         (get_odr_type): Always look for main variant.
1005         (register_odr_type): Simplify.
1007 2018-10-25  Richard Biener  <rguenther@suse.de>
1009         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
1010         Initialize ng to silence error with release checking bootstrap.
1012 2018-10-25  Richard Biener  <rguenther@suse.de>
1014         * tree-if-conv.c: Include tree-ssa-sccvn.h.
1015         (tree_if_conversion): Run CSE on the if-converted loop body.
1017 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1019         * config/s390/constraints.md (ZL): New constraint.
1020         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
1021         operands.
1022         * config/s390/s390.md (movdi_larl): Remove.
1023         (movdi_64): Add the LARL alternative.
1025 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1027         PR bootstrap/87747
1028         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
1029         (RTX_CODE_HWINT_P): New macro.
1030         (rtx_code_size): Use RTX_CODE_HWINT_P ().
1032 2018-10-25  Jan Hubicka  <jh@suse.cz>
1034         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
1035         is anonymous.
1037 2018-10-25  Richard Biener  <rguenther@suse.de>
1039         PR tree-optimization/87665
1040         PR tree-optimization/87745
1041         * tree-vectorizer.h (get_earlier_stmt): Remove.
1042         (get_later_stmt): Pick up UID from the original non-pattern stmt.
1044 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
1046         * options.texi (Deprecated): Move list to Var section.
1048 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
1049             Jinsong Ji <jji@us.ibm.com>
1051         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
1052         __vector long to __vector long long.
1053         (_mm_cvtpd_ps): Likewise.
1054         (_mm_cvttpd_epi32): Likewise.
1055         (_mm_cvtpi32_pd): Likewise.
1056         (_mm_unpackhi_epi64): Likewise.
1057         (_mm_unpacklo_epi64): Likewise.
1059 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
1061         PR rtl-optimization/87720
1062         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
1064 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
1066         * gimple-ssa-isolate-paths.c
1067         (find_implicit_erroneous_behavior): Do not change code if the
1068         pass is running for warnings only.
1069         (find_explicit_erroneous_behavior): Likewise.
1071 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
1073         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
1074         Define as rs6000_mangle_decl_assembler_name.
1075         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
1076         long double to IEEE long double, switch the names of the long
1077         double built-in functions to be <func>f128 instead of <func>l.
1079 2018-10-24  Martin Sebor  <msebor@redhat.com>
1081         * doc/extend.texi (nonnull): List no-argument form.  Reference
1082         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
1084 2018-10-24  Richard Biener  <rguenther@suse.de>
1086         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
1088 2018-10-24  Martin Liska  <mliska@suse.cz>
1090         PR tree-optimization/84436
1091         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
1092         Remove.
1093         (switch_conversion::contains_linear_function_p): New.
1094         (switch_conversion::build_one_array): Support linear
1095         transformation on input.
1096         * tree-switch-conversion.h (struct switch_conversion): Add
1097         contains_linear_function_p declaration.
1099 2018-10-24  Richard Biener  <rguenther@suse.de>
1101         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
1102         if its argument is CONSTANT_CLASS_P.
1104 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1106         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
1107         it is wrong for forward declarations.
1109 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1111         * config/s390/s390.c (s390_check_qrst_address): Add the missing
1112         SYMBOL_REF_P () check.
1114 2018-10-24  Richard Biener  <rguenther@suse.de>
1116         PR tree-optimization/87105
1117         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
1118         dump classification.
1119         (vect_analyze_data_ref_accesses): Handle duplicate loads and
1120         stores by splitting the affected group after the fact.
1121         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
1122         fail the SLP build because of size constraints.
1124 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1126         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
1127         * configure: Regenerate.
1128         * config.in: Regenerate.
1129         * varasm.c (mergeable_string_section): Use readonly_data_section
1130         if linker doesn't support SHF_MERGE with alignment > 8.
1131         (mergeable_constant_section): Likewise.
1133 2018-10-24  Richard Biener  <rguenther@suse.de>
1135         PR tree-optimization/84013
1136         * tree-ssa-structalias.c (struct msdi_data): New struct for
1137         marshalling data to walk_stmt_load_store_ops.
1138         (maybe_set_dependence_info): Refactor as callback for
1139         walk_stmt_load_store_ops.
1140         (compute_dependence_clique): Set restrict info on all stmt kinds.
1142 2018-10-24  Martin Liska  <mliska@suse.cz>
1144         * cgraph.c (cgraph_node::dump):
1145         Remove reduntant dumps and make tp_first_run dump more compact.
1147 2018-10-24  Richard Biener  <rguenther@suse.de>
1149         PR tree-optimization/87665
1150         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
1151         to reflect reality.
1153 2018-10-12  Jeff Law  <law@redhat.com>
1155         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
1156         for H8/S.
1158 2018-10-23  Richard Biener  <rguenther@suse.de>
1160         * tree-vrp.c (add_assert_info): Guard dump_printf with
1161         dump_enabled_p.
1162         * gimple-ssa-evrp-analyze.c
1163         (evrp_range_analyzer::record_ranges_from_incoming_edge):
1164         Use value_range::ignore_equivs_equal_p.
1166 2018-10-23  Richard Biener  <rguenther@suse.de>
1168         PR tree-optimization/87105
1169         PR tree-optimization/87608
1170         * passes.def (pass_all_early_optimizations): Add early phi-opt
1171         after dce.
1172         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
1173         addition to debug stmts.
1174         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
1175         and abs replacement early.
1176         * tree-cfg.c (gimple_empty_block_p): Likewise.
1178 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
1180         PR target/86383
1181         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
1182         specified to configure.
1183         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
1185 2018-10-23  Richard Biener  <rguenther@suse.de>
1187         PR tree-optimization/87700
1188         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
1190 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
1192         PR target/87674
1193         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
1194         second argument from __mmask16 to __mmask8.
1195         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
1196         _mm_mask_packs_epi32): Likewise.
1197         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
1198         Likewise.
1199         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
1201 2018-10-23  Richard Biener  <rguenther@suse.de>
1203         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
1205 2018-10-23  Richard Biener  <rguenther@suse.de>
1207         PR tree-optimization/86144
1208         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
1209         over simd attribute.
1211 2018-10-23  Richard Biener  <rguenther@suse.de>
1213         PR tree-optimization/87693
1214         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
1215         the case we do not find the taken edge.
1217 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
1218             Jinsong Ji  <jji@us.ibm.com>
1220         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
1221         (_mm_store_pd): Use unaligned vector type for pointer cast.
1222         (_mm_maskmoveu_si128): Likewise.
1223         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
1224         (_mm_store_ps): Use unaligned vector type for pointer cast.
1226 2018-10-22  Paul Koning  <ni1d@arrl.net>
1228         * symtab.c (symtab_node::increase_alignment): Correct max
1229         alignment check.
1231 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
1233         PR tree-optimization/87633
1234         * match.pd: Do not generate unordered integer comparisons.
1236 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1238         PR rtl-optimization/87600
1239         * combine.c: Add include of expr.h.
1240         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
1241         register to a pseudo.
1242         (make_more_copies): New function, add a copy to a new pseudo after
1243         the moves from hard registers into pseudos.
1244         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
1245         later.  Call make_more_copies.
1247 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
1249         * lra-constraints.c (process_alt_operands): New local array,
1250         matching_early_clobber.  Check matching_early_clobber before
1251         decrementing reject, and set matching_early_clobber after.
1253 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1255         PR target/87598
1256         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
1257         call output_addr_const and hope for the best.
1259 2018-10-22  Richard Biener  <rguenther@suse.de>
1261         * gimple-ssa-evrp-analyze.c
1262         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
1263         smarter about what ranges to use.
1264         * tree-vrp.c (add_assert_info): Dump here.
1265         (register_edge_assert_for_2): Instead of here at multiple but
1266         not all places.
1268         * gcc.dg/tree-ssa/evrp12.c: New testcase.
1269         * gcc.dg/predict-6.c: Adjust.
1270         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
1271         * gcc.dg/tree-ssa/vrp02.c: Likewise.
1272         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
1274 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
1275         Richard Biener  <rguenther@suse.de>
1277         * bitmap.h: Update data structure documentation, including a
1278         description of bitmap views as either linked-lists or splay trees.
1279         (struct bitmap_element_def): Update comments for splay tree bitmaps.
1280         (struct bitmap_head_def): Likewise.
1281         (bitmap_list_view, bitmap_tree_view): New prototypes.
1282         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
1283         tree_form fields.
1284         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
1285         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
1286         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
1287         released bitmap element here.
1288         (bitmap_element_free): Remove.
1289         (bitmap_elt_clear_from): Work on splay tree bitmaps.
1290         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
1291         this function similar ones such that linked-list bitmap implementation
1292         functions are grouped.
1293         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
1294         and moved for grouping.
1295         (bitmap_list_insert_element_after): Renamed from
1296         bitmap_elt_insert_after, and moved for grouping.
1297         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
1298         (bitmap_tree_link_left, bitmap_tree_link_right,
1299         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
1300         bitmap_tree_link_element, bitmap_tree_unlink_element,
1301         bitmap_tree_find_element): New functions for splay-tree bitmap
1302         implementation.
1303         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
1304         Renamed and moved, see above entries.
1305         (bitmap_tree_listify_from): New function to convert part of a splay
1306         tree bitmap to a linked-list bitmap.
1307         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
1308         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
1309         (bitmap_find_bit): Remove.
1310         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
1311         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
1312         Handle splay tree bitmaps.
1313         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
1314         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
1315         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
1316         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
1317         bitmap_intersect_compl_p, bitmap_ior_and_compl,
1318         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
1319         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
1320         corresponding changes to use linked-list specific bitmap_element
1321         manipulation functions as applicable for efficiency.
1322         (bitmap_tree_to_vec): New function.
1323         (debug_bitmap_elt_file): New function split out from ...
1324         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
1325         (bitmap_print): Likewise.
1327         PR tree-optimization/63155
1328         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
1329         SSA edge worklists.
1330         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
1331         in tree-view.
1333 2018-10-22  Martin Liska  <mliska@suse.cz>
1335         PR tree-optimization/87686
1336         Revert
1337         2018-08-29  Martin Liska  <mliska@suse.cz>
1339         * tree-switch-conversion.c (switch_conversion::expand):
1340         Strenghten assumption about gswitch statements.
1342 2018-10-22  Martin Liska  <mliska@suse.cz>
1344         * ipa-icf.c (sem_item::compare_attributes): Remove.
1345         (sem_item::compare_referenced_symbol_properties): Use
1346         attribute_list_equal instead.
1347         (sem_function::equals_wpa): Likewise.
1348         * ipa-icf.h: Remove compare_attributes.
1350 2018-10-22  Richard Biener  <rguenther@suse.de>
1352         PR middle-end/87682
1353         * mem-stats.h (mem_usage::operator==): Fix pasto.
1355 2018-10-22  Richard Biener  <rguenther@suse.de>
1357         PR tree-optimization/87640
1358         * tree-vrp.c (set_value_range_with_overflow): Decompose
1359         incomplete result.
1360         (extract_range_from_binary_expr_1): Adjust.
1362 2018-10-22  Martin Jambor  <mjambor@suse.cz>
1364         * tree-eh.h (stmt_could_throw_p): Add function parameter.
1365         (stmt_can_throw_external): Likewise.
1366         (stmt_can_throw_internal): Likewise.
1367         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
1368         (lower_eh_constructs_2): Likewise.
1369         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
1370         (stmt_can_throw_external): Likewise.
1371         (stmt_can_throw_internal): Likewise.
1372         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
1373         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
1374         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
1375         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
1376         (pass_lower_eh_dispatch::execute): Pass cfun to
1377         stmt_can_throw_external.
1378         (cleanup_empty_eh): Likewise.
1379         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
1380         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
1381         stmt_can_throw_external instead of pushing it to cfun.
1382         (symbol_table::create_edge): Likewise.
1383         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
1384         stmt_can_throw_internal.
1385         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
1386         to stmt_could_throw_p.
1387         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
1388         stmt_can_throw_internal.
1389         (pass_store_merging::execute): Likewise.
1390         * gimple-ssa-strength-reduction.c
1391         (find_candidates_dom_walker::before_dom_children): Pass cfun to
1392         stmt_could_throw_p.
1393         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
1394         stmt_can_throw_internal.
1395         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1396         to stmt_can_throw_external.
1397         (check_stmt): Pass cfun to stmt_could_throw_p.
1398         (check_stmt): Pass cfun to stmt_can_throw_external.
1399         (pass_nothrow::execute): Likewise.
1400         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1401         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1402         stmt_can_throw_internal.
1403         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1404         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1405         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1406         * tree-complex.c (expand_complex_libcall): Pass cfun to
1407         stmt_could_throw_p and to stmt_can_throw_internal.
1408         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1409         * tree-inline.c (copy_edges_for_bb): Likewise.
1410         (maybe_move_debug_stmts_to_successors): Likewise.
1411         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1412         stmt_could_throw_p.
1413         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1414         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1415         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1416         stmt_can_throw_internal.
1417         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1418         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1419         stmt_could_throw_p.
1420         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1421         stmt_can_throw_internal.
1422         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1423         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1424         stmt_could_throw_p.
1425         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1426         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1427         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1428         (convert_mult_to_fma_1): Likewise.
1429         (convert_to_divmod): Likewise.
1430         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1431         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1432         * tree-ssa-propagate.c
1433         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1434         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1435         (maybe_optimize_range_tests): Likewise.
1436         (linearize_expr_tree): Likewise.
1437         (reassociate_bb): Likewise.
1438         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1439         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1440         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1441         (handle_char_store): Likewise.
1442         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1443         stmt_can_throw_internal.
1444         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1445         stmt_could_throw_p.
1446         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1447         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1448         (vectorizable_simd_clone_call): Likewise.
1449         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1450         (gimple_stringop_fixed_value): Likewise.
1452 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
1454         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1455         literal pool references.
1456         (s390_check_qrst_address): Adapt to the new behavior of
1457         s390_loadrelative_operand_p ().
1459 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1461         PR target/72782
1462         * config/i386/sse.md (*andnot<mode>3_bcst): New.
1464 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1466         PR target/72782
1467         * config/i386/sse.md (*<code><mode>3_bcst): New.
1469 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1471         PR target/72782
1472         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1473         V4DI, V16SI and V8DI.
1474         (*sub<mode>3<mask_name>_bcst): New.
1475         (*add<mode>3<mask_name>_bcst): Likewise.
1477 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
1478             Jinsong Ji  <jji@us.ibm.com>
1480         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1481         __m64 with __vector unsigned long long for compatibility.
1482         (_mm_movemask_epi8): Likewise.
1483         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1484         (_mm_cvttps_pi32): Likewise.
1485         (_mm_cvtpi32_ps): Likewise.
1486         (_mm_cvtps_pi16): Likewise.
1487         (_mm_loadh_pi): Likewise.
1488         (_mm_storeh_pi): Likewise.
1489         (_mm_movehl_ps): Likewise.
1490         (_mm_movelh_ps): Likewise.
1491         (_mm_loadl_pi): Likewise.
1492         (_mm_storel_pi): Likewise.
1493         (_mm_movemask_ps): Likewise.
1494         (_mm_shuffle_pi16): Likewise.
1496 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1498         PR target/72782
1499         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1500         __builtin_ia32_vfnmsubpd512_mask.
1501         (_mm512_mask_fnmsub_round_pd): Likewise.
1502         (_mm512_fnmsub_pd): Likewise.
1503         (_mm512_mask_fnmsub_pd): Likewise.
1504         (_mm512_maskz_fnmsub_round_pd): Use
1505         __builtin_ia32_vfnmsubpd512_maskz.
1506         (_mm512_maskz_fnmsub_pd): Likewise.
1507         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1508         (_mm512_mask_fnmsub_round_ps): Likewise.
1509         (_mm512_fnmsub_ps): Likewise.
1510         (_mm512_mask_fnmsub_ps): Likewise.
1511         (_mm512_maskz_fnmsub_round_ps): Use
1512         __builtin_ia32_vfnmsubps512_maskz.
1513         (_mm512_maskz_fnmsub_ps): Likewise.
1514         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1515         __builtin_ia32_vfnmsubpd256_mask.
1516         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1517         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1518         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1519         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1520         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1521         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1522         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1523         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1524         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1525         __builtin_ia32_vfnmsubpd.
1526         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1527         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1528         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1529         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1530         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1531         * config/i386/i386-builtin.def: Add
1532         __builtin_ia32_vfnmsubpd256_mask,
1533         __builtin_ia32_vfnmsubpd256_maskz,
1534         __builtin_ia32_vfnmsubpd128_mask,
1535         __builtin_ia32_vfnmsubpd128_maskz,
1536         __builtin_ia32_vfnmsubps256_mask,
1537         __builtin_ia32_vfnmsubps256_maskz,
1538         __builtin_ia32_vfnmsubps128_mask,
1539         __builtin_ia32_vfnmsubps128_maskz,
1540         __builtin_ia32_vfnmsubpd512_mask,
1541         __builtin_ia32_vfnmsubpd512_maskz,
1542         __builtin_ia32_vfnmsubps512_mask,
1543         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1544         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1545         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1546         __builtin_ia32_vfnmsubpd256.
1547         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1548         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1549         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1550         Likewise.
1551         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1552         Likewise.
1553         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1554         Likewise.
1555         (fmai_vmfnmsub_<mode><round_name>): Likewise.
1557 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1559         PR target/72782
1560         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1561         __builtin_ia32_vfnmaddpd512_mask.
1562         (_mm512_mask_fnmadd_round_pd): Likewise.
1563         (_mm512_fnmadd_pd): Likewise.
1564         (_mm512_mask_fnmadd_pd): Likewise.
1565         (_mm512_maskz_fnmadd_round_pd): Use
1566         __builtin_ia32_vfnmaddpd512_maskz.
1567         (_mm512_maskz_fnmadd_pd): Likewise.
1568         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1569         (_mm512_mask_fnmadd_round_ps): Likewise.
1570         (_mm512_fnmadd_ps): Likewise.
1571         (_mm512_mask_fnmadd_ps): Likewise.
1572         (_mm512_maskz_fnmadd_round_ps): Use
1573         __builtin_ia32_vfnmaddps512_maskz.
1574         (_mm512_maskz_fnmadd_ps): Likewise.
1575         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1576         __builtin_ia32_vfnmaddpd256_mask.
1577         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1578         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1579         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1580         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1581         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1582         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1583         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1584         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1585         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1586         __builtin_ia32_vfnmaddpd.
1587         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1588         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1589         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1590         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1591         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1592         * config/i386/i386-builtin.def: Add
1593         __builtin_ia32_vfnmaddpd256_mask,
1594         __builtin_ia32_vfnmaddpd256_maskz,
1595         __builtin_ia32_vfnmaddpd128_mask,
1596         __builtin_ia32_vfnmaddpd128_maskz,
1597         __builtin_ia32_vfnmaddps256_mask,
1598         __builtin_ia32_vfnmaddps256_maskz,
1599         __builtin_ia32_vfnmaddps128_mask,
1600         __builtin_ia32_vfnmaddps128_maskz,
1601         __builtin_ia32_vfnmaddpd512_mask,
1602         __builtin_ia32_vfnmaddpd512_maskz,
1603         __builtin_ia32_vfnmaddps512_mask,
1604         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1605         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1606         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1607         __builtin_ia32_vfnmaddpd256.
1608         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1609         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1610         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1611         Likewise.
1612         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1613         Likewise.
1614         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1615         Likewise.
1616         (fmai_vmfnmadd_<mode><round_name>): Likewise.
1618 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1620         PR target/72782
1621         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1622         __builtin_ia32_vfmsubpd512_mask.
1623         (_mm512_mask_fmsub_round_pd): Likewise.
1624         (_mm512_fmsub_pd): Likewise.
1625         (_mm512_mask_fmsub_pd): Likewise.
1626         (_mm512_maskz_fmsub_round_pd): Use
1627         __builtin_ia32_vfmsubpd512_maskz.
1628         (_mm512_maskz_fmsub_pd): Likewise.
1629         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1630         (_mm512_mask_fmsub_round_ps): Likewise.
1631         (_mm512_fmsub_ps): Likewise.
1632         (_mm512_mask_fmsub_ps): Likewise.
1633         (_mm512_maskz_fmsub_round_ps): Use
1634         __builtin_ia32_vfmsubps512_maskz.
1635         (_mm512_maskz_fmsub_ps): Likewise.
1636         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1637         __builtin_ia32_vfmsubpd256_mask.
1638         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1639         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1640         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1641         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1642         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1643         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1644         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1645         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1646         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1647         __builtin_ia32_vfmsubpd.
1648         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1649         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1650         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1651         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1652         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1653         * config/i386/i386-builtin.def: Add
1654         __builtin_ia32_vfmsubpd256_mask,
1655         __builtin_ia32_vfmsubpd256_maskz,
1656         __builtin_ia32_vfmsubpd128_mask,
1657         __builtin_ia32_vfmsubpd128_maskz,
1658         __builtin_ia32_vfmsubps256_mask,
1659         __builtin_ia32_vfmsubps256_maskz,
1660         __builtin_ia32_vfmsubps128_mask,
1661         __builtin_ia32_vfmsubps128_maskz,
1662         __builtin_ia32_vfmsubpd512_mask,
1663         __builtin_ia32_vfmsubpd512_maskz,
1664         __builtin_ia32_vfmsubps512_mask,
1665         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1666         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1667         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1668         __builtin_ia32_vfmsubpd256.
1669         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1670         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1671         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1672         Likewise.
1673         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1674         Likewise.
1675         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1676         Likewise.
1677         (fmai_vmfmsub_<mode><round_name>): Likewise.
1679 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1681         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1682         Remove plus.  Renamed to ...
1683         (*sub<mode>3<mask_name>_bcst): This.
1684         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1685         (*add<mode>3<mask_name>_bcst): This.
1687 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1689         PR target/72782
1690         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1692 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1694         PR target/87662
1695         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1696         (_mm_or_epi32): Likewise.
1697         (_mm256_xor_epi32): Likewise.
1698         (_mm_xor_epi32): Likewise.
1699         (_mm256_or_epi64): Likewise.
1700         (_mm_or_epi64): Likewise.
1701         (_mm256_xor_epi64): Likewise.
1702         (_mm_xor_epi64): Likewise.
1704 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
1706         PR target/72782
1707         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1709 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
1711         PR middle-end/87647
1712         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1714 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
1716         * doc/ux.texi: Move @section directly after @node.
1718 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
1720         PR middle-end/85488
1721         PR middle-end/87649
1722         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1723         depend closely nested inside of loop with ordered clause with
1724         a parameter.
1726 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
1728         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1729         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1730         * doc/ux.texi: New file.
1732 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
1734         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1735         be the first CR field allocated.
1737 2018-10-19  Richard Biener  <rguenther@suse.de>
1739         PR target/87657
1740         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1741         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1743 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1745         PR target/72782
1746         * config/i386/sse.md
1747         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1748         (*add<mode>3<mask_name>_bcst_2): Likewise.
1750 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1752         * config/i386/sse.md
1753         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1754         Replace nonimmediate_operand with register_operand.
1755         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1756         Likewise.
1757         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1758         Likewise.
1760 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
1762         PR rtl-optimization/87596
1763         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1764         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1765         for instructions in FROM..TO range.
1767 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
1769         * cfgexpand.c (expand_one_var): Use specific wording in error message
1770         for non-local frame variables.
1771         * stor-layout.c (layout_decl): Do not issue a warning for them.
1773 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
1775         * haifa-sched.c (priority): Add force_recompute parameter.
1776         (apply_replacement): Call priority () with force_recompute = true.
1777         (restore_pattern): Likewise.
1779 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1781         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
1782         HOST_BITS_PER_WIDE_INT.
1783         (test_vector_ops_duplicate): Likewise.
1785 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1787         PR target/72782
1788         * config/i386/sse.md (VF_AVX512): New.
1789         (avx512bcst): Likewise.
1790         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1791         Likewise.
1792         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1793         Likewise.
1794         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1795         Likewise.
1797 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
1799         * doc/invoke.texi (-dumpversion): Improve grammar.
1800         (-dumpfullversion): Make more consistent with -dumpversion.
1802 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
1804         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
1805         Set JUMP_LABEL to the jump insn.
1806         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
1807         Predict emitted jump and add label to jump insn.
1809 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
1811         PR tree-optimization/87562
1812         * input.c (get_substring_ranges_for_loc): Use
1813         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
1814         getting the linemap for the endpoint.  Verify that it's either
1815         in the same linemap as the start point's spelling location, or
1816         at least in the same file.
1818 2018-10-18  Richard Biener  <rguenther@suse.de>
1820         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
1821         feed width-specific load/store costs through ix86_vec_cost.
1822         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
1823         (k8_cost): Likewise.
1824         (bdver_cost): Likewise.
1825         (znver1_cost): Likewise.
1826         (btver1_cost): Likewise.
1827         (btver2_cost): Likewise.
1829 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1831         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
1832         to simplify subreg of vec_merge.
1834 2018-10-18  Richard Biener  <rguenther@suse.de>
1836         * config/i386/i386.c: Fix costing of vector FMA.
1838 2018-10-18  Richard Biener  <rguenther@suse.de>
1840         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
1841         and argument.
1842         (ix86_builtin_vectorization_cost): For vec_construct properly
1843         cost insertion into SSE regs.
1844         (...): Adjust calls to ix86_vec_cost.
1846 2018-10-18  Richard Biener  <rguenther@suse.de>
1848         PR middle-end/87087
1849         Revert
1850         2018-02-07  Richard Biener  <rguenther@suse.de>
1852         PR tree-optimization/84204
1853         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1854         this place.
1856 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1858         PR target/87537
1859         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
1860         of vec_duplicate.
1861         (test_vector_ops_duplicate): Add test for a scalar subreg of a
1862         VEC_MERGE of a VEC_DUPLICATE.
1864 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1866         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
1867         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
1868         * doc/standards.texi (C Language): Document C2X.
1869         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
1870         (rl78_option_override): Handle "GNU C2X" language name.
1872 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1874         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
1875         Document C17 as published in 2018.
1877 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
1879         PR middle-end/87623
1880         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
1881         bail out if both sides do not have the same storage order.
1883 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1885         * bitmap.c (bitmap_head::dump): New.
1886         * bitmap.h (bitmap_head): Add dump().
1887         * gimple-ssa-evrp-analyze.c
1888         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
1889         (evrp_range_analyzer::set_ssa_range_info): Same.
1890         (evrp_range_analyzer::record_ranges_from_phis): Same.
1891         (evrp_range_analyzer::record_ranges_from_stmt): Same.
1892         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
1893         * gimple-ssa-sprintf.c (get_int_range): Same.
1894         (format_integer): Same.
1895         (sprintf_dom_walker::handle_gimple_call): Same.
1896         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
1897         (ipcp_vr_lattice::top_p): Same.
1898         (ipcp_vr_lattice::bottom_p): Same.
1899         (ipcp_vr_lattice::set_to_bottom): Same.
1900         (ipa_vr_operation_and_type_effects): Same.
1901         (propagate_vr_across_jump_function): Same.
1902         (ipcp_store_vr_results): Same.
1903         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
1904         (ipa_print_node_jump_functions_for_edge): Same.
1905         (ipa_get_value_range): Same.
1906         (ipa_compute_jump_functions_for_edge): Same.
1907         (ipa_write_jump_function): Same.
1908         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
1909         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1910         Same.
1911         * vr-values.c (set_value_range_to_nonnegative): Same.
1912         (set_value_range_to_truthvalue): Same.
1913         (vr_values::get_value_range): Same.
1914         (vr_values::set_defs_to_varying): Same.
1915         (vr_values::update_value_range): Same.
1916         (symbolic_range_based_on_p): Same.
1917         (vr_values::op_with_boolean_value_range_p): Same.
1918         (vr_values::extract_range_for_var_from_comparison_expr): Same.
1919         (vr_values::extract_range_from_ssa_name): Same.
1920         (vr_values::extract_range_from_binary_expr): Same.
1921         (vr_values::extract_range_from_unary_expr): Same.
1922         (vr_values::extract_range_from_cond_expr): Same.
1923         (vr_values::extract_range_from_comparison): Same.
1924         (vr_values::check_for_binary_op_overflow): Same.
1925         (vr_values::extract_range_basic): Same.
1926         (vr_values::extract_range_from_assignment): Same.
1927         (compare_ranges): Same.
1928         (compare_range_with_value): Same.
1929         (vr_values::adjust_range_with_scev): Same.
1930         (vrp_valueize): Same.
1931         (vrp_valueize_1): Same.
1932         (vr_values::get_vr_for_comparison): Same.
1933         (vr_values::compare_name_with_value): Same.
1934         (vr_values::compare_names): Same.
1935         (vr_values::vrp_evaluate_conditional): Same.
1936         (find_case_label_ranges): Same.
1937         (vr_values::vrp_visit_switch_stmt): Same.
1938         (vr_values::extract_range_from_phi_node): Same.
1939         (vr_values::simplify_div_or_mod_using_ranges): Same.
1940         (vr_values::simplify_bit_ops_using_ranges): Same.
1941         (test_for_singularity): Same.
1942         (range_fits_type_p): Same.
1943         (vr_values::simplify_cond_using_ranges_1): Same.
1944         (vr_values::simplify_switch_using_ranges): Same.
1945         (vr_values::simplify_float_conversion_using_ranges): Same.
1946         (vr_values::two_valued_val_range_p): Same.
1947         (vr_values::add_equivalence): Move to value_range::equiv_add.
1948         * vr-values.h (vr_values::add_equivalence): Remove.
1949         (VR_INITIALIZER): Remove.
1950         * tree-vrp.c (value_range::set): New.
1951         (value_range::equiv_add): New.
1952         (value_range::value_range): New.
1953         (value_range::deep_copy): New.
1954         (value_range::check): New.
1955         (value_range::equal_p): New.
1956         (value_range::ignore_equivs_equal_p): New.
1957         (value_range::operator==): New.
1958         (value_range::operator!=): New.
1959         (value_range::symbolic_p): New.
1960         (value_range::numeric_p): New.
1961         (value_range::set_undefined): New.
1962         (value_range::set_varying): New.
1963         (value_range::may_contain_p): New.
1964         (value_range::equiv_clear): New.
1965         (value_range::singleton_p): New.
1966         (value_range::intersect): New.
1967         (value_range::dump): New.
1968         (value_range::set_and_canonicalize): New.
1969         (set_value_range): Adjust for value_range API.
1970         (set_value_range_to_undefined): Same.
1971         (set_value_range_to_varying): Same.
1972         (set_and_canonicalize_value_range): Same.
1973         (set_value_range_to_nonnull): Same.
1974         (set_value_range_to_null): Same.
1975         (range_is_null): Same.
1976         (range_is_nonnull): Same.
1977         (range_int_cst_p): Same.
1978         (range_int_cst_singleton_p): Same.
1979         (symbolic_range_p): Same.
1980         (range_includes_zero_p): Same.
1981         (value_range_constant_singleton): Same.
1982         (vrp_set_zero_nonzero_bits): Same.
1983         (ranges_from_anti_range): Same.
1984         (extract_range_into_wide_ints): Same.
1985         (extract_range_from_multiplicative_op): Same.
1986         (set_value_range_with_overflow): Same.
1987         (extract_range_from_binary_expr_1): Same.
1988         (extract_range_from_unary_expr): Same.
1989         (dump_value_range): Same.
1990         (debug_value_range): Same.
1991         (vrp_prop::check_array_ref): Same.
1992         (vrp_prop::check_mem_ref): Same.
1993         (vrp_prop::vrp_initialize): Same.
1994         (vrp_prop::visit_stmt): Same.
1995         (intersect_ranges): Same.
1996         (vrp_prop::visit_phi): Same.
1997         (vrp_prop::vrp_finalize): Same.
1998         (determine_value_range_1): Same.
1999         (determine_value_range): Same.
2000         (vrp_intersect_ranges_1): Rename to...
2001         (vrp_intersect_1): this.
2002         (vrp_intersect_ranges): Rename to...
2003         (value_range::intersect_helper): ...this.
2004         (vrp_meet_1): Rename to...
2005         (value_range::union_helper): ...this.
2006         (vrp_meet): Rename to...
2007         (value_range::union_): ...this.
2008         (copy_value_range): Remove.
2009         * tree-vrp.h (struct value_range): Rewrite into a proper class.
2010         (value_range::vrtype): New.
2011         (value_range::type): New.
2012         (value_range::equiv): New.
2013         (value_range::min): New.
2014         (value_range::max): New.
2015         (value_range::varying_p): New.
2016         (value_range::undefined_p): New.
2017         (value_range::null_p): New.
2018         (value_range::equiv_add): New.
2019         (copy_value_range): Remove.
2021 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
2023         * Makefile.in (SELFTEST_TARGETS): New.
2024         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
2025         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
2026         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
2027         c/Make-lang.in.
2028         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
2029         (selftest-c++-gdb, selftest-c++-valgrind): Move to
2030         cp/Make-lang.in.
2031         * configure: Regenerate.
2032         * configure.ac (selftest_languages): New.
2034 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2036         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
2037         overflow wraps argument.
2038         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
2039         wide_int_range_multiplicative_op.
2040         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
2041         overflow wraps argument.
2042         (wide_int_range_multiplicative_op): Same.
2043         (wide_int_range_lshift): Same.
2044         (wide_int_range_div): Same.
2045         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
2046         (wide_int_range_lshift): Same.
2047         (wide_int_range_div): Same.
2049 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2051         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
2052         use sign as argument.
2053         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
2054         wide_int_range_shift_undefined_p.
2056 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2058         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
2059         Rename to...
2060         (@despeculate_copy<ALLI_TI:mode>): ... This.
2061         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
2062         switch statement.
2064 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2066         * config.gcc: Obsolete *-*-solaris2.10*.
2067         * doc/install.texi (Specific, *-*-solaris2*): Document it.
2069 2018-10-12  Jeff Law  <law@redhat.com>
2071         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
2072         reg + sym +- const_int addressing modes.
2074 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2076         * common.opt (fdiagnostics-minimum-margin-width=): New option.
2077         * diagnostic-show-locus.c (layout::layout): Apply the minimum
2078         margin width.
2079         (layout::start_annotation_line): Only print up to 3 of the
2080         margin character, to avoid touching the left-hand side.
2081         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
2082         minimum margin width, as set by test_diagnostic_context's ctor.
2083         (selftest::test_fixit_insert_containing_newline): Likewise.
2084         (selftest::test_fixit_insert_containing_newline_2): Likewise.
2085         (selftest::test_line_numbers_multiline_range): Clear
2086         dc.min_margin_width.
2087         * diagnostic.c (diagnostic_initialize): Initialize
2088         min_margin_width.
2089         * diagnostic.h (struct diagnostic_context): Add field
2090         "min_margin_width".
2091         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
2092         * opts.c (common_handle_option): Handle
2093         OPT_fdiagnostics_minimum_margin_width_.
2094         * selftest-diagnostic.c
2095         (selftest::test_diagnostic_context::test_diagnostic_context):
2096         Initialize min_margin_width to 6.
2097         * toplev.c (general_init): Initialize global_dc->min_margin_width.
2099 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2101         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2102         Fix usage of "error_at_rich_loc" in the comment.
2104 2018-10-15  Renlin Li  <renlin.li@arm.com>
2106         PR target/87563
2107         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
2108         if-conversioned loop when it contains ifn with types not
2109         supported by backend.
2110         * internal-fn.c (expand_direct_optab_fn): Add an assert.
2111         (direct_internal_fn_supported_p): New helper function.
2112         * internal-fn.h (direct_internal_fn_supported_p): Declare.
2114 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
2116         PR target/87572
2117         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2118         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2119         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2121 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
2123         PR tree-optimization/87022
2124         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
2125         bits in dist vector rather than the first one.
2127 2018-10-15  Richard Biener  <rguenther@suse.de>
2129         PR middle-end/87610
2130         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2131         (visit_loadstore): When a used restrict tag escaped verify that
2132         the points-to solution of "other" pointers do not include
2133         escaped.
2134         (compute_dependence_clique): If a used restrict tag escaped
2135         communicated that down to visit_loadstore.
2137 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
2139         * config/s390/s390.c (s390_expand_vec_init): Force vector element
2140         into reg if it isn't a general operand.
2142 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2144         PR target/87599
2145         * config/i386/sse.md (*vec_dupv2di): Add register source to
2146         movddup.
2148 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2150         PR target/87572
2151         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2152         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
2153         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2154         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2156 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
2158         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
2159         (notice_args_size): Set it in the current trace if no insn that can
2160         throw internally has been seen yet.
2161         (connect_traces): When connecting args_size between traces, allow the
2162         incoming values not to match if there is an insn setting it before the
2163         first insn that can throw internally; in that case, force the creation
2164         of a CFI note on this latter insn.
2166 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
2168         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
2169         the base class, because using the injected-class-name was not clearly
2170         specified until DR 176.
2172 2018-10-12  Paul Koning  <ni1d@arrl.net>
2174         * config/pdp11/pdp11.md (doloop_end): New expander.
2175         (doloop_end_insn): renamed from "doloop_end".
2176         (addqi3): New pattern.
2177         (subqi3): New pattern.
2178         * config/pdp11/predicates.md (incdec_operand): New predicate.
2180 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
2182         PR middle-end/81376
2183         * real.c (format_helper::can_represent_integral_type_p): New function
2184         * real.h (format_helper::can_represent_integral_type_p): Ditto.
2185         * match.pd: New pattern.
2187 2018-10-12  Alexandre Oliva <oliva@adacore.com>
2189         * configure.ac: Introduce --enable-large-address-aware
2190         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
2191         * doc/install.texi: Document it.
2192         * configure, config.in: Rebuilt.
2193         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
2194         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
2195         (LINK_SPEC): Insert it.
2196         * config/i386/mingw-w64.h: Likewise.
2198         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
2200 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
2202         PR rtl-optimization/87600
2203         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
2205 2018-10-12  Paul Koning  <ni1d@arrl.net>
2207         * doc/md.texi (doloop_end): Document that the pattern code may
2208         need to check operand mode.
2210 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
2212         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
2213         to zero-extend between int and floating-point registers.
2214         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
2215         ldp into floating-point registers.  Add type and arch attributes.
2216         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
2217         Use f_loads for type attribute.
2219 2018-10-11  Martin Sebor  <msebor@redhat.com>
2221         * doc/extend.texi (attribute packed): Correct typos.
2223 2018-10-11  Martin Sebor  <msebor@redhat.com>
2225         * doc/extend.texi (attribute flatten): Mention interaction with
2226         noinline.
2228 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
2230         PR target/87156
2231         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2232         Set new_decl virtual flag to zero.
2234 2018-10-11  Martin Sebor  <msebor@redhat.com>
2236         PR middle-end/87593
2237         * doc/extend.texi (attribute format_arg): Discuss using multiple
2238         attributes on a single function.
2240 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
2242         PR tree-optimization/86829
2243         * match.pd (sin (atan (x))): New simplification rules.
2244         (cos (atan (x))): Likewise.
2245         * real.c (build_sinatan_real): New function.
2246         * real.h (build_sinatan_real): Prototype.
2248 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
2250         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
2251         function.
2252         (fold_mergeeo_helper): New helper function.
2253         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
2254         intrinsics.  Correct some whitespace indentation issues.
2256 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
2258         PR target/87511
2259         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2260         Use HOST_WIDE_INT_1U for shift.
2262 2018-10-11  Doug Rupp  <rupp@adacore.com>
2263             Olivier Hainque  <hainque@adacore.com>
2265         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
2266         Pass --relax to the linker for RTPs.
2267         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
2269 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
2270             Jan Hubicka  <jh@suse.cz>
2271             Martin Jambor  <mjambor@suse.cz>
2273         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
2274         the same elements are repeated rather than printing all of them.
2275         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
2276         "repeated" elements.
2277         * read-rtl-function.c (test_loading_repeat): New function.
2278         (read_rtl_function_c_tests): Call test_loading_repeat.
2279         * rtl-tests.c (test_dumping_repeat): New function.
2280         (rtl_tests_c_tests): Call test_dumping_repeat.
2282 2018-10-11  Richard Biener  <rguenther@suse.de>
2284         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
2285         bdver?_cost): Unify to ...
2286         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
2287         * config/i386/i386.c (processor_cost_table): Adjust.
2289 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
2291         PR middle-end/87574
2292         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
2293         the thunk when expanding to GIMPLE.
2295 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2297         * varasm.c (mergeable_string_section): Don't try to move zero-length
2298         strings to the merge section.
2300 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
2302         PR target/87573
2303         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
2305 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
2307         PR target/87550
2308         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2309         to special_args set.
2311 2018-10-10  Richard Biener  <rguenther@suse.de>
2313         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
2314         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
2315         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
2316         and recursing and pattern terminating the recursion on SSE
2317         vector width using ix86_expand_reduc.
2318         (reduc_sminmax_scal_<mode>): Split into part reducing to half
2319         width and recursing and SSE2 vector variant doing the final
2320         reduction with ix86_expand_reduc.
2321         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
2322         with terminating the recursion at AVX level, splitting that
2323         to SSE there.
2325 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
2327         * genmatch.c (error_cb): Rename to...
2328         (diagnostic_cb): ...this, converting int params to enums.
2329         (fatal_at): Update for renaming.
2330         (warning_at): Likewise.
2331         (main): Likewise.
2332         * input.c (selftest::ebcdic_execution_charset::apply):
2333         Update for renaming of...
2334         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
2335         to...
2336         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
2337         converting level and reason to enums.
2338         (class selftest::lexer_error_sink): Rename to...
2339         (class selftest::lexer_test_options): ...this, renaming field
2340         "m_errors" to "m_diagnostics".
2341         (selftest::lexer_test_options::apply): Update for renaming of...
2342         (selftest::lexer_test_options::on_error): ...this, renaming to...
2343         (selftest::lexer_test_options::on_diagnostic): ...this
2344         converting level and reason to enums.
2345         (selftest::test_lexer_string_locations_raw_string_unterminated):
2346         Update for renamings.
2347         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
2348         "reason".
2350 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
2352         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
2353         * config/rs6000/pmmintrin.h: New file.
2355 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
2357         PR tree-optimization/86659
2358         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
2360 2018-10-09  Richard Biener  <rguenther@suse.de>
2362         PR tree-optimization/63155
2363         * tree-ssa-structalias.c: Include tree-ssa.h.
2364         (get_constraint_for_ssa_var): For undefs return nothing_id.
2365         (find_func_aliases): Cleanup PHI handling.
2367 2018-10-09  Richard Biener  <rguenther@suse.de>
2369         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
2370         replacements.
2372 2018-10-09  Martin Liska  <mliska@suse.cz>
2374         * asan.c (asan_emit_stack_protection): If a stack variable
2375         is located in a same file as current function, then emit
2376         line info into variable definition string.
2378 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2380         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
2381         information.
2383 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2385         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
2386         on the thunk.
2388 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2390         PR tree-optimization/86659
2391         * gimple-match.h (struct gimple_match_op): Add reverse field.
2392         (gimple_match_op::set_op): New overloaded method.
2393         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
2394         the REF_REVERSE_STORAGE_ORDER flag on the value.
2395         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2396         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2397         
2398 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
2400         PR middle-end/63155
2401         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2402         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2404 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
2406         PR target/87517
2407         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2408         Defined with __builtin_ia32_vfmaddsubpd512_mask.
2410 2018-10-08  Richard Biener  <rguenther@suse.de>
2412         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2413         cost the same as AVX128 ones.
2415 2018-10-08  Paul Koning  <ni1d@arrl.net>
2417         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2418         (expand_block_move): New function.
2419         * config/pdp11/pdp11.c (output_block_move): Remove.
2420         (expand_block_move): New function.
2421         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2422         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2423         (*movmemhi1): Remove.
2425 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
2427         * config/s390/2827.md: Increase latencies for some FP instructions.
2429 2018-10-08  Richard Biener  <rguenther@suse.de>
2431         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2432         Open a dump scope.
2433         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2434         * tree-vectorizer.h (dump_stmt_cost): Adjust.
2435         (add_stmt_cost): Dump return value of the hook.
2437 2018-10-08  Richard Biener  <rguenther@suse.de>
2439         PR tree-optimization/63155
2440         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2441         (ssa_propagation_engine::ssa_propagate): Remove redundant
2442         bitmap bit clearing.
2444 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
2446         PR rtl-optimization/86939
2447         PR rtl-optimization/87479
2448         * ira.h (non_conflicting_reg_copy_p): New prototype.
2449         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2450         (make_hard_regno_dead): Don't add conflicts for register
2451         ignore_reg_for_conflicts.
2452         (make_object_dead): Likewise.
2453         (non_conflicting_reg_copy_p): New function.
2454         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2455         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2456         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2457         (make_hard_regno_dead): Don't add conflicts for register
2458         ignore_reg_for_conflicts.  Remove special conflict handling of
2459         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
2460         check_pic_pseudo_p and update callers.
2461         (mark_pseudo_dead): Don't add conflicts for register
2462         ignore_reg_for_conflicts.
2463         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2465 2018-10-05  Andrew Waterman  <andrew@sifive.com>
2466             Jim Wilson  <jimw@sifive.com>
2468         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2469         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
2470         new pattern using HONOR_SNANS that emits one extra instruction.
2472 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2474         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2475         patterns): Merge SI and DI patterns to a GPR pattern.
2476         (unnamed define_insn and define_split for record form of that): Merge
2477         to a single define_insn_and_split pattern.
2479 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
2481         PR c++/56856
2482         * input.c (expand_location_to_spelling_point): Add param "aspect"
2483         and use rather than hardcoding LOCATION_ASPECT_CARET.
2484         (get_substring_ranges_for_loc): Handle the case of a single token
2485         within a macro expansion.
2486         * input.h (expand_location_to_spelling_point): Add "aspect" param,
2487         defaulting to LOCATION_ASPECT_CARET.
2489 2018-10-05  Paul Koning  <ni1d@arrl.net>
2491         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2492         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2493         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2494         (pdp11_guard_type): New function.
2496 2018-10-05  Paul Koning  <ni1d@arrl.net>
2498         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2499         * config/pdp11/pdp11.opt (mfloat32): Remove.
2500         (mfloat64): Remove.
2501         * doc/invoke.texi (pdp11 -mfloat32): Remove:
2502         (pdp11 -mfloat64): Remove.
2504 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2506         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2507         (*cmp<mode>_cc_i387): Ditto.
2508         (*cmpu<mode>_cc_i387): Ditto.
2509         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2510         * config/i386/i386.c (ix86_expand_fp_compare): Remove
2511         "scratch" argument.
2512         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2513         Emit x86_sahf_1 pattern.
2514         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2515         (ix86_expand_carry_flag_compare): Ditto.
2517 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2519         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2520         to reg_or_0_operand.  Add "C" constraint.
2521         (*cmpxf_cc_i387): Ditto.
2522         (*cmp<mode>_i387): Change operand 2 predicate
2523         to nonimm_or_0_operand.  Add "C" constraint.
2524         (*cmp<mode>_cc_i387): Ditto.
2525         (*cmp<mode>_0_i387): Remove insn pattern.
2526         (*cmp<mode>_0_cc_i387): Ditto.
2528 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2530         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2531         * config/i386/predicates.md (nonimm_or_0_operand): Rename
2532         from vector_move_operand.  Update all uses.
2534 2018-10-05  Martin Sebor  <msebor@redhat.com>
2536         PR tree-optimization/87490
2537         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2538         consistently.
2540 2018-10-05  Richard Biener  <rguenther@suse.de>
2542         PR tree-optimization/63155
2543         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2544         vertical space in dumpfiles.
2545         * tree-ssa-propagate.h
2546         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2547         * tree-ssa-propagate.c (cfg_blocks_back): New global.
2548         (ssa_edge_worklist_back): Likewise.
2549         (curr_order): Likewise.
2550         (cfg_blocks_get): Remove abstraction.
2551         (cfg_blocks_add): Likewise.
2552         (cfg_blocks_empty_p): Likewise.
2553         (add_ssa_edge): Add to current or next worklist based on
2554         RPO index.
2555         (add_control_edge): Likewise.
2556         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2557         into ...
2558         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
2559         iteration from CFG and SSA edge worklist so we process
2560         everything in RPO order, prioritizing forward progress
2561         over iteration.
2562         (ssa_prop_init): Allocate new worklists, do not dump
2563         immediate uses.
2564         (ssa_prop_fini): Free new worklists.
2566 2018-10-05  Richard Biener  <rguenther@suse.de>
2568         * tree-core.h (tree_block::abstract_flag): Remove.
2569         (tree_block::block_num): Make full 32bits.
2570         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2571         * tree.h (BLOCK_ABSTRACT): Remove.
2572         * dwarf2out.c (gen_lexical_block_die): Remove dead code
2573         resulting from BLOCK_ABSTRACT being always false.
2574         (gen_inlined_subroutine_die): Likewise.
2575         (gen_block_die): Likewise.
2576         * tree.c (block_ultimate_origin): Likewise.
2577         * tree-pretty-print.c (dump_block_node): Remove code dealing
2578         with BLOCK_ABSTRACT.
2579         * tree-ssa-live.c (dump_scope_block): Likewise.
2580         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2581         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2583 2018-10-05   Richard Biener  <rguenther@suse.de>
2585         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2586         is asked for initialize mode to the component mode of the
2587         vector type.
2589 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
2591         PR target/87522
2592         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2593         assembler for -mavx.
2594         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2596 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2598         PR target/87509
2599         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2600         RS6000_BTM_DFP.
2601         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2602         to be DImode.  When using mffscrn, force the operand to a register.
2604 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
2606         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2607         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2608         X87MODEF mode iterator.
2609         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2610         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2611         X87MODEF mode iterator.
2613 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
2615         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2616         -Wno-prio-ctor-dtor.
2618 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2620         * Makefile.in (OBJS): Add opt-problem.o.
2621         * dump-context.h: Include "selftest.h.
2622         (selftest::temp_dump_context): New forward decl.
2623         (class dump_context): Make friend of class
2624         selftest::temp_dump_context.
2625         (dump_context::dump_loc_immediate): New decl.
2626         (class dump_pretty_printer): Move here from dumpfile.c.
2627         (class temp_dump_context): Move to namespace selftest.
2628         (temp_dump_context::temp_dump_context): Add param
2629         "forcibly_enable_dumping".
2630         (selftest::verify_dumped_text):
2631         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2632         (selftest::verify_item):
2633         (ASSERT_IS_TEXT): Move here from dumpfile.c.
2634         (ASSERT_IS_TREE): Likewise.
2635         (ASSERT_IS_GIMPLE): Likewise.
2636         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2637         to...
2638         (dump_context::dump_loc_immediate): ...this new function.
2639         (class dump_pretty_printer): Move to dump-context.h.
2640         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2641         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2642         (temp_dump_context::temp_dump_context): Move to "selftest"
2643         namespace.  Add param "forcibly_enable_dumping", and use it to
2644         conditionalize the use of m_pp;
2645         (selftest::verify_dumped_text): Make non-static.
2646         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2647         (selftest::verify_item): Make non-static.
2648         (ASSERT_IS_TEXT): Move to dump-context.h.
2649         (ASSERT_IS_TREE): Likewise.
2650         (ASSERT_IS_GIMPLE): Likewise.
2651         (selftest::test_capture_of_dump_calls): Pass "true" for new
2652         param of temp_dump_context.
2653         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2654         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
2655         TDF_COMPARE_DEBUG.
2656         * opt-problem.cc: New file.
2657         * opt-problem.h: New file.
2658         * optinfo-emit-json.cc
2659         (selftest::test_building_json_from_dump_calls): Pass "true" for
2660         new param of temp_dump_context.
2661         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2662         (optinfo::emit_for_opt_problem): New function.
2663         (optinfo::emit): Clarity which emit_item is used.
2664         * optinfo.h (optinfo::get_dump_location): New accessor.
2665         (optinfo::emit_for_opt_problem): New decl.
2666         (optinfo::emit): Make const.
2667         * selftest-run-tests.c (selftest::run_tests): Call
2668         selftest::opt_problem_cc_tests.
2669         * selftest.h (selftest::opt_problem_cc_tests): New decl.
2670         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2671         bool to opt_result, converting fprintf messages to
2672         opt_result::failure_at calls.  Add "stmt" param for use by the
2673         failure_at calls.
2674         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2675         (runtime_alias_check_p): Convert return type from bool to
2676         opt_result, converting dump_printf calls to
2677         opt_result::failure_at, using the statement DDR_A for their
2678         location.
2679         (find_data_references_in_stmt): Convert return type from bool to
2680         opt_result, converting "return false" to opt_result::failure_at
2681         with a new message.
2682         * tree-data-ref.h: Include "opt-problem.h".
2683         (dr_analyze_innermost): Convert return type from bool to opt_result,
2684         and add a const gimple * param.
2685         (find_data_references_in_stmt): Convert return type from bool to
2686         opt_result.
2687         (runtime_alias_check_p): Likewise.
2688         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2689         dr_analyze_innermost.
2690         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2691         Convert return type from bool to opt_result, adding a message for
2692         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2693         (vect_analyze_data_ref_dependence): Convert return type from bool
2694         to opt_result.  Change sense of return type from "false"
2695         effectively meaning "no problems" to "false" meaning a problem,
2696         so that "return false" becomes "return opt_result::success".
2697         Convert "return true" calls to opt_result::failure_at, using
2698         the location of statement A rather than vect_location.
2699         (vect_analyze_data_ref_dependences): Convert return type from bool
2700         to opt_result.
2701         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2702         calls to opt_result::failure_at, using the stmt location rather
2703         than vect_location.
2704         (vect_verify_datarefs_alignment): Convert return type from bool
2705         to opt_result.
2706         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
2707         into multiple more-tightly-scoped copies.
2708         (vect_analyze_data_refs_alignment): Convert return type from bool
2709         to opt_result.
2710         (vect_analyze_data_ref_accesses): Likewise, converting a
2711         "return false" to a "return opt_result::failure_at", adding a
2712         new message.
2713         (vect_prune_runtime_alias_test_list): Convert return type from
2714         bool to opt_result, converting dump_printf_loc to
2715         opt_result::failure_at.  Add a %G to show the pertinent statement,
2716         and use the stmt's location rather than vect_location.
2717         (vect_find_stmt_data_reference): Convert return type from
2718         bool to opt_result, converting dump_printf_loc to
2719         opt_result::failure_at, using stmt's location.
2720         (vect_analyze_data_refs):  Convert return type from bool to
2721         opt_result.  Convert "return false" to "return
2722         opt_result::failure_at", adding messages as needed.
2723         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2724         type from bool to opt_result.
2725         (vect_determine_vf_for_stmt): Likewise.
2726         (vect_determine_vectorization_factor): Likewise, converting
2727         dump_printf_loc to opt_result::failure_at, using location of phi
2728         rather than vect_location.
2729         (vect_analyze_loop_form_1): Convert return type from bool to
2730         opt_result, converting dump_printf_loc calls, retaining the use of
2731         vect_location.
2732         (vect_analyze_loop_form): Convert return type from loop_vec_info
2733         to opt_loop_vec_info.
2734         (vect_analyze_loop_operations): Convert return type from bool to
2735         opt_result, converting dump_printf_loc calls, using the location
2736         of phi/stmt rather than vect_location where available.  Convert
2737         various "return false" to "return opt_result::failure_at" with
2738         "unsupported phi" messages.
2739         (vect_get_datarefs_in_loop): Convert return type from bool to
2740         opt_result.  Add a message for the
2741         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2742         (vect_analyze_loop_2): Convert return type from bool to
2743         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
2744         each "goto again;", adding new messages where needed.
2745         Add "unsupported grouped {store|load}" messages.
2746         (vect_analyze_loop): Convert return type from loop_vec_info to
2747         opt_loop_vec_info.
2748         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2749         bool to opt_result.
2750         * tree-vect-stmts.c (process_use): Likewise, converting
2751         dump_printf_loc call and using stmt location, rather than
2752         vect_location.
2753         (vect_mark_stmts_to_be_vectorized): Likeise.
2754         (vect_analyze_stmt): Likewise, adding a %G.
2755         (vect_get_vector_types_for_stmt): Convert return type from bool to
2756         opt_result, converting dump_printf_loc calls and using stmt
2757         location, rather than vect_location.
2758         (vect_get_mask_type_for_stmt): Convert return type from tree to
2759         opt_tree, converting dump_printf_loc calls and using stmt location.
2760         * tree-vectorizer.c: Include "opt-problem.h.
2761         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2762         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
2763         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
2764         enabled, use it to report at the top level "couldn't vectorize
2765         loop" followed by the problem.
2766         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2767         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2768         to opt_result.
2769         (vect_analyze_stmt): Likewise.
2770         (vect_get_vector_types_for_stmt): Likewise.
2771         (tree vect_get_mask_type_for_stmt): Likewise.
2772         (vect_analyze_data_ref_dependences): Likewise.
2773         (vect_enhance_data_refs_alignment): Likewise.
2774         (vect_analyze_data_refs_alignment): Likewise.
2775         (vect_verify_datarefs_alignment): Likewise.
2776         (vect_analyze_data_ref_accesses): Likewise.
2777         (vect_prune_runtime_alias_test_list): Likewise.
2778         (vect_find_stmt_data_reference): Likewise.
2779         (vect_analyze_data_refs): Likewise.
2780         (vect_analyze_loop): Convert return type from loop_vec_info to
2781         opt_loop_vec_info.
2782         (vect_analyze_loop_form): Likewise.
2783         (vect_analyze_slp): Convert return type from bool to opt_result.
2785 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2787         * doc/invoke.texi (-fopt-info): Document new "internals"
2788         sub-option.
2789         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
2790         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
2791         MSG_ALL_KINDS.
2792         (optinfo_verbosity_options): Add "internals".
2793         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
2794         (dump_context::apply_dump_filter_p): New member function.
2795         (dump_context::dump_loc): Use apply_dump_filter_p rather than
2796         explicitly masking the dump_kind.
2797         (dump_context::begin_scope): Increment the scope depth first.  Use
2798         apply_dump_filter_p rather than explicitly masking the dump_kind.
2799         (dump_context::emit_item): Use apply_dump_filter_p rather than
2800         explicitly masking the dump_kind.
2801         (dump_dec): Likewise.
2802         (dump_hex): Likewise.
2803         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
2804         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
2805         (opt_info_switch_p): Update handling of default
2806         MSG_OPTIMIZED_LOCATIONS to cope with default of
2807         MSG_PRIORITY_USER_FACING.
2808         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
2809         masking the dump_kind.
2810         (selftest::test_capture_of_dump_calls): Update test_dump_context
2811         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
2812         than MSG_ALL.  Generalize scope test to be run at all four
2813         combinations of with/without MSG_PRIORITY_USER_FACING and
2814         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
2815         for each of the two values.
2816         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
2817         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
2818         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
2819         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
2820         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
2821         with MSG_PRIORITY_*.
2822         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
2823         dump messages as MSG_PRIORITY_USER_FACING.
2824         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
2825         about the interaction with MSG_PRIORITY_*.
2827 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2829         * varasm.c (output_constant): Add new parameter merge_strings.
2830         Make strings properly zero terminated in merge string sections.
2831         (mergeable_string_section): Don't fail if the last char is non-zero.
2832         (assemble_variable_contents): Handle merge string sections.
2833         (assemble_variable): Likewise.
2834         (assemble_constant_contents): Likewise.
2835         (output_constant_def_contents): Likewise.
2836         (output_constructor_array_range,
2837         output_constructor_regular_field): Adjust call to output_constant.
2838         (output_object_block): Adjust call to assemble_constant_contents
2839         and assemble_variable_contents.
2841 2018-10-04  Martin Liska  <mliska@suse.cz>
2843         PR c/87483
2844         * cgraphunit.c (process_function_and_variable_attributes):
2845         Warn about a function with alias attribute and a body.
2847 2018-10-04  Martin Liska  <mliska@suse.cz>
2849         PR ipa/82625
2850         * multiple_target.c (redirect_to_specific_clone): New function.
2851         (ipa_target_clone): Use it.
2852         * tree-inline.c: Fix comment.
2854 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2856         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
2857         fields.
2858         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
2859         (gcc::dump_manager::register_pass): New member function, adapted
2860         from loop body in gcc::pass_manager::register_pass, adding a
2861         call to update_dfi_for_opt_info.
2862         (gcc::dump_manager::opt_info_enable_passes): Store the
2863         -fopt-info options into the new fields.  Move the loop
2864         bodies into...
2865         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
2866         function.
2867         * dumpfile.h (struct opt_pass): New forward decl.
2868         (gcc::dump_manager::register_pass): New decl.
2869         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
2870         (class gcc::dump_manager): Add fields "m_optgroup_flags",
2871         "m_optinfo_flags", and "m_optinfo_filename".
2872         * passes.c (gcc::pass_manager::register_pass): Move all of the
2873         dump-handling code to gcc::dump_manager::register_pass.
2875 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
2877         PR rtl-optimization/87466
2878         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
2879         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
2880         * doc/tm.texi: Regenerate.
2881         * ira-lives.c (process_bb_node_lives): Use the new target hook.
2882         * lra-lives.c (process_bb_lives): Likewise.
2883         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
2884         Define.
2886 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
2888         * params.c (add_params): Fix initialization.
2890 2018-10-04  Martin Liska  <mliska@suse.cz>
2892         PR gcov-profile/84107
2893         * tree-profile.c (init_ic_make_global_vars):
2894         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
2895         Come up with new ic_tuple* variables.  Emit
2896         __gcov_indirect_call{,_topn} variables.
2897         (gimple_gen_ic_profiler): Access the variable
2898         and emit gimple.
2899         (gimple_gen_ic_func_profiler): Access
2900         __gcov_indirect_call.callee field.
2901         (gimple_init_gcov_profiler): Use ptr_type_node.
2902         * value-prof.c (gimple_ic): Use ptr_type_node.
2904 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2906         PR tree-optimization/85787
2907         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
2908         into this function and add support for detecting multiple phis.
2909         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
2911 2018-10-04  Martin Liska  <mliska@suse.cz>
2913         PR ipa/87491
2914         * ipa-inline.c (inline_to_all_callers_1):
2915         Call ultimate_alias_target for node being inlined.
2917 2018-10-03  Jeff Law  <law@redhat.com>
2919         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
2920         target's wchar_t.
2921         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
2922         * tree.h (get_typenode_from_name): Prototype.
2924 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2926         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
2927         Change operand 2 predicate to nonimmediate_operand.
2928         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2930 2018-10-03  Martin Sebor  <msebor@redhat.com>
2931             Jeff Law  <law@redhat.com>
2933         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
2934         initialize it.
2935         (get_string_length): Detect unterminated arrays.
2936         (format_string): Same.
2937         (format_directive): Warn about unterminated arrays.
2938         (handle_gimple_call): Mark statements with no_warning as needed.
2940 2018-10-03  Jim Wilson  <jimw@sifive.com>
2942         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
2943         also define __riscv_abi_rve.  Delete trailing white space.
2945 2018-10-03  Paul Koning  <ni1d@arrl.net>
2947         Enable LRA register allocator for PDP11.
2948         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
2949         (R): Likewise.
2950         (D): Likewise.
2951         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
2952         * config/pdp11/pdp11.opt (-mlra): New option.
2953         * doc/invoke.texi (PDP-11 Options): Document -mlra.
2955 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2957         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
2958         (*<absneg:code>extend<mode>xf2): Ditto.
2960 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2962         PR tree-optimization/87415
2963         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
2964         precision fields.
2966 2018-10-02  Jeff Law  <law@redhat.com>
2968         * gimple-fold.c (get_range_strlen): Only set *nonstr when
2969         an unterminated string is discovered.  Bubble up range
2970         even for unterminated strings.
2971         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
2972         indicates the string was not terminated via NONSTR.
2974 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2976         * tree-vrp.c (extract_range_from_unary_expr): Special case all
2977         pointer conversions.
2978         Do not do anything special for anti-ranges.
2980 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
2982         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
2983         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
2985 2018-10-03  Martin Liska  <mliska@suse.cz>
2987         PR gcov-profile/86109
2988         * coverage.c (coverage_begin_function): Do not
2989         mark lambdas as artificial.
2990         * tree-core.h (struct GTY): Remove tm_clone_flag
2991         and introduce new lambda_function.
2992         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2994 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2996         PR target/87474
2997         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
2998         P8_VECTOR and VSX are enabled.
3000 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3002         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
3003         0x3907 as CPU model number.
3005 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3007         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
3008         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
3009         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
3010         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
3011         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
3012         * config/s390/s390.md: Likewise. Rename also the cpu attribute
3013         value from arch12 to z14.
3015 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3017         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
3018         (isinfxf2): Ditto.
3019         (isinf<mode>2): Ditto.
3021 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3023         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
3024         before emitting fxam.  Perform calculations in XFmode.
3026 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3028         * match.pd (((X /[ex] A) +- B) * A): New transformation.
3030 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3032         PR middle-end/87319
3033         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
3034         * tree.c (signed_or_unsigned_type_for): Handle complex.
3036 2018-10-02  Jeff Law  <law@redhat.com>
3038         * gimple-fold.c (get_range_strlen): Remove dead code.
3040 2018-10-02  Martin Sebor  <msebor@redhat.com>
3041             Jeff Law  <law@redhat.com>
3043         * builtins.c (unterminated_array): Add new arguments.
3044         If argument is not terminated, bubble up size and exact
3045         state to callers.
3046         (expand_builtin_strnlen): Detect, avoid expanding
3047         and diagnose unterminated arrays.
3048         (c_strlen): Fill in offset of start of unterminated strings.
3049         * builtins.h (unterminated_array): Update prototype.
3051 2018-10-02  Richard Biener  <rguenther@suse.de>
3053         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
3054         of haddv4df, first reduce to SSE width and exploit the fact
3055         that we only need element zero with the reduction result.
3056         (reduc_plus_scal_v2df): Likewise.
3058 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3060         * dojump.h (do_jump): Delete.
3061         (do_jump_1): Likewise.
3062         (split_comparison): Move around.
3063         * dojump.c (do_jump): Make static.
3064         (do_jump_1): Likewise.
3065         (jumpifnot): Move around.
3066         (jumpifnot_1): Likewise.
3067         (jumpif): Likewise.
3068         (jumpif_1): Likewise.
3069         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
3071 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3073         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
3074         insns in the delay slot and add_insn_after for the jump insn.
3076 2018-10-02  Richard Biener  <rguenther@suse.de>
3078         * tree-inline.c (expand_call_inline): Use the location of
3079         the callee declaration for the inline-entry marker.
3080         * final.c (notice_source_line): Remove special-casing of
3081         NOTE_INSN_INLINE_ENTRY.
3083 2018-10-01  Carl Love  <cel@us.ibm.com>
3085         PR 69431
3086         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
3087         (__builtin_mtfsb0): New.
3088         (__builtin_mtfsb1): New.
3089         ( __builtin_set_fpscr_rn): New.
3090         (__builtin_set_fpscr_drn): New.
3091         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
3092         (rs6000_expand_set_fpscr_rn_builtin): Add.
3093         (rs6000_expand_set_fpscr_drn_builtin): Add.
3094         (rs6000_expand_builtin): Add case statement entries for
3095         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
3096         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
3097         RS6000_BUILTIN_MFFSL.
3098         (rs6000_init_builtins): Add ftype initialization and def_builtin
3099         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
3100         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
3101         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
3102         rs6000_mffscdrn): Add define_insn.
3103         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
3104         * doc/extend.texi: Add documentation for the builtins.
3106 2018-10-01  Richard Biener  <rguenther@suse.de>
3108         PR tree-optimization/87465
3109         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
3110         causing branch miscounts.
3112 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3114         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
3115         aarch64_option_default_param):  New.
3116         (params.h): Include.
3117         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
3118         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
3119         stack-clash protection validation code.
3121 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3123         * params.c (validate_param): New.
3124         (add_params): Use it.
3125         (set_param_value): Refactor param validation into validate_param.
3126         (diagnostic.h): Include.
3127         * diagnostic.h (diagnostic_ready_p): New.
3129 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3131         * params.c (set_param_value):
3132         Add index of parameter being validated.
3133         * common/common-target.def (option_validate_param): New.
3134         * common/common-targhooks.h (default_option_validate_param): New.
3135         * common/common-targhooks.c (default_option_validate_param): New.
3136         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
3137         * doc/tm.texi: Regenerate.
3139 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3141         PR target/86486
3142         * config/aarch64/aarch64.c (aarch64_override_options_internal):
3143         Add validation for stack-clash parameters and set defaults.
3145 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3147         PR target/86486
3148         * configure.ac: Add stack-clash-protection-guard-size.
3149         * doc/install.texi: Document it.
3150         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
3151         * params.def: Update comment for guard-size.
3152         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
3153         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
3154         * configure: Regenerate.
3156 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3158         PR target/86486
3159         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
3160         STACK_DYNAMIC_OFFSET): New.
3161         * config/aarch64/aarch64.c (aarch64_layout_frame):
3162         Update outgoing args size.
3163         (aarch64_stack_clash_protection_alloca_probe_range,
3164         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3166 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3168         PR target/86486
3169         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
3170         probe ranges.
3171         * target.def (stack_clash_protection_alloca_probe_range): New.
3172         (stack_clash_protection_final_dynamic_probe): Remove.
3173         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
3174         (default_stack_clash_protection_final_dynamic_probe): Remove.
3175         * targhooks.c: Likewise.
3176         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3177         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
3178         * doc/tm.texi: Regenerate.
3180 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3182         PR target/86486
3183         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
3184         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
3185         aarch64_clamp_to_uimm12_shift): New.
3186         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
3187         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
3189 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3191         PR target/86486
3192         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
3194 2018-10-01  Jeff Law  <law@redhat.com>
3195             Richard Sandiford <richard.sandiford@linaro.org>
3196             Tamar Christina  <tamar.christina@arm.com>
3198         PR target/86486
3199         * config/aarch64/aarch64.md
3200         (probe_stack_range): Add k (SP) constraint.
3201         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
3202         STACK_CLASH_MAX_UNROLL_PAGES): New.
3203         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
3204         stack probes for stack clash.
3205         (aarch64_allocate_and_probe_stack_space): New.
3206         (aarch64_expand_prologue): Use it.
3207         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
3208         (aarch64_sub_sp): Add emit_move_imm optional param.
3210 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
3212         PR tree-optimization/87261
3213         * match.pd: Remove trailing whitespace.
3214         Add (x & y) | ~(x | y) -> ~(x ^ y),
3215         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
3217 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3219         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
3220         constraints.
3221         (ashlsi3_insn): Update instruction constraints.
3222         (ashrsi3_insn): Likewise.
3223         (rotrsi3): Likewise.
3224         (add_shift): Likewise.
3225         * config/arc/constraints.md (Csz): New 32 bit constraint. It
3226         avoids placing in the limm field small constants which, otherwise,
3227         could end into a small instruction.
3229 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3231         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
3232         destination register is not odd-even.
3233         (umaddsidi4_split): Likewise.
3235 2018-10-01  Richard Biener  <rguenther@suse.de>
3237         * tree-inline.c (expand_call_inline): Store origin of fn
3238         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
3239         * tree.c (block_ultimate_origin): Simplify and do some
3240         checking.
3242 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
3244         * config/i386/mmx.md (EMMS): New int iterator.
3245         (emms): New int attribute.
3246         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
3247         EMMS int iterator.  Explicitly declare clobbers.
3248         (mmx_emms): Remove expander.
3249         (mmx_femms): Ditto.
3250         * config/i386/predicates.md (emms_operation): Remove predicate.
3251         (vzeroall_pattern): New predicate.
3252         (vzeroupper_pattern): Rename from vzeroupper_operation.
3253         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
3254         vzeroupper_pattern and vzeroall_pattern predicates.
3256 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
3258         PR rtl-optimization/86939
3259         * ira-lives.c (make_hard_regno_born): Rename from this...
3260         (make_hard_regno_live): ... to this.  Remove update to conflict
3261         information.  Update function comment.
3262         (make_hard_regno_dead): Add conflict information update.  Update
3263         function comment.
3264         (make_object_born): Rename from this...
3265         (make_object_live): ... to this.  Remove update to conflict information.
3266         Update function comment.
3267         (make_object_dead):  Add conflict information update.  Update function
3268         comment.
3269         (mark_pseudo_regno_live): Call make_object_live.
3270         (mark_pseudo_regno_subword_live): Likewise.
3271         (mark_hard_reg_dead): Update function comment.
3272         (mark_hard_reg_live): Call make_hard_regno_live.
3273         (process_bb_node_lives): Likewise.
3274         * lra-lives.c (make_hard_regno_born): Rename from this...
3275         (make_hard_regno_live): ... to this.  Remove update to conflict
3276         information.  Remove now uneeded check_pic_pseudo_p argument.
3277         Update function comment.
3278         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
3279         to conflict information.  Update function comment.
3280         (mark_pseudo_live): Remove update to conflict information.  Update
3281         function comment.
3282         (mark_pseudo_dead): Add conflict information update.
3283         (mark_regno_live): Call make_hard_regno_live.
3284         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
3285         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
3287 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
3289         PR target/87370
3290         * config/i386/i386.c (construct_container): Use TImode for
3291         BLKmode values in 2 integer registers.
3293 2018-09-29  Jeff Law  <law@redhat.com>
3295         * builtins.c (unterminated_array): Pass in c_strlen_data * to
3296         c_strlen rather than just a tree *.
3297         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
3298         Update recursive calls appropriately.  If caller did not provide a
3299         suitable data pointer, create a local one.  When a non-terminated
3300         string is discovered, bubble up information about the string via the
3301         c_strlen_data object.
3302         * builtins.h (c_strlen): Update prototype.
3303         (c_strlen_data): New structure.
3304         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
3305         For a type 2 call, if c_strlen indicates a non-terminated string
3306         use the length of the non-terminated string.
3307         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
3309 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
3311         PR target/87467
3312         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
3313         __m512d type for __A argument rather than __m512.
3315 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
3317         * match.pd (simple_comparison): Don't optimize if either operand is
3318         a function pointer when target needs function pointer canonicalization.
3320 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3322         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
3323         power5 .. power9 to remove indirection.
3324         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
3325         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
3326         ASM_CPU_476_SPEC): Delete.
3327         (ASM_CPU_SPEC): Adjust.
3328         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
3329         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
3331 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3333         * config.in: Delete HAVE_AS_DCI.
3334         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
3335         * config/rs6000/rs6000.h: Ditto.
3336         * configure.ac: Delete HAVE_AS_DCI.
3337         * configure: Regenerate.
3339 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3341         * config.in (HAVE_AS_LWSYNC): Delete.
3342         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3343         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
3344         do it as a .long .
3345         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3346         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
3347         as a .long .
3348         * configure.ac: Delete HAVE_AS_LWSYNC.
3349         * configure: Regenerate.
3351 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3352             Pierre-Marie de Rodat  <derodat@adacore.com>
3354         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
3355         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
3356         (cgraph_node::create_thunk): Add indirect_offset parameter.
3357         (thunk_adjust): Likewise.
3358         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
3359         and initialize the corresponding field with it.
3360         (cgraph_node::dump): Dump indirect_offset field.
3361         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
3362         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
3363         (thunk_adjust): Add indirect_offset parameter and deal with it.
3364         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
3365         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
3366         or if the thunk is external or local.  Fix formatting.  Do not chain
3367         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
3368         if any, in the GIMPLE representation.
3369         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
3370         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
3371         (input_node): Read indirect_offset field.
3372         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
3373         call to thunk_adjust.
3374         * tree-nested.c (struct nesting_info): Add thunk_p field.
3375         (create_nesting_tree): Set it.
3376         (convert_all_function_calls): Copy static chain from targets to thunks.
3377         (finalize_nesting_tree_1): Return early for thunks.
3378         (unnest_nesting_tree_1): Do not finalize thunks.
3379         (gimplify_all_functions): Do not gimplify thunks.
3381 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3383         * opt-suggestions.c (option_proposer::build_option_suggestions):
3384         Release "option_values".
3386 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3388         * coverage.c (get_coverage_counts): Convert problem-reporting dump
3389         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
3390         * dumpfile.c (kind_as_string): New function.
3391         (dump_loc): Rather than a hardcoded prefix of "note: ", use
3392         kind_as_string to vary the prefix based on dump_kind.
3393         (selftest::test_capture_of_dump_calls): Update for above.
3395 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3397         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3398         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
3400 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3402         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3403         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3404         INVALID_REGNUM instead of FPSR_REG.
3405         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3406         * config/i386/i386.md: Update comment of FP compares.
3407         (fldenv): Do not clobber FPSR_REG.
3409 2018-09-28  Richard Biener  <rguenther@suse.de>
3411         * tree.h (BLOCK_ORIGIN): New.
3412         * omp-expand.c (grid_expand_target_grid_body): Assign
3413         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3414         * tree-inline.c (remap_block): Likewise.
3415         * auto-profile.c (get_function_decl_from_block): Simplify
3416         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3417         * langhooks.c (lhd_print_error_function): Likewise.
3418         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3419         Likewise.
3420         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3421         * tree.c (block_nonartificial_location): Likewise.
3422         (block_ultimate_origin): Likewise.
3423         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
3424         no longer needed LTO case.
3426 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
3427             Jan Hubicka  <jh@suse.cz>
3428             Martin Jambor  <mjambor@suse.cz>
3430         * simplify-rtx.c (simplify_merge_mask): New function.
3431         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3432         same masks are used in op1 or op2.
3433         (test_vec_merge): New function.
3434         (test_vector_ops): Call test_vec_merge.
3436 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3438         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3439         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3440         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3441         (cypress_costs): Set it.
3442         (supersparc_costs): Likewise.
3443         (hypersparc_costs): Likewise.
3444         (leon_cost): Likewise.
3445         (leon3_costs): Likewise.
3446         (sparclet_costs): Likewise.
3447         (ultrasparc_costs): Likewise.
3448         (ultrasparc_costs): Likewise.
3449         (niagara_costs): Likewise.
3450         (niagara2_costs): Likewise.
3451         (niagara3_costs): Likewise.
3452         (niagara4_costs): Likewise.
3453         (niagara7_costs): Likewise.
3454         (m8_costs): Likewise.
3455         (TARGET_CAN_FOLLOW_JUMP): Define.
3456         (pass_work_around_errata::gate): Minor tweak.
3457         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3458         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3459         Automaitcally clear MASK_FSMULD mask for V7 processors.
3460         (sparc_can_follow_jump): New static function.
3461         (output_ubranch): Deal with CROSSING_JUMP_P.
3462         (sparc_use_sched_lookahead): Rewrite using switch statement.
3463         (sparc_issue_rate): Reorder.
3464         (sparc_branch_cost): New function.
3466 2018-09-27  Martin Sebor  <msebor@redhat.com>
3468         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3469         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3470         (int_fits_type_p): Same.
3472 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
3474         * config/i386/i386.md (FPCR_REG): Remove.
3475         (UNSPEC_FLDCW): Remove.
3476         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3477         (x86_fldcw_1): Remove insn pattern.
3478         (fnstenv): Do not clobber FPCR_REG.
3479         (fldenv): Ditto.
3480         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3481         (CALL_USED_REGISTERS): Ditto.
3482         (REG_ALLOC_ORDER): Ditto.
3483         (REG_CLASS_CONTENTS): Ditto.
3484         (HI_REGISTER_NAMES): Ditto.
3485         (ADDITIONAL_REGISTER_NAMES): Use defines instead
3486         of numerical constants.
3487         * config/i386/i386.c (regclass_map): Remove fpsr register.
3488         (dbx_register_map): Ditto.
3489         (dbx64_register_map): Ditto.
3490         (svr4_dbx_register_map): Ditto.
3491         (print_reg): Do not handle FPCR_REG.
3493 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
3495         PR target/87149
3496         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3497         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3498         Delete, always treat as true.
3499         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3500         Ditto.  Simplify remaining code.
3501         * config/powerpcspe/powerpcspe.h: Ditto.
3502         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3503         Simplify remaining code.
3504         (rs6000_expand_builtin): Ditto.
3505         * config/rs6000/rs6000.h: Ditto.
3506         * configure.ac: Ditto.
3507         * configure: Regenerate.
3509 2018-09-27  Martin Liska  <mliska@suse.cz>
3511         * coverage.c (get_coverage_counts): Revert the formatting
3512         of missing profile opt info.
3514 2018-09-27  Richard Biener  <rguenther@suse.de>
3516         PR debug/37801
3517         PR debug/87440
3518         * dwarf2out.c (set_block_origin_self): Do not mark outermost
3519         block as we do not output that.
3520         (gen_inlined_subroutine_die): Elide the originally outermost
3521         block, matching what we do for concrete instances.
3522         (decls_for_scope): Add parameter specifying whether to recurse
3523         to subblocks.
3525 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
3526             Tom de Vries  <tom@codesourcery.com>
3528         PR 82089
3530         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3531         STORE_FLAG_VALUE == 1.
3533 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
3535         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3536         constant definitions.
3537         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3538         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3539         ("speculation_barrier"): New expander definition.
3541 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
3543         PR gcov-profile/86957
3544         * common.opt: New warning option -Wmissing-profile.
3545         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3546         * doc/invoke.texi: Document -Wmissing-profile.
3548 2018-09-26  Jim Wilson  <jimw@sifive.com>
3550         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3551         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3553 2018-09-26  Martin Sebor  <msebor@redhat.com>
3555         * tree.c (zerop): Change return type to bool.
3556         (integer_zerop, integer_onep, integer_each_onep): Same.
3557         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3558         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3559         (real_onep, real_minus_onep, chain_index): Same.
3560         (print_type_hash_statistics, type_list_equal): Same.
3561         * tree.h (zerop): Same.
3562         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
3563         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3564         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3565         (real_onep, real_minus_onep, chain_index): Same.
3566         (print_type_hash_statistics, type_list_equal): Same.
3568 2018-09-26  Jim Wilson  <jimw@sifive.com>
3570         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3572 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
3574         PR target/87414
3575         * config/i386/i386.c: Include debug.h and dwarf2out.h.
3576         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3577         call.
3579 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
3581         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3583 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3585         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3586         and FP_SECOND_SSE_REGS.
3587         (REG_CLASS_NAMES): Ditto.
3588         (REG_CLASS_CONTENTS): Ditto.
3589         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3590         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3591         (ix86_preferred_output_reload_class): Ditto.
3592         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3593         clobber constraint to "=&f".
3594         (fix_truncdi_i387): Ditto.
3595         (lrintxfdi2): Ditto.
3596         (fistdi2_<rounding>): Ditto.
3597         (fpremxf4_i387): Change "=u" constraint to "=f".
3598         (fprem1xf4_i387): Ditto.
3599         (sincosxf3): Ditto.
3600         (fptanxf4_i387): Ditto.
3601         (fxtractxf3_i387): Ditto.
3602         (fscalexf4_i387): Ditto.
3603         (atan2xf3): Change "u" constraint to "f".
3604         (fyl2xxf3_i387): Ditto.
3605         (fyl2xp1xf3_i387): Ditto.
3607 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3609         PR target/87439
3610         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3611         for removed I387_MASK_PM entity.
3614 2018-09-26  Jeff Law  <law@redhat.com>
3615         Revert
3616         2018-09-26  Alexey Neyman  <stilor@att.net>
3618         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3619         headers are no longer pulled in by <isl/val.h>.
3621 2018-09-26  Richard Biener  <rguenther@suse.de>
3623         PR debug/87443
3624         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3625         or concrete instance DIE to the tree.  Create abstract origin
3626         attributes also for concrete instances.
3628 2018-09-26  Alexey Neyman  <stilor@att.net>
3630         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3631         headers are no longer pulled in by <isl/val.h>.
3633 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
3635         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3636         Use new helper functions.
3637         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3638         Use new helper functions.
3639         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3640         aarch_mm_needs_release): New declarations.
3641         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3642         aarch_mm_needs_release): New.
3644 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
3646         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3647         (arm32_output_mi_thunk): Deal with long calls.
3649 2018-09-26  Richard Biener  <rguenther@suse.de>
3651         PR debug/87428
3652         PR debug/87362
3653         * tree-inline.c (expand_call_inline): When the location
3654         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3655         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3656         the inserted BLOCK to make inlined_function_outer_scope_p
3657         recognize it.
3658         * dwarf2out.c (add_call_src_coords_attributes): Do not add
3659         coords for reserved locations.
3661 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3663         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3664         (*call_indirect_nonlocal_sysv<mode>): Ditto.
3665         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3666         (*sibcall_nonlocal_sysv<mode>): Ditto.
3667         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3668         (<bd>_<mode>): Ditto.
3669         (<bd>tf_<mode>): Ditto.
3671 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3673         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3674         control string as a list of templates instead of as C code.
3675         (*altivec_movti): Ditto.
3676         * config/rs6000/darwin.md (movdf_low_di): Ditto.
3678 2018-09-25  Jim Wilson  <jimw@sifive.com>
3680         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3681         when target symbol is weak.
3683 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3685         PR c/87387
3686         * builtins.c (unterminated_array): Simplify.
3687         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
3688         where pointer arithmetic is safe.
3690 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3692         PR target/86987
3693         * config/rs6000/altivec.md (altivec_vspltb): Use
3694         const_0_to_15_operand instead of u5bit_cint_operand.
3695         (*altivec_vspltb_internal): Ditto.
3696         (altivec_vspltb_direct): Ditto.
3697         (altivec_vsplth): Use const_0_to_7_operand instead of
3698         u5bit_cint_operand.
3699         (*altivec_vsplth_internal): Ditto.
3700         (altivec_vsplth_direct): Ditto.
3701         (altivec_vspltw): Use const_0_to_3_operand instead of
3702         u5bit_cint_operand.
3703         (*altivec_vspltw_internal): Ditto.
3704         (altivec_vspltw_direct): Ditto.
3705         (altivec_vspltsf): Ditto.
3706         (*altivec_vspltsf_internal): Ditto.
3707         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3708         various splats with the proper size immediate.  Reorder the various
3709         cases by ascending size of immediate, and put all such together.
3711 2018-09-25  Richard Biener  <rguenther@suse.de>
3713         PR debug/83941
3714         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3715         GC-ification.
3716         (maybe_create_die_with_external_ref): Do not create
3717         DW_TAG_imported_unit here.
3718         (add_abstract_origin_attribute): Handle external BLOCK refs.
3719         (dwarf2out_abstract_function): Simplify LTO case.
3720         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3721         rather than using maybe_create_die_with_external_ref.
3723 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
3725         PR target/71278
3726         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3727         (frndintxf2_mask_pm_i387): Ditto.
3728         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3729         Enable for !flag_trapping_math.
3730         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3731         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3732         Change operand 1 predicate to nonimmediate_operand.
3733         (attr "i387_cw"): Remove mask_pm.
3734         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3735         (enum ix86_entity): Remove I387_MASK_PM.
3736         * config/i386/i386.c (ix86_i387_mode_needed): Do not
3737         handle I387_MASK_PM.
3738         (ix86_mode_needed): Ditto.
3739         (ix86_mode_after): Ditto.
3740         (ix86_mode_entry): Ditto.
3741         (ix86_mode_exit): Ditto.
3742         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3744 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
3746         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3747         to_update_switch_stmts to vNULL instead of calling create on them
3748         immediately.
3750 2018-09-25  Richard Biener  <rguenther@suse.de>
3752         PR tree-optimization/87402
3753         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3754         (visit_phi): Re-instantiate handling of supposed to be VARYING
3755         but non-VARYING backedge value.
3757 2018-09-25  Richard Biener  <rguenther@suse.de>
3759         PR debug/83941
3760         * dwarf2out.c (struct sym_off_pair): New.
3761         (external_die_map): New global.
3762         (lookup_decl_die): When in LTO create DIEs lazily from the
3763         external_die_map.
3764         (lookup_block_die): New function, create DIEs lazily in LTO.
3765         (equate_block_to_die): New function.
3766         (dwarf2out_die_ref_for_decl): During WPA get the association
3767         from the external DIE map.
3768         (dwarf2out_register_external_die): Record mapping into the
3769         external DIE map.
3770         (maybe_create_die_with_external_ref): New function split out from
3771         DIE generation part of old dwarf2out_register_external_die.
3772         (add_abstract_origin_attribute): Do not return the DIE.  When
3773         in LTO reference externals directly.
3774         (dwarf2out_abstract_function): When in LTO ignore calls for
3775         decls with external DIEs (already present abstract instances).
3776         (gen_call_site_die): Adjust.
3777         (add_high_low_attributes): Likewise.
3778         (gen_lexical_block_die): Likewise.
3779         (gen_inlined_subroutine_die): Likewie.
3780         (gen_block_die): Likewise.
3781         (dwarf2out_inline_entry): Likewise.
3782         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
3783         DIEs.
3785 2018-09-25  Martin Liska  <mliska@suse.cz>
3787         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
3788         integers and not by a float value.
3790 2018-09-25  Martin Liska  <mliska@suse.cz>
3792         PR fortran/87394
3793         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
3794         instead of NULL.
3795         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
3796         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
3797         * godump.c (go_format_type): Remove extra parenthesis.
3799 2018-09-25  Martin Liska  <mliska@suse.cz>
3801         * alias.c (set_dest_equal_p): Remove unused function.
3802         * config/i386/i386.c (def_builtin_pure2): Likewise.
3803         * diagnostic-show-locus.c (class layout): Remove
3804         unused field.
3805         (layout::layout): Likewise here.
3806         * dump-context.h (class temp_dump_context): Likewise.
3807         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
3808         (add_AT_loclistsptr): Likewise.
3809         (add_AT_offset): Likewise.
3810         (get_AT_hi_pc): Likewise.
3811         (is_comdat_die): Likewise.
3812         (type_is_enum): Likewise.
3813         (ceiling): Likewise.
3814         (add_AT_vms_delta): Likewise.
3815         (is_class_die): Likewise.
3816         * edit-context.c (class line_event): Remove unused field.
3817         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
3818         unused function.
3819         * ipa-cp.c (ipa_get_vr_lat): Likewise.
3820         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
3821         (ok_for_base_p_nonstrict): Likewise.
3822         * tree-chrec.c (is_not_constant_evolution): Likewise.
3823         (chrec_fold_poly_cst): Likewise.
3824         * tree-if-conv.c (has_pred_critical_p): Likewise.
3825         * tree-ssa-coalesce.c (print_exprs): Likewise.
3826         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
3827         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
3828         * tree-vrp.c (value_ranges_intersect_p): Likewise.
3829         (value_range_nonnegative_p): Likewise.
3831 2018-09-25  Martin Liska  <mliska@suse.cz>
3833         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3834         Do not handle "GNU Pascal".
3835         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3836         Likewise.
3837         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
3838         from documentation. Likewise.
3839         * dbxout.c (dbxout_range_type): Likewise.
3840         * doc/cpp.texi: Likewise.
3841         * doc/extend.texi: Likewise.
3842         * doc/frontends.texi: Likewise.
3843         * doc/invoke.texi: Remove Pascal entry.
3844         * tree.def (CLEANUP_POINT_EXPR): Likewise.
3845         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
3847 2018-09-25  Martin Liska  <mliska@suse.cz>
3849         PR middle-end/86078
3850         * doc/invoke.texi: Document all parameters and remove default
3851         of the parameters.
3853 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3855         PR bootstrap/87417
3856         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
3857         contains HOST_WIDE_INTs when computing its size.
3859 2018-09-24  Jim Wilson  <jimw@sifive.com>
3861         PR target/87391
3862         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
3863         not TARGET_RVE.
3864         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
3866 2018-09-24  Andrew Pinski  <apinski@marvell.com>
3868         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3869         access prev before checking it for NULLness in the
3870         AARCH64_FUSE_CMP_BRANCH case.
3872 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
3874         PR target/82699
3875         * config/i386/i386.c (rest_of_insert_endbranch): Set
3876         endbr_queued_at_entrance to true and don't insert ENDBR if
3877         x86_function_profiler will be called.
3878         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3879         is true.
3880         * config/i386/i386.h (machine_function): Add
3881         endbr_queued_at_entrance.
3883 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3885         * genattrtab.c (mk_attr_alt): Use alternative_mask.
3886         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
3887         types.
3888         (check_attr_test): Use alternative_mask.
3889         (get_attr_value): Likewise.
3890         (compute_alternative_mask): Use alternative_mask and XWINT.
3891         (make_alternative_compare): Use alternative_mask.
3892         (attr_alt_subset_p): Use XWINT.
3893         (attr_alt_subset_of_compl_p): Likewise.
3894         (attr_alt_intersection): Use alternative_mask and XWINT.
3895         (attr_alt_union): Likewise.
3896         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
3897         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
3898         (simplify_test_exp): Use alternative_mask and XWINT.
3899         (write_test_expr): Use alternative_mask and XWINT, adjust bit
3900         number calculation to support 64 bits.  Generate code that
3901         checks 64-bit masks.
3902         (main): Use alternative_mask.
3903         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
3905 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3907         PR target/80080
3908         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
3909         RETURN+USE when returning via %r14.
3911 2018-09-24  Martin Liska  <mliska@suse.cz>
3913         * gcov.c (output_lines): Print colorization legend
3914         for both flag_use_colors and flag_use_hotness_colors.
3915         Reword the help.
3917 2018-09-24  Martin Liska  <mliska@suse.cz>
3919         * coverage.c (get_coverage_counts): Use warning_at
3920         with current_function_decl location. Use %qD in warning
3921         message.
3923 2018-09-24  Martin Liska  <mliska@suse.cz>
3925         * memory-block.h (memory_block_pool::release): Annotate with
3926         valgrind that the memory is not accessible.
3928 2018-09-24  Martin Liska  <mliska@suse.cz>
3930         PR sanitizer/85774
3931         * asan.c: Make asan_handled_variables extern.
3932         * asan.h: Likewise.
3933         * cfgexpand.c (expand_stack_vars): Make sure
3934         a representative is unpoison if another
3935         variable in the partition is handled by
3936         use-after-scope sanitization.
3938 2018-09-24  Richard Biener  <rguenther@suse.de>
3940         PR tree-optimization/63155
3941         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
3942         the worklist when the edge of the respective argument isn't
3943         executable.
3945 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3947         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
3948         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
3949         (MASK_CLASS_P): Update for rename.
3950         (MAYBE_MASK_CLASS_P): Ditto.
3951         (REG_CLASS_NAMES): Update.
3952         (REG_CLASS_CONTENT): Update.
3953         * config/i386/i386.c (regclass_map): Update for MASK_REG
3954         and ALL_MASK_REGS rename.
3955         * config/i386/constraints.md (Yk): Update for rename.
3956         (k): Ditto.
3958 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3960         * config/i386/i386.h (enum reg_class): Remove
3961         EVEX_SSE_REGS and MOD4_SSE_REGS.
3962         (REG_CLASS_NAMES): Update.
3963         (REG_CLASS_CONTENT): Update.
3964         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
3965         registers as ALL_SSE_REGS.
3966         (ix86_additional_allocno_class_p): Remove.
3967         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
3968         (ix86_register_priority): Lower priority of EVEX SSE registers.
3969         Use IN_RANGE macro where appropriate.
3970         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
3971         AVX-5124VNNIW checks.
3972         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
3973         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
3974         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
3975         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
3976         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
3977         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
3978         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
3979         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
3980         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
3981         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
3982         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
3983         * config/i386/constraints.md (Yh): Remove.
3985 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3987         * config/i386/i386.c (regclass_map): Declare integer REX registers
3988         as GENERAL_REGS.
3990 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
3992         * doc/service.texi (Service): Switch the fsf.org link to https.
3994 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3996         PR target/86798
3997         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3998         Define to speculation_safe_value_not_needed.
4000 2018-09-21  Florian Weimer  <fweimer@redhat.com>
4002         PR middle-end/81035
4003         * doc/extend.texi (Common Function Attributes): Mention that
4004         noreturn suppresses tail call optimization.
4006 2018-09-21  Jeff Law  <law@redhat.com>
4008         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
4009         vr_values::cleanup_edges_and_switches.
4010         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
4011         vr_values class.
4012         (identify_jump_threads): Remove EDGE_IGNORE handling.
4013         (execute_vrp): Move handling of to_remove_edges and
4014         to_update_switch_stmts into vr_values class member functions.
4015         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
4016         (to_update_switch_stmts): Likewise.
4017         * vr-values.c: Include cfghooks.h.
4018         (vr_values::vr_values): Initialize to_remove_edges and
4019         to_update_switch_stmts.
4020         (vr_values::~vr_values): Verify to_remove_edges and
4021         to_update_switch_stmts are empty.
4022         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
4023         (vr_values::cleanup_edges_and_switches): New member function.
4024         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
4025         function.  Add new data members.
4027 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
4029         PR tree-optimization/87309
4030         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
4031         calls with pflags and alt_flags.
4032         (selftest::test_capture_of_dump_calls): Add test of interaction of
4033         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
4035 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4037         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
4039 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4041         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
4043 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4045         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
4046         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
4047         (VXWORKS_LIBS_RTP): Minor reordering.
4049 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4051         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
4052         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
4054 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4056         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
4057         (PTRDIFF_TYPE): Likewise.
4059 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4061         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
4062         triplet, similar to support for VxWorks7.
4063         * config/vxworks-dummy.h: Provide a default definition
4064         of TARGET_VXWORKS64 to 0.
4066 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4068         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
4069         * config/vxworks-dummy.h: here.
4071 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4073         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
4075 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
4076             Bo Zhou  <zbo.zhou@hisilicon.com>
4078         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
4079         * config/aarch64/aarch64-tune.md: Regenerated.
4080         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
4081         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
4082         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
4084 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
4085             Julian Brown  <julian@codesourcery.com>
4087         * builtins.c (get_builtin_sync_mem): Handle address spaces.
4089 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4091         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
4092         if the call takes a static chain.
4094 2018-09-21  Martin Liska  <mliska@suse.cz>
4096         * auto-profile.c (autofdo_source_profile::read): Do not
4097         set sum_all.
4098         (read_profile): Do not add working sets.
4099         (read_autofdo_file): Remove sum_all.
4100         (afdo_callsite_hot_enough_for_early_inline): Remove const
4101         qualifier.
4102         * coverage.c (struct counts_entry): Remove gcov_summary.
4103         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
4104         do not support GCOV_TAG_PROGRAM_SUMMARY.
4105         (get_coverage_counts): Remove summary and expected
4106         arguments.
4107         * coverage.h (get_coverage_counts): Likewise.
4108         * doc/gcov-dump.texi: Remove -w option.
4109         * gcov-dump.c (dump_working_sets): Remove.
4110         (main): Do not support '-w' option.
4111         (print_usage): Likewise.
4112         (tag_summary): Likewise.
4113         * gcov-io.c (gcov_write_summary): Do not dump
4114         histogram.
4115         (gcov_read_summary): Likewise.
4116         (gcov_histo_index): Remove.
4117         (gcov_histogram_merge): Likewise.
4118         (compute_working_sets): Likewise.
4119         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
4120         it not obsolete.
4121         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
4122         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
4123         (GCOV_HISTOGRAM_SIZE): Remove.
4124         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
4125         (struct gcov_summary): Simplify rapidly just
4126         to runs and sum_max fields.
4127         (gcov_histo_index): Remove.
4128         (NUM_GCOV_WORKING_SETS): Likewise.
4129         (compute_working_sets): Likewise.
4130         * gcov-tool.c (print_overlap_usage_message): Remove
4131         trailing empty line.
4132         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
4133         (output_lines): Remove program related line.
4134         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
4135         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
4136         histogram.
4137         (input_profile_summary): Do not read it.
4138         (merge_profile_summaries): And do not merge it.
4139         (input_symtab): Do not call removed function.
4140         * modulo-sched.c (sms_schedule): Do not print sum_max.
4141         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
4142         removed when histogram method was invented.
4143         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
4144         mode.
4145         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
4146         GCOV coding style.
4147         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
4148         and dump selected value.
4149         * profile.c (add_working_set): Remove.
4150         (get_working_sets): Likewise.
4151         (find_working_set): Likewise.
4152         (get_exec_counts): Do not work with working sets.
4153         (read_profile_edge_counts): Do not inform as sum_max is removed.
4154         (compute_branch_probabilities): Likewise.
4155         (compute_value_histograms): Remove argument for call of
4156         get_coverage_counts.
4157         * profile.h: Do not make gcov_summary const.
4159 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
4161         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
4163 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4165         PR tree-optimization/86990
4166         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
4167         Check that the entire merged store group is made of constants only for
4168         overlapping stores.
4170 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
4172         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
4173         (VTABLE_VERIFICATION_SPEC): Likewise.
4174         (SANITIZER_EARLY_SPEC): Likewise.
4175         (SANITIZER_SPEC): Likewise.
4176         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
4177         * doc/invoke.texi (Link Options): Document -r.
4179 2018-09-20  Richard Biener <rguenther@suse.de>
4181         PR middle-end/87054
4182         * gimplify.c (gimplify_expr): Retain alignment of
4183         addressable lvalue in dereference.
4185 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
4187         PR bootstrap/87013
4188         * configure.ac: Check for .loc is_stmt support.
4189         * configure, config.in: Rebuilt.
4190         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
4191         if not supported.
4193 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4195         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
4196         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
4197         -misel=no.
4199 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4201         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4202         VECTOR_OTHER.
4203         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
4204         case VECTOR_OTHER.
4206 2018-09-20  Marek Polacek  <polacek@redhat.com>
4208         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
4210 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4212         PR tree-optimization/87288
4213         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
4214         into account when determining PEELING_FOR_NITERS.
4216 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4218         PR tree-optimization/86877
4219         * tree-vect-loop.c (vect_analyze_loop_2): Call
4220         vect_verify_datarefs_alignment.
4222 2018-09-19  Marek Polacek  <polacek@redhat.com>
4224         * doc/invoke.texi: Document -Wclass-conversion.
4226 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
4228         * config/pa/pa.c (pa_adjust_priority): Delete.
4229         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
4231         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
4232         (atomic_storehi): Likewise.
4233         (atomic_storesi): Likewise.
4234         (atomic_loaddi): Restore compare and swap exchange loop code.
4236 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
4238         PR rtl-optimization/86902
4239         * combine.c (try_combine): When changing the CC mode used, don't change
4240         an unrelated mode in other_insn to that new CC mode.
4242 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
4244         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
4245         with %T in place of calls to dump_generic_expr.
4246         (prune_runtime_alias_test_list): Likewise.
4247         (create_runtime_alias_checks): Likewise.
4248         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
4249         (vect_analyze_data_ref_dependence): Likewise.
4250         (vect_slp_analyze_data_ref_dependence): Likewise.
4251         (vect_record_base_alignment): Likewise.  Use %G in place of call
4252         to dump_gimple_stmt.
4253         (vect_compute_data_ref_alignment): Likewise.
4254         (verify_data_ref_alignment): Likewise.
4255         (vect_find_same_alignment_drs): Likewise.
4256         (vect_analyze_group_access_1): Likewise.
4257         (vect_analyze_data_ref_accesses): Likewise.
4258         (dependence_distance_ge_vf): Likewise.
4259         (dump_lower_bound): Likewise.
4260         (vect_prune_runtime_alias_test_list): Likewise.
4261         (vect_find_stmt_data_reference): Likewise.
4262         (vect_analyze_data_refs): Likewise.
4263         (vect_create_addr_base_for_vector_ref): Likewise.
4264         (vect_create_data_ref_ptr): Likewise.
4265         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
4266         (vect_can_advance_ivs_p): Likewise.
4267         (vect_update_ivs_after_vectorizer): Likewise.
4268         (vect_gen_prolog_loop_niters): Likewise.
4269         (vect_prepare_for_masked_peels): Likewise.
4270         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4271         (vect_determine_vectorization_factor): Likewise.
4272         (vect_is_simple_iv_evolution): Likewise.
4273         (vect_analyze_scalar_cycles_1): Likewise.
4274         (vect_analyze_loop_operations): Likewise.
4275         (report_vect_op): Likewise.
4276         (vect_is_slp_reduction): Likewise.
4277         (check_reduction_path): Likewise.
4278         (vect_is_simple_reduction): Likewise.
4279         (vect_create_epilog_for_reduction): Likewise.
4280         (vect_finalize_reduction:): Likewise.
4281         (vectorizable_induction): Likewise.
4282         (vect_transform_loop_stmt): Likewise.
4283         (vect_transform_loop): Likewise.
4284         (optimize_mask_stores): Likewise.
4285         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
4286         (vect_split_statement): Likewise.
4287         (vect_recog_over_widening_pattern): Likewise.
4288         (vect_recog_average_pattern): Likewise.
4289         (vect_determine_min_output_precision_1): Likewise.
4290         (vect_determine_precisions_from_range): Likewise.
4291         (vect_determine_precisions_from_users): Likewise.
4292         (vect_mark_pattern_stmts): Likewise.
4293         (vect_pattern_recog_1): Likewise.
4294         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4295         (vect_record_max_nunits): Likewise.
4296         (vect_build_slp_tree_1): Likewise.
4297         (vect_build_slp_tree_2): Likewise.
4298         (vect_print_slp_tree): Likewise.
4299         (vect_analyze_slp_instance): Likewise.
4300         (vect_detect_hybrid_slp_stmts): Likewise.
4301         (vect_detect_hybrid_slp_1): Likewise.
4302         (vect_slp_analyze_operations): Likewise.
4303         (vect_slp_analyze_bb_1): Likewise.
4304         (vect_transform_slp_perm_load): Likewise.
4305         (vect_schedule_slp_instance): Likewise.
4306         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4307         (vect_mark_stmts_to_be_vectorized): Likewise.
4308         (vect_init_vector_1): Likewise.
4309         (vect_get_vec_def_for_operand): Likewise.
4310         (vect_finish_stmt_generation_1): Likewise.
4311         (vect_check_load_store_mask): Likewise.
4312         (vectorizable_call): Likewise.
4313         (vectorizable_conversion): Likewise.
4314         (vectorizable_operation): Likewise.
4315         (vectorizable_load): Likewise.
4316         (vect_analyze_stmt): Likewise.
4317         (vect_is_simple_use): Likewise.
4318         (vect_get_vector_types_for_stmt): Likewise.
4319         (vect_get_mask_type_for_stmt): Likewise.
4320         * tree-vectorizer.c (increase_alignment): Likewise.
4322 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
4324         * doc/rtl.texi: Adjust vec_select description.
4325         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
4326         non-constant selectors.
4328 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
4330         * config/aarch64/aarch64-protos.h
4331         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
4332         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
4333         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
4334         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
4335         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
4336         (AARCH64_FL_PROFILE): Move index so flags are ordered.
4337         (AARCH64_ISA_RCPC8_4): New flag.
4338         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
4339         to aarch64_offset_9bit_signed_unscaled_p.
4340         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
4341         and use stlur.
4342         * config/aarch64/constraints.md (Ust): New constraint.
4343         * config/aarch64/predicates.md.
4344         (aarch64_9bit_offset_memory_operand): New predicate.
4345         (aarch64_rcpc_memory_operand): New predicate.
4347 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
4349         PR rtl-optimization/87361
4350         * rtlanal.c (nonzero_bits1): Revert accidental change.
4352 2018-09-19  Richard Biener  <rguenther@suse.de>
4354         PR tree-optimization/87349
4355         PR tree-optimization/87342
4356         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
4358 2018-09-18  Marek Polacek  <polacek@redhat.com>
4360         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
4361         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
4363 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4365         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
4367 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4369         PR rtl-optimization/86882
4370         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
4372 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
4374         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
4375         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
4377 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
4379         PR other/87353
4380         * doc/invoke.texi (Link Options): Fix formatting and grammar.
4382 2018-09-18  Richard Biener  <rguenther@suse.de>
4384         PR middle-end/63155
4385         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
4386         (compute_samebase_partition_bases): Likewise.
4387         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
4388         (gimple_can_coalesce_p): Simplify.
4390 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
4392         Handle a library implementation of ffs calling __builtin_ffs.
4393         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
4394         (mmix_init_libfuncs): New function: make __builtin_ffs expand
4395         to __ffsdi2.
4397 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4399         * diagnostic-show-locus.c (class layout_range): Add field
4400         "m_original_idx".
4401         (layout_range::layout_range): Add "original_idx" param and use it
4402         to initialize new field.
4403         (make_range): Use 0 for original_idx.
4404         (layout::layout): Pass in index to calls to
4405         maybe_add_location_range.
4406         (layout::maybe_add_location_range): Add param "original_idx" and
4407         pass it on to layout_range.
4408         (layout::print_any_labels): Pass on range->m_original_idx to
4409         get_text call.
4410         (gcc_rich_location::add_location_if_nearby): Use 0 for
4411         original_idx.
4412         * gcc-rich-location.h (text_range_label::get_text): Update for new
4413         param.
4414         (range_label_for_type_mismatch::get_text): Likewise.
4416 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4418         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4420 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4422         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4423         format_string_diagnostic_t.
4424         (fmtwarn_n): Likewise.
4425         * substring-locations.c
4426         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4427         (format_warning_n_va): Convert to...
4428         (format_string_diagnostic_t::emit_warning_n_va): ...this.
4429         (format_warning_va): Convert to...
4430         (format_string_diagnostic_t::emit_warning_va): ...this.
4431         (format_warning_at_substring): Convert to...
4432         (format_string_diagnostic_t::emit_warning): ...this.
4433         (format_warning_at_substring_n): Convert to...
4434         (format_string_diagnostic_t::emit_warning_n): ...this.
4435         * substring-locations.h (class format_string_diagnostic_t): New
4436         class.
4437         (format_warning_va): Convert to
4438         format_string_diagnostic_t::emit_warning_va.
4439         (format_warning_n_va): Convert to
4440         format_string_diagnostic_t::emit_warning_n_va.
4441         (format_warning_at_substring): Convert to
4442         format_string_diagnostic_t::emit_warning.
4443         (format_warning_at_substring_n): Convert to
4444         format_string_diagnostic_t::emit_warning_n.
4446 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
4447             Bernd Schmidt <bernds_cb1@t-online.de>
4449         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4450         SImode args.
4452 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4454         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4455         operand 0 predicate to nonimmediate operand.
4456         (rint<mode>2_frndint): Remove insn pattern.
4457         (rint<mode>2): Change operand 1 predicate to general_operand.
4458         Extend operand 1 to XFmode and generate rintxf2 insn.
4459         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4460         Do not use X87MODEF mode macro.
4461         (frndintxf2_<rounding>_i387): Rename from
4462         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
4463         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4464         to XFmode and generate significandxf3 insn.
4466 2018-09-17  Richard Biener  <rguenther@suse.de>
4468         PR tree-optimization/87328
4469         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4470         visiting unexecutable backedges when not iterating.
4471         (do_rpo_vn): Mark all edges not executable even when not
4472         iterating.
4474 2018-09-17  Martin Jambor  <mjambor@suse.cz>
4476         PR c/63886
4477         * doc/invoke.texi (Warning Options): Likewise.
4479 2018-09-17  Richard Biener  <rguenther@suse.de>
4481         PR tree-optimization/87301
4482         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4483         clean EH info from leftover copy assignments.
4485 2018-09-17  Martin Liska  <mliska@suse.cz>
4487         PR gcov-profile/85871
4488         * gcov.c (output_intermediate_file): Fix out of bounds
4489         access.
4491 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
4493         * config/arc/arc.c: Object attributes for core4 not reflected
4494         correctly.
4495         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4496         core3).
4498 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
4500         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4502 2018-09-17  Martin Liska  <mliska@suse.cz>
4504         * doc/gcov.texi: Document new option --use-hotness-colors.
4505         * gcov.c (struct source_info): Declare new field.
4506         (source_info::source_info): Set default for maximum_count.
4507         (print_usage): Add new -q option.
4508         (process_args): Process it.
4509         (accumulate_line_info): Save src->maximum_count.
4510         (output_line_beginning): Make color line number if
4511         flag_use_hotness_colors is set.
4512         (output_line_details): Pass default argument value.
4513         (output_lines): Pass src->maximum_count.
4515 2018-09-17  Martin Liska  <mliska@suse.cz>
4517         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4518         Use processor_names table.
4519         * config/i386/i386.c (ix86_default_align): Use
4520         processor_cost_table for alignment values.
4521         (ix86_option_override_internal): Use processor_names.
4522         (ix86_function_specific_print): Likewise.
4523         * config/i386/i386.h (struct processor_costs):
4524         Add alignment values.
4525         (struct ptt): Remove and replace with const char *.
4526         * config/i386/x86-tune-costs.h (struct processor_costs):
4527         Declare default alignments for all costs.
4529 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
4531         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4532         symbolics or VR_VARYING ranges for ABS_EXPR.
4533         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4534         when range will wrap.
4536 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
4538         PR middle-end/86864
4539         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4540         before and after a JUMP_TABLE_DATA.
4542 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
4544         PR middle-end/87188
4545         * dojump.c (do_compare_and_jump): Canonicalize function pointers
4546         when one operand is a function pointer.  Use POINTER_TYPE_P and
4547         FUNC_OR_METHOD_TYPE_P.
4548         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4549         * fold-const.c (build_range_check): Likewise.
4550         * match.pd (simple_comparison): Likewise.
4552 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
4554         PR c/82967
4555         * spellcheck.c (get_edit_distance_cutoff): New function.
4556         (selftest::test_edit_distance_unit_test_oneway): Rename to...
4557         (selftest::test_get_edit_distance_one_way): ...this.
4558         (selftest::test_get_edit_distance_unit): Rename to...
4559         (selftest::test_get_edit_distance_both_ways): ...this.
4560         (selftest::test_edit_distances): Move tests to this new function,
4561         and test some more pairs of strings.  Update for above renaming.
4562         (selftest::get_old_cutoff): New function.
4563         (selftest::test_get_edit_distance_cutoff): New function.
4564         (selftest::assert_suggested_for): New function.
4565         (ASSERT_SUGGESTED_FOR): New macro.
4566         (selftest::assert_not_suggested_for): New function.
4567         (ASSERT_NOT_SUGGESTED_FOR): New macro.
4568         (selftest::test_suggestions): New function.
4569         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4570         tests to selftest::test_edit_distances and call it.  Add calls to
4571         selftest::test_get_edit_distance_cutoff and
4572         selftest::test_suggestions.
4573         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4574         (best_match::consider): Replace hard-coded cutoff calculation with
4575         a call to...
4576         (best_match::get_cutoff): New declaration.
4577         (best_match::get_best_meaningful_candidate): Likewise.
4579 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4581         * builtins.c (fold_builtin_strlen): Remove TODO comment.
4583 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4585         revert:
4586         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4588         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4589         terminated string literal.
4591 2018-09-14  Martin Sebor  <msebor@redhat.com>
4593         * builtins.c (unterminated_array): Handle ARRAY_REF.
4594         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4595         * builtins.h (unterminated_array): Declare extern.
4596         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4597         arrays.
4598         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4599         calls.
4601 2018-09-14  Martin Sebor  <msebor@redhat.com>
4602             Jeff Law  <law@redhat.com>
4604         * builtins.c (unterminated_array): New.
4605         (expand_builtin_strcpy): Adjust.
4606         (expand_builtin_strcpy_args): Detect unterminated arrays.
4607         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
4608         unterminated arrays.
4609         * gimple-fold.h (get_maxval_strlen): Add argument.
4610         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4612         * gimple-fold.c (get_range_strlen): Add argument.
4613         (get_maxval_strlen): Adjust.
4614         * gimple-fold.h (get_range_strlen): Add argument.
4616 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
4618         * config/i386/movdirintrin.h: Fix copyright year.
4620 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
4622         * reg-stack.c: Include regs.h.
4623         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4624         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4625         FIRST_STACK_REG, not DFmode.
4626         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4627         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
4628         (change stack): Default register mode to the reg_raw_mode of
4629         FIRST_STACK_REG, not DFmode.
4630         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4631         (*swapxf): Rename from swapxf.
4633 2018-09-14  Carl Love  <cel@us.ibm.com>
4635         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4636         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4638 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
4640         PR target/87224
4641         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4642         alternatives.
4644 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
4646         PR target/85628
4647         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4649 2018-09-14  Jason Merrill  <jason@redhat.com>
4651         Fix --enable-gather-detailed-mem-stats.
4652         * hash-table.c (hash_table_usage): Change from variable to function.
4653         * hash-table.h: Adjust.
4654         * Makefile.in: Add missing dependencies on hash-table.h.
4656 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4658         PR tree-optimization/87259
4659         PR lto/87283
4660         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4661         execute_cse_reciprocals_1 has tried transforming.
4663 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
4665         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4666         VR_VARYING for PLUS/MINUS_EXPR.
4668 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4670         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4671         formatting.
4673 2018-09-14  Richard Biener  <rguenther@suse.de>
4675         PR middle-end/63155
4676         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4677         bits for the merged partition.
4679 2018-09-13  Martin Sebor  <msebor@redhat.com>
4680             Bernd Edlinger  <bernd.edlinger@hotmail.de>
4682         * builtins.h (c_srlen): Add argument.
4683         * builtins.c (warn_string_no_nul): New function.
4684         (c_strlen): Add argument and use it.  Update recursive calls.
4685         Pass DECL argument to string_constant to get info on non
4686         terminated strings.  Update *NONSTR as needed.
4687         (fold_builtin_strlen): Add argument to calls to c_strlen.
4688         Warn for unterminated arrays.
4689         (warn_string_no_null): Add prototype.
4690         * expr.c (string_constant): Update arguments.  Update recursive
4691         calls appropriately.  Detect missing NUL terminator and outermost
4692         declaration its missing in.
4693         Improve checks for arrays with nonzero lower bound or elements
4694         that are not a single byte.  Simplify offset computation.
4695         Simplify checks for non-NUL terminated strings.
4696         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4697         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4699 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4701         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4702         correctly.
4703         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
4704         argument.  Fix range checks.
4705         * fold-const.h (c_getstr): Adjust protoype.
4706         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4707         string is constant but contains no NUL byte.
4709         * expr.c (string_constant): Adjust function comment.
4710         Remove bogus check for zero termination.
4712         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4714         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4715         (get_constant_size): Don't make STRING_CSTs larger than they are.
4716         (check_string_literal): New check function for STRING_CSTs.
4717         (output_constant): Use it.
4719 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
4721         PR target/86812
4722         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4724 2018-09-13  Richard Biener  <rguenther@suse.de>
4726         PR tree-optimization/87263
4727         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4728         (struct unwind_state): Add max_rpo field.
4729         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4730         Compute max_rpo, the max RPO number a block can be backwards reached
4731         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
4732         separating it from the iterating mode.
4734 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4736         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4737         (rfs_decision): New scheduling decision.
4739 2018-09-13  Richard Biener  <rguenther@suse.de>
4741         PR bootstrap/87134
4742         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4743         (vn_nary_op_insert_pieces_predicated): Do not write useless
4744         valid_dominated_by_p entry outside of the allocated storage.
4746 2018-09-13  Omar Sandoval  <osandov@osandov.com>
4747             Tom de Vries  <tdevries@suse.de>
4749         PR debug/86985
4750         * dwarf2out.c (is_c): New function.
4751         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4753 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
4755         PR target/85628
4756         * config/aarch64/aarch64.md (*aarch64_bfxil):
4757         Define.
4758         * config/aarch64/constraints.md (Ulc): Define.
4759         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4760         Define.
4761         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4762         New function.
4764 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4766         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4767         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4768         aarch64_layout_frame call.
4769         (aarch64_expand_epilogue): Likewise.
4770         (aarch64_initial_elimination_offset): Likewise.
4771         (aarch64_get_separate_components): Likewise.
4772         (aarch64_use_return_insn_p): Likewise.
4773         (aarch64_layout_frame): Remove unneeded check.
4775 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4777         * configure.ac: Only append
4778         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
4779         gcc_config_arguments if it was never reconfigured or last reconfigure
4780         was with different arguments.
4781         * configure: Regenerated.
4783 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4784             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4786         PR middle-end/87290
4787         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
4788         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
4790 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4792         PR tree-optimization/87287
4793         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
4794         X % C == 0 to X % (unsigned) C == 0 optimization to ...
4795         * match.pd (X % C == 0): ... here.  New optimization.
4797 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4799         PR middle-end/82853
4800         * expr.h (maybe_optimize_mod_cmp): Declare.
4801         * expr.c (mod_inv): New function.
4802         (maybe_optimize_mod_cmp): New function.
4803         (do_store_flag): Use it.
4804         * cfgexpand.c (expand_gimple_cond): Likewise.
4806 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
4807             Julian Brown  <julian@codesourcery.com>
4809         PR middle-end/86336
4810         * gimplify.c (gimplify_scan_omp_clauses): Set
4811         target_firstprivatize_array_bases in OpenACC parallel and kernels
4812         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
4813         OpenACC data regions.
4815 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
4817         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
4818         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
4820 2018-09-12  Richard Biener  <rguenther@suse.de>
4822         PR tree-optimization/87280
4823         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
4824         edge but unreachable target.
4825         (do_rpo_vn): For conservatively handling a PHI only mark
4826         the backedge executable but not the block reachable.
4828 2018-09-12  Richard Biener  <rguenther@suse.de>
4830         PR tree-optimization/87266
4831         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
4832         visited blocks.
4834 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
4836         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
4837         constants.
4838         ("trunc<BFP:mode><DFP_ALL:mode>2")
4839         ("trunc<DFP_ALL:mode><BFP:mode>2")
4840         ("extend<BFP:mode><DFP_ALL:mode>2")
4841         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
4842         according to the target operand type.
4844 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4845             Andreas Krebbel  <krebbel@linux.ibm.com>
4847         PR tree-optimization/86844
4848         * gimple-ssa-store-merging.c
4849         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
4850         there are any overlapping stores in between them, make sure they are
4851         also coalesced or we give up completely.
4853 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4855         PR middle-end/87248
4856         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
4857         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
4859 2018-09-12  Tom de Vries  <tdevries@suse.de>
4861         * common.opt (gdescribe-dies): Add option.
4862         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
4863         attribute for artifical and nameless decls.
4864         (dwarf2out_register_external_die): Add description attribute to
4865         external reference die.
4866         (add_desc_attribute): New functions.
4867         (gen_subprogram_die): Add description attribute to
4868         DW_TAG_call_site_parameter.
4869         * tree-pretty-print.c (print_generic_expr_to_str): New function.
4870         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
4871         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
4872         -gno-describe-dies.
4873         (@item -gdescribe-dies): Add.
4875 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4877         * tree-vrp.c (vrp_shift_undefined_p): Remove.
4878         (extract_range_from_binary_expr_1: Call
4879         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
4880         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
4881         depend on sign.
4883 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4885         * gimple-ssa-warn-alloca.c
4886         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
4887         field for ALLOCA_BOUND_*_LARGE.
4889 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4891         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
4893 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4895         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
4896         for clobbers.  Remove obsolete comment.
4898 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4900         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
4901         mpxchk, mpxld and mpxst types.
4902         (define_attr length_immediate): Remove all processing of mpx types.
4903         (define_attr prefix_0f): Ditto.
4904         (define_attr memory): Ditto.
4906 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4908         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
4909         (log<mode>2): Change operand 1 predicate to general_operand.
4910         Extend operand 1 to XFmode and generate logxf3 insn.
4911         (log10<mode>2): Change operand 1 predicate to general_operand.
4912         Extend operand 1 to XFmode and generate log10xf3 insn.
4913         (log2<mode>2): Change operand 1 predicate to general_operand.
4914         Extend operand 1 to XFmode and generate log2xf3 insn.
4915         (fyl2xp1_extend<mode>xf3_i387): Remove.
4916         (log1p<mode>2): Change operand 1 predicate to general_operand.
4917         Extend operand 1 to XFmode and generate log1pxf3 insn.
4918         (fxtract_extend<mode>xf3_i387): Remove.
4919         (logb<mode>2): Change operand 1 predicate to general_operand.
4920         Extend operand 1 to XFmode and generate logbxf3 insn.
4921         (ilogb<mode>2): Change operand 1 predicate to general_operand.
4922         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
4923         (significand<mode>2): Change operand 1 predicate to general_operand.
4924         Extend operand 1 to XFmode and generate significandxf3 insn.
4926 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4928         * gcc.c (perror_with_name, pfatal_with_name): Delete.
4929         (load_specs): Use fatal_error.
4930         (DELETE_IF_ORDINARY, process_command): Use error.
4931         (execute, run_attempt): Use fatal_error.
4933 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
4935         * diagnostic-core.h (sorry_at): New prototype.
4936         * diagnostic.c (sorry_at): New function.
4938 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
4940         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
4941         by zero as VR_UNDEFINED.
4943 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4945         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
4946         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
4947         (<sincos>mode2): New expander.
4948         (sincos_extend<mode>xf3_i387): Remove insn pattern.
4949         (sincos -> sin, cos splitters): Remove splitter patterns.
4950         (sincos<mode>3): Change operand 2 predicate to general_operand.
4951         Extend operand 2 to XFmode and generate sincosxf3 insn.
4952         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
4953         Change operand 3 predicate to const1_operand.
4954         (fptan_extend<mode>xf4_i387): Remove insn pattern.
4955         (tanxf2): Update operands in the call to fptanxf4_i387.
4956         (tan<mode>2): Change operand 1 predicate to general_operand.
4957         Extend operand 1 to XFmode and generate tanxf3 insn.
4958         (atan2xf3): Rename from *fpatanxf3_i387.
4959         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
4960         (atan2xf3): Remove expander.
4961         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
4962         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
4963         (atan<mode>2): Change operand 1 predicate to general_operand.
4964         Extend operand 1 to XFmode and generate atanxf3 insn.
4966 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4968         * config/i386/i386.md (x87/SSE constant load splitter): Use
4969         memory_operand instead of nonimmediate_operand for input operand
4970         predicate.
4972 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
4974         * config/i386/i386.md (float partial SSE register stall splitter): Move
4975         splitter near its instruction pattern.
4976         (float_extend partial SSE register stall splitter): Ditto.
4977         (float_truncate partial SSE register stall splitter): Ditto.
4979 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
4981         PR target/86794
4982         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4983         to speculation_safe_value_not_needed.
4985         PR target/85666
4986         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
4987         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
4988         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
4989         leaf_function_p, instead use has_hard_reg_initial_val.
4991 2018-09-09  Nathan Sidwell  <nathan@acm.org>
4993         * gcc.h (pfatal_with_name): Don't declare here.
4994         * gcc.c (pfatal_with_name): Make static.
4996 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4998         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
4999         earlyclobber.
5001 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
5003         PR rtl-optimization/85458
5004         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
5005         priority hook to reduce the priority of EXPR.
5007 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
5009         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
5010         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
5011         with X87_ENABLE_FLOAT.  Remove preparation code.
5012         (*float<SWI48:mode><MODEF:mode>2): Rename from
5013         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
5014         math using "enabled" attribute.
5015         (*floatdi<MODEF:mode>2_i387): Rename from
5016         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
5017         enable for 32bit targets only.
5018         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
5019         splitter.
5020         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
5021         as operand 1 predicate.  Rewrite as define_insn_and_split.
5022         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
5024 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5026         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
5027         to fallthru to FLOAT case.
5029 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5031         PR target/86731
5032         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
5033         around folding of vec_sl to handle out of range shift values.
5035 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5037         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
5038         Update callers to gen_fix_trunc<mode>_i387_fisttp
5039         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
5040         nonimmediate_operand.
5041         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
5042         and corresponding splitters.
5043         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
5044         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
5045         (fix_truncdi_i387_with_temp): Remove insn pattern
5046         and corresponding splitters.
5047         (fix_trunc<mode>_i387): Change operand 0 predicate to
5048         nonimmediate_operand.
5049         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
5050         and corresponding splitters.
5051         (*fistdi2_1): Remove.
5052         (fistdi2): Ditto.
5053         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
5054         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
5055         (*fist<mode>2_1): Remove.
5056         (fist<mode>2): Ditto.
5057         (fist<mode>2_with_temp): Remove insn pattern and corresponding
5058         splitters.
5059         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
5060         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
5061         (fistdi2_<rounding>): Change operand 0 predicate to
5062         nonimmediate_operand.
5063         (fistdi2_<rounding>_with_temp): Remove insn pattern
5064         and corresponding splitters.
5065         (fist<mode>2_<rounding>): Change operand 0 predicate to
5066         nonimmediate_operand.
5067         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
5068         and corresponding splitters.
5070         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
5072 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5074         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
5075         the init value.
5077 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5079         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
5080         early gimple folding of vec_splat().
5081         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
5082         * gimple-fold.h: Add an extern define for tree_vec_extract().
5084 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5086         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
5087         wrappers around TREE_TYPE comparisons.
5089 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5091         PR target/80080
5092         * config/s390/predicates.md: Add nonsym_memory_operand.
5093         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
5094         contains a SYMBOL_REF, load it into an intermediate pseudo.
5095         (s390_emit_compare_and_swap): Legitimize operand.
5096         * config/s390/s390.md: Use the new nonsym_memory_operand
5097         with UNSPECV_CAS patterns.
5099 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5101         PR target/80080
5102         * config/s390/s390-passes.def: New file.
5103         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
5104         declaration.
5105         (make_pass_s390_early_mach): Add declaration.
5106         * config/s390/s390.c (make_pass_s390_early_mach):
5107         (s390_option_override): Remove dynamic registration.
5108         * config/s390/t-s390: Add s390-passes.def.
5110 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5112         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5113         Remove UNSPEC_LTREL_BASE check.
5114         (annotate_constant_pool_refs): Likewise.
5115         (find_constant_pool_ref): Likewise.
5116         (find_ltrel_base): Removed.
5117         (replace_ltrel_base): Removed.
5118         (s390_mainpool_finish): Remove replace_ltrel_base call.
5119         (s390_chunkify_start): Remove pending LTREL_BASE logic.
5120         (s390_chunkify_finish): Remove replace_ltrel_base call.
5121         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
5123 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
5125         PR target/86779
5126         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5127         to speculation_safe_value_not_needed.
5129 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
5130             Bernd Schmidt  <bernds_cb1@t-online.de>
5132         * config/nvptx/nvptx-opts.h: New file.
5133         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
5134         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
5135         (ASM_SPEC): Define.
5136         (TARGET_SM35): New macro.
5137         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
5138         correct predicate.
5139         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
5140         values.
5141         (misa=): New option.
5142         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
5144 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
5146         * config/i386/i386.md (truncdfsf2): Remove expander.
5147         (truncdfsf2_with_temp): Ditto.
5148         (truncxf<mode>2): Ditto.
5149         (*truncdfsf_fast_mixed): Remove insn pattern.
5150         (*truncdfsf_fast_i387): Ditto.
5151         (*truncdfsf_mixed): Ditto.
5152         (*truncdfsf_i387): Ditto.
5153         (*truncdfsf2_i387_1): Ditto.
5154         (*truncxfsf2_mixed): Ditto.
5155         (*truncxfdf2_mixed): Ditto.
5156         (*truncxf<mode>2_i387_noop): Ditto. Update callers
5157         to call gen_truncxf<mode>2 instead.
5158         (*truncxf<mode>2_i387): Remove.
5159         (reg->reg splitters): Remove splitter pattern.
5160         (reg->mem splitters): Ditto.
5162         (truncdfsf2): New insn pattern.
5163         (truncxf<mode>2): Ditto.
5165 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5167         * tree-ssa-math-opts.c (is_mult_by): New function.
5168         (is_square_of): Use the above.
5169         (optimize_recip_sqrt): New function.
5170         (pass_cse_reciprocals::execute): Use the above.
5172 2018-09-05  Richard Biener  <rguenther@suse.de>
5174         PR bootstrap/87134
5175         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
5176         to zero-init the emplaced vec.
5178 2018-09-05  Martin Liska  <mliska@suse.cz>
5180         PR tree-optimization/87205
5181         * tree-switch-conversion.c (pass_lower_switch::execute):
5182         Group cases for switch statements.
5184 2018-09-05  Richard Biener  <rguenther@suse.de>
5186         PR tree-optimization/87217
5187         * tree-ssa-sccvn.c (vuse_valueize): New.
5188         (vn_reference_lookup_pieces): Use it.
5189         (vn_reference_lookup): Likewise.
5191 2018-09-05  Nathan Sidwell  <nathan@acm.org>
5193         PR c++/87137
5194         * stor-layout.c (place_field): Scan forwards to check last
5195         bitfield when ms_bitfield_placement is in effect.
5197 2018-09-05  Richard Biener  <rguenther@suse.de>
5199         PR bootstrap/87225
5200         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
5201         return.
5203 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5204             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
5206         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
5207         * config.gcc (extra_objs): Build it.
5208         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
5209         Likewise.
5210         * config/aarch64/aarch64-passes.def
5211         (pass_tag_collision_avoidance): New pass.
5212         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
5213         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
5214         (aarch64_classify_address): Remove static qualifier.
5215         (aarch64_address_info, aarch64_address_type): Move to...
5216         * config/aarch64/aarch64-protos.h: ... here.
5217         (make_pass_tag_collision_avoidance): New function.
5218         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
5219         New tuning flag.
5221 2018-09-05  Martin Liska  <mliska@suse.cz>
5223         * doc/gcov.texi: Update documentation of humar
5224         readable mode.
5225         * gcov.c (format_count): Print one decimal place, it provides
5226         more fine number of situations like '1G' vs. '1.4G'.
5228 2018-09-05  Martin Liska  <mliska@suse.cz>
5230         PR target/87164
5231         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
5232         * optc-gen.awk: Allow 'Var' for Deprecated options in order
5233         to generate a MASK value.
5235 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
5237         PR debug/86593
5238         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
5239         if frame pointer isn't used.
5240         (compute_frame_pointer_to_fb_displacement): Likewise.
5242 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
5244         PR target/87198
5245         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
5246         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
5247         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
5248         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
5249         and OPTION_MASK_ISA_XSAVEC_UNSET.
5251 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
5253         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
5254         XOR operations in NAND case.
5256 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5258         * wide-int-range.cc (wide_int_range_convert): New.
5259         * wide-int-range.h (wide_int_range_convert): New.
5260         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
5261         code into wide_int_range_convert.
5262         (extract_range_into_wide_ints): Do not munge anti range constants
5263         into the entire domain.  Just return the range back.
5265 2018-09-04  Martin Liska  <mliska@suse.cz>
5267         * genmatch.c (output_line_directive): Add new argument
5268         fnargs.
5269         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
5271 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5273         * doc/invoke.texi (Option Summary): Add whitespace.
5275         * doc/invoke.texi (Option Summary): Add -Waligned-new.
5277 2018-09-04  Richard Biener  <rguenther@suse.de>
5279         PR tree-optimization/87211
5280         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
5281         backedge value we're supposed to treat as VARYING also number
5282         the PHI to VARYING in case it got a different value-number already.
5284 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5286         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
5287         (extract_range_from_binary_expr_1): Do not call
5288         vrp_can_optimize_bit_op.
5289         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
5290         static.
5291         (wide_int_range_get_mask_and_bounds): New.
5292         (wide_int_range_optimize_bit_op): New.
5293         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
5294         (wide_int_range_bit_and): Same.
5295         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
5296         (wide_int_range_optimize_bit_op): New.
5297         (wide_int_range_get_mask_and_bounds): New.
5299 2018-09-04  Richard Biener  <rguenther@suse.de>
5301         PR tree-optimization/87176
5302         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
5303         variable.  When value-numbering a virtual PHI node make sure
5304         to not value-number to the backedge value.
5306 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5308         * doc/extend.texi (Long Long, Hex Floats): Document support for
5309         long long and hex floats in more recent versions of ISO C++.
5311 2018-09-03  Richard Biener  <rguenther@suse.de>
5313         PR tree-optimization/87177
5314         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
5315         cleanup.
5317 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5319         * bb-reorder.c (edge_order): Convert to C-qsort-style
5320         tri-state comparator.
5321         (reorder_basic_blocks_simple): Change std::stable_sort to
5322         gcc_stablesort.
5324 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5326         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
5327         tri-state comparator.
5328         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
5330 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5332         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
5333         (mergesort): ... here as maximum count for using netsort.
5334         (gcc_qsort): Set nlim to 3 if stable sort is requested.
5335         (gcc_stablesort): New.
5336         * system.h (gcc_stablesort): Declare.
5338 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5340         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
5341         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
5342         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
5344 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
5346         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
5347         lxsdx and stxsdx alternatives.
5348         (*mov<mode>_hardfloat64): Ditto.
5349         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
5351 2018-09-03  Richard Biener  <rguenther@suse.de>
5353         PR tree-optimization/87200
5354         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
5355         simplify result.
5357 2018-09-03  Martin Liska  <mliska@suse.cz>
5359         PR tree-optimization/87201
5360         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5361         Fix parenthesis in an expression.
5363 2018-09-03  Richard Biener  <rguenther@suse.de>
5365         PR tree-optimization/87197
5366         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
5367         visited.  CSE the VN_INFO hashtable lookup.
5369         PR tree-optimization/87169
5370         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
5371         iterating make sure there's no extra backedges from irreducible
5372         regions feeding the header.  Mark the destination block
5373         executable.
5375 2018-09-03  Martin Liska  <mliska@suse.cz>
5377         PR driver/83193
5378         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
5379         * common/common-targhooks.c (default_get_valid_option_values):
5380         New function.
5381         * common/common-targhooks.h (default_get_valid_option_values):
5382         Likewise.
5383         * common/config/i386/i386-common.c: Move processor_target_table
5384         from i386.c.
5385         (ix86_get_valid_option_values): New function.
5386         (TARGET_GET_VALID_OPTION_VALUES): New macro.
5387         * config/i386/i386.c (struct ptt): Move to i386-common.c.
5388         (PTA_*): Move all defined masks into i386-common.c.
5389         (ix86_function_specific_restore): Use new processor_cost_table.
5390         * config/i386/i386.h (struct ptt): Moved from i386.c.
5391         (struct pta): Likewise.
5392         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
5393         * doc/tm.texi.in: Likewise.
5394         * opt-suggestions.c (option_proposer::suggest_option):
5395         Pass prefix to build_option_suggestions.
5396         (option_proposer::get_completions): Likewise.
5397         (option_proposer::build_option_suggestions): Use the new target
5398         hook.
5399         * opts.c (struct option_help_tuple): New struct.
5400         (print_filtered_help): Use the new target hook.
5402 2018-09-03  Martin Liska  <mliska@suse.cz>
5404         PR middle-end/59521
5405         * predict.c (set_even_probabilities): Add likely_edges
5406         argument and handle cases where we have precisely one
5407         likely edge.
5408         (combine_predictions_for_bb): Catch also likely_edges.
5409         (tree_predict_by_opcode): Handle gswitch statements.
5410         * tree-cfg.h (find_case_label_for_value): New declaration.
5411         (find_taken_edge_switch_expr): Likewise.
5412         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5413         Find pivot in decision tree based on probabily, not by number of
5414         nodes.
5416 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
5418         * doc/standards.texi (Standards): Update Objective-C reference.
5420 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5422         * doc/install.texi (Prerequisites): Update link for MPC.
5424 2018-09-01  Michael Matz  <matz@suse.de>
5426         PR tree-optimization/87074
5427         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5428         PHIs for outer-loop uses.
5430 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5432         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5433         * doc/invoke.texi (C Dialect Options): Ditto.
5435 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5437         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5439 2018-08-31  Richard Biener  <rguenther@suse.de>
5441         PR tree-optimization/87168
5442         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5443         (rpo_elim::eliminate_avail): When OP was not visited it must
5444         be available.
5446 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
5448         * tree-vrp.c (copy_value_range): Convert param "from" from
5449         "value_range *" to "const value_range *".
5450         (range_is_null): Likewise for param "vr".
5451         (range_int_cst_p): Likewise.
5452         (range_int_cst_singleton_p): Likewise.
5453         (symbolic_range_p): Likewise.
5454         (value_ranges_intersect_p): Likewise for both params.
5455         (value_range_nonnegative_p): Likewise for param "vr".
5456         (value_range_constant_singleton): Likewise.
5457         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5458         (extract_range_into_wide_ints): Likewise for param "vr".
5459         (extract_range_from_multiplicative_op): Likewise for params "vr0"
5460         and "vr1".
5461         (vrp_can_optimize_bit_op): Likewise.
5462         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5463         "vr1_".
5464         (extract_range_from_unary_expr): Likewise.
5465         (debug_value_range): Likewise for param "vr".
5466         (value_range::dump): Add "const" qualifier.
5467         (vrp_prop::check_array_ref): Convert local "vr" from
5468         "value_range *" to "const value_range *".
5469         (vrp_prop::check_mem_ref): Likewise.
5470         (vrp_prop::visit_stmt): Likewise for local "old_vr".
5471         (vrp_intersect_ranges_1): Likewise for param "vr_1".
5472         (vrp_intersect_ranges): Likewise.
5473         (simplify_stmt_for_jump_threading): Likewise for local "vr".
5474         (vrp_prop::vrp_finalize): Likewise.
5475         * tree-vrp.h (value_range::dump): Add "const" qualifier.
5476         (vrp_intersect_ranges): Add "const" qualifier to params as above.
5477         (extract_range_from_unary_expr): Likewise.
5478         (value_range_constant_singleton): Likewise.
5479         (symbolic_range_p): Likewise.
5480         (copy_value_range): Likewise.
5481         (extract_range_from_binary_expr_1): Likewise.
5482         (range_int_cst_p): Likewise.
5483         (vrp_set_zero_nonzero_bits): Likewise.
5484         (range_int_cst_singleton_p): Likewise.
5486 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
5488         * config/aarch64/arm_neon.h (vabsd_s64): New.
5489         (vnegd_s64): Likewise.
5491 2018-08-31  Martin Jambor  <mjambor@suse.cz>
5493         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5495 2018-08-31  Martin Liska  <mliska@suse.cz>
5497         * ipa-icf.c (sem_item::add_type): Use
5498         sem_item::m_type_hash_cache.
5499         * ipa-icf.h: Move the cache from sem_item_optimizer
5500         to sem_item.
5502 2018-08-31  Nathan Sidwell  <nathan@acm.org>
5504         * doc/extend.texi (Backwards Compatibility): Remove implicit
5505         extern C leeway of () being (...).
5507 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5509         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5511 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
5513         PR target/86684
5514         PR target/87149
5515         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5517 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
5519         PR middle-end/87138
5520         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5521         gen_int_mode.  Formatting fixes.
5523 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
5525         * target.def (custom_function_descriptors): Improve documentation.
5526         * doc/tm.texi.in (Trampolines): Expand discussion of function
5527         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5528         beginning of the section.
5529         * doc/tm.texi: Regenerated.
5531 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
5533         * cfg.h (class auto_edge_flag): Spell out the template-id of the
5534         base class in the initializer list.  This is a workaround for
5535         building with older GCC.
5536         (class auto_bb_flag): Likewise.
5538 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5540         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5541         (altivec_vcmpequ<VI_char>_p): Remove star.
5542         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5543         vector load modes.
5544         (expand_strncmp_vec_sequence): New function.
5545         (emit_final_str_compare_vec): New function.
5546         (expand_strn_compare): Add support for vector strncmp.
5547         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5548         length specification to bytes.
5549         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5550         (vcmpnezb_p): New pattern.
5551         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5552         for option -mstring-compare-inline-limit.
5554 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
5556         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5557         (PTA_SKYLAKE): Add PTA_AES.
5558         (PTA_GOLDMONT): Likewise.
5560 2018-08-29  Jan Hubicka  <jh@suse.cz>
5562         PR lto/86517
5563         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5564         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5566 2018-08-29  Jan Hubicka  <jh@suse.cz>
5568         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5569         TYPE_STUB_DECL.
5570         (hash_tree): Do not visit TYPE_STUB_DECL.
5571         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5572         stream TYPE_STUB_DECL.
5573         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5574         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5575         after free_lang_data.
5576         (type_in_anonymous_namespace_p): Likewise.
5578 2018-08-29  Jan Hubicka  <jh@suse.cz>
5580         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5581         comment that it has to be even number.
5582         (class sreal): Change m_sig type to int32_t.
5583         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5584         int64_t for temporary calculations.
5585         (sreal_verify_basics): Drop one bit from minimum and maximum.
5587 2018-08-30  Richard Biener  <rguenther@suse.de>
5589         PR tree-optimization/87147
5590         * tree-ssa-sccvn.c (SSA_VISITED): New function.
5591         (visit_phi): When the degenerate result is from the backedge and
5592         we didn't visit its definition yet drop to VARYING.
5593         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5595 2018-08-29  Jan Hubicka  <jh@suse.cz>
5597         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5598         DECL_VINDEX.
5599         (hash_tree): Likewise.
5601 2018-08-29  Jan Hubicka  <jh@suse.cz>
5603         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5604         and TYPE_NEXT_REF_TO.
5606 2018-08-29  Jan Hubicka  <jh@suse.cz>
5608         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5609         comment that it has to be even number.
5610         (class sreal): Change m_sig type to int32_t.
5611         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5612         int64_t for temporary calculations.
5613         (sreal_verify_basics): Drop one bit from minimum and maximum.
5615 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
5617         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5619 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
5621         PR middle-end/86995
5622         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5623         if to_add is negative.
5625 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5627         PR middle-end/87053
5628         * builtins.c (c_strlen): Improve range checks.
5630 2018-08-29  Martin Sebor  <msebor@redhat.com>
5631             Jeff Law  <law@redhat.com>
5633         PR tree-optimization/86714
5634         PR tree-optimization/86711
5635         * builtins.c (c_strlen): Add arguments to call to string_constant.
5636         * expr.c (string_constant): Add argument.  Detect missing nul
5637         terminator and outermost declaration it's missing in.
5638         * expr.h (string_constant): Add argument.
5639         * fold-const.c (read_from_constant_string): Add arguments to call to
5640         string_constant.
5641         (c_getstr): Likewise.
5642         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5643         to string_constant.
5644         * tree-ssa-strlen.c (get_stridx): Likewise.
5646 2018-08-29  Jan Hubicka  <jh@suse.cz>
5648         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5649         Do not stream DECL_VINDEX.
5650         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5651         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5652         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5654 2018-08-29  Richard Biener  <rguenther@suse.de>
5656         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5657         virtual operands that are not default defs to honor region
5658         boundaries.
5659         (rpo_vn_valueize): Remove ineffective code here.
5661 2018-08-29  Richard Biener  <rguenther@suse.de>
5663         PR tree-optimization/87132
5664         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5665         when skipping defs reachable over backedges.
5667 2018-08-29  Richard Biener  <rguenther@suse.de>
5669         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5670         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5671         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5672         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5673         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5674         (vn_reference_lookup_3): Remove use of const_parms.
5675         (free_rpo_vn): Do not free const_parms.
5676         (do_rpo_vn): Do not call init_const_parms.
5677         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5678         SSA_NAME_POINTS_TO_READONLY_MEMORY.
5679         (call_may_clobber_ref_p_1): Likewise.
5681 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
5683         PR other/86726
5684         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5685         (-O): Ditto.
5686         (-ftree-scev-cprop): Document.
5688 2018-08-29  Jan Hubicka  <jh@suse.cz>
5690         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5691         parameters.
5692         (sreal constructor): Update.
5693         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5694         sreal:operator/): Update.
5696 2018-08-29  Martin Liska  <mliska@suse.cz>
5698         * tree-switch-conversion.c (switch_conversion::expand):
5699         Strenghten assumption about gswitch statements.
5701 2018-08-29  Richard Biener  <rguenther@suse.de>
5703         PR tree-optimization/87117
5704         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5705         re-value-number released SSA VDEFs.
5707 2018-08-29  Richard Biener  <rguenther@suse.de>
5709         PR tree-optimization/87126
5710         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5712 2018-08-28  Jim Wilson  <jimw@sifive.com>
5714         * config/riscv/pic.md: Rewrite.
5715         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5716         invalid address.
5717         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5718         (SOFTF, default_load, softload, softstore): New.
5720 2018-08-28  Jeff Law  <law@redhat.com>
5722         * fold-const.c (fold_binary_loc): Remove recently added assert.
5724 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
5726         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5727         to OP parmeter of generated function.
5729 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
5731         PR tree-optimization/87009
5732         * match.pd: Add boolean optimizations.
5734 2018-08-28  Martin Sebor  <msebor@redhat.com>
5736         PR middle-end/86631
5737         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5738         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5739         (pass_walloca::gate): Use it.
5740         (alloca_call_type): Same.
5741         (pass_walloca::execute): Same.
5742         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5744 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
5746         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5747         GCC_VERSION for usage of "__gcc_dump_printf__" format from
5748         >= 3005 to >= 9000.
5750 2018-08-28  Richard Biener  <rguenther@suse.de>
5752         PR tree-optimization/87124
5753         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5754         constants before looking up avail.
5756 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
5758         PR middle-end/87099
5759         * calls.c (maybe_warn_nonstring_arg): Punt early if
5760         warn_stringop_overflow is zero.  Don't call get_range_strlen
5761         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5762         Swap comparison operands to have constants on rhs.  Only use
5763         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
5764         increment lenrng[0].
5766 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
5768         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5769         use of tree_to_shwi.  Remove duplicated test for the size being
5770         a whole number of bytes.
5772 2018-08-28  Richard Biener  <rguenther@suse.de>
5774         PR tree-optimization/87117
5775         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5776         Handle removed stmt without LHS (GIMPLE_NOP).
5778 2018-08-28  Richard Biener  <rguenther@suse.de>
5780         PR tree-optimization/87117
5781         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
5782         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
5784 2018-08-28  Richard Biener  <rguenther@suse.de>
5786         PR tree-optimization/87117
5787         * tree-ssa-pre.c (compute_avail): Do not make expressions
5788         with predicated values available.
5789         (get_expr_value_id): Assert we do not run into predicated value
5790         expressions.
5792 2018-08-28  Richard Biener  <rguenther@suse.de>
5794         PR tree-optimization/87117
5795         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
5796         get virtual operands.
5797         (get_expr_operands): Handle STRING_CST like other decls.
5799 2018-08-28  Martin Liska  <mliska@suse.cz>
5801         * tree.h: Update documentation of fndecl_built_in_p
5802         functions.
5805 2018-08-27  Jeff Law  <law@redhat.com>
5806         PR tree-optimization/87110
5807         * tree-ssa-dse.c (compute_trims): Handle non-constant
5808         TYPE_SIZE_UNIT.
5810 2018-08-27  Martin Sebor  <msebor@redhat.com>
5812         PR tree-optimization/86914
5813         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
5815 2018-08-27  Martin Sebor  <msebor@redhat.com>
5817         PR tree-optimization/87112
5818         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
5819         the type of the bound argument.
5821 2018-08-27  Jeff Law  <law@redhat.com>
5823         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
5824         type does not have a TYPE_SIZE_UNIT.
5826 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
5828         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
5829         with include of backend.h.
5831 2018-08-27  Richard Biener  <rguenther@suse.de>
5833         PR tree-optimization/86927
5834         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
5835         use const cond reduction code.
5837 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
5839         PR tree-optimization/85758
5840         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
5842 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5844         PR c++/87091
5845         * diagnostic-show-locus.c (class layout_range): Update for
5846         conversion of show_caret_p to a tri-state.
5847         (layout_range::layout_range): Likewise.
5848         (make_range): Likewise.
5849         (layout::maybe_add_location_range): Likewise.
5850         (layout::should_print_annotation_line_p): Don't show annotation
5851         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
5852         (layout::get_state_at_point): Update for conversion of
5853         show_caret_p to a tri-state.  Bail out early for
5854         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
5855         underlining or source colorization.
5856         (gcc_rich_location::add_location_if_nearby): Update for conversion
5857         of show_caret_p to a tri-state.
5858         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
5859         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
5860         Likewise.
5861         (selftest::test_one_liner_labels): Likewise.
5862         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
5863         conversion of show_caret_p to a tri-state.
5864         * pretty-print.c (text_info::set_location): Likewise.
5865         * pretty-print.h (text_info::set_location): Likewise.
5866         * substring-locations.c (format_warning_n_va): Likewise.
5867         * tree-diagnostic.c (default_tree_printer): Likewise.
5868         * tree-pretty-print.c (newline_and_indent): Likewise.
5870 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5872         PR c++/87091
5873         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
5874         line above for line-insertion fix-it hints.
5875         (selftest::test_fixit_insert_containing_newline): Update the
5876         expected results, and add a test with line-numbering enabled.
5878 2018-08-27  Martin Liska  <mliska@suse.cz>
5880         PR sanitizer/86962
5881         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
5882         params with DECL_HAS_VALUE_EXPR_P.
5884 2018-08-27  Martin Liska  <mliska@suse.cz>
5886         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
5887         selected expansion strategy.
5889 2018-08-27  Martin Liska  <mliska@suse.cz>
5891         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
5892         * builtins.c (is_builtin_fn): Likewise.
5893         * attribs.c (diag_attr_exclusions): Use new function
5894         fndecl_built_in_p and remove check for FUNCTION_DECL if
5895         possible.
5896         (builtin_mathfn_code): Likewise.
5897         (fold_builtin_expect): Likewise.
5898         (fold_call_expr): Likewise.
5899         (fold_builtin_call_array): Likewise.
5900         (fold_call_stmt): Likewise.
5901         (set_builtin_user_assembler_name): Likewise.
5902         (is_simple_builtin): Likewise.
5903         * calls.c (gimple_alloca_call_p): Likewise.
5904         (maybe_warn_nonstring_arg): Likewise.
5905         * cfgexpand.c (expand_call_stmt): Likewise.
5906         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
5907         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
5908         (cgraph_node::verify_node): Likewise.
5909         * cgraphclones.c (build_function_decl_skip_args): Likewise.
5910         (cgraph_node::create_clone): Likewise.
5911         * config/arm/arm.c (arm_insert_attributes): Likewise.
5912         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
5913         * dse.c (scan_insn): Likewise.
5914         * expr.c (expand_expr_real_1): Likewise.
5915         * fold-const.c (operand_equal_p): Likewise.
5916         (fold_binary_loc): Likewise.
5917         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
5918         * gimple-low.c (lower_stmt): Likewise.
5919         * gimple-pretty-print.c (dump_gimple_call): Likewise.
5920         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
5921         Likewise.
5922         * gimple.c (gimple_build_call_from_tree): Likewise.
5923         (gimple_call_builtin_p): Likewise.
5924         (gimple_call_combined_fn): Likewise.
5925         * gimplify.c (gimplify_call_expr): Likewise.
5926         (gimple_boolify): Likewise.
5927         (gimplify_modify_expr): Likewise.
5928         (gimplify_addr_expr): Likewise.
5929         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
5930         * ipa-cp.c (determine_versionability): Likewise.
5931         * ipa-fnsummary.c (compute_fn_summary): Likewise.
5932         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
5933         * ipa-split.c (visit_bb): Likewise.
5934         (split_function): Likewise.
5935         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5936         * lto-cgraph.c (input_node): Likewise.
5937         * lto-streamer-out.c (write_symbol): Likewise.
5938         * omp-low.c (setjmp_or_longjmp_p): Likewise.
5939         (lower_omp_1): Likewise.
5940         * predict.c (strip_predict_hints): Likewise.
5941         * print-tree.c (print_node): Likewise.
5942         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
5943         * trans-mem.c (is_tm_irrevocable): Likewise.
5944         (is_tm_load): Likewise.
5945         (is_tm_simple_load): Likewise.
5946         (is_tm_store): Likewise.
5947         (is_tm_simple_store): Likewise.
5948         (is_tm_abort): Likewise.
5949         (tm_region_init_1): Likewise.
5950         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
5951         * tree-cfg.c (verify_gimple_call): Likewise.
5952         (move_stmt_r): Likewise.
5953         (stmt_can_terminate_bb_p): Likewise.
5954         * tree-eh.c (lower_eh_constructs_2): Likewise.
5955         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
5956         * tree-inline.c (remap_gimple_stmt): Likewise.
5957         (copy_bb): Likewise.
5958         (estimate_num_insns): Likewise.
5959         (fold_marked_statements): Likewise.
5960         * tree-sra.c (scan_function): Likewise.
5961         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
5962         (optimize_stack_restore): Likewise.
5963         (pass_fold_builtins::execute): Likewise.
5964         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
5965         (mark_all_reaching_defs_necessary_1): Likewise.
5966         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
5967         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5968         (pass_forwprop::execute): Likewise.
5969         * tree-ssa-loop-im.c (stmt_cost): Likewise.
5970         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
5971         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
5972         * tree-ssa-strlen.c (get_string_length): Likewise.
5973         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
5974         (find_func_aliases_for_call): Likewise.
5975         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
5976         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
5977         * tree-tailcall.c (find_tail_calls): Likewise.
5978         * tree.c (need_assembler_name_p): Likewise.
5979         (free_lang_data_in_decl): Likewise.
5980         (get_call_combined_fn): Likewise.
5981         * ubsan.c (is_ubsan_builtin_p): Likewise.
5982         * varasm.c (incorporeal_function_p): Likewise.
5983         * tree.h (DECL_BUILT_IN): Remove and replace with
5984         fndecl_built_in_p.
5985         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
5986         (fndecl_built_in_p): New.
5988 2018-08-27  Martin Liska  <mliska@suse.cz>
5990         PR tree-optimization/86847
5991         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
5992         Dump also subtree probability.
5993         (switch_decision_tree::do_jump_if_equal): New function.
5994         (switch_decision_tree::emit_case_nodes): Handle special
5995         situations in balanced tree that can be emitted much simpler.
5996         Fix calculation of probabilities that happen in tree expansion.
5997         * tree-switch-conversion.h (struct cluster): Add
5998         is_single_value_p.
5999         (struct simple_cluster): Likewise.
6000         (struct case_tree_node): Add new function has_child.
6001         (do_jump_if_equal): New.
6003 2018-08-27  Martin Liska  <mliska@suse.cz>
6005         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
6006         Add new argument to bit_test_cluster constructor.
6007         (bit_test_cluster::emit): Set bits really number of values
6008         handlel by a test.
6009         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
6010         probability argument.
6011         * tree-switch-conversion.h (struct bit_test_cluster):
6012         Add m_handles_entire_switch.
6014 2018-08-27  Martin Liska  <mliska@suse.cz>
6016         PR tree-optimization/86702
6017         * tree-switch-conversion.c (jump_table_cluster::emit):
6018         Make probabilities even for values in jump table
6019         according to number of cases handled.
6020         (switch_decision_tree::compute_cases_per_edge): Pass
6021         argument to reset_out_edges_aux function.
6022         (switch_decision_tree::analyze_switch_statement): Likewise.
6023         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
6024         Make it static.
6026 2018-08-27  Martin Liska  <mliska@suse.cz>
6028         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
6029         cfun argument explicitly.
6030         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
6031         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
6032         function gimple_switch_default_bb.
6033         (convert_switch_statements):
6034         (expand_builtins):
6035         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
6036         * stmt.c (label_to_block_fn): Use label_to_block and pass
6037         cfun argument explicitly and use gimple_switch_label_bb.
6038         (expand_case): Likewise.
6039         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
6040         cfun argument explicitly. Likewise.
6041         (make_edges_bb): Likewise.
6042         (make_cond_expr_edges): Likewise.
6043         (get_cases_for_edge): Likewise.
6044         (make_gimple_switch_edges): Likewise.
6045         (label_to_block_fn): Likewise.
6046         (label_to_block): Likewise.
6047         (make_goto_expr_edges): Likewise.
6048         (make_gimple_asm_edges): Likewise.
6049         (main_block_label): Likewise.
6050         (group_case_labels_stmt): Likewise.
6051         (find_taken_edge_computed_goto): Likewise.
6052         (find_taken_edge_switch_expr): Likewise.
6053         (gimple_verify_flow_info): Likewise.
6054         (gimple_redirect_edge_and_branch): Likewise.
6055         (gimple_switch_label_bb): New function.
6056         (gimple_switch_default_bb): Likewise.
6057         (gimple_switch_edge): Likewise.
6058         (gimple_switch_default_edge): Likewise.
6059         * tree-cfg.h (label_to_block_fn): Remove and replace ...
6060         (label_to_block): ... with this.
6061         (gimple_switch_label_bb): New.
6062         (gimple_switch_default_bb): Likewise.
6063         (gimple_switch_edge): Likewise.
6064         (gimple_switch_default_edge): Likewise.
6065         * tree-cfgcleanup.c (convert_single_case_switch): Use
6066         new gimple functions and pass new argument to label_to_block.
6067         (cleanup_control_flow_bb):
6068         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
6069         cfun argument explicitly.
6070         (make_eh_edges): Likewise.
6071         (redirect_eh_dispatch_edge): Likewise.
6072         (lower_resx): Likewise.
6073         (lower_eh_dispatch): Likewise.
6074         (maybe_remove_unreachable_handlers): Likewise.
6075         (unsplit_eh): Likewise.
6076         (cleanup_empty_eh): Likewise.
6077         (verify_eh_edges): Likewise.
6078         (verify_eh_dispatch_edge): Likewise.
6079         * tree-ssa-dom.c (record_edge_info): Likewise.
6080         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
6081         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
6082         (thread_through_normal_block): Likewise.
6083         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
6084         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
6085         * tree-switch-conversion.c (switch_conversion::collect): Use new
6086         gimple functions.
6087         (switch_conversion::check_final_bb): Likewise.
6088         (switch_conversion::gather_default_values): Pass new argument
6089         to label_to_block.
6090         (switch_conversion::build_constructors): Likewise.
6091         (switch_decision_tree::compute_cases_per_edge): Use new
6092         gimple_switch_edge function.
6093         (switch_decision_tree::analyze_switch_statement): Pass new argument
6094         to label_to_block.
6095         (switch_decision_tree::try_switch_expansion): Use
6096         gimple_switch_default_edge.
6097         * tree-vrp.c (find_switch_asserts): Pass new argument
6098         to label_to_block.
6099         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
6100         (vr_values::simplify_switch_using_ranges): Likewise.
6102 2018-08-27  Richard Biener  <rguenther@suse.de>
6104         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
6105         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
6107         * tree-ssa-sccvn.h (struct vn_pval): New structure.
6108         (struct vn_nary_op_s): Add unwind_to member.  Add
6109         predicated_values flag and put result into a union together
6110         with a linked list of vn_pval.
6111         (struct vn_ssa_aux): Add name member to make maintaining
6112         a map of SSA name to vn_ssa_aux possible.  Remove no longer
6113         needed info, dfsnum, low, visited, on_sccstack, use_processed
6114         and range_info_anti_range_p members.
6115         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
6116         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
6117         New functions.
6118         (vn_valueize): New global.
6119         (vn_context_bb): Likewise.
6120         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
6121         VN_INFO_PTR_INFO): Remove.
6122         * tree-ssa-sccvn.c: ... (rewrite)
6123         (pass_fre::execute): For -O2+ initialize loops and run
6124         RPO VN in optimistic mode (iterating).  For -O1 and -Og
6125         run RPO VN in non-optimistic mode.
6126         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
6127         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
6128         * doc/invoke.texi (sccvn-max-scc-size): Remove.
6129         (rpo-vn-max-loop-depth): Document.
6130         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
6131         when valuezing the VUSE signals we walked out of the region.
6132         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
6133         (phi_translate): Set VN context block to use for availability
6134         lookup.
6135         (compute_avail): Likewise.
6136         (pre_valueize): New function.
6137         (pass_pre::execute): Adjust to the RPO VN API.
6139         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
6140         (propagate_constants_for_unrolling): Remove.
6141         (tree_unroll_loops_completely): Perform value-numbering
6142         on the unrolled bodies loop parent.
6144 2018-08-27  Richard Biener  <rguenther@suse.de>
6146         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
6147         for partial antic compute.
6149 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
6151         PR rtl-optimization/87065
6152         * combine.c (simplify_if_then_else): Formatting fix.
6153         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
6154         check.
6155         (known_cond): Don't return const_true_rtx for vector modes.  Use
6156         CONST0_RTX instead of const0_rtx.  Formatting fixes.
6158 2018-08-27  Martin Liska  <mliska@suse.cz>
6160         PR gcov-profile/87069
6161         * gcov.c (process_file): Record files already processed
6162         and warn about a file being processed multiple times.
6164 2018-08-27  Martin Liska  <mliska@suse.cz>
6166         PR driver/83193
6167         * config/aarch64/aarch64.c (aarch64_override_options_internal):
6168         Set default values for x_aarch64_*_string strings.
6169         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
6170         prefix.  For -mabi do not print '=ABI' in help and use
6171         <option_value> format for -msve-vector-bits and -moverride
6172         options.
6174 2018-08-26  Jeff Law  <law@redhat.com>
6176         * config/mips/frame-header-opt.c: Include "backend.h" rather than
6177         "cfg.h"
6179 2018-08-26  Marek Polacek  <polacek@redhat.com>
6181         PR c++/87029, Implement -Wredundant-move.
6182         * doc/invoke.texi: Document -Wredundant-move.
6184 2018-08-25  Martin Sebor  <msebor@redhat.com>
6186         PR tree-optimization/87059
6187         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
6188         to the same type as the other.
6189         * fold-const.c (fold_binary_loc): Assert expectation.
6191 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
6193         * config/darwin.c (machopic_legitimize_pic_address): Clean up
6194         extraneous parentheses, dead code section and formatting.
6196 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
6198         PR c++/87091
6199         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
6200         wide enough for jumps in the line-numbering to be visible.
6201         (layout::print_gap_in_line_numbering): New member function.
6202         (layout::calculate_line_spans): When using line numbering, merge
6203         line spans that are only 1 line apart.
6204         (diagnostic_show_locus): When printing line numbers, show gaps in
6205         line numbering directly, rather than printing headers.
6206         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
6207         line-numbering with multiple line spans.
6208         (selftest::test_fixit_insert_containing_newline_2): Add test of
6209         line-numbering, in which the spans are close enough to be merged.
6211 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6213         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
6214         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
6215         * tree-vrp.c (range_is_nonnull): Remove.
6216         (range_includes_zero_p): Accept value_range instead of min/max.
6217         (extract_range_from_binary_expr_1): Do not early bail on
6218         POINTER_PLUS_EXPR.
6219         Use range_includes_zero_p instead of range_is_nonnull.
6220         (extract_range_from_unary_expr): Use range_includes_zero_p instead
6221         of range_is_nonnull.
6222         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
6223         special case VR_ANTI_RANGE.
6224         (vrp_finalize): Same.
6225         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
6226         instead of min/max.
6227         (range_is_nonnull): Remove.
6228         * vr-values.c (vrp_stmt_computes_nonzero): Use
6229         range_includes_zero_p instead of range_is_nonnull.
6230         (extract_range_basic): Pass value_range to range_includes_zero_p
6231         instead of range_is_nonnull.
6233 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
6235         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
6236         * emit-rtl.h (rtl_data): Remove return_bnd.
6237         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
6238         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
6239         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
6240         (POINTER_BOUNDS_MODE): Remove definition.
6241         (make_pointer_bounds_mode): Remove.
6242         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
6243         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
6244         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
6245         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
6246         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
6247         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
6248         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
6250         * config/i386/i386-modes.def (BND32, BND64): Remove.
6251         * config/i386/i386.c (dbx_register_map): Remove bound registers.
6252         (dbx64_register_map): Ditto.
6253         (svr4_dbx_register_map): Ditto.
6254         (indirect_thunk_bnd_needed): Remove.
6255         (indirect_thunks_bnd_used): Ditto.
6256         (indirect_return_bnd_needed): Ditto.
6257         (indirect_return_via_cx_bnd): Ditto.
6258         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
6259         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
6260         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
6261         (output_indirect_thunk_function): Remove handling of
6262         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
6263         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
6264         (ix86_save_reg): Remove handling of crtl->return_bnd.
6265         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
6266         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
6267         and UNSPEC_BNDLX_ADDR.
6268         (ix86_output_indirect_branch_via_reg): Remove handling of
6269         indirect_thunk_prefix_bnd.
6270         (ix86_output_indirect_branch_via_push): Ditto.
6271         (ix86_output_function_return): Ditto.
6272         (ix86_output_indirect_function_return): Ditto.
6273         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
6274         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
6275         (CALL_USED_REGISTERS): Ditto.
6276         (REG_ALLOC_ORDER): Update for removal of bound registers.
6277         (HI_REGISTER_NAMES): Ditto.
6278         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
6279         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
6280         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
6281         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
6282         (FIRST_PSEUDO_REG): Update.
6283         (BND): Remove mode iterator.
6284         * config/i386/predicates.md (bnd_mem_operator): Remove.
6286 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6288         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
6289         vectors.
6291 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6293         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
6294         the case in which the permute needs only a single element and
6295         repeats for every vector of the result.  Extend that case to
6296         handle variable-length vectors.
6297         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6299 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
6301         PR debug/79342
6302         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
6303         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
6305 2018-08-24  Richard Biener  <rguenther@suse.de>
6307         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
6308         bb_flags_allocated members.
6309         (auto_flag): New RAII class for allocating flags.
6310         (auto_edge_flag): New RAII class for allocating edge flags.
6311         (auto_bb_flag): New RAII class for allocating bb flags.
6312         * cfgloop.c (verify_loop_structure): Allocate temporary edge
6313         flag dynamically.
6314         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
6315         in favor of temporarily allocated BB flag.
6316         * hsa-brig.c: Re-order includes.
6317         * hsa-dump.c: Likewise.
6318         * hsa-regalloc.c: Likewise.
6319         * print-rtl.c: Likewise.
6320         * profile-count.c: Likewise.
6322 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
6324         PR target/86989
6325         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
6326         the TOC register.
6328 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6330         PR 87073/bootstrap
6331         * wide-int-range.cc (wide_int_range_div): Do not ignore result
6332         from wide_int_range_multiplicative_op.
6334 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6336         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
6337         "permutaion".
6339 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
6341         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
6342         to 'expanded'.
6344 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
6346         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
6347         full GENERIC expression used for replacement.
6349 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
6351         * tree-vrp.c (abs_extent_range): Remove.
6352         (extract_range_into_wide_ints): Pass wide ints by reference.
6353         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
6354         Pass wide ints by reference in all calls to
6355         extract_range_into_wide_ints.
6356         * wide-int-range.cc (wide_int_range_div): New.
6357         * wide-int-range.h (wide_int_range_div): New.
6358         (wide_int_range_includes_zero_p): New.
6359         (wide_int_range_zero_p): New.
6361 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
6363         * config/aarch64/aarch64.md (arches): New enum.
6364         (arch): New enum attr.
6365         (arch_enabled): New attr.
6366         (enabled): Now uses arch_enabled only.
6367         (simd, sve, fp16): Removed attribute.
6368         (fp): Attr now defined in terms of 'arch'.
6369         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
6370         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
6371         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
6372         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
6373         attributes into 'arch'.
6374         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
6375         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
6376         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
6377         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
6378         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
6379         'simd' attribute into 'arch'.
6380         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
6381         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
6382         Convert use of 'fp' attribute to 'arch'.
6383         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
6384         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
6385         into 'arch'.
6386         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
6387         (different modes) Merge 'fp' and 'simd' into 'arch'.
6388         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
6389         'simd' into 'arch'.
6391 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
6393         PR rtl-optimization/87026
6394         * expmed.c (canonicalize_comparison): If we can no longer create
6395         pseudoregisters, don't.
6397 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
6399         PR target/86951
6400         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6401         prototype.
6402         * config/arm/arm.c (speculation_barrier_libfunc): New static
6403         variable.
6404         (arm_init_libfuncs): Initialize it.
6405         (arm_emit_speculation_barrier): New function.
6406         * config/arm/arm.md (speculation_barrier): Call
6407         arm_emit_speculation_barrier for architectures that do not have
6408         DSB or ISB.
6409         (speculation_barrier_insn): Only match on Armv7 or later.
6411 2018-08-23  Richard Biener  <rguenther@suse.de>
6413         PR middle-end/87024
6414         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6415         calls.
6417 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6419         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6420         of single-vector TBLs.
6421         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6422         one input is given.
6424 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6426         PR target/85910
6427         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6428         aarch64_evpc_tbl guard.
6430 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6432         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6433         behaviour.
6435 2018-08-22  Martin Sebor  <msebor@redhat.com>
6437         PR middle-end/87052
6438         * tree-pretty-print.c (pretty_print_string): Add argument.
6439         (dump_generic_node): Call to pretty_print_string with string size.
6441 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
6443         PR rtl-optimization/86771
6444         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6445         of two SETs into those two SETs, one to be placed at i2, if that SETs
6446         destination is modified between i2 and i3.
6448 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6450         PR tree-optimization/86725
6451         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6452         function.
6453         (vect_analyze_scalar_cycles_1): Check it.
6455 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6457         PR tree-optimization/86725
6458         * tree-vect-loop.c (vect_is_simple_reduction): When treating
6459         an outer loop phi as a double reduction, make sure that the
6460         single user of the phi result is an inner loop phi.
6462 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6464         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6465         grouped stores with gaps to a strided group.
6467 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6469         * tree-vect-stmts.c (get_group_load_store_type)
6470         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6471         first statement in a group.
6473 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6475         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6476         the sequence used in gcc/gcc.c.
6478 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6480         PR other/704
6481         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6482         building it.
6484 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6486         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6487         Darwin10-specific unwinder-shim.
6488         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6489         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6490         New to cater for Darwin10 Rosetta.
6492 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6494         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6495         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6497 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6499         PR bootstrap/81033
6500         PR target/81733
6501         PR target/52795
6502         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6503         (dwarf2out_switch_text_section): Generate a local label for the
6504         second function sub-section and apply it as the second FDE start
6505         label.
6506         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6507         second sub-section start.
6509 2018-08-22  Richard Biener  <rguenther@suse.de>
6511         PR tree-optimization/86988
6512         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6514 2018-08-22  Richard Biener  <rguenther@suse.de>
6516         PR tree-optimization/86945
6517         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6519 2018-08-22  Alexandre Oliva <oliva@adacore.com>
6521         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6522         a comment about how uses of r2 for .sdata2 come about.
6524 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
6526         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6528 2018-08-21  Marek Polacek  <polacek@redhat.com>
6530         PR c++/86981, Implement -Wpessimizing-move.
6531         * doc/invoke.texi: Document -Wpessimizing-move.
6533 2018-08-21  Jan Hubicka  <jh@suse.cz>
6535         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6536         * tree.h (is_redundant_typedef): Remove.
6537         * dwarf2out.c (is_redundant_typedef): Turn into static function.
6539 2018-08-21  Jan Hubicka  <jh@suse.cz>
6541         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6542         when possible.
6544 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
6546         * expmed.c (extract_low_bits): Reject invalid subregs early.
6548 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6550         PR middle-end/86121
6551         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6552         behaviour.
6554 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6556         * config/vxworks.h: Guard vxworks_asm_out_constructor and
6557         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6558         * config/vxworks.c: Likewise.
6560 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6562         * config/vxworks.c: Set targetm.have_ctors_dtors
6563         if HAVE_INITFINI_ARRAY_SUPPORT.
6564         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6565         if HAVE_INITFINI_ARRAY_SUPPORT.
6567 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6569         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6570         default search path for VxWorks < 7.
6572 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6574         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6575         (get_format_string): Refer to c_getstr.
6577 2018-08-21  Tom de Vries  <tdevries@suse.de>
6579         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6580         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6581         (debuginfo_early_stop): Declare.
6582         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6583         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6584         (debuginfo_early_stop): New function.
6585         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6586         and debuginfo_early_stop.
6587         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6588         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6589         (general_init): Call debuginfo_early_init.
6590         (finalize): Call debuginfo_fini.
6591         (do_compile): Call debuginfo_init.
6592         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6593         -fdump-early-debug.
6594         (@item -fdump-debug, @item -fdump-earlydebug): Add.
6596 2018-08-21  Tom de Vries  <tdevries@suse.de>
6598         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6599         flag_dump_noaddr and flag_dump_unnumbered.
6601 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
6603         * wide-int-range.cc (wide_int_range_abs): New.
6604         (wide_int_range_order_set): Rename from wide_int_range_min_max.
6605         * wide-int-range.h (wide_int_range_abs): New.
6606         (wide_int_range_min_max): New.
6607         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6608         case to call wide_int_range_abs.
6609         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6610         (extract_range_from_abs_expr): Delete.
6612 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
6614         PR target/87033
6615         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6616         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6617         for indexed loads.
6619 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6620             Jeff Law <law@redhat.com>
6622         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6623         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6625 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
6627         PR other/84889
6628         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6629         (decl_attributes): Likewise.
6630         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6631         instance.
6632         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6633         * diagnostic-core.h (class auto_diagnostic_group): New class.
6634         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6635         (diagnostic_report_diagnostic): Handle the first diagnostics within
6636         a group.
6637         (emit_diagnostic): Add auto_diagnostic_group instance.
6638         (inform): Likewise.
6639         (inform_n): Likewise.
6640         (warning): Likewise.
6641         (warning_at): Likewise.
6642         (warning_n): Likewise.
6643         (pedwarn): Likewise.
6644         (permerror): Likewise.
6645         (error): Likewise.
6646         (error_n): Likewise.
6647         (error_at): Likewise.
6648         (sorry): Likewise.
6649         (fatal_error): Likewise.
6650         (internal_error): Likewise.
6651         (internal_error_no_backtrace): Likewise.
6652         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6653         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6654         * diagnostic.h (struct diagnostic_context): Add fields
6655         "diagnostic_group_nesting_depth",
6656         "diagnostic_group_emission_count", "begin_group_cb",
6657         "end_group_cb".
6658         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6659         Add auto_diagnostic_group instance(s).
6660         (find_explicit_erroneous_behavior): Likewise.
6661         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6662         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6663         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6664         (gimplify_va_arg_expr): Likewise.
6665         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6666         (HSA_SORRY_AT): Likewise.
6667         * ipa-devirt.c (compare_virtual_tables): Likewise.
6668         (warn_odr): Likewise.
6669         * multiple_target.c (expand_target_clones): Likewise.
6670         * opts-common.c (cmdline_handle_error): Likewise.
6671         * reginfo.c (globalize_reg): Likewise.
6672         * substring-locations.c (format_warning_n_va): Likewise.
6673         * tree-inline.c (expand_call_inline): Likewise.
6674         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6675         * tree-ssa-loop-niter.c
6676         (do_warn_aggressive_loop_optimizations): Likewise.
6677         * tree-ssa-uninit.c (warn_uninit): Likewise.
6678         * tree.c (warn_deprecated_use): Likewise.
6680 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
6682         PR target/87014
6683         * config/i386/i386.md (eh_return): Always update EH return
6684         address in word_mode.
6686 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
6688         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6689         TARGET_SPLIT_COMPLEX_ARG is defined.
6691 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6693         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6695 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6697         PR target/86984
6698         * expr.c (expand_assignment): Assert that bitpos is positive.
6699         (store_field): Likewise
6700         (expand_expr_real_1): Make sure that bitpos is positive.
6701         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6702         integer overflow.
6704 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6706         * Makefile.in (CPP_ID_DATA_H): Delete.
6707         (CPP_INTERNAL_H): Don't add it.
6708         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6709         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6711 2018-08-20  Richard Biener  <rguenther@suse.de>
6713         PR tree-optimization/78655
6714         * tree-vrp.c (extract_range_from_binary_expr_1): Make
6715         pointer + offset nonnull if either operand is nonnull work.
6717 2018-08-20  Tom de Vries  <tdevries@suse.de>
6719         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6720         unless the referenced die describes the added property using
6721         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
6722         Otherwise, add a DW_AT_location to the referenced die.
6724 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
6726         PR target/86994
6727         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6728         register_operand when calling ix86_set_reg_reg_cost.
6729         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6730         Set *total to 0 for operands that satisfy x86_64_immediate_operand
6731         predicate and to 1 otherwise.
6733 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
6735         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6736         emit a diagnostic that it is not supported and reset the option.
6737         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6738         supported and consume the option.  (ASM_FINAL_SPEC): New.
6740 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6742         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6743         a sentence.
6745 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
6747         C-SKY port: Documentation
6749         * doc/extend.texi (C-SKY Function Attributes): New section.
6750         * doc/invoke.texi (Option Summary): Add C-SKY options.
6751         (C-SKY Options): New section.
6752         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6754 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6755             Huibin Wang  <huibin_wang@c-sky.com>
6756             Sandra Loosemore  <sandra@codesourcery.com>
6757             Chung-Lin Tang  <cltang@codesourcery.com>
6759         C-SKY port: Backend implementation
6761         * config/csky/*: New.
6762         * common/config/csky/*: New.
6764 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6765             Huibin Wang  <huibin_wang@c-sky.com>
6766             Sandra Loosemore  <sandra@codesourcery.com>
6767             Chung-Lin Tang  <cltang@codesourcery.com>
6768             Andrew Jenner  <andrew@codesourcery.com>
6770         C-SKY port: Configury
6772         * config.gcc (csky-*-*): New.
6773         * configure.ac: Add csky to targets for dwarf2 debug_line support.
6774         * configure: Regenerated.
6776 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
6778         * dump-context.h: Include "dumpfile.h".
6779         (dump_context::dump_printf_va): Convert final param from va_list
6780         to va_list *.  Convert from ATTRIBUTE_PRINTF to
6781         ATTRIBUTE_GCC_DUMP_PRINTF.
6782         (dump_context::dump_printf_loc_va): Likewise.
6783         * dumpfile.c: Include "stringpool.h".
6784         (make_item_for_dump_printf_va): Delete.
6785         (make_item_for_dump_printf): Delete.
6786         (class dump_pretty_printer): New class.
6787         (dump_pretty_printer::dump_pretty_printer): New ctor.
6788         (dump_pretty_printer::emit_items): New member function.
6789         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
6790         function.
6791         (dump_pretty_printer::emit_item): New member function.
6792         (dump_pretty_printer::stash_item): New member function.
6793         (dump_pretty_printer::format_decoder_cb): New member function.
6794         (dump_pretty_printer::decode_format): New member function.
6795         (dump_context::dump_printf_va): Reimplement in terms of
6796         dump_pretty_printer.
6797         (dump_context::dump_printf_loc_va): Convert final param from va_list
6798         to va_list *.
6799         (dump_context::begin_scope): Reimplement call to
6800         make_item_for_dump_printf.
6801         (dump_printf): Update for change to dump_printf_va.
6802         (dump_printf_loc): Likewise.
6803         (selftest::test_capture_of_dump_calls): Convert "stmt" from
6804         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
6805         with %T, %E, and %G.
6806         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
6807         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
6808         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
6809         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
6810         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
6811         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
6812         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
6813         within a dump_printf_loc call to "%wu".
6814         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
6815         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
6816         missing space after "=".
6817         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
6818         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
6819         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
6820         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
6821         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
6822         duplicate "vectorized" from message.
6824 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
6826         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
6827         polyNxK_t element's TYPE_STRING_FLAG.
6829 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6831         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
6832         (they were unnamed before).  Fix comments.
6834 2018-08-17  Nathan Sidwell  <nathan@acm.org>
6836         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
6838 2018-08-17  Richard Biener  <rguenther@suse.de>
6840         PR tree-optimization/86841
6841         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6843 2018-08-17  Martin Liska  <mliska@suse.cz>
6845         * common.opt: Remove Warn, Init and Report for options with
6846         Ignore/Deprecated flag. Warning is done automatically for
6847         Deprecated flags.
6848         * config/i386/i386.opt: Likewise.
6849         * config/ia64/ia64.opt: Likewise.
6850         * config/rs6000/rs6000.opt: Likewise.
6851         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6852         Remove usage of flag_check_pointer_bounds.
6853         * lto-wrapper.c (merge_and_complain): Do not handle
6854         OPT_fcheck_pointer_bounds.
6855         (append_compiler_options): Likewise.
6856         * opt-functions.awk: Do not handle Deprecated.
6857         * optc-gen.awk: Check that Var, Report and Init are not
6858         used for an option with Ignore/Deprecated flag.
6859         * opts-common.c (decode_cmdline_option): Do not report
6860         CL_ERR_DEPRECATED.
6861         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
6862         options.
6863         * opts.h (struct cl_option): Remove cl_deprecated flag.
6864         (CL_ERR_DEPRECATED): Remove error enum value.
6866 2018-08-17  Richard Biener  <rguenther@suse.de>
6868         PR middle-end/86505
6869         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
6870         across a va-arg-pack using call adjust its return value accordingly.
6872 2018-08-16  Martin Sebor  <msebor@redhat.com>
6874         PR tree-optimization/86853
6875         * gimple-ssa-sprintf.c (struct format_result): Rename member.
6876         (struct fmtresult): Add member and initialize it in ctors.
6877         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
6878         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
6879         (format_directive): Set POSUNDER4K when MAYFAIL is set.
6880         (parse_directive): Handle %C same as %c.
6881         (sprintf_dom_walker::compute_format_length): Adjust.
6882         (is_call_safe): Adjust.
6884 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6886         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
6887         for determining how to count the elements.
6888         * builtins.h (c_strlen): Adjust prototype.
6889         * expr.c (string_constant): Add new parameter mem_size.
6890         Set *mem_size appropriately.
6891         * expr.h (string_constant): Adjust protoype.
6892         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
6893         * gimple-fold.h (get_range_strlen): Adjust prototype.
6894         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
6895         (format_string): Call get_string_length with eltsize.
6897 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6899         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
6900         to emit the span, rather than setting it as the prefix.
6902 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6904         * diagnostic-show-locus.c (layout::start_annotation_line): Add
6905         "margin_char" parameter, defaulting to space.  Use it in place
6906         of pp_space for the initial part of the margin.
6907         (layout::print_leading_fixits): Use '+' when filling the margin
6908         of line-insertion fix-it hints.
6910 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6912         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
6913         Delete.
6915 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6917         * config/rs6000/altivec.md: Don't set length attribute to the default
6918         value.
6919         * config/rs6000/darwin.md: Ditto.
6920         * config/rs6000/dfp.md: Ditto.
6921         * config/rs6000/htm.md: Ditto.
6922         * config/rs6000/rs6000.md: Ditto.
6923         * config/rs6000/sync.md: Ditto.
6924         * config/rs6000/vsx.md: Ditto.
6926 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6928         * config/rs6000/altivec.md: Don't set length attribute to the default
6929         value, for branch instructions.
6930         * config/rs6000/darwin.md: Ditto.
6931         * config/rs6000/rs6000.md: Ditto.
6933 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6935         * config/rs6000/rs6000.md (length): Always define as const_int 4.
6936         (unnamed conditional branch define_insn): Set length to 4 or 8
6937         depending on offset.
6938         (<bd>_<mode>): Similar, for alternative 0.
6939         (<bd>tf_<mode>): Ditto.
6941 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6943         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
6945 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
6947         * doc/rtl.texi: Replace old RTX class names with new names.
6950 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
6952         * expmed.h (canonicalize_comparison): New declaration.
6953         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
6954         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
6955         * optabs.c (prepare_cmp_insn): Likewise.
6956         * rtl.h (unsigned_condition_p): New function which checks if a
6957         comparison operator is unsigned.
6959 2018-08-16  Nathan Sidwell  <nathan@acm.org>
6961         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
6962         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
6964 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6966         PR target/84711
6967         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
6968         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
6969         (mov<mov>): ..this and enable unconditionally.
6971 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6973         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
6975 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
6977         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
6978         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
6979         "Common" with "Target".
6981 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
6983         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
6984         * doc/invoke.texi (mmitigate-rop): Remove.
6985         * config/i386/i386.c: Do not include "regrename.h".
6986         (ix86_rop_should_change_byte_p, reg_encoded_number)
6987         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
6988         Remove.
6989         (ix86_reorg): Remove call to ix86_mitigate_rop.
6990         * config/i386/i386.md (attr "modrm_class"): Remove.
6991         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
6992         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
6993         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
6995 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
6997         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
6998         allow folding of mergeh() and mergel() for the float and double types.
6999         (fold_mergehl_helper): Rework to handle building a permute tree
7000         for float vectors.
7002 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
7004         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
7005         for TARGET_SSE.
7007 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
7009         * common.opt (fdiagnostics-show-labels): New option.
7010         * diagnostic-show-locus.c (class layout_range): Add field
7011         "m_label".
7012         (class layout): Add field "m_show_labels_p".
7013         (layout_range::layout_range): Add param "label" and use it to
7014         initialize m_label.
7015         (make_range): Pass in NULL for new "label" param of layout_range's
7016         ctor.
7017         (layout::layout): Initialize m_show_labels_p.
7018         (layout::maybe_add_location_range): Pass in loc_range->m_label
7019         when constructing layout_range instances.
7020         (struct line_label): New struct.
7021         (layout::print_any_labels): New member function.
7022         (layout::print_line): Call it if label-printing is enabled.
7023         (selftest::test_one_liner_labels): New test.
7024         (selftest::test_diagnostic_show_locus_one_liner): Call it.
7025         * diagnostic.c (diagnostic_initialize): Initialize
7026         context->show_labels_p.
7027         * diagnostic.h (struct diagnostic_context): Add field
7028         "show_labels_p".
7029         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7030         -fno-diagnostics-show-labels.
7031         * dwarf2out.c (gen_producer_string): Add
7032         OPT_fdiagnostics_show_labels to the ignored options.
7033         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
7034         param.
7035         (gcc_rich_location::maybe_add_expr): Likewise.
7036         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
7037         label" param, defaulting to NULL.
7038         (gcc_rich_location::add_expr): Add "label" param.
7039         (gcc_rich_location::maybe_add_expr): Likewise.
7040         (class text_range_label): New class.
7041         (class range_label_for_type_mismatch): New class.
7042         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
7043         of format_warning_va.
7044         (fmtwarn_n): Likewise for new params of format_warning_n_va.
7045         * lto-wrapper.c (merge_and_complain): Add
7046         OPT_fdiagnostics_show_labels to the "pick one setting" options.
7047         (append_compiler_options): Likewise to the dropped options.
7048         (append_diag_options): Likewise to the passed-on options.
7049         * opts.c (common_handle_option): Handle the new option.
7050         * selftest-diagnostic.c
7051         (test_diagnostic_context::test_diagnostic_context): Enable
7052         show_labels_p.
7053         * substring-locations.c: Include "gcc-rich-location.h".
7054         (format_warning_n_va): Add "fmt_label" and "param_label" params
7055         and use them as appropriate.
7056         (format_warning_va): Add "fmt_label" and "param_label" params,
7057         passing them on to format_warning_n_va.
7058         (format_warning_at_substring): Likewise.
7059         (format_warning_at_substring_n): Likewise.
7060         * substring-locations.h (format_warning_va): Add "fmt_label" and
7061         "param_label" params.
7062         (format_warning_n_va): Likewise.
7063         (format_warning_at_substring): Likewise.
7064         (format_warning_at_substring_n): Likewise.
7065         * toplev.c (general_init): Initialize global_dc->show_labels_p.
7067 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
7069         PR testsuite/86519
7070         * builtins.c (expand_builtin_memcmp): Do not expand the call
7071         when overflow is detected.
7073 2018-08-15  Martin Sebor  <msebor@redhat.com>
7075         PR tree-optimization/71625
7076         * config/aarch64/aarch64-builtins.c
7077         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
7079 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
7081         * config/s390/s390.c (s390_reorg): Remove loop.
7083 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7085         * config/darwin.c
7086          (darwin_function_switched_text_sections): Delete.
7087         * gcc/config/darwin.h
7088          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
7090 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7092         PR target/81685
7093         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
7094         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
7095         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
7097 2018-08-15  Martin Liska  <mliska@suse.cz>
7099         PR tree-optimization/86925
7100         * predict.c (expr_expected_value_1): When taking
7101         later predictor, assign also probability.
7102         Use fold_build2_initializer_loc in order to fold
7103         the expression in -frounding-math.
7105 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
7107         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
7108         patterns.
7109         (expand_vec_perm_1): Try the new method.
7111 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7113         PR target/86547
7114         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
7115         Check whether lra_live_max_point is 0 before dividing.
7117 2018-08-14  Martin Sebor  <msebor@redhat.com>
7119         PR tree-optimization/86650
7120         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
7121         (vrp_prop::check_mem_ref): Same.
7123 2018-08-13  Liu Hao <lh_mouse@126.com>
7125         * pretty-print.c (eat_esc_sequence): Swap the foreground and
7126         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
7127         and clear it thereafter, as it only works for DBCS.
7129 2018-08-13  Liu Hao <lh_mouse@126.com>
7131         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
7132         handle returned by _get_osf_handle().
7134 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7136         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
7137         for folding vec_perm.
7139 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7141         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
7142         Add support for gimple-folding of vec_pack() and vec_unpack()
7143         intrinsics.
7145 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
7147         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
7148         vec_xst variants to the list.
7149         (rs6000_gimple_fold_builtin): Add support for folding unaligned
7150         vector loads and stores.
7152 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
7154         * config.gcc (rs6000-ibm-aix4.x): Delete.
7155         (rs6000-ibm-aix5.1): Delete.
7156         (rs6000-ibm-aix5.2): Delete.
7157         (rs6000-ibm-aix5.3): Delete.
7158         * config/rs6000/aix43.h: Delete.
7159         * config/rs6000/aix51.h: Delete.
7160         * config/rs6000/aix52.h: Delete.
7161         * config/rs6000/t-aix43: Delete.
7163 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
7165         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7166         New function.
7167         (s390_decompose_address): Factor out constant pool ref
7168         decomposition.
7170 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7172         * config/nds32/nds32-predicates.c
7173         (nds32_can_use_bclr_p): Change return type as bool.
7174         (nds32_can_use_bset_p): Ditto.
7175         (nds32_can_use_btgl_p): Ditto.
7176         (nds32_can_use_bitci_p): Ditto.
7177         * config/nds32/nds32-protos.h
7178         (nds32_can_use_bclr_p): Change declaration.
7179         (nds32_can_use_bset_p): Ditto.
7180         (nds32_can_use_btgl_p): Ditto.
7181         (nds32_can_use_bitci_p): Ditto.
7183 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7185         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
7186         Support -msched-prolog-epilog option.
7187         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
7189 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7191         * common/config/nds32/nds32-common.c
7192         (nds32_option_optimization_table): Enalbe -malways-align.
7194 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7196         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
7197         extra_headers.
7198         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
7199         OPT_misr_secure_ case.
7200         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
7201         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
7202         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
7203         secure attribute.
7204         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
7205         (nds32_isr_info): New field security_level.
7206         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
7207         * config/nds32/nds32.md (return_internal): Consider critical attribute.
7208         * config/nds32/nds32.opt (misr-secure): New option.
7209         * config/nds32/nds32_init.inc: New file.
7210         * config/nds32/nds32_isr.h: New file.
7212 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
7214         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
7215         Update comment for atomic instructions.
7216         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
7217         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
7218         Remove.
7219         (atomic_loaddi): Revise fence expansion to only emit fence prior to
7220         load for __ATOMIC_SEQ_CST model.
7221         (atomic_loaddi_1): Remove float register target.
7222         (atomic_storedi): Handle CONST_INT values.
7223         (atomic_storedi_1): Remove float register source.  Add special case
7224         for zero value.
7225         (memory_barrier): New expander and insn.
7227 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
7229         PR tree-optimization/86835
7230         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
7231         new_stmt after def_gsi, make sure to insert new_square_stmt after
7232         that stmt, not 2 stmts before it.
7234 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
7236         PR target/82418
7237         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
7238         instead of SWI48.
7240 2018-08-10  Martin Liska  <mliska@suse.cz>
7242         PR target/83610
7243         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
7244         function type.
7245         * builtins.c (expand_builtin_expect_with_probability):
7246         New function.
7247         (expand_builtin_expect_with_probability): New function.
7248         (build_builtin_expect_predicate): Add new argumnet probability
7249         for BUILT_IN_EXPECT_WITH_PROBABILITY.
7250         (fold_builtin_expect):
7251         (fold_builtin_2):
7252         (fold_builtin_3):
7253         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
7254         * builtins.h (fold_builtin_expect): Set new argument.
7255         * doc/extend.texi: Document __builtin_expect_with_probability.
7256         * doc/invoke.texi: Likewise.
7257         * gimple-fold.c (gimple_fold_call): Pass new argument.
7258         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
7259         also BUILT_IN_EXPECT_WITH_PROBABILITY.
7260         * predict.c (get_predictor_value): New function.
7261         (expr_expected_value): Add new argument probability. Assume
7262         that predictor and probability are always non-null.
7263         (expr_expected_value_1): Likewise.  For __builtin_expect and
7264         __builtin_expect_with_probability set probability.  Handle
7265         combination in binary expressions.
7266         (tree_predict_by_opcode): Simplify code by simply calling
7267         get_predictor_value.
7268         (pass_strip_predict_hints::execute): Add handling of
7269         BUILT_IN_EXPECT_WITH_PROBABILITY.
7270         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
7271         new predictor.
7272         * tree.h (DECL_BUILT_IN_P): New function.
7274 2018-08-10  Martin Liska  <mliska@suse.cz>
7276         PR tree-optimization/85799
7277         * passes.def: Add argument for pass_strip_predict_hints.
7278         * predict.c (class pass_strip_predict_hints): Add new argument
7279         early_p.
7280         (strip_predictor_early): New function.
7281         (pass_strip_predict_hints::execute): Call the function to
7282         strip predictors.
7283         (strip_predict_hints): New function.
7284         * predict.def: Fix comment.
7286 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
7288         * Makefile.in: Clarify which tm.texi to copy over to assert the
7289         right to grant a GFDL license for all.
7291 2018-08-09  Jeff Law  <law@redhat.com>
7293         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
7294         unused variable.
7296 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
7298         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
7299         prototype.
7301 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7303         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
7304         reductions for variable-length vectors.
7306 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
7308         PR other/84889
7309         * common.opt (fdiagnostics-show-line-numbers): New option.
7310         * diagnostic-show-locus.c (class layout): Add fields
7311         "m_show_line_numbers_p" and "m_linenum_width";
7312         (num_digits): New function.
7313         (test_num_digits): New function.
7314         (layout::layout): Initialize new fields.  Update m_x_offset
7315         logic to handle any left margin.
7316         (layout::print_source_line): Print line number when requested.
7317         (layout::start_annotation_line): New member function.
7318         (layout::print_annotation_line): Call it.
7319         (layout::print_leading_fixits): Likewise.
7320         (layout::print_trailing_fixits): Likewise.  Update calls to
7321         move_to_column for new parameter.
7322         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
7323         it to potentially call start_annotation_line.
7324         (layout::show_ruler): Call start_annotation_line.
7325         (selftest::test_line_numbers_multiline_range): New selftest.
7326         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
7327         and selftest::test_line_numbers_multiline_range.
7328         * diagnostic.c (diagnostic_initialize): Initialize
7329         show_line_numbers_p.
7330         * diagnostic.h (struct diagnostic_context): Add field
7331         "show_line_numbers_p".
7332         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7333         -fno-diagnostics-show-line-numbers.
7334         * dwarf2out.c (gen_producer_string): Add
7335         OPT_fdiagnostics_show_line_numbers to the ignored options.
7336         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
7337         one setting" options.
7338         (append_compiler_options): Likewise to the dropped options.
7339         (append_diag_options): Likewise to the passed-on options.
7340         * opts.c (common_handle_option): Handle the new option.
7341         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
7343 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7345         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
7346         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
7347         third argument of type "const signed char" to descriptions of
7348         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
7349         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
7350         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
7351         __builtin_bcdsub_ov functions.
7353 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7355         PR tree-optimization/86858
7356         * tree-vect-loop.c (vect_is_simple_reduction): Restore
7357         flow_bb_inside_loop_p calls.
7359 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7361         PR tree-optimization/86871
7362         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
7363         instead of gimple_assign_lhs.
7365 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
7367         PR target/86887
7368         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
7369         register constraint to operand 0.
7370         (add<mode>3_carryinC): Likewise.
7371         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
7373 2018-08-09  Martin Liska  <mliska@suse.cz>
7375         PR c/86895
7376         * common.opt: Remove extra line.
7378 2018-08-09  Martin Liska  <mliska@suse.cz>
7380         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
7381         at the end of a line, make first letter capital and end up
7382         a sentence with a dot.
7383         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
7384         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
7385         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
7386         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
7387         (PARAM_MAX_ISL_OPERATIONS): Likewise.
7388         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
7389         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
7390         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
7391         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
7392         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
7393         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
7394         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
7395         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7396         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7397         (PARAM_TREE_REASSOC_WIDTH): Likewise.
7398         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7399         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7400         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7402 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
7404         PR target/84332
7405         * config/s390/s390.c (s390_option_override_internal): Reduce the
7406         stack-clash-protection-probe-interval param if it would be too big
7407         for z900.
7409 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
7411         PR target/46179
7412         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7413         * config/m68k/m68k.c (handle_move_double): Don't call
7414         m68k_final_prescan_insn.
7415         (m68k_adjust_decorated_operand): Renamed from
7416         m68k_final_prescan_insn, remove first and third operand and
7417         simplify.
7418         (print_operand): Call it.
7419         (print_operand_address): Call it.
7421 2018-08-08  Nathan Sidwell  <nathan@acm.org>
7423         * diagnostic.c (diagnostic_report_current_module): Use
7424         linemap_included_from & linemap_included_from_linemap.
7426 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
7428         * config/aarch64/aarch64-cores.def: Add phecda core.
7429         * config/aarch64/aarch64-tune.md: Regenerate.
7430         * doc/invoke.texi: Add phecda core.
7432 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7434         PR target/85295
7435         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7436         definitions.
7437         * config/s390/s390.md ("movti"): Add more alternatives for
7438         constant to GPR copies.
7440 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7442         * config/s390/s390.c: Fix whitespace damage throughout the file.
7443         * config/s390/s390.h: Likewise.
7444         * config/s390/tpf.h: Likewise.
7446 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7448         * config/s390/s390.c (s390_loadrelative_operand_p):
7449         Remove TARGET_CPU_ZARCH usages.
7450         (s390_rtx_costs): Likewise.
7451         (s390_legitimate_constant_p): Likewise.
7452         (s390_cannot_force_const_mem): Likewise.
7453         (legitimate_reload_constant_p): Likewise.
7454         (s390_preferred_reload_class): Likewise.
7455         (legitimize_pic_address): Likewise.
7456         (legitimize_tls_address): Likewise.
7457         (s390_split_branches): Removed.
7458         (s390_add_execute): Removed.
7459         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7460         (s390_mainpool_start): Likewise.
7461         (s390_mainpool_finish): Likewise.
7462         (s390_mainpool_cancel): Removed.
7463         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7464         (s390_chunkify_cancel): Likewise.
7465         (s390_return_addr_rtx): Likewise.
7466         (s390_register_info): Remove split_branches_pending_p uages.
7467         (s390_optimize_register_info): Likewise.
7468         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7469         split_branches_pending_p usages.
7470         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7471         (s390_load_got): Likewise.
7472         (s390_expand_split_stack_prologue): Likewise.
7473         (output_asm_nops): Likewise.
7474         (s390_function_profiler): Likewise.
7475         (s390_emit_call): Likewise.
7476         (s390_conditional_register_usage): Likewise.
7477         (s390_optimize_prologue): Likewise.
7478         (s390_reorg): Remove TARGET_CPU_ZARCH and
7479         split_branches_pending_p usages.
7480         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7481         usages.
7482         (s390_output_indirect_thunk_function): Likewise.
7483         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7484         (TARGET_CPU_ZARCH_P): Removed.
7485         (struct machine_function): Remove split_branches_pending_p.
7486         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7488 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7490         * common/config/s390/s390-common.c (processor_flags_table):
7491         Remove flags.
7492         * config.gcc: Remove with_arch/with_tune support.
7493         * config/s390/2064.md: Remove cpu attribute comparisons.
7494         * config/s390/driver-native.c (s390_host_detect_local_cpu):
7495         Remove MTN.
7496         * config/s390/linux.h (ASM_SPEC):
7497         Remove -march support.
7498         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7499         Use a table to get an arch level.
7500         * config/s390/s390-opts.h (enum processor_type):
7501         Remove enum values.
7502         * config/s390/s390.c
7503         (processor_table): Remove entries, add arch_level values.
7504         (s390_issue_rate): Remove cases.
7505         (s390_option_override): Adjust
7506         s390_option_override_internal() call.
7507         (s390_option_override_internal): Remove deprecation warning.
7508         (s390_valid_target_attribute_tree): Adjust
7509         s390_option_override_internal() call.
7510         * config/s390/s390.h (struct s390_processor):
7511         Share with s390-c.c, add arch_level field.
7512         * config/s390/s390.md:
7513         Remove occurrences in cpu attribute.
7514         * config/s390/s390.opt: Remove -march/-mtune support.
7515         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7516         * doc/invoke.texi: Remove deprecation warning.
7518 2018-08-08  Luis Machado  <luis.machado@linaro.org>
7520         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7521         global.
7522         (qdf24xx_tunings): Set vector cost structure to
7523         qdf24xx_vector_cost.
7525         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7526         <register_sextend>: Set to 3.
7528 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
7530         PR target/86838
7531         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7532         * config/aarch64/aarch64-simd.md
7533         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7534         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7535         (aarch64_frecpx<mode>): ...this new pattern.
7536         * config/aarch64/aarch64-simd-builtins.def: Remove comment
7537         about aarch64_frecp<FRECP:frecp_suffix><mode>.
7539 2018-08-07  Martin Liska  <mliska@suse.cz>
7541         PR middle-end/83023
7542         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7543         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7544         * predict.def (PRED_MALLOC_NONNULL): New predictor.
7545         * doc/extend.texi: Document that malloc attribute adds
7546         hit to compiler.
7548 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
7550         PR target/86785
7551         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7552         Define to speculation_safe_value_not_needed.
7554 2018-08-06  Jeff Law  <law@redhat.com>
7556         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7557         the vr_values instance to cprop_into_stmt.
7558         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7559         (cprop_operand): Also query EVRP to determine if OP is a constant.
7561 2018-08-06  Nathan Sidwell  <nathan@acm.org>
7563         * diagnostic.c (diagnostic_report_current_module): Reroll
7564         included-at loop.  Translate text.
7566 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
7568         * function-tests.c (selftest::test_expansion_to_rtl): Call
7569         free_after_compilation.
7571 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7573         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7575 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
7577         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7578         loops with memory block operations from getting unrolled.
7580 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
7582         PR target/86807
7583         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7584         Define to speculation_safe_value_not_needed.
7586 2018-08-06  Jeff Law  <law@redhat.com>
7588         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7589         assert.
7591 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7593         PR target/86662
7594         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7595         with all enabled __intN types.
7597         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7599 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7601         * alias.c (record_set): Check for clobber high.
7602         * cfgexpand.c (expand_gimple_stmt): Likewise.
7603         * combine-stack-adj.c (single_set_for_csa): Likewise.
7604         * combine.c (find_single_use_1): Likewise.
7605         (set_nonzero_bits_and_sign_copies): Likewise.
7606         (get_combine_src_dest): Likewise.
7607         (is_parallel_of_n_reg_sets): Likewise.
7608         (try_combine): Likewise.
7609         (record_dead_and_set_regs_1): Likewise.
7610         (reg_dead_at_p_1): Likewise.
7611         (reg_dead_at_p): Likewise.
7612         * dce.c (deletable_insn_p): Likewise.
7613         (mark_nonreg_stores_1): Likewise.
7614         (mark_nonreg_stores_2): Likewise.
7615         * df-scan.c (df_find_hard_reg_defs): Likewise.
7616         (df_uses_record): Likewise.
7617         (df_get_call_refs): Likewise.
7618         * dwarf2out.c (mem_loc_descriptor): Likewise.
7619         * haifa-sched.c (haifa_classify_rtx): Likewise.
7620         * ira-build.c (create_insn_allocnos): Likewise.
7621         * ira-costs.c (scan_one_insn): Likewise.
7622         * ira.c (equiv_init_movable_p): Likewise.
7623         (rtx_moveable_p): Likewise.
7624         (interesting_dest_for_shprep): Likewise.
7625         * jump.c (mark_jump_label_1): Likewise.
7626         * postreload-gcse.c (record_opr_changes): Likewise.
7627         * postreload.c (reload_cse_simplify): Likewise.
7628         (struct reg_use): Add source expr.
7629         (reload_combine): Check for clobber high.
7630         (reload_combine_note_use): Likewise.
7631         (reload_cse_move2add): Likewise.
7632         (move2add_note_store): Likewise.
7633         * print-rtl.c (print_pattern): Likewise.
7634         * recog.c (decode_asm_operands): Likewise.
7635         (store_data_bypass_p): Likewise.
7636         (if_test_bypass_p): Likewise.
7637         * regcprop.c (kill_clobbered_value): Likewise.
7638         (kill_set_value): Likewise.
7639         * reginfo.c (reg_scan_mark_refs): Likewise.
7640         * reload1.c (maybe_fix_stack_asms): Likewise.
7641         (eliminate_regs_1): Likewise.
7642         (elimination_effects): Likewise.
7643         (mark_not_eliminable): Likewise.
7644         (scan_paradoxical_subregs): Likewise.
7645         (forget_old_reloads_1): Likewise.
7646         * reorg.c (find_end_label): Likewise.
7647         (try_merge_delay_insns): Likewise.
7648         (redundant_insn): Likewise.
7649         (own_thread_p): Likewise.
7650         (fill_simple_delay_slots): Likewise.
7651         (fill_slots_from_thread): Likewise.
7652         (dbr_schedule): Likewise.
7653         * resource.c (update_live_status): Likewise.
7654         (mark_referenced_resources): Likewise.
7655         (mark_set_resources): Likewise.
7656         * rtl.c (copy_rtx): Likewise.
7657         * rtlanal.c (reg_referenced_p): Likewise.
7658         (single_set_2): Likewise.
7659         (noop_move_p): Likewise.
7660         (note_stores): Likewise.
7661         * sched-deps.c (sched_analyze_reg): Likewise.
7662         (sched_analyze_insn): Likewise.
7664 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7666         * cse.c (invalidate_reg): New function extracted from...
7667         (invalidate): ...here.
7668         (canonicalize_insn): Check for clobber high.
7669         (invalidate_from_clobbers): invalidate clobber highs.
7670         (invalidate_from_sets_and_clobbers): Likewise.
7671         (count_reg_usage): Check for clobber high.
7672         (insn_live_p): Likewise.
7673         * cselib.c (cselib_expand_value_rtx_1):Likewise.
7674         (cselib_invalidate_regno): Check for clobber in setter.
7675         (cselib_invalidate_rtx): Pass through setter.
7676         (cselib_invalidate_rtx_note_stores):
7677         (cselib_process_insn): Check for clobber high.
7678         * cselib.h (cselib_invalidate_rtx): Add operand.
7680 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7682         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7683         (mark_not_eliminable): Likewise.
7684         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7685         * lra-lives.c (process_bb_lives): Check for clobber high.
7686         * lra.c (new_insn_reg): Remember clobber highs.
7687         (collect_non_operand_hard_regs): Check for clobber high.
7688         (lra_set_insn_recog_data): Likewise.
7689         (add_regs_to_insn_regno_info): Likewise.
7690         (lra_update_insn_regno_info): Likewise.
7692 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7694         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7695         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7697 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7699         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7700         (copy_insn_1): Likewise.
7701         (gen_hard_reg_clobber_high): New gen function.
7702         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7703         * genemit.c (gen_exp): Likewise.
7704         (gen_emit_seq): Pass through info.
7705         (gen_insn): Check for CLOBBER_HIGH.
7706         (gen_expand): Pass through info.
7707         (gen_split): Likewise.
7708         (output_add_clobbers): Likewise.
7709         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7710         (remove_clobbers): Likewise.
7711         * rtl.h (gen_hard_reg_clobber_high): New declaration.
7713 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7715         * doc/rtl.texi (clobber_high): Add.
7716         (parallel): Add in clobber high
7717         * rtl.c (rtl_check_failed_code3): Add function.
7718         * rtl.def (CLOBBER_HIGH): Add expression.
7719         * rtl.h (RTL_CHECKC3): Add macro.
7720         (rtl_check_failed_code3): Add declaration.
7721         (XC3EXP): Add macro.
7723 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
7725         PR target/86386
7726         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7727         cfun->machine->max_used_stack_alignment if needed.
7729 2018-08-04  Martin Sebor  <msebor@redhat.com>
7731         PR tree-optimization/86571
7732         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7733         NaN output to 4.
7735 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
7737         PR target/86799
7738         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7739         Define.
7741 2018-08-03  Jeff Law  <law@redhat.com>
7743         PR target/86795
7744         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7745         Define to speculation_safe_value_not_needed.
7747 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
7749         * doc/gcov.texi (-x): Remove duplicate "to".
7750         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7751         (-Wif-not-aligned): Remove duplicate "is".
7752         (-flto): Remove duplicate "the".
7753         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7754         duplicate "v5.00.b".
7755         (MSP430 Options): Remove duplicate "and" from the description
7756         of "-mgprel-sec=regexp".
7757         (x86 Options): Remove duplicate copies of "vmldLog102" and
7758         vmlsLog104 from description of "-mveclibabi=type".
7760 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
7762         * internal-fn.h (first_commutative_argument): Declare.
7763         * internal-fn.c (first_commutative_argument): New function.
7764         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7765         restrictions for pattern statements.  Use first_commutative_argument
7766         to look for commutative operands in calls to internal functions.
7768 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
7770         * Makefile.in (wide-int-range.o): New.
7771         * tree-vrp.c: Move all the wide_int_* functions to...
7772         * wide-int-range.cc: ...here.
7773         * tree-vrp.h: Move all the wide_int_* prototypes to...
7774         * wide-int-range.h: ...here.
7776 2018-08-03  Tom de Vries  <tdevries@suse.de>
7778         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
7779         UI_NONE.
7780         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
7781         * except.c (output_function_exception_table): Do early exit if
7782         targetm_common.except_unwind_info (&global_options) == UI_NONE.
7784 2018-08-03  Martin Liska  <mliska@suse.cz>
7786         * predict.c (dump_prediction): Change to 2 digits
7787         in fraction part.
7789 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7791         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
7792         neon_dup_q to...
7793         (falkor_am_1_gtov_gtov): ... a new insn reservation.
7795 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
7797         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
7798         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
7799         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
7800         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
7801         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
7803 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7805         * diagnostic-show-locus.c (diagnostic_show_locus): Use
7806         pp_take_prefix when saving the existing prefix.
7807         * diagnostic.c (diagnostic_append_note): Likewise.
7808         * langhooks.c (lhd_print_error_function): Likewise.
7809         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
7810         param's type.  Free the existing prefix.
7811         (pp_take_prefix): New function.
7812         (pretty_printer::pretty_printer): Drop the prefix parameter.
7813         Rename the length parameter to match the comment.
7814         (pretty_printer::~pretty_printer): Free the prefix.
7815         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
7816         parameter.
7817         (struct pretty_printer): Drop the "const" from "prefix" field's
7818         type and clarify memory management.
7819         (pp_set_prefix): Drop the "const" from the 2nd param.
7820         (pp_take_prefix): New decl.
7822 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
7824         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
7825         for word_mode_ok here instead of passing as argument.
7826         (expand_block_compare): Change select_block_compare_mode() call.
7827         (expand_strncmp_gpr_sequence): New function.
7828         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
7830 2018-08-02  Jeff Law  <law@redhat.com>
7832         PR target/86790
7833         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7834         Define to speculation_safe_value_not_needed.
7836         PR target/86784
7837         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7838         Define to speculation_safe_value_not_needed.
7840 2018-08-02  Tom de Vries  <tdevries@suse.de>
7842         PR target/86660
7843         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
7844         function.  Return UI_TARGET unconditionally.
7845         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
7846         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
7848 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7850         * genemit.c (print_overload_test): Fix typo.
7852 2018-08-02  Richard Biener  <rguenther@suse.de>
7854         PR tree-optimization/86816
7855         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
7856         which checks for value availability before querying it.
7857         (gvn_uses_equal): Use it.
7858         (same_succ_hash): Likewise.
7859         (gimple_equal_p): Likewise.
7861 2018-08-02  Nick Clifton  <nickc@redhat.com>
7863         PR target/86813
7864         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7865         Define to speculation_safe_value_not_needed.
7867         PR target/86810
7868         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7869         Define to speculation_safe_value_not_needed.
7871         PR target/86810
7872         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7873         Define to speculation_safe_value_not_needed.
7875         PR target/86803
7876         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7877         Define to speculation_safe_value_not_needed.
7879         PR target/86797
7880         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7881         Define to speculation_safe_value_not_needed.
7883         PR target/86791
7884         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7885         Define to speculation_safe_value_not_needed.
7887         PR target/86789
7888         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7889         Define to speculation_safe_value_not_needed.
7891         PR target/86787
7892         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7893         Define to speculation_safe_value_not_needed.
7895         PR target/86782
7896         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7897         speculation_safe_value_not_needed.
7899         PR target/86781
7900         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
7901         to speculation_safe_value_not_needed.
7903 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7905         * doc/md.texi: Expand the documentation of instruction names
7906         to mention port-local uses.  Document '@' in pattern names.
7907         * read-md.h (overloaded_instance, overloaded_name): New structs.
7908         (mapping): Declare.
7909         (md_reader::handle_overloaded_name): New member function.
7910         (md_reader::get_overloads): Likewise.
7911         (md_reader::m_first_overload): New member variable.
7912         (md_reader::m_next_overload_ptr): Likewise.
7913         (md_reader::m_overloads_htab): Likewise.
7914         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
7915         m_next_overload_ptr and m_overloads_htab.
7916         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
7917         (get_mode_token, get_code_token, get_int_token): New functions.
7918         (map_attr_string): Add an optional argument that passes back
7919         the associated iterator.
7920         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
7921         (md_reader::handle_overloaded_name, add_overload_instance): New
7922         functions.
7923         (apply_iterators): Handle '@' names.  Report an error if '@'
7924         is used without iterators.
7925         (initialize_iterators): Initialize the new iterator_group fields.
7926         * genopinit.c (handle_overloaded_code_for)
7927         (handle_overloaded_gen): New functions.
7928         (main): Use them to print declarations of maybe_code_for_* and
7929         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
7930         * genemit.c (print_overload_arguments, print_overload_test)
7931         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
7932         (main): Use it to print definitions of maybe_code_for_* and
7933         maybe_gen_* functions.
7934         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
7935         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
7936         instead of explicit mode checks.
7937         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
7938         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
7939         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
7940         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
7941         (aarch64_expand_compare_and_swap): Likewise
7942         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
7943         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
7944         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
7945         (aarch64_constant_pool_reload_icode): Delete.
7946         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
7947         instead of aarch64_constant_pool_reload_icode.  Use
7948         code_for_aarch64_reload_mov instead of explicit mode checks.
7949         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
7950         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
7951         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
7952         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
7953         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
7954         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
7955         (aarch64_atomic_load_op_code): Delete.
7956         (aarch64_emit_atomic_load_op): Likewise.
7957         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
7958         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
7959         instead of aarch64_emit_atomic_load_op.
7960         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
7961         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
7962         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
7963         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
7964         character before the pattern name.
7965         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
7966         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
7967         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
7968         (aarch64_frecps<mode>): Likewise.
7969         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
7970         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
7971         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
7972         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
7973         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
7975 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7977         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
7978         Allow HFmode constants if TARGET_FP_F16INST.
7980 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
7982         PR target/86014
7983         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
7984         No longer check last store for clobber of address register.
7986 2018-08-02  Martin Liska  <mliska@suse.cz>
7988         PR gcov-profile/86817
7989         * gcov.c (process_all_functions): New function.
7990         (main): Call it.
7991         (process_file): Move functions processing to
7992         process_all_functions.
7994 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7996         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
7997         "const" to the "gimple *" and "rtx_insn *" parameters.
7998         * dumpfile.h (dump_user_location_t::dump_user_location_t):
7999         Likewise.
8000         (dump_location_t::dump_location_t): Likewise.
8002 2018-08-01  Martin Sebor  <msebor@redhat.com>
8004         PR tree-optimization/86650
8005         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
8006         rather than a "gcall *".  Directly pass the data of interest
8007         to percent_K_format, rather than building a temporary CALL_EXPR
8008         to hold it.
8009         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
8010         (gimple_fold_builtin_strncat): Adjust.
8011         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
8012         gcall* argument with gimple*.
8013         * gimple-ssa-warn-restrict.c (check_call): Same.
8014         (wrestrict_dom_walker::before_dom_children): Same.
8015         (builtin_access::builtin_access): Same.
8016         (check_bounds_or_overlap): Same
8017         (maybe_diag_overlap): Same.
8018         (maybe_diag_offset_bounds): Same.
8019         * tree-diagnostic.c (default_tree_printer): Move usage of
8020         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
8021         to this callsite.
8022         * tree-pretty-print.c (percent_K_format): Add argument.
8023         * tree-pretty-print.h: Add argument.
8024         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
8025         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
8026         (maybe_diag_stxncpy_trunc): Same.
8027         (handle_builtin_stxncpy): Same.
8028         (handle_builtin_strcat): Same.
8030 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8032         * match.pd: Optimise pointer range checks.
8034 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8036         PR tree-optimization/86758
8037         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
8038         to remove pattern statements.
8040 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8042         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
8043         result of dfs_enumerate_from when constructing stmt_vec_infos,
8044         instead of additionally calling get_loop_body.
8046 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8048         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
8049         parameter.
8050         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
8051         When creating an iv, assert that the step is not known to be zero.
8052         (vect_setup_realignment): Update call accordingly.
8053         * tree-vect-stmts.c (vectorizable_store): Likewise.
8054         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
8056 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8058         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
8059         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
8060         (vectorizable_reduction): Likewise.
8061         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8062         (vect_detect_hybrid_slp_stmts): Likewise.
8063         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8065 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
8067         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
8068         (wide_int_set_zero_nonzero_bits): ...this.
8069         (zero_nonzero_bits_from_vr): Rename to...
8070         (vrp_set_zero_nonzero_bits): ...this.
8071         (extract_range_from_multiplicative_op_1): Abstract wide int
8072         code...
8073         (wide_int_range_multiplicative_op): ...here.
8074         (extract_range_from_binary_expr_1): Extract wide int binary
8075         operations into their own functions.
8076         (wide_int_range_lshift): New.
8077         (wide_int_range_can_optimize_bit_op): New.
8078         (wide_int_range_shift_undefined_p): New.
8079         (wide_int_range_bit_xor): New.
8080         (wide_int_range_bit_ior): New.
8081         (wide_int_range_bit_and): New.
8082         (wide_int_range_trunc_mod): New.
8083         (extract_range_into_wide_ints): New.
8084         (vrp_shift_undefined_p): New.
8085         (extract_range_from_multiplicative_op): New.
8086         (vrp_can_optimize_bit_op): New.
8087         * tree-vrp.h (value_range::dump): New.
8088         (wide_int_range_multiplicative_op): New.
8089         (wide_int_range_lshift):New.
8090         (wide_int_range_shift_undefined_p): New.
8091         (wide_int_range_bit_xor): New.
8092         (wide_int_range_bit_ior): New.
8093         (wide_int_range_bit_and): New.
8094         (wide_int_range_trunc_mod): New.
8095         (zero_nonzero_bits_from_bounds): Rename to...
8096         (wide_int_set_zero_nonzero_bits): ...this.
8097         (zero_nonzero_bits_from_vr): Rename to...
8098         (vrp_set_zero_nonzero_bits): ...this.
8099         (range_easy_mask_min_max): Rename to...
8100         (wide_int_range_can_optimize_bit_op): this.
8101         * vr-values.c (simplify_bit_ops_using_ranges): Rename
8102         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
8104 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8106         * tree-vectorizer.h (vect_orig_stmt): New function.
8107         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
8108         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
8109         (vect_create_epilog_for_reduction): Likewise.
8110         (vectorizable_live_operation): Likewise.
8111         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8112         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
8113         * tree-vect-stmts.c (vectorizable_call): Likewise.
8114         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
8116 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8118         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
8119         argument.
8120         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
8121         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
8122         (vect_transform_loop): Likewise.
8123         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8125 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8127         * tree-vectorizer.h (vect_schedule_slp): Return void.
8128         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8129         (vect_schedule_slp): Likewise.
8131 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8133         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
8134         argument.
8135         (vect_transform_loop): Update calls accordingly.  Schedule SLP
8136         instances before the main loop, if any exist.
8138 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8140         PR tree-optimization/86749
8141         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
8142         If the lhs is used in a COND_EXPR, check that it is being used
8143         as the "then" or "else" value.
8145 2018-08-01  Tom de Vries  <tdevries@suse.de>
8147         PR target/86800
8148         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8149         speculation_safe_value_not_needed.
8151 2018-08-01  Richard Biener  <rguenther@suse.de>
8153         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
8154         as base and offset.
8156 2018-08-01  Martin Liska  <mliska@suse.cz>
8158         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
8159         format how successful transformation is dumped.
8160         (gimple_mod_pow2_value_transform): Likewise.
8161         (gimple_mod_subtract_transform): Likewise.
8162         (gimple_stringops_transform): Likewise.
8164 2018-08-01  Martin Liska  <mliska@suse.cz>
8166         PR value-prof/35543
8167         * value-prof.c (interesting_stringop_to_profile_p):
8168         Simplify the code and add BUILT_IN_MEMMOVE.
8169         (gimple_stringops_transform): Likewise.
8171 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
8173         * config/aarch64/aarch64-simd.md
8174         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
8175         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
8176         use GPI iterator instead of SI mode.
8178 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8180         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
8181         rs6000_speculation_barrier.
8182         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
8183         new barrier pattern name.
8185 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8187         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
8188         (speculation_barrier): New insn.
8190 2018-08-01  Richard Biener  <rguenther@suse.de>
8192         PR bootstrap/86724
8193         * graphite.h: Include isl/id.h and isl/space.h to allow build
8194         with ISL 0.20.
8196 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
8198         PR target/86651
8199         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8200         mode for COFF targets.
8201         * defaults.h (TARGET_COFF): Define.
8202         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
8203         TARGET_COFF): Define.
8204         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
8205         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
8206         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
8208 2018-07-31  Alexandre Oliva <oliva@adacore.com>
8210         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
8211         * gimple-streamer-out.c (output_bb): Save it.
8212         * lto-streamer-in.c (input_struct_function_base): Restore
8213         instance discriminator if available.  Create map on demand.
8214         * lto-streamer-out.c (output_struct_function_base): Save it if
8215         available.
8216         * final.c (decl_to_instance_map): Document LTO strategy.
8218 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
8219             Olivier Hainque  <hainque@adacore.com>
8221         * debug.h (decl_to_instance_map_t): New type.
8222         (decl_to_instance_map): Declare.
8223         (maybe_create_decl_to_instance_map): New inline function.
8224         * final.c (bb_discriminator, last_bb_discriminator): New statics,
8225         to track basic block discriminators.
8226         (final_start_function_1): Initialize them.
8227         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
8228         bb_discriminator.
8229         (decl_to_instance_map): New variable.
8230         (map_decl_to_instance, maybe_set_discriminator): New functions.
8231         (notice_source_line): Set discriminator.
8233 2018-07-31  Ian Lance Taylor  <iant@golang.org>
8235         * targhooks.c (default_have_speculation_safe_value): Add
8236         ATTRIBUTE_UNUSED.
8238 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8240         * dump-context.h: Include "pretty-print.h".
8241         (dump_context::refresh_dumps_are_enabled): New decl.
8242         (dump_context::emit_item): New decl.
8243         (class dump_context): Add fields "m_test_pp" and
8244         "m_test_pp_flags".
8245         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8246         (temp_dump_context::get_dumped_text): New decl.
8247         (class temp_dump_context): Add field "m_pp".
8248         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
8249         (dump_context::refresh_dumps_are_enabled): ...and add a test for
8250         m_test_pp.
8251         (set_dump_file): Update for above change.
8252         (set_alt_dump_file): Likewise.
8253         (dump_loc): New overload, taking a pretty_printer *.
8254         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
8255         to any test pretty-printer.
8256         (make_item_for_dump_gimple_stmt): New function, adapted from
8257         optinfo::add_gimple_stmt.
8258         (dump_context::dump_gimple_stmt): Call it, and use the result,
8259         eliminating the direct usage of dump_file and alt_dump_file in
8260         favor of indirectly using them via emit_item.
8261         (make_item_for_dump_gimple_expr): New function, adapted from
8262         optinfo::add_gimple_expr.
8263         (dump_context::dump_gimple_expr): Call it, and use the result,
8264         eliminating the direct usage of dump_file and alt_dump_file in
8265         favor of indirectly using them via emit_item.
8266         (make_item_for_dump_generic_expr): New function, adapted from
8267         optinfo::add_tree.
8268         (dump_context::dump_generic_expr): Call it, and use the result,
8269         eliminating the direct usage of dump_file and alt_dump_file in
8270         favor of indirectly using them via emit_item.
8271         (make_item_for_dump_printf_va): New function, adapted from
8272         optinfo::add_printf_va.
8273         (make_item_for_dump_printf): New function.
8274         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
8275         and use the result, eliminating the direct usage of dump_file and
8276         alt_dump_file in favor of indirectly using them via emit_item.
8277         (make_item_for_dump_dec): New function.
8278         (dump_context::dump_dec): Call it, and use the result,
8279         eliminating the direct usage of dump_file and alt_dump_file in
8280         favor of indirectly using them via emit_item.
8281         (make_item_for_dump_symtab_node): New function, adapted from
8282         optinfo::add_symtab_node.
8283         (dump_context::dump_symtab_node): Call it, and use the result,
8284         eliminating the direct usage of dump_file and alt_dump_file in
8285         favor of indirectly using them via emit_item.
8286         (dump_context::begin_scope): Reimplement, avoiding direct usage
8287         of dump_file and alt_dump_file in favor of indirectly using them
8288         via emit_item.
8289         (dump_context::emit_item): New member function.
8290         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8291         Set up test pretty-printer on the underlying context.  Call
8292         refresh_dumps_are_enabled.
8293         (temp_dump_context::~temp_dump_context): Call
8294         refresh_dumps_are_enabled.
8295         (temp_dump_context::get_dumped_text): New member function.
8296         (selftest::verify_dumped_text): New function.
8297         (ASSERT_DUMPED_TEXT_EQ): New macro.
8298         (selftest::test_capture_of_dump_calls): Run all tests twice, with
8299         and then without optinfo enabled.  Add uses of
8300         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
8301         * dumpfile.h: Update comment for the dump_* API.
8302         * optinfo-emit-json.cc
8303         (selftest::test_building_json_from_dump_calls): Update for new
8304         param for temp_dump_context ctor.
8305         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
8306         and "m_owned" field.
8307         (optinfo_item::~optinfo_item): Likewise.
8308         (optinfo::add_item): New member function.
8309         (optinfo::emit): Update comment.
8310         (optinfo::add_string): Delete.
8311         (optinfo::add_printf): Delete.
8312         (optinfo::add_printf_va): Delete.
8313         (optinfo::add_gimple_stmt): Delete.
8314         (optinfo::add_gimple_expr): Delete.
8315         (optinfo::add_tree): Delete.
8316         (optinfo::add_symtab_node): Delete.
8317         (optinfo::add_dec): Delete.
8318         * optinfo.h (class dump_context): New forward decl.
8319         (optinfo::add_item): New decl.
8320         (optinfo::add_string): Delete.
8321         (optinfo::add_printf): Delete.
8322         (optinfo::add_printf_va): Delete.
8323         (optinfo::add_gimple_stmt): Delete.
8324         (optinfo::add_gimple_expr): Delete.
8325         (optinfo::add_tree): Delete.
8326         (optinfo::add_symtab_node): Delete.
8327         (optinfo::add_dec): Delete.
8328         (optinfo::add_poly_int): Delete.
8329         (optinfo_item::optinfo_item): Remove "owned" param.
8330         (class optinfo_item): Remove field "m_owned".
8332 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
8334         PR middle-end/86705
8335         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
8336         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
8337         requested variable alignment.
8338         (expand_one_ssa_partition): Likewise.
8339         (expand_one_var): Likewise.
8341 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8343         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8344         to speculation_safe_value_not_needed.
8346 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8348         * targhooks.h (speculation_safe_value_not_needed): New prototype.
8349         * targhooks.c (speculation_safe_value_not_needed): New function.
8350         * target.def (have_speculation_safe_value): Update documentation.
8351         * doc/tm.texi: Regenerated.
8353 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8355         * config/aarch64/iterators.md (ALLI_TI): New iterator.
8356         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
8357         expand.
8358         (despeculate_copy<ALLI:mode>_insn): New insn.
8359         (despeculate_copyti_insn): New insn.
8360         (despeculate_simple<ALLI:mode>): New insn
8361         (despeculate_simpleti): New insn.
8362         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
8363         function.
8364         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
8365         aarch64_speculation_safe_value.
8366         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
8368 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8370         * config/aarch64/aarch64-speculation.cc: New file.
8371         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
8372         before pass_reorder_blocks.
8373         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
8374         prototype.
8375         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
8376         X14 and X15 when tracking speculation.
8377         * config/aarch64/aarch64.md (register name constants): Add
8378         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
8379         (unspec): Add UNSPEC_SPECULATION_TRACKER.
8380         (speculation_barrier): New insn attribute.
8381         (cmp<mode>): Allow SP in comparisons.
8382         (speculation_tracker): New insn.
8383         (speculation_barrier): Add speculation_barrier attribute.
8384         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
8385         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
8386         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
8388 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8390         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
8391         aarch64_track_speculation is true.
8392         (tb<optab><mode>1): Likewise.
8393         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
8394         generate CB[N]Z when tracking speculation.
8395         (aarch64_split_compare_and_swap): Likewise.
8396         (aarch64_split_atomic_op): Likewise.
8398 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8400         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8402 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8404         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8405         (speculation_barrier): New insn.
8407 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8409         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8410         * config/arm/arm.md (speculation_barrier): New expand.
8411         (speculation_barrier_insn): New pattern.
8413 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8415         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8416         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8417         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8418         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8419         list.
8420         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8421         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8422         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8423         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8424         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8425         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8426         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8427         * builtins.c (expand_speculation_safe_value): New function.
8428         (expand_builtin): Call it.
8429         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8430         * doc/extend.texi: Document __builtin_speculation_safe_value.
8431         * doc/md.texi: Document "speculation_barrier" pattern.
8432         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8433         TARGET_HAVE_SPECULATION_SAFE_VALUE.
8434         * doc/tm.texi: Regenerated.
8435         * target.def (have_speculation_safe_value, speculation_safe_value): New
8436         hooks.
8437         * targhooks.c (default_have_speculation_safe_value): New function.
8438         (default_speculation_safe_value): New function.
8439         * targhooks.h (default_have_speculation_safe_value): Add prototype.
8440         (default_speculation_safe_value): Add prototype.
8442 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8444         * dump-context.h (dump_context::dump_loc): New decl.
8445         * dumpfile.c (dump_context::dump_loc): New member function.
8446         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8447         and dump_gimple_stmt.
8448         (dump_context::dump_gimple_expr_loc): Likewise, using
8449         dump_gimple_expr.
8450         (dump_context::dump_generic_expr_loc): Likewise, using
8451         dump_generic_expr.
8452         (dump_context::dump_printf_loc_va): Likewise, using
8453         dump_printf_va.
8454         (dump_context::begin_scope): Explicitly using the global function
8455         "dump_loc", rather than the member function.
8457 2018-07-31  Martin Sebor  <msebor@redhat.com>
8459         PR tree-optimization/86741
8460         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8462 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
8464         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8465         depend on whether prefetch instructions will be emitted or not.
8466         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8467         will be emitted or not.
8468         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8469         (TARGET_SETMEM_PFD): New macros.
8471 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8473         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8474         (NULL_STMT_VEC_INFO): Delete.
8475         (stmt_vec_info::operator*): Likewise.
8476         (stmt_vec_info::operator gimple *): Likewise.
8477         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8478         of NULL_STMT_VEC_INFO.
8479         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8480         (vect_reassociating_reduction_p): Likewise.
8481         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8482         (vectorizable_store): Likewise.
8483         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8484         (vec_info::free_stmt_vec_infos): Likewise.
8486 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8488         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8489         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8491 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8493         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8494         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8495         (vec_info::free_stmt_vec_info): New private member functions.
8496         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8497         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8498         * tree-parloops.c (gather_scalar_reductions): Remove calls to
8499         set_stmt_vec_info_vec and free_stmt_vec_infos.
8500         * tree-vect-loop.c (_loop_vec_info): Remove call to
8501         set_stmt_vec_info_vec.
8502         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8503         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8504         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8505         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8506         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
8507         assignments in {vec_info::,}new_stmt_vec_info that are redundant
8508         with the clearing in the xcalloc.
8509         (stmt_vec_info_vec): Delete.
8510         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8511         (vectorize_loops): Likewise.
8512         (vec_info::~vec_info): Remove argument from call to
8513         free_stmt_vec_infos.
8514         (vec_info::add_stmt): Remove vinfo argument from call to
8515         new_stmt_vec_info.
8517 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8519         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8520         rather than a gimple stmt.
8521         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
8522         information for pattern statements when passed the original
8523         statement; instead wait to be passed the pattern statement itself.
8524         Don't call set_vinfo_for_stmt here.
8525         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8526         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8527         stmt_vec_infos here.
8528         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8529         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8530         stmt_vec_infos entry.
8532 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8534         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8535         * tree-vectorizer.c (vec_info::replace_stmt): New function.
8536         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8537         * tree-vect-stmts.c (vectorizable_call): Likewise.
8538         (vectorizable_simd_clone_call): Likewise.
8540 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8542         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8543         * tree-vectorizer.c (vec_info::remove_stmt): New function.
8544         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8545         * tree-vect-loop.c (vect_transform_loop): Likewise.
8546         * tree-vect-slp.c (vect_schedule_slp): Likewise.
8547         * tree-vect-stmts.c (vect_remove_stores): Likewise.
8549 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8551         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8552         (vect_dr_stmt): Delete.
8553         * tree-vectorizer.c (vec_info::lookup_dr): New function.
8554         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8555         of DR_VECT_AUX.
8556         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8557         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8558         (vect_verify_datarefs_alignment, vect_peeling_supportable)
8559         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8560         (vect_analyze_data_refs): Likewise.
8561         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
8562         argument.
8563         (vect_find_same_alignment_drs): Likewise.
8564         (vect_slp_analyze_node_dependences): Update calls accordingly.
8565         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8566         instead of DR_VECT_AUX.
8567         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8568         of a vector data references.  Use vec_info::lookup_dr instead of
8569         DR_VECT_AUX.
8570         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8571         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8572         instead of DR_VECT_AUX.
8574 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8576         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8577         dr_vec_info.
8578         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8579         accordingly.
8580         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8581         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8582         (vect_gen_prolog_loop_niters): Likewise.
8584 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8586         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8587         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8588         (vect_known_alignment_in_bytes, vect_dr_behavior)
8589         (vect_get_scalar_dr_size): Take references as dr_vec_infos
8590         instead of data_references.  Update calls to other routines for
8591         which the same change has been made.
8592         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8593         dr_vec_infos instead of stmt_vec_infos.
8594         (vect_analyze_data_ref_dependence): Update call accordingly.
8595         (vect_slp_analyze_data_ref_dependence)
8596         (vect_record_base_alignments): Use DR_VECT_AUX.
8597         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8598         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8599         (vector_alignment_reachable_p, vect_get_data_access_cost)
8600         (vect_peeling_supportable, vect_analyze_group_access_1)
8601         (vect_analyze_group_access, vect_analyze_data_ref_access)
8602         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8603         (vect_compile_time_alias, vect_small_gap_p)
8604         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8605         (vect_supportable_dr_alignment): Take references as dr_vec_infos
8606         instead of data_references.  Update calls to other routines for
8607         which the same change has been made.
8608         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8609         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8610         (vect_slp_analyze_and_verify_node_alignment)
8611         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8612         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8613         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
8614         above changes.
8615         (_vect_peel_info::dr): Replace with...
8616         (_vect_peel_info::dr_info): ...this new field.
8617         (vect_peeling_hash_get_most_frequent)
8618         (vect_peeling_hash_choose_best_peeling): Update accordingly.
8619         (vect_peeling_hash_get_lowest_cost):
8620         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
8621         routines for which the same change has been made.
8622         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
8623         data_reference.
8624         * tree-vect-loop-manip.c (get_misalign_in_elems)
8625         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
8626         above changes.
8627         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8628         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8629         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8630         (get_group_load_store_type, get_negative_load_store_type)
8631         (vect_get_data_ptr_increment, vectorizable_store)
8632         (vectorizable_load): Likewise.
8633         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8634         Update calls to other routines for which the same change has been made.
8636 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8638         * tree-vectorizer.h (vec_info::move_dr): New member function.
8639         (dataref_aux): Rename to...
8640         (dr_vec_info): ...this and add "dr" and "stmt" fields.
8641         (_stmt_vec_info::dr_aux): Update accordingly.
8642         (_stmt_vec_info::data_ref_info): Delete.
8643         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8644         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8645         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8646         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8647         of data_ref.
8648         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
8649         (STMT_VINFO_DR_INFO): New macro.
8650         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8651         (set_dr_misalignment): Update after rename of dataref_aux.
8652         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
8653         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8654         initialization of STMT_VINFO_DATA_REF.
8655         * tree-vectorizer.c (vec_info::move_dr): New function.
8656         * tree-vect-patterns.c (vect_recog_bool_pattern)
8657         (vect_recog_mask_conversion_pattern)
8658         (vect_recog_gather_scatter_pattern): Use it.
8659         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8660         the "dr" and "stmt" fields of dr_vec_info instead of
8661         STMT_VINFO_DATA_REF.
8663 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8665         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8666         (is_pattern_stmt_p): Use it.
8667         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8668         on pattern statements.
8670 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8672         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8673         original stmt as a stmt_vec_info rather than a gimple stmt.
8674         (vect_pattern_recog_1): Take the statement directly as a
8675         stmt_vec_info, rather than via a gimple_stmt_iterator.
8676         Update call to vect_mark_pattern_stmts.
8677         (vect_pattern_recog): Update calls accordingly.
8679 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8681         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8682         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8683         a vect_def_type for the first argument.
8684         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8685         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
8686         operand if it isn't defined by a vectorized statement.
8687         (vect_build_gather_load_calls): Remove the mask_dt argument and
8688         update calls to vect_get_vec_def_for_stmt_copy.
8689         (vectorizable_bswap): Likewise the dt argument.
8690         (vectorizable_call): Update calls to vectorizable_bswap and
8691         vect_get_vec_def_for_stmt_copy.
8692         (vectorizable_simd_clone_call, vectorizable_assignment)
8693         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8694         (vectorizable_comparison): Update calls to
8695         vect_get_vec_def_for_stmt_copy.
8696         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
8697         vect_is_simple_use.
8698         (vect_get_loop_based_defs): Remove dt argument and update call
8699         to vect_get_vec_def_for_stmt_copy.
8700         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8701         and vect_get_vec_def_for_stmt_copy.
8702         (vectorizable_load): Update calls to vect_build_gather_load_calls
8703         and vect_get_vec_def_for_stmt_copy.
8704         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8705         (vectorizable_reduction, vectorizable_live_operation): Update calls
8706         to vect_get_vec_def_for_stmt_copy.
8708 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8710         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8711         and gimple stmt arguments with a stmt_vec_info.
8712         (vect_record_base_alignments): Update calls accordingly.
8713         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8714         and gimple stmt arguments with a stmt_vec_info.
8715         (vect_build_slp_tree_1): Remove vinfo argument and update call
8716         to vect_record_max_nunits.
8717         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8718         and vect_record_max_nunits.
8720 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8722         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8723         file and take a stmt_vec_info instead of a gimple stmt.
8724         (supportable_widening_operation, vect_finish_replace_stmt)
8725         (vect_finish_stmt_generation, vect_get_store_rhs)
8726         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8727         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8728         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8729         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8730         (vect_create_data_ref_ptr, bump_vector_ptr)
8731         (vect_permute_store_chain, vect_setup_realignment)
8732         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8733         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8734         (vectorizable_reduction, vectorizable_induction)
8735         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8736         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8737         than gimple stmts as arguments.
8738         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8739         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8740         (can_group_stmts_p, vect_check_gather_scatter)
8741         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8742         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8743         (vect_permute_load_chain, vect_shift_permute_load_chain)
8744         (vect_transform_grouped_load)
8745         (vect_record_grouped_load_vectors): Likewise.
8746         * tree-vect-loop.c (vect_fixup_reduc_chain)
8747         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8748         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8749         (vectorizable_reduction, vectorizable_induction)
8750         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8751         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8752         (vect_get_load_store_mask): Likewise.
8753         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8754         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8755         * tree-vect-stmts.c (vect_mark_relevant)
8756         (is_simple_and_all_uses_invariant)
8757         (exist_non_indexing_operands_for_use_p, process_use)
8758         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8759         (vect_get_vec_def_for_operand, vect_get_vec_defs)
8760         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8761         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8762         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8763         (get_negative_load_store_type, get_load_store_type)
8764         (vect_check_load_store_mask, vect_check_store_rhs)
8765         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8766         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8767         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8768         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8769         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8770         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8771         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8772         (supportable_widening_operation): Likewise.
8774 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8776         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8777         a stmt_vec_info instead of a gcall.
8778         (vect_check_gather_scatter): Update call accordingly.
8779         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
8780         of a gphi.
8781         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
8782         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
8783         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
8784         instead of a gimple stmt.
8785         (vect_transform_loop): Update calls accordingly.
8786         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
8787         stmt_vec_infos instead of gimple stmts.
8788         (vect_analyze_slp_instance): Update use accordingly.
8789         * tree-vect-stmts.c (read_vector_array, write_vector_array)
8790         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
8791         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
8792         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
8793         (vect_gen_widened_results_half, vect_get_loop_based_defs)
8794         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
8795         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
8796         down to subroutines.
8798 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8800         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
8801         of the worklist from a vector of gimple stmts to a vector of
8802         stmt_vec_infos.
8803         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8804         (vect_mark_stmts_to_be_vectorized): Likewise
8806 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8808         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
8809         statement before passing it to vect_analyze_stmt.
8810         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
8811         the chain of phi vector definitions.  Track the exit phi via its
8812         stmt_vec_info.
8813         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
8814         STMT_VINFO_REDUC_DEF.
8815         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
8816         stmt_vec_infos to handle the statement chains.
8817         (vect_get_slp_defs): Record the first statement in the node
8818         using a stmt_vec_info.
8819         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
8820         statements here and pass their stmt_vec_info down to subroutines.
8821         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
8822         down to vect_finish_stmt_generation.
8823         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
8824         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
8825         stmt_vec_infos to subroutines.
8826         (vect_remove_stores): Use stmt_vec_infos to handle the statement
8827         chains.
8829 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8831         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
8832         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
8833         (vect_permute_store_chain, vect_setup_realignment)
8834         (vect_permute_load_chain, vect_shift_permute_load_chain)
8835         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
8836         stmts internally, and when passing values to other vectorizer routines.
8837         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8838         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
8839         (vect_analyze_loop_operations, get_initial_def_for_reduction)
8840         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
8841         (vectorizable_reduction, vectorizable_induction)
8842         (vectorizable_live_operation, vect_transform_loop_stmt)
8843         (vect_transform_loop): Likewise.
8844         * tree-vect-patterns.c (vect_reassociating_reduction_p)
8845         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
8846         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
8847         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8848         (vect_slp_analyze_node_operations_1): Likewise.
8849         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8850         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
8851         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
8852         (vect_finish_stmt_generation_1, get_group_load_store_type)
8853         (get_load_store_type, vect_build_gather_load_calls)
8854         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8855         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8856         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8857         (vectorizable_store, vectorizable_load, vectorizable_condition)
8858         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
8859         (supportable_widening_operation): Likewise.
8860         (vect_get_vector_types_for_stmt): Likewise.
8861         * tree-vectorizer.h (vect_dr_behavior): Likewise.
8863 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8865         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8866         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
8867         (vect_permute_store_chain, vect_permute_load_chain)
8868         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
8869         repeated stmt_vec_info lookups.
8870         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8871         (vect_update_ivs_after_vectorizer): Likewise.
8872         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
8873         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8874         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
8875         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8876         (vect_bb_slp_scalar_cost): Likewise.
8877         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
8879 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8881         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
8882         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
8883         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
8884         to get gassigns and gcalls, rather than operating on generc gimple
8885         stmts.
8886         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
8887         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
8888         (vectorizable_load, vect_analyze_stmt): Likewise.
8889         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
8891 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8893         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
8894         return stmt_vec_infos rather than gimple stmts.  Do not accept
8895         null arguments.
8896         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8897         of a gimple stmt.
8898         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8899         Update use of get_later_stmt.
8900         (vect_get_constant_vectors): Update call accordingly.
8901         (vect_schedule_slp_instance): Likewise
8902         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
8903         (vect_slp_analyze_instance_dependence): Likewise.
8904         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
8906 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8908         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
8909         (stmt_info_for_cost::stmt_info): ...this new field.
8910         (add_stmt_costs): Update accordingly.
8911         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8912         (vect_get_known_peeling_cost): Likewise.
8913         (vect_estimate_min_profitable_iters): Likewise.
8914         * tree-vect-stmts.c (record_stmt_cost): Likewise.
8916 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8918         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
8919         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8920         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8921         accordingly.
8922         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
8924 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8926         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
8927         a gimple stmt to a stmt_vec_info.
8928         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8930 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8932         * tree-vectorizer.h (vec_info::grouped_stores): Change from
8933         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8934         (_loop_vec_info::reduction_chains): Likewise.
8935         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
8936         accordingly.
8937         * tree-vect-slp.c (vect_analyze_slp): Likewise.
8939 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8941         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
8942         a gimple stmt to a stmt_vec_info.
8943         (_stmt_vec_info::next_element): Likewise.
8944         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
8945         (vect_slp_analyze_and_verify_node_alignment)
8946         (vect_analyze_group_access_1, vect_analyze_group_access)
8947         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
8948         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
8949         (vect_supportable_dr_alignment): Update accordingly.
8950         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
8951         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
8952         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
8953         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8954         * tree-vect-slp.c (vect_build_slp_tree_1)
8955         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
8956         (vect_split_slp_store_group, vect_analyze_slp_instance)
8957         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
8958         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
8959         (get_group_load_store_type, get_load_store_type)
8960         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
8961         (vect_transform_stmt, vect_remove_stores): Likewise.
8963 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8965         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
8966         than a gimple stmt.
8967         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8968         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
8969         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
8970         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
8971         (vector_alignment_reachable_p, vect_get_data_access_cost)
8972         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
8973         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
8974         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8975         (vect_analyze_group_access_1, vect_analyze_group_access)
8976         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
8977         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
8978         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
8979         result of vect_dr_stmt and use the stmt_vec_info instead of
8980         the associated gimple stmt.
8981         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8982         (vect_gen_prolog_loop_niters): Likewise.
8983         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8985 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8987         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
8988         to a vec<stmt_vec_info>.
8989         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
8990         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
8991         vec<stmt_vec_info>.
8992         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
8993         to a vec<stmt_vec_info>.
8994         (bst_traits::value_type, bst_traits::value_type): Likewise.
8995         (bst_traits::hash): Update accordingly.
8996         (vect_get_and_check_slp_defs): Change the stmts parameter from
8997         a vec<gimple *> to a vec<stmt_vec_info>.
8998         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
8999         (vect_build_slp_tree): Likewise.
9000         (vect_build_slp_tree_2): Likewise.  Update uses of
9001         SLP_TREE_SCALAR_STMTS.
9002         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
9003         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
9004         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
9005         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
9006         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
9007         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
9008         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
9009         (vect_get_constant_vectors, vect_get_slp_defs)
9010         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
9011         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
9012         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
9013         instead of gimple stmts.
9014         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
9015         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
9016         (vect_slp_analyze_instance_dependence): Update uses of
9017         SLP_TREE_SCALAR_STMTS.
9018         (vect_slp_analyze_and_verify_node_alignment): Likewise.
9019         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
9020         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
9021         (get_initial_defs_for_reduction): Likewise.
9022         (vect_create_epilog_for_reduction): Likewise.
9023         (vectorize_fold_left_reduction): Likewise.
9024         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
9025         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
9026         (can_vectorize_live_stmts): Likewise.
9028 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9030         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
9031         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9032         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
9033         than gimple stmts.
9034         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
9035         of a gimple stmt.
9036         (gather_scalar_reductions): Update after above interface changes.
9037         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
9038         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
9039         than gimple stmts.
9040         (vect_force_simple_reduction): Likewise.
9041         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
9042         LOOP_VINFO_REDUCTIONS.
9043         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9045 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9047         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
9048         a gimple stmt to a stmt_vec_info.
9049         * tree-vect-loop.c (vect_active_double_reduction_p)
9050         (vect_force_simple_reduction, vectorizable_reduction): Update
9051         accordingly.
9053 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9055         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
9056         vec<gimple *> to a vec<stmt_vec_info>.
9057         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
9058         the reduction_phis argument from a vec<gimple *> to a
9059         vec<stmt_vec_info>.
9060         (vectorizable_reduction): Likewise the phis local variable that
9061         is passed to vect_create_epilog_for_reduction.  Update for new type
9062         of SLP_TREE_VEC_STMTS.
9063         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
9064         (vectorizable_live_operation): Likewise.
9065         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
9066         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
9068 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9070         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
9071         a gimple stmt to a stmt_vec_info.
9072         (vectorizable_condition, vectorizable_live_operation)
9073         (vectorizable_reduction, vectorizable_induction): Pass back the
9074         vectorized statement as a stmt_vec_info.
9075         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
9076         use of STMT_VINFO_VEC_STMT.
9077         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
9078         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
9079         as stmt_vec_infos rather than gimple stmts.
9080         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
9081         to a stmt_vec_info.
9082         (vectorizable_live_operation): Likewise.
9083         (vectorizable_reduction, vectorizable_induction): Likewise,
9084         updating use of STMT_VINFO_VEC_STMT.
9085         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
9086         of STMT_VINFO_VEC_STMT.
9087         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
9088         (vectorizable_simd_clone_call, vectorizable_conversion)
9089         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9090         (vectorizable_store, vectorizable_load, vectorizable_condition)
9091         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
9092         from a gimple stmt to a stmt_vec_info.
9093         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
9094         pointer to a stmt_vec_info to the vectorizable_* routines.
9096 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9098         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
9099         a gimple stmt to a stmt_vec_info.
9100         (is_pattern_stmt_p): Update accordingly.
9101         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
9102         (vect_record_grouped_load_vectors): Likewise.
9103         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
9104         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
9105         (vect_model_reduction_cost): Likewise.
9106         (vect_create_epilog_for_reduction): Likewise.
9107         (vectorizable_reduction, vectorizable_induction): Likewise.
9108         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9109         Return the stmt_vec_info for the pattern statement.
9110         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
9111         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
9112         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9113         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
9114         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
9115         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
9116         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
9117         (free_stmt_vec_info, vect_is_simple_use): Likewise.
9119 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9121         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
9122         (vect_finish_stmt_generation): Likewise.
9123         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9124         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
9125         (vect_build_gather_load_calls): Use the return value of the above
9126         functions instead of a separate call to vinfo_for_stmt.  Use narrow
9127         scopes for the input gimple stmt and wider scopes for the associated
9128         stmt_vec_info.  Use vec_info::lookup_def when setting these
9129         stmt_vec_infos from an SSA_NAME definition.
9130         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9131         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9132         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9133         (vectorizable_store, vectorizable_load, vectorizable_condition)
9134         (vectorizable_comparison): Likewise.
9135         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
9136         (vectorizable_reduction): Likewise.
9138 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9140         * tree-vectorizer.h (vect_is_simple_use): Add an optional
9141         stmt_vec_info * parameter before the optional gimple **.
9142         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9143         (process_use, vect_get_vec_def_for_operand_1): Update callers.
9144         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
9145         * tree-vect-loop.c (vectorizable_reduction): Likewise.
9146         (vectorizable_live_operation): Likewise.
9147         * tree-vect-patterns.c (type_conversion_p): Likewise.
9148         (vect_look_through_possible_promotion): Likewise.
9149         (vect_recog_rotate_pattern): Likewise.
9150         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9152 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9154         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
9155         a typedef to a wrapper class.
9156         (NULL_STMT_VEC_INFO): New macro.
9157         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
9158         (stmt_vec_info::operator*): New function.
9159         (stmt_vec_info::operator gimple *): Likewise.
9160         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
9161         (add_stmt_costs): Likewise.
9162         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
9163         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9164         (vect_get_known_peeling_cost): Likewise.
9165         (vect_estimate_min_profitable_iters): Likewise.
9166         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9167         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
9168         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9169         (vectorizable_store, free_stmt_vec_infos): Likewise.
9170         (new_stmt_vec_info): Change return type of xcalloc to
9171         _stmt_vec_info *.
9173 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9175         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
9176         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
9177         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
9178         a single_imm_use-based sequence.
9179         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
9181 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9183         * tree-vectorizer.h (vec_info::lookup_def): Declare.
9184         * tree-vectorizer.c (vec_info::lookup_def): New function.
9185         * tree-vect-patterns.c (vect_get_internal_def): Use it.
9186         (vect_widened_op_tree): Likewise.
9187         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9188         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
9189         (vectorizable_reduction): Likewise.
9190         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
9191         of a gimple *.
9192         (vect_is_slp_reduction): Update calls accordingly.  Use
9193         vec_info::lookup_def.
9194         (vect_is_simple_reduction): Likewise
9195         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
9197 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9199         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
9200         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
9201         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
9202         of vinfo_for_stmt.
9203         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
9204         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
9205         (vect_update_vf_for_slp, vect_analyze_loop_operations)
9206         (vect_is_slp_reduction, vectorizable_induction)
9207         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
9208         * tree-vect-patterns.c (vect_init_pattern_stmt):
9209         (vect_determine_min_output_precision_1, vect_determine_precisions)
9210         (vect_pattern_recog): Likewise.
9211         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
9212         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
9213         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
9214         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9215         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
9216         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
9217         info field from a loop to a loop_vec_info.
9219 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9221         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
9222         (vec_info::add_stmt): Declare.
9223         * tree-vectorizer.c (vec_info::add_stmt): New function.
9224         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
9225         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
9226         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9227         (vectorizable_induction): Likewise.
9228         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
9229         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9230         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
9231         (vectorizable_load): Likewise.
9232         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9233         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
9234         (vect_recog_gather_scatter_pattern): Likewise.
9235         (append_pattern_def_seq): Likewise.  Remove a check that is
9236         performed by add_stmt itself.
9238 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9240         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
9241         which make_ssa_name was called with new_stmt before new_stmt
9242         had been created.
9244 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9246         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
9247         split out from...
9248         (vect_is_slp_reduction): ...here...
9249         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
9250         that are already known to be false.
9252 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9254         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
9255         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
9256         STMT_VINFO_NUM_SLP_USES when it's true.
9257         (vect_free_slp_instance): Add a final_p parameter and pass it to
9258         vect_free_slp_tree.
9259         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
9260         (vect_analyze_slp_instance): Likewise.
9261         (vect_slp_analyze_operations): Likewise.
9262         (vect_slp_analyze_bb_1): Likewise.
9263         * tree-vectorizer.c (vec_info): Likewise.
9264         * tree-vect-loop.c (vect_transform_loop): Likewise.
9266 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9268         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9269         function is not called for second and subsequent members of
9270         a reduction group.
9272 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9274         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
9275         cases for nested loops from here to ...
9276         (vect_create_epilog_for_reduction): ...here.  Only call
9277         vect_is_simple_use for inner-loop reductions.
9279 2018-07-31  Martin Liska  <mliska@suse.cz>
9281         PR gcov-profile/85338
9282         PR gcov-profile/85350
9283         PR gcov-profile/85372
9284         * profile.c (struct location_triplet): New.
9285         (struct location_triplet_hash): Likewise.
9286         (output_location): Do not output a BB that
9287         is already recorded for a line.
9288         (branch_prob): Use streamed_locations.
9290 2018-07-31  Martin Liska  <mliska@suse.cz>
9292         PR gcov-profile/85370
9293         * coverage.c (coverage_begin_function): Do not mark target
9294         clones as artificial functions.
9296 2018-07-31  Martin Liska  <mliska@suse.cz>
9298         PR gcov-profile/83813
9299         PR gcov-profile/84758
9300         PR gcov-profile/85217
9301         PR gcov-profile/85332
9302         * profile.c (branch_prob): Do not record GOTO expressions
9303         for GIMPLE statements which locations are already streamed.
9305 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9307         * gcc.c (handle_spec_function): Accept a soft_matched_part
9308         argument, as do_spec_1.  Pass it down to ...
9309         (eval_spec_function): Accept a soft_matched_part argument,
9310         and pass it down to ...
9311         (do_spec_2): Accept a soft_matched_part argument, and pass
9312         it down to do_spec_1.
9313         (do_spec_1): Pass soft_matched_part to handle_spec_function.
9314         (handle_braces): Update call to handle_spec_function.
9315         (driver::set_up_specs): Update calls to do_spec_2.
9316         (compare_debug_dump_opt_spec_function): Likewise.
9317         (compare_debug_self_opt_spec_function): Likewise.
9319 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9321         * common.opt (nolibc): New option.
9322         * doc/invoke.texi (Link Options): Document it.
9323         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
9324         * config/alpha/linux.h: Likewise.
9325         * config/arc/elf.h: Likewise.
9326         * config/arm/uclinux-elf.h: Likewise.
9327         * config/arm/unknown-elf.h: Likewise.
9328         * config/avr/avrlibc.h: Likewise.
9329         * config/bfin/bfin.h: Likewise.
9330         * config/bfin/linux.h: Likewise.
9331         * config/bfin/uclinux.h: Likewise.
9332         * config/darwin.h: Likewise.
9333         * config/darwin10.h: Likewise.
9334         * config/darwin12.h: Likewise.
9335         * config/gnu-user.h: Likewise.
9336         * config/lm32/uclinux-elf.h: Likewise.
9337         * config/pa/pa-hpux11.h: Likewise.
9338         * config/pa/pa64-hpux.h: Likewise.
9339         * config/sparc/sparc.h: Likewise.
9341 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9343         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
9344         undefined variables.
9346 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9348         PR target/86640
9349         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
9350         instead of GEN_INT.
9352 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9354         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9355         terminated string literal.
9357 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9359         PR rtl-optimization/85160
9360         * combine.c (is_just_move): New function.
9361         (try_combine): Allow combining two instructions into two if neither of
9362         the original instructions was a move.
9364 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
9366         PR target/86673
9367         * doc/extend.texi (Global Register Variables): Discourage use of type
9368         qualifiers.
9369         (Local Register Variables): Likewise.
9371 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
9373         PR tree-optimization/86506
9374         * hwint.h (ceil_log2): Resync with hwint.c implementation.
9376 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
9378         PR target/86547
9379         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
9380         hard_regno, make sure no insn between `from` and `to` clobbers it.
9382 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
9383             Tom de Vries  <tdevries@suse.de>
9385         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
9386         (PTX_DEFAULT_RUNTIME_DIM): ... this.
9387         (nvptx_goacc_validate_dims): Set default worker and gang dims to
9388         PTX_DEFAULT_RUNTIME_DIM.
9389         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
9391 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
9393         * config/pa/pa.c (pa_output_addr_vec): Align address table.
9394         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
9395         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9397 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
9399         * config/rs6000/constraints.md (wG constraint): Delete, no longer
9400         used.
9401         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9402         predicate to reflect toc fusion has been deleted.
9403         (toc_fusion_mem_raw): Delete, no longer used.
9404         (toc_fusion_mem_wrapped): Likewise.
9405         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9406         fusion mask bit.
9407         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9408         Delete, no longer used.
9409         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9410         meant to be used for toc fusion.
9411         (rs6000_debug_print_mode): Delete toc fusion debugging.
9412         (rs6000_debug_reg_global): Likewise.
9413         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9414         fusion and secondary reload support that were never used.
9415         (rs6000_option_override_internal): Delete TOC fusion, that was only
9416         partially defined, and it did not work unless you also used the
9417         -mcmodel= switch.
9418         (rs6000_legitimate_address_p): Delete TOC fusion support.
9419         (rs6000_opt_masks): Likewise.
9420         (fusion_wrap_memory_address): Delete function, no longer used.
9421         (fusion_split_address); Delete TOC fusion support.
9422         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9423         longer used with toc fusion being deleted.
9424         (TARGET_TOC_FUSION_FP): Likewise.
9425         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9426         UNSPEC.
9427         (toc fusion spliter): Delete TOC fusion support.
9428         (toc_fusionload_<mode>): Likewise.
9429         (toc_fusionload_di): Likewise.
9430         (fusion_gpr_load_<mode>): Delete generator function, this insn no
9431         longer needs to be named.  Rename predicate to delete TOC fusion.
9432         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9433         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9434         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9435         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9436         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9438 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9440         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9441         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
9442         __int128 in built-in function prototypes.
9443         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9444         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9446 2018-07-27  Martin Sebor  <msebor@redhat.com>
9448         PR tree-optimization/86696
9449         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9450         types, including enums.
9451         (handle_char_store): Be prepared for the above function to fail.
9453 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
9455         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9456         when optimization level is lower than 2 or optimize for size.
9458 2018-07-26  Martin Sebor  <msebor@redhat.com>
9460         PR tree-optimization/86043
9461         PR tree-optimization/86042
9462         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9463         (get_string_cst_length): Rename...
9464         (get_min_string_length): ...to this.  Add argument.
9465         (handle_char_store): Extend to handle multi-character stores by
9466         MEM_REF.
9467         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
9468         * tree.h (initializer_zerop): Add argument.
9470 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
9472         PR middle-end/86660
9473         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9474         declare target to variables if they have always,{to,from,tofrom} map
9475         kinds.
9477 2018-07-26  Martin Liska  <mliska@suse.cz>
9479         PR lto/86548
9480         * lto-wrapper.c: Add linker_output as prefix
9481         for ltrans_output_file.
9483 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
9485         PR rtl-optimization/85805
9486         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9487         value for hard registers if that was written in the same mode.
9489 2018-07-26  Martin Liska  <mliska@suse.cz>
9491         PR gcov-profile/86536
9492         * gcov.c (format_gcov): Use printf format %.*f directly
9493         and do not handle special values.
9495 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9497         * common/config/arc/arc-common.c (arc_option_optimization_table):
9498         Update default optimizations for size.
9500 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9502         * config/arc/arc.md (movsf_insn): Add short instruction selection.
9503         * config/arc/constraints.md (CfZ): New constraint.
9504         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9505         (subsf3_fpu): Likewise.
9506         (cmpsf_fpu): Likewise.
9507         (cmpsf_fpu_uneq): Likewise.
9509 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9511         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9512         accesses as well.
9513         (arc_is_uncached_mem_p): uncached applies to both the variable and
9514         the pointer.
9516 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9518         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9519         register names.
9521 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
9523         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9524         field "m_scopes" from vec to auto_vec.
9526 2018-07-25  Martin Liska  <mliska@suse.cz>
9528         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9529         return type.
9531 2018-07-25  Richard Biener  <rguenther@suse.de>
9533         PR debug/86654
9534         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9535         special wrt context_die late.
9536         (gen_subprogram_die): Re-use DIEs in local scope.
9538 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
9540         PR tree-optimization/86644
9541         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
9543 2018-07-25  Martin Liska  <mliska@suse.cz>
9545         PR middle-end/86645
9546         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9547         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9549 2018-07-25  Martin Liska  <mliska@suse.cz>
9551         PR sanitizer/79635
9552         * params.def: Explain ASan abbreviation and provide
9553         a documentation link.
9555 2018-07-24  Martin Sebor  <msebor@redhat.com>
9557         PR tree-optimization/86622
9558         PR tree-optimization/86532
9559         * builtins.h (string_length): Declare.
9560         * builtins.c (c_strlen): Correct handling of non-constant offsets.
9561         (check_access): Be prepared for non-constant length ranges.
9562         (string_length): Make extern.
9563         * expr.c (string_constant): Only handle the minor non-constant
9564         array index.  Use string_constant to compute the length of
9565         a generic string constant.
9567 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
9569         PR tree-optimization/86618
9570         * tree-vect-stmts.c (vectorizable_call): Don't take the address
9571         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9573 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
9575         PR tree-optimization/86636
9576         * json.cc (json::object::set): Fix comment.  Add assertions.
9577         (json::array::append): Move here from json.h.  Add comment and an
9578         assertion.
9579         (json::string::string): Likewise.
9580         * json.h (json::array::append): Move to json.cc.
9581         (json::string::string): Likewise.
9582         * optinfo-emit-json.cc
9583         (optrecord_json_writer::impl_location_to_json): Assert that we
9584         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9585         wrapper around it.  Expand the location once, rather than three
9586         times.
9587         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9588         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9589         wrappers.
9590         (optrecord_json_writer::optinfo_to_json): Likewise, in four
9591         places.  Fix some overlong lines.
9593 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
9595         * config/aarch64/aarch64-simd.md
9596         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9597         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9598         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9599         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9600         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9601         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9602         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9603         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9604         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9606 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
9608         PR middle-end/86627
9609         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9610         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
9611         and abs_d == d, do the power of two handling if profitable.
9613 2018-07-24  Richard Biener  <rguenther@suse.de>
9615         * match.pd: Add BIT_FIELD_REF canonicalizations.
9617 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9619         PR c/86617
9620         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9622 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9624         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9625         terminated STRING_CST object.
9627 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9629         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9631 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9633         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9634         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9635         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9636         the elements into a register.
9637         (rs6000_split_v4si_init_di_reg): Delete.
9638         (rs6000_split_v4si_init): Delete.
9639         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9640         (vsx_init_v4si): Rewrite as a define_expand.
9642 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9644         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9645         zero_extend argument from memory): New.
9647 2018-07-22  Martin Sebor  <msebor@redhat.com>
9649         PR bootstrap/86621
9650         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9651         diagnosing calls with unknown arguments unless -Walloca-larger-than
9652         is restricted to less than PTRDIFF_MAX bytes.
9654 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
9656         * doc/gcov.texi (Invoking Gcov): Editorial changes.
9658 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9660         * pretty-print.c (text_info::set_location): Remove redundant
9661         "line_table" parameter from call to rich_location::set_range.
9663 2018-07-20  Martin Sebor  <msebor@redhat.com>
9665         PR middle-end/82063
9666         * builtins.c (expand_builtin_alloca): Adjust.
9667         * calls.c (alloc_max_size): Simplify.
9668         * cgraphunit.c (cgraph_node::expand): Adjust.
9669         * common.opt (larger_than_size, warn_frame_larger_than): Remove
9670         variables.
9671         (frame_larger_than_size): Same.
9672         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9673         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9674         Initialize.
9675         * doc/invoke.texi (GCC Command Options): Document option arguments.
9676         Explain byte-size arguments and suffixes.
9677         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9678         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9679         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9680         * doc/options.texi (UInteger): Expand.
9681         (Host_Wide_Int, ByteSize): Document new properties.
9682         * final.c (final_start_function_1): Include sizes in an error message.
9683         * function.c (frame_offset_overflow): Same.
9684         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9685         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9686         Diagnose unbounded alloca calls only for limits of less than
9687         PTRDIFF_MAX.
9688         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
9689         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
9690         for alloca(0).
9691         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
9692         only for limits of less than PTRDIFF_MAX.
9693         * langhooks-def.h (lhd_handle_option): Change function argument
9694         to HOST_WIDE_INT.
9695         * langhooks.c (lhd_handle_option): Same.
9696         * langhooks.h (handle_option): Same.
9697         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9698         ByteSize flags.
9699         (var_type, var_type_struct): Same.
9700         (var_set): Handle ByteSize flag.
9701         * optc-gen.awk: Add comments to output to ease debugging.  Make
9702         use of HOST_WIDE_INT where appropriate.
9703         * opts-gen-save.awk:  Use %lx to format unsigned long.
9704         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9705         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9706         arguments.  Parse bytes-size suffixes.
9707         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9708         (enum_value_to_arg): Same.
9709         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
9710         (handle_option): Adjust.
9711         (generate_option): Change function argument to HOST_WIDE_INT.
9712         (cmdline_handle_error): Adjust.
9713         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9714         (set_option): Change function argument to HOST_WIDE_INT.
9715         (option_enabled): Handle cl_host_wide_int.
9716         (get_option_state): Handle CLVC_SIZE.
9717         (control_warning_option): Same.
9718         * opts.c (common_handle_option): Change function argument to
9719         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
9720         OPT_Wvla_larger_than_.
9721         * opts.h (enum cl_var_type): Add an enumerator.
9722         * stor-layout.c (layout_decl): Print a more meaningful warning.
9723         * toplev.c (output_stack_usage): Adjust.
9725 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
9727         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9728         call to inline_expand_builtin_string_cmp.
9729         (expand_builtin_strcmp): Likewise.
9730         (expand_builtin_strncmp): Likewise.
9731         (inline_string_cmp): Delete the last parameter, change char_type_node
9732         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9733         two operands.
9734         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9735         the inlining expansion on target where the type of the call has same or
9736         narrower precision than unsigned char.
9738 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9740         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9741         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9742         * common.opt (fsave-optimization-record): New option.
9743         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9744         * doc/invoke.texi (-fsave-optimization-record): New option.
9745         * dumpfile.c: Include "optinfo-emit-json.h".
9746         (struct kv_pair): Move to coretypes.h.
9747         (optgroup_options): Make non-static.
9748         (dump_context::end_scope): Call
9749         optimization_records_maybe_pop_dump_scope.
9750         * dumpfile.h (optgroup_options): New decl.
9751         * json.cc: New file.
9752         * json.h: New file.
9753         * optinfo-emit-json.cc: New file.
9754         * optinfo-emit-json.h: New file.
9755         * optinfo.cc: Include "optinfo-emit-json.h".
9756         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9757         (optinfo_enabled_p): Check optimization_records_enabled_p.
9758         (optinfo_wants_inlining_info_p): Likewise.
9759         * optinfo.h: Update comment.
9760         * profile-count.c (profile_quality_as_string): New function.
9761         * profile-count.h (profile_quality_as_string): New decl.
9762         (profile_count::quality): New accessor.
9763         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9764         and optinfo_emit_json_cc_tests.
9765         * selftest.h (selftest::json_cc_tests): New decl.
9766         (selftest::optinfo_emit_json_cc_tests): New decl.
9767         * toplev.c: Include "optinfo-emit-json.h".
9768         (compile_file): Call optimization_records_finish.
9769         (do_compile): Call optimization_records_start.
9770         * tree-ssa-live.c: Include optinfo.h.
9771         (remove_unused_scope_block_p): Retain inlining information if
9772         optinfo_wants_inlining_info_p returns true.
9774 2018-07-20  Richard Biener  <rguenther@suse.de>
9776         PR debug/86585
9777         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9778         to cover -flto-partition=none.
9780 2018-07-20  Martin Liska  <mliska@suse.cz>
9782         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
9783         (get_decl_source_range): Remove unused function.
9785 2018-07-20  Richard Biener  <rguenther@suse.de>
9787         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
9788         (struct vn_phi_s): Likewise.
9789         (struct vn_reference_s): Likewise.
9790         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
9791         for searching the slot of an entry known to be in the hash itself.
9792         (vn_phi_hasher::equal): Likewise.
9793         (vn_reference_hasher::equal): Likewise.
9794         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
9795         globals.
9796         (optimistic_info, current_info): Remove, keeping only valid_info.
9797         (vn_reference_lookup_1): Remove fallback lookup.
9798         (vn_reference_lookup_2): Likewise.
9799         (vn_nary_op_lookup_1): Likewise.
9800         (vn_phi_lookup): Likewise.
9801         (vn_nary_build_or_lookup_1): Make sure to not chain the built
9802         hash element.
9803         (vn_reference_insert): Adjust, chain the inserted hash element
9804         at last_inserted_ref.
9805         (vn_reference_insert_pieces): Likewise.
9806         (visit_reference_op_call): Likewise.
9807         (vn_nary_op_insert_into): Chain the inserted hash element at
9808         last_inserted_nary.
9809         (vn_nary_op_insert_pieces): Adjust.
9810         (vn_nary_op_insert): Likewise.
9811         (vn_nary_op_insert_stmt): Likewise.
9812         (vn_phi_insert): Adjust, chain the inserted hash element at
9813         last_inserted_phi.
9814         (process_scc): Remove clearing and copying the optimistic
9815         table.  Instead remove elements inserted during an optimistic
9816         iteration from the single table we maintain.
9817         (init_scc_vn): Adjust.
9818         (free_scc_vn): Likewise.
9819         (sccvn_dom_walker::record_cond): Likewise.
9820         (sccvn_dom_walker::after_dom_children): Likewise.
9822 2018-07-19  Martin Sebor  <msebor@redhat.com>
9824         PR tree-optimization/84047
9825         PR tree-optimization/83776
9826         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
9827         (check_array_bounds): Call it.
9829 2018-07-19  Martin Sebor  <msebor@redhat.com>
9831         * align.h (align_flags): Use member initialization.
9833 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
9835         * Makefile.in (OBJS): Add optinfo.o.
9836         * coretypes.h (class symtab_node): New forward decl.
9837         (struct cgraph_node): New forward decl.
9838         (class varpool_node): New forward decl.
9839         * dump-context.h: New file.
9840         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
9841         "tree-pass.h".
9842         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
9843         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
9844         (set_alt_dump_file): Likewise.
9845         (dump_context::~dump_context): New dtor.
9846         (dump_gimple_stmt): Move implementation to...
9847         (dump_context::dump_gimple_stmt): ...this new member function.
9848         Add the stmt to any pending optinfo, creating one if need be.
9849         (dump_gimple_stmt_loc): Move implementation to...
9850         (dump_context::dump_gimple_stmt_loc): ...this new member function.
9851         Start a new optinfo and add the stmt to it.
9852         (dump_gimple_expr): Move implementation to...
9853         (dump_context::dump_gimple_expr): ...this new member function.
9854         Add the stmt to any pending optinfo, creating one if need be.
9855         (dump_gimple_expr_loc): Move implementation to...
9856         (dump_context::dump_gimple_expr_loc): ...this new member function.
9857         Start a new optinfo and add the stmt to it.
9858         (dump_generic_expr): Move implementation to...
9859         (dump_context::dump_generic_expr): ...this new member function.
9860         Add the tree to any pending optinfo, creating one if need be.
9861         (dump_generic_expr_loc): Move implementation to...
9862         (dump_context::dump_generic_expr_loc): ...this new member
9863         function.  Add the tree to any pending optinfo, creating one if
9864         need be.
9865         (dump_printf): Move implementation to...
9866         (dump_context::dump_printf_va): ...this new member function.  Add
9867         the text to any pending optinfo, creating one if need be.
9868         (dump_printf_loc): Move implementation to...
9869         (dump_context::dump_printf_loc_va): ...this new member function.
9870         Start a new optinfo and add the stmt to it.
9871         (dump_dec): Move implementation to...
9872         (dump_context::dump_dec): ...this new member function.  Add the
9873         value to any pending optinfo, creating one if need be.
9874         (dump_context::dump_symtab_node): New member function.
9875         (dump_context::get_scope_depth): New member function.
9876         (dump_context::begin_scope): New member function.
9877         (dump_context::end_scope): New member function.
9878         (dump_context::ensure_pending_optinfo): New member function.
9879         (dump_context::begin_next_optinfo): New member function.
9880         (dump_context::end_any_optinfo): New member function.
9881         (dump_context::s_current): New global.
9882         (dump_context::s_default): New global.
9883         (dump_scope_depth): Delete global.
9884         (dumpfile_ensure_any_optinfo_are_flushed): New function.
9885         (dump_symtab_node): New function.
9886         (get_dump_scope_depth): Reimplement in terms of dump_context.
9887         (dump_begin_scope): Likewise.
9888         (dump_end_scope): Likewise.
9889         (selftest::temp_dump_context::temp_dump_context): New ctor.
9890         (selftest::temp_dump_context::~temp_dump_context): New dtor.
9891         (selftest::verify_item): New function.
9892         (ASSERT_IS_TEXT): New macro.
9893         (ASSERT_IS_TREE): New macro.
9894         (ASSERT_IS_GIMPLE): New macro.
9895         (selftest::test_capture_of_dump_calls): New test.
9896         (selftest::dumpfile_c_tests): Call it.
9897         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
9898         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
9899         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
9900         descriptive comment.
9901         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
9902         (dump_node, dump_bb): Move these unrelated decls.
9903         (class dump_manager): Add leading comment.
9904         * optinfo.cc: New file.
9905         * optinfo.h: New file.
9907 2018-07-19  Michael Collison  <michael.collison@arm.com>
9908             Richard Henderson <rth@redhat.com>
9910         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
9911         (subti): Handle op1 zero.
9912         (subvti4, usub4ti4): New.
9913         (*sub<GPI>3_compare1_imm): New.
9914         (sub<GPI>3_carryinCV): New.
9915         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
9916         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
9918 2018-07-19  Michael Collison  <michael.collison@arm.com>
9919             Richard Henderson <rth@redhat.com>
9921         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
9922         (addti3): Create simpler code if low part is already known to be 0.
9923         (addvti4, uaddvti4): New.
9924         (*add<GPI>3_compareC_cconly_imm): New.
9925         (*add<GPI>3_compareC_cconly): New.
9926         (*add<GPI>3_compareC_imm): New.
9927         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
9928         handle constants within this pattern..
9929         (*add<GPI>3_compareV_cconly_imm): New.
9930         (*add<GPI>3_compareV_cconly): New.
9931         (*add<GPI>3_compareV_imm): New.
9932         (add<GPI>3_compareV): New.
9933         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
9934         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
9935         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
9936         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
9937         with 'comparison' operator.
9938         (*add<GPI>3_compareV_cconly_imm): Ditto.
9939         (*add<GPI>3_compareV_cconly): Ditto.
9940         (*add<GPI>3_compareV_imm): Ditto.
9941         (add<GPI>3_compareV): Ditto.
9942         (add<mode>3_carryinC): Ditto.
9943         (*add<mode>3_carryinC_zero): Ditto.
9944         (*add<mode>3_carryinC): Ditto.
9945         (add<mode>3_carryinV): Ditto.
9946         (*add<mode>3_carryinV_zero): Ditto.
9947         (*add<mode>3_carryinV): Ditto.
9949 2018-07-19  Michael Collison  <michael.collison@arm.com>
9950             Richard Henderson <rth@redhat.com>
9952         * config/aarch64/aarch64-modes.def (CC_V): New.
9953         * config/aarch64/aarch64-protos.h
9954         (aarch64_addti_scratch_regs): Declare
9955         (aarch64_subvti_scratch_regs): Declare.
9956         (aarch64_expand_subvti): Declare.
9957         (aarch64_gen_unlikely_cbranch): Declare
9958         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
9959         for signed overflow using CC_Vmode.
9960         (aarch64_get_condition_code_1): Handle CC_Vmode.
9961         (aarch64_gen_unlikely_cbranch): New function.
9962         (aarch64_addti_scratch_regs): New function.
9963         (aarch64_subvti_scratch_regs): New function.
9964         (aarch64_expand_subvti): New function.
9966 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9968         * config/aarch64/aarch64-option-extensions.def: New entry for profile
9969         extension.
9970         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
9971         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
9972         extension.
9974 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9976         PR target/83009
9977         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9978         address check not strict.
9980 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9982         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
9983         Umq with Umn.
9984         (store_pair_lanes<mode>): Likewise.
9985         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
9986         enum value 'ADDR_QUERY_LDP_STP_N'.
9987         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
9988         (aarch64_print_address_internal): Add declaration.
9989         (aarch64_print_ldpstp_address): Remove.
9990         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
9991         (aarch64_print_operand): Change printing of 'y'.
9992         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
9993         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
9994         'true' rather than '1'.
9995         * config/aarch64/constraints.md (Uml): Likewise.
9996         (Uml): Rename to Umn.
9997         (Umq): Remove.
9999 2018-07-19  Richard Biener  <rguenther@suse.de>
10001         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
10002         a trailing array.
10003         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
10004         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
10005         (vn_reference_hasher): Likewise.
10006         (struct vn_tables_s): Remove obstack and alloc-pool members.
10007         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
10008         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
10009         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
10010         (vn_reference_insert_pieces): Likewise.
10011         (alloc_vn_nary_op_noinit): Adjust.
10012         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
10013         (vn_phi_eq): Adjust.
10014         (shared_lookup_phiargs): Remove.
10015         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
10016         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
10017         (visit_reference_op_call): Likewise.
10018         (copy_nary, copy_phi, copy_reference): Remove.
10019         (process_scc): Rewind the obstack when iterating.  Do not
10020         copy the elements to valid_info but just move them from one
10021         hashtable to the other.
10022         (allocate_vn_table): Adjust.
10023         (free_vn_table): Likewise.
10024         (init_scc_vn): Likewise.
10025         (free_scc_vn): Likewise.
10027 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
10029         PR target/86560
10030         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
10031         indirect_return as function type attribute.
10032         (ix86_attribute_table): Change indirect_return to function
10033         type attribute.
10034         * doc/extend.texi: Update indirect_return attribute.
10036 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
10038         * wide-int.h (widest2_int): New.
10039         * gimple-fold.c (arith_overflowed_p): Use it.
10040         * tree.h (widest2_int_cst): New.
10041         * tree-vrp.c (wide_int_binop_overflow): Rename from
10042         vrp_int_const_binop.
10043         Rewrite to work on trees.
10044         (extract_range_from_multiplicative_op_1): Abstract code to...
10045         (wide_int_range_min_max): ...here.
10046         (wide_int_range_cross_product): ...and here.
10047         (extract_range_from_binary_expr_1): Abstract overflow code to...
10048         (wide_int_range_mult_wrapping): ...here.
10049         * tree-vrp.h (wide_int_range_cross_product): New.
10050         (wide_int_range_mult_wrapping): New.
10052 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
10053             Julia Koval  <julia.koval@intel.com>
10055         * config/i386/x86-tune-costs.h (skylake_memcpy,
10056         skylake_memset): Replace rep_prefix with unrolling for size 512.
10058 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
10060         PR middle-end/86544
10061         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
10062         comparision with EQ_EXPR in last stmt.
10064 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10066         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
10067         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
10068         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
10069         previously known as "PowerPC AltiVec Built-in Functions".  Move
10070         some material to new subsubsections "PowerPC AltiVec Built-in
10071         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
10072         ISA 2.07".
10073         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
10074         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
10075         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
10076         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
10078 2018-07-18  Richard Biener  <rguenther@suse.de>
10080         PR tree-optimization/86557
10081         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
10082         EXACT_DIV_EXPR.
10084 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
10086         * config/s390/s390.c (s390_function_profiler): Generate CFI.
10088 2018-07-17  Jeff Law  <law@redhat.com>
10090         * config/arm/arm.c (get_label_padding): Update for recent
10091         changes to label_to_alignment.
10093         PR tree-optimization/86010
10094         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
10096         * config/mips/mips.c (vr4130_align_insns): Update for recent
10097         changes to label_to_alignment.
10099         * config/frv/frv.c (frv_label_align): Update for recent changes
10100         to label_to_alignment.
10102         * config/nios2/nios2.c (nios2_label_align): Update for recent
10103         changes which dropped ALIGN_LABELS_LOG.
10105 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
10107         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
10108         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
10110 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
10112         * config/arc/arc.c (arc_label_align): Use align_labels instead of
10113         deprecated align_labels_log.
10115 2018-07-17  Richard Biener  <rguenther@suse.de>
10117         PR lto/86456
10118         * dwarf2out.c (init_sections_and_labels): Always generate
10119         a debug_line_str_section for early LTO debug.
10120         (dwarf2out_finish): Reset debug_line_str_hash output early.
10121         Bump counter for extra dwarf5 .debug_loc labels to not conflict
10122         with fat LTO part.
10123         (dwarf2out_early_finish): Output debug_line_str.
10125 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10127         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
10128         index register on z196 or later.
10130 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10132         * config/s390/s390.c (s390_default_align): Set default function
10133         alignment to 16.
10134         (s390_override_options_after_change): Call s390_default align.
10135         (s390_option_override_internal): Call s390_default_align.
10136         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
10138 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10140         PR middle-end/86542
10141         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
10142         to _looptemp_ clauses, other than the first two.
10144 2018-07-17  Martin Liska  <mliska@suse.cz>
10146         * opts.c: Do not enable OPT_falign_* for -Os.
10148 2018-07-17  Martin Liska  <mliska@suse.cz>
10150         * align.h (MAX_CODE_ALIGN): New.
10151         (MAX_CODE_ALIGN_VALUE): New.
10152         * common/config/i386/i386-common.c (ix86_handle_option):
10153         (MAX_CODE_ALIGN): Moved to align.h.
10154         * final.c (MAX_CODE_ALIGN): Likewise.
10155         * opts.c (parse_and_check_align_values):
10156         (MAX_CODE_ALIGN): Likewise.
10157         (MAX_CODE_ALIGN_VALUE): Likewise.
10159 2018-07-17  Martin Liska  <mliska@suse.cz>
10161         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
10162         in order to fulfil coding style.
10163         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
10164         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10165         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10166         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
10167         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
10168         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
10169         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
10170         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
10171         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10173 2018-07-17  Martin Liska  <mliska@suse.cz>
10175         * align.h: New file.
10176         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
10177         directly.
10178         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
10179         align_flags of label_to_alignment.
10180         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
10181         align_flags class.
10182         * config/m68k/m68k.c: Do not use removed align_labels_value and
10183         align_loops_value.
10184         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
10185         (LOOP_ALIGN): Likewise.
10186         (LABEL_ALIGN): Likewise.
10187         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10188         Remove not used macro.
10189         (rs6000_loop_align): Change return type to align_flags.
10190         (rs6000_loop_align_max_skip): Remove.
10191         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
10192         Change return type to align_flags.
10193         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10194         Remove not used macro.
10195         (rs6000_loop_align):  Change return type to align_flags.
10196         (rs6000_loop_align_max_skip): Remove.
10197         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
10198         * config/rx/rx-protos.h (rx_align_for_label): Make it
10199         static function.
10200         * config/rx/rx.c (rx_align_for_label): Change return type
10201         to align_flags.
10202         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
10203         macro definitions.
10204         into align_flags class.
10205         (LABEL_ALIGN): Likewise.
10206         (LOOP_ALIGN): Likewise.
10207         * config/s390/s390.c (s390_label_align): Use align_flags
10208         class member.
10209         (s390_asm_output_function_label): Likewise.
10210         * config/sh/sh.c (sh_override_options_after_change):
10211         Use align_flags class directly without macros.
10212         (find_barrier): Likewise.
10213         (barrier_align): Likewise.
10214         (sh_loop_align): Likewise.
10215         * config/spu/spu.c (spu_option_override):
10216         Use align_flags_tuple::get_value instead of removed macros.
10217         (spu_sched_init): Likewise.
10218         * config/spu/spu.h (GTY): Likewise.
10219         * config/visium/visium.c (visium_option_override):
10220         Set "8" as default secondary alignment.
10221         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
10222         in order to guarantee secondary alignment of 8.
10223         * coretypes.h: Include align.h header file.
10224         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
10225         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
10226         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
10227         * doc/tm.texi.in: Likewise.
10228         * final.c (struct label_alignment): Remove not used structure.
10229         (LABEL_ALIGN): Change type to align_flags.
10230         (LOOP_ALIGN): Likewise.
10231         (JUMP_ALIGN): Likewise.
10232         (default_loop_align_max_skip): Remove.
10233         (default_label_align_max_skip): Likewise.
10234         (default_jump_align_max_skip): Likewise.
10235         (default_label_align_after_barrier_max_skip):
10236         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
10237         (LABEL_TO_MAX_SKIP): Remove.
10238         (label_to_alignment): Return align_flags type instead of integer.
10239         (label_to_max_skip): Remove.
10240         (align_fuzz): Use align_flags type.
10241         (compute_alignments): Use align_flags type and use align_flags::max
10242         to combine multiple alignments.
10243         (grow_label_align): Grow vec instead of C array.
10244         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
10245         (shorten_branches):  Use align_flags type and use align_flags::max
10246         to combine multiple alignments.
10247         (final_scan_insn_1): Remove usage of secondary alignment that comes
10248         from label alignment, but instead use proper secondary alignment
10249         which is computed in grow_label_align.
10250         * flags.h (struct align_flags_tuple): Move to align.h.
10251         (struct align_flags): Likewise.
10252         (state_align_loops): Rename to align_loops.
10253         (state_align_jumps): Rename to align_jumps.
10254         (state_align_labels): Rename to align_labels.
10255         (state_align_functions): Rename to align_functions.
10256         (align_loops_log): Remove.
10257         (align_jumps_log): Remove.
10258         (align_labels_log): Remove.
10259         (align_functions_log): Remove.
10260         (align_loops_max_skip): Remove.
10261         (align_jumps_max_skip): Remove.
10262         (align_labels_max_skip): Remove.
10263         (align_functions_max_skip): Remove.
10264         (align_loops_value): Remove.
10265         (align_jumps_value): Remove.
10266         (align_labels_value): Remove.
10267         (align_functions_value): Remove.
10268         * output.h (label_to_alignment): Change return type to align_flags.
10269         (label_to_max_skip): Remove.
10270         * target.def: Remove loop_align_max_skip, label_align_max_skip,
10271         jump_align_max_skip macros.
10272         * targhooks.h (default_loop_align_max_skip): Remove.
10273         (default_label_align_max_skip): Likewise.
10274         (default_jump_align_max_skip): Likewise.
10275         (default_label_align_after_barrier_max_skip): Remove.
10276         * toplev.c (read_log_maxskip): Use ::normalize function.
10277         (parse_N_M): Remove not used argument and also call ::normalize.
10278         (parse_alignment_opts): Do not pass unused arguments.
10279         * varasm.c (assemble_start_function): Use directly align_functions
10280         instead of removed macros.
10281         * system.h: Do not poison removed macros.
10283 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10285         PR middle-end/86539
10286         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
10287         and cond temporaries don't have reference type if iterator has
10288         pointer type.  For init use &for_pre_body instead of pre_p if
10289         for_pre_body is non-empty.
10291 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10293         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
10294         double-double modes to SFmode directly directly.
10295         (trunc<mode>sf2_fprs): Delete.
10297 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10299         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
10300         for conversions between IFmode and the decimal floating point modes.
10301         (init_float128_ieee): Use the correct names for conversions between
10302         KFmode and the decimal floating point modes.
10304 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10306         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
10307         for the conversions between TDmode and IFmode.
10308         (init_float128_ieee): Use more correct names for the conversions
10309         between TDmode and KFmode.
10311 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
10313         PR tree-optimization/86526
10314         * builtins.c (expand_builtin_memcmp): Formatting fixes.
10315         (inline_expand_builtin_string_cmp): Likewise.
10316         (inline_string_cmp): Likewise.  Use c_readstr instead of
10317         builtin_memcpy_read_str.  Add unit_mode temporary.
10319 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10321         PR middle-end/86528
10322         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
10323         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
10325 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10327         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
10328         Alphabetize prototypes of built-in functions, separating out
10329         built-in functions that are listed in this section but should be
10330         described elsewhere.
10332 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
10334         PR target/86511
10335         * expmed.c (emit_store_flag): Do not emit setcc followed by a
10336         conditional move when trapping comparison was split to a
10337         non-trapping one (and vice versa).
10339 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10341         * config/s390/s390.c (s390_function_profiler): Generate nops
10342         instead of profiler call sequences.
10343         * config/s390/s390.opt: Add the new option.
10345 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10347         * config/s390/s390.c (s390_function_profiler): Generate
10348         __mcount_loc section.
10349         * config/s390/s390.opt: Add the new option.
10351 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10353         * common.opt: Add the new warning.
10354         * config/s390/s390.c (s390_function_profiler): Emit "brasl
10355         %r0,__fentry__" when -mfentry is specified.
10356         (s390_option_override_internal): Disallow -mfentry for 31-bit
10357         CPUs.
10358         * config/s390/s390.opt: Add the new option.
10360 2018-07-16  Richard Biener  <rguenther@suse.de>
10362         PR lto/86523
10363         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
10364         for function-local FUNCTION_DECL and RESULT_DECL immediately.
10366 2018-07-16  Martin Liska  <mliska@suse.cz>
10368         PR ipa/86529
10369         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
10370         to ::get_create.
10372 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10374         * config/arc/arcHS.md: Update ARCHS scheduling rules.
10376 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10378         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
10379         for ARCHS4x.
10380         * config/arc/arc-cpus.def (hs4x): New cpu.
10381         (hs4xd): Likewise.
10382         * config/arc/arc-tables.opt: Regenerate.
10383         * config/arc/arc.c (arc_sched_issue_rate): New function.
10384         (TARGET_SCHED_ISSUE_RATE): Define.
10385         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
10386         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
10387         fpu_cvt.
10388         (attr tune): Add ARCHS4x tune values.
10389         (attr tune_dspmpy): Define.
10390         (*tst): Correct instruction type.
10391         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
10392         * config/arc/arcHS4x.md: New file.
10393         * config/arc/fpu.md: Update instruction type attributes.
10394         * config/arc/t-multilib: Regenerate.
10396 2018-07-16  Tom de Vries  <tdevries@suse.de>
10398         PR debug/86455
10399         * var-tracking.c (vt_initialize): Fix pre_dec handling.
10401 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
10403         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10404         early clobber.
10406 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
10408         PR tree-optimization/86514
10409         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10410         conversion to a boolean type from a type with greater precision.
10412 2018-07-16  Tom de Vries  <tdevries@suse.de>
10414         * var-tracking.c (vt_initialize): Print adjusted insn slim if
10415         dump_flags request TDF_SLIM.
10417 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
10419         * fold-const.c (int_const_binop_1): Abstract...
10420         (wide_int_binop): ...wide int code here.
10421         (poly_int_binop): ...poly int code here.
10422         Abstract the rest of int_const_binop_1 into int_const_binop.
10423         * fold-const.h (wide_int_binop): New.
10424         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10425         Remove useless PLUS/MINUS_EXPR case.
10426         (zero_nonzero_bits_from_vr): Move wide int code...
10427         (zero_nonzero_bits_from_bounds): ...here.
10428         (extract_range_from_binary_expr_1): Move mask optimization code...
10429         (range_easy_mask_min_max): ...here.
10430         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10431         (range_easy_mask_min_max): New.
10433 2018-07-15  Jeff Law  <law@redhat.com>
10435         PR target/85993
10436         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10437         block.
10439 2018-07-14  Jim Wilson  <jimw@sifive.com>
10441         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10443 2018-07-14  Paul Koning  <ni1d@arrl.net>
10445         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10447 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
10449         * lto-streamer-out.c (copy_function_or_variable): Dump info about
10450         copying section.
10452 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
10453             Steve Munroe  <munroesj52@gmail.com>
10455         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10456         (_mm_andnot_si128): Likewise.
10457         (_mm_or_si128): Likewise.
10458         (_mm_xor_si128): Likewise.
10460 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
10462         PR middle-end/78809
10463         * builtins.c (expand_builtin_memcmp): Inline the calls first
10464         when result_eq is false.
10465         (expand_builtin_strcmp): Inline the calls first.
10466         (expand_builtin_strncmp): Likewise.
10467         (inline_string_cmp): New routine. Expand a string compare
10468         call by using a sequence of char comparison.
10469         (inline_expand_builtin_string_cmp): New routine. Inline expansion
10470         a call to str(n)cmp/memcmp.
10471         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10472         New option.
10473         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10475 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
10477         * config/arm/driver-arm.c: Include arm-native.h.
10478         (host_detect_local_cpu): Use auto-generated data tables.
10479         (vendors, arm_cpu_table): Delete.  Move part information to ...
10480         * config/arm/arm-cpus.in: ... here.
10481         * config/arm/parsecpu.awk (gen_native): New function.
10482         (vendor, part): New CPU fields.
10483         (END): Add support for building the native CPU detection tables.
10484         * config/arm/t-arm (arm-native.h): Add build rule.
10485         (driver-arm.o): Add dependency on arm-native.h.
10487 2018-07-13  Richard Biener  <rguenther@suse.de>
10489         PR middle-end/85974
10490         * match.pd (addr1 - addr2): Allow either of the operand to
10491         have a conversion.
10493 2018-07-13  Tom de Vries  <tdevries@suse.de>
10495         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10496         in remap_ssa_name.
10498 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
10500         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10501         arrays instead of numbered variables.
10503 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
10505         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10506         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10508 2018-07-13  Richard Biener  <rguenther@suse.de>
10510         PR debug/86452
10511         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10512         instead of get_context_die.
10514 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
10515             Richard Biener  <rguenther@suse.de>
10517         PR middle-end/86489
10518         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10519         that the loop latch destination where phi is defined.
10521 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
10523         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10524         (riscv_expand_epilogue): Add assertion to check interrupt mode.
10525         (riscv_set_current_function): Extract getting interrupt type to new
10526         function.
10527         (riscv_get_interrupt_type): New function.
10528         (riscv_merge_decl_attributes): New function, checking interrupt type is
10529         same.
10530         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10532 2018-07-12  Paul Koning  <ni1d@arrl.net>
10534         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10535         directive.
10537 2018-07-12  Paul Koning  <ni1d@arrl.net>
10539         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10540         zero reference, add doloop_end instead.
10541         * doc/md.texi (decrement_and_branch_until_zero): Remove.
10542         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
10543         detail for doloop_end.
10545 2018-07-12  Martin Sebor  <msebor@redhat.com>
10547         PR c/86453
10548         * attribs.c (decl_attributes): Reject conflicting attributes before
10549         calling attribute handlers.
10551 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10553         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10554          parameter.
10555         (gcc::dump_manager::get_dump_file_name): likewise.
10556         (dump_begin): Likewise.
10557         * dumpfile.h (dump_begin): Update prototype.
10558         (gcc::dump_manager::get_dump_file_name,
10559         gcc::dump_manager::get_dump_file_name): Update prototype.
10561 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10563         * internal-fn.h (vectorizable_internal_fn_p): New function.
10564         * tree-vect-slp.c (compatible_calls_p): Likewise.
10565         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
10566         to internal functions.
10567         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10569 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10571         * fold-const.h (inverse_conditions_p): Declare.
10572         * fold-const.c (inverse_conditions_p): New function.
10573         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
10574         that test the inverse condition of a conditional internal function.
10575         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10576         * internal-fn.c (internal_fn_mask_index): Handle conditional
10577         internal functions.
10578         (vectorized_internal_fn_supported_p): New function.
10579         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10580         (any_pred_load_store): Replace with...
10581         (need_to_predicate): ...this new variable.
10582         (redundant_ssa_names): New variable.
10583         (ifcvt_can_use_mask_load_store): Move initial checks to...
10584         (ifcvt_can_predicate): ...this new function.  Handle tree codes
10585         for which a conditional internal function exists.
10586         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10587         instead of ifcvt_can_use_mask_load_store.  Update after variable
10588         name change.
10589         (predicate_load_or_store): New function, split out from
10590         predicate_mem_writes.
10591         (check_redundant_cond_expr): New function.
10592         (value_available_p): Likewise.
10593         (predicate_rhs_code): Likewise.
10594         (predicate_mem_writes): Rename to...
10595         (predicate_statements): ...this.  Use predicate_load_or_store
10596         and predicate_rhs_code.
10597         (combine_blocks, tree_if_conversion): Update after above name changes.
10598         (ifcvt_local_dce): Handle redundant_ssa_names.
10599         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10600         general conditional functions.
10601         * tree-vect-stmts.c (vectorizable_call): Likewise.
10603 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10604             Alan Hayward  <alan.hayward@arm.com>
10605             David Sherwood  <david.sherwood@arm.com>
10607         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10608         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10609         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10610         plus and minus and convert them into IFN_COND_FMA-based sequences.
10611         (convert_mult_to_fma): Handle conditional plus and minus.
10613 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10615         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10616         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10617         (cond_fnms_optab): New optabs.
10618         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10619         internal functions.
10620         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10621         * internal-fn.h (get_conditional_internal_fn): Declare.
10622         (get_unconditional_internal_fn): Likewise.
10623         * internal-fn.c (cond_ternary_direct): New macro.
10624         (expand_cond_ternary_optab_fn): Likewise.
10625         (direct_cond_ternary_optab_supported_p): Likewise.
10626         (FOR_EACH_COND_FN_PAIR): Likewise.
10627         (get_conditional_internal_fn): New function.
10628         (get_unconditional_internal_fn): Likewise.
10629         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10630         (gimple_match_op::gimple_match_op): Add a new overload for 5
10631         operands.
10632         (gimple_match_op::set_op): Likewise.
10633         (gimple_resimplify5): Declare.
10634         * genmatch.c (decision_tree::gen): Generate simplifications for
10635         5 operands.
10636         * gimple-match-head.c (gimple_simplify): Define an overload for
10637         5 operands.  Handle calls with 5 arguments in the top-level overload.
10638         (convert_conditional_op): Handle conversions from unconditional
10639         internal functions to conditional ones.
10640         (gimple_resimplify5): New function.
10641         (build_call_internal): Pass a fifth operand.
10642         (maybe_push_res_to_seq): Likewise.
10643         (try_conditional_simplification): Try converting conditional
10644         internal functions to unconditional internal functions.
10645         Handle 3-operand unconditional forms.
10646         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10647         Define ternary equivalents of the current rules for binary conditional
10648         internal functions.
10649         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10650         ternary operations.
10651         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10652         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10653         (optab): Handle them.
10654         (SVE_COND_FP_TERNARY): New int iterator.
10655         (sve_fmla_op, sve_fmad_op): New int attributes.
10656         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10657         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10658         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10660 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10662         * target.def (preferred_else_value): New target hook.
10663         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10664         * doc/tm.texi: Regenerate.
10665         * targhooks.h (default_preferred_else_value): Declare.
10666         * targhooks.c (default_preferred_else_value): New function.
10667         * internal-fn.h (conditional_internal_fn_code): Declare.
10668         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10669         (get_conditional_internal_fn): Use it.
10670         (conditional_internal_fn_code): New function.
10671         * gimple-match.h (gimple_match_cond): New struct.
10672         (gimple_match_op): Add a cond member function.
10673         (gimple_match_op::gimple_match_op): Update all forms to take a
10674         gimple_match_cond.
10675         * genmatch.c (expr::gen_transform): Use the same condition as res_op
10676         for the suboperation, but don't specify a particular else_value.
10677         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10678         (visit_nary_op, visit_reference_op_load): Pass
10679         gimple_match_cond::UNCOND to the gimple_match_op constructor.
10680         * gimple-match-head.c: Include tree-eh.h
10681         (convert_conditional_op): New function.
10682         (maybe_resimplify_conditional_op): Likewise.
10683         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10684         (gimple_resimplify2): Likewise.
10685         (gimple_resimplify3): Likewise.
10686         (gimple_resimplify4): Likewise.
10687         (maybe_push_res_to_seq): Return null for conditional operations.
10688         (try_conditional_simplification): New function.
10689         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
10690         constructor.
10691         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10692         IFN_COND_* call.
10693         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10694         function.
10695         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10697 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10699         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10700         DECL_FCONTEXT
10701         (hash_tree): Do not hash DECL_FCONTEXT
10702         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10703         Do not stream DECL_FCONTEXT.
10704         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10705         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10707 2018-07-12  Richard Biener  <rguenther@suse.de>
10709         PR debug/86462
10710         * dwarf2out.c (gen_block_die): Only output blocks when they have
10711         at least one !DECL_IGNORED_P variable.
10713 2018-07-12  Richard Biener  <rguenther@suse.de>
10715         PR target/84829
10716         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10717         Remove -mieee-fp handling.
10719 2018-07-12  Richard Biener  <rguenther@suse.de>
10721         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10722         left-over from last patch.
10724 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
10726         PR tree-optimization/86492
10727         * gimple-ssa-store-merging.c
10728         (imm_store_chain_info::coalesce_immediate_stores): Call
10729         check_no_overlap even for the merge_overlapping case.  Formatting fix.
10731 2018-07-12  Richard Biener  <rguenther@suse.de>
10733         PR middle-end/86479
10734         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10735         move possibly trapping operations into the conditional.
10737 2018-07-12  Richard Biener  <rguenther@suse.de>
10739         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10740         (vn_lookup_simplify_result): Remove recursion limit applied
10741         here.
10742         (vn_nary_build_or_lookup_1): Adjust.
10743         (try_to_simplify): Likewise.
10744         * gimple-match-head.c (gimple_resimplify1): Instead apply one
10745         here.
10746         (gimple_resimplify2): Likewise.
10747         (gimple_resimplify3): Likewise.
10748         (gimple_resimplify4): Likewise.
10750 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
10752         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10753         Use __mmask64 type instead of __mmask8 for __M argument.
10754         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10755         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10756         __U argument.
10757         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10758         __mmask16 for __M argument.
10759         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10760         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10761         to __mmask16 instead of __mmask8.
10762         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10763         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10764         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10765         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10766         instead of __mmask16 for __U argument.
10767         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10768         __mmask16 instead of __mmask8 for __U argument.
10769         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10770         __U argument.
10771         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10772         __mmask16.
10773         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10774         argument.
10775         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10776         __U argument.
10777         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10778         __mmask16.
10779         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
10780         of __mmask16.
10781         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
10782         __U argument.
10783         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
10784         __U argument.
10785         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
10786         __U argument.
10787         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
10788         __U argument.
10789         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
10790         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
10791         return type as well as __M argument type and all casts from __mmask8
10792         to __mmask32.
10793         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
10794         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
10795         return type as well as __M argument type and all casts from __mmask8
10796         to __mmask16.
10797         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
10798         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
10799         return type as well as __M argument type and all casts from __mmask8
10800         to __mmask32.
10801         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
10802         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
10803         return type as well as __M argument type and all casts from __mmask8
10804         to __mmask16.
10805         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
10806         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
10807         __mmask16.
10809 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
10811         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
10812         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
10813         for __U argument.
10815 2018-07-11  Paul Koning  <ni1d@arrl.net>
10817         * doc/md.texi (define_subst): Document how multiple occurrences of
10818         the same argument in the replacement pattern are handled.
10820 2018-07-11  Paul Koning  <ni1d@arrl.net>
10822         * doc/extend.texi (Common Variable Attributes): Move "mode" into
10823         alphabetical order.
10824         (Common Type Attributes): Add "mode" attribute.
10826 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
10828         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
10829         stream DECL_ORIGINAL_TYPE.
10830         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
10831         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
10832         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
10833         Do not walk original type.
10834         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
10835         external decls.
10836         (write_ts_decl_non_common_tree_pointers): Do not stream
10837         DECL_ORIGINAL_TYPE
10838         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
10839         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
10841 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10843         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
10844         thread twice from the same starting edge.
10846 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10848         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
10849         * gimple.c (gimple_call_nonnull_result_p): ...here...
10850         (gimple_call_nonnull_arg): ...and here.
10851         * gimple.h (gimple_call_nonnull_result_p): New.
10852         (gimple_call_nonnull_arg): New.
10854 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
10856         * config/arm/arm-cpus.in: Move information from fpu field of each
10857         cpu definition to the isa field.
10858         * config/arm/parsecpu.awk (fpu): Delete match rule.
10859         (gen_comm_data): Don't add bits from the CPU's FPU entry.
10861 2018-07-11  Richard Biener  <rguenther@suse.de>
10863         PR debug/86457
10864         * dwarf2out.c (init_sections_and_labels): Use
10865         output_asm_line_debug_info consistently.
10866         (dwarf2out_early_finish): Likewise.
10867         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
10868         type units.
10870 2018-07-11  Richard Biener  <rguenther@suse.de>
10872         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
10873         Rework father_bb setting in a way to avoid propagating constants
10874         multiple times on a loop body.
10876 2018-07-10  Mark Wielaard  <mark@klomp.org>
10878         PR debug/86459
10879         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
10881 2018-07-10  Richard Biener  <rguenther@suse.de>
10883         * hash-map.h (hash_map::iterator::operator*): Return
10884         references to key and value.
10886 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
10888         PR c++/86443
10889         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10890         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
10891         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
10892         (gimplify_omp_for): For composite loops, move outer
10893         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
10894         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
10895         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
10896         TREE_LIST for both the original class iterator and the "last" helper
10897         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
10898         loop, remember has_decl_expr from outer composite loops for the
10899         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
10901 2018-07-09  Martin Sebor  <msebor@redhat.com>
10903         PR middle-end/77357
10904         PR middle-end/86428
10905         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
10906         accessing implicitly initialized array elements.
10907         * expr.c (string_constant): Handle string initializers of
10908         character arrays within aggregates.
10909         * gimple-fold.c (fold_array_ctor_reference): Add argument.
10910         Store element offset.  As a special case, handle zero size.
10911         (fold_nonarray_ctor_reference): Same.
10912         (fold_ctor_reference): Add argument.  Store subobject offset.
10913         * gimple-fold.h (fold_ctor_reference): Add argument.
10915 2018-07-09  Paul Koning  <ni1d@arrl.net>
10917         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
10918         (pdp11_insn_cost): New function.
10919         (pdp11_md_asm_adjust): New function.
10920         (TARGET_INVALID_WITHIN_DOLOOP): Define.
10921         (pdp11_rtx_costs): Update to match machine better.
10922         (output_addr_const_pdp11): Correct format mismatch warnings.
10923         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
10924         * config/pdp11/pdp11.md: General change to add base_cost and/or
10925         length attributes for use by new pdp11_insn_cost function.
10926         (MIN_BRANCH): Correct definition.
10927         (MIN_SOB): Ditto.
10928         (doloop_end): Use standard pattern name for looping pattern.
10929         (doloop_end_nocc): New.
10930         (movsf): Add another constraint alternative.
10931         (zero_extendqihi2): Add constraint alternatives for not in place
10932         extend.
10933         (zero_extendhisi2): Remove.
10934         (shift patterns): Add CC handling variants.
10935         (bswaphi2): New.
10936         (bswapsi2): New.
10937         (rothi3): New.
10938         (define_peephole2): New peephole to recognize mov that sets CC for
10939         subsequent test.
10941 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10943         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
10944         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
10945         wi::add.
10947 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
10949         PR c/86420
10950         * real.c (real_nextafter): Return true if result is denormal.
10952 2018-07-09  Martin Liska  <mliska@suse.cz>
10954         * common.opt: Add back wrongly removed attribute.
10956 2018-07-09  Richard Biener  <rguenther@suse.de>
10958         PR debug/86413
10959         * dwarf2out.c (gen_block_die): For an early generated DIE
10960         always output high/low PC attributes.
10962 2018-07-09  Tom de Vries  <tdevries@suse.de>
10964         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
10965         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
10966         onto VAR_DECL with abstract origin.
10968 2018-07-07  Jim Wilson  <jimw@sifive.com>
10970         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
10972 2018-07-07  Tom de Vries  <tdevries@suse.de>
10974         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
10976 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
10978         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
10979         overflow_type.
10980         (combine_bound): Use wide-int overflow calculation instead of
10981         rolling our own.
10982         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
10983         overflow_type.
10984         * fold-const.c (int_const_binop_2): Same.
10985         (extract_muldiv_1): Same.
10986         (fold_div_compare): Same.
10987         (fold_abs_const): Same.
10988         * match.pd: Same.
10989         * poly-int.h (add): Same.
10990         (sub): Same.
10991         (neg): Same.
10992         (mul): Same.
10993         * predict.c (predict_iv_comparison): Same.
10994         * profile-count.c (slow_safe_scale_64bit): Same.
10995         * simplify-rtx.c (simplify_const_binary_operation): Same.
10996         * tree-chrec.c (tree_fold_binomial): Same.
10997         * tree-data-ref.c (split_constant_offset_1): Same.
10998         * tree-if-conv.c (idx_within_array_bound): Same.
10999         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
11000         * tree-ssa-phiopt.c (minmax_replacement): Same.
11001         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
11002         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
11003         * vr-values.c (vr_values::adjust_range_with_scev): Same.
11004         * wide-int.cc (wi::add_large): Same.
11005         (wi::mul_internal): Same.
11006         (wi::sub_large): Same.
11007         (wi::divmod_internal): Same.
11008         * wide-int.h: Change overflow type to overflow_type for neg, add,
11009         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
11010         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
11011         mul_internal, divmod_internal.
11012         (overflow_type): New enum.
11013         (accumulate_overflow): New.
11015 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
11017         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
11018         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
11020 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11022         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
11023         argument is checked for zero before entering loop, avoid checking again.
11025 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11027         * gimplify.h (generic_expr_could_trap_p): Set as global function.
11028         * gimplify.h (generic_expr_could_trap_p): Likwise.
11029         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
11031 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
11033         PR tree-optimization/86401
11034         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
11035         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
11036         (fold_bit_and_mask): ... here.  New helper function for match.pd.
11037         * fold-const.h (fold_bit_and_mask): Declare.
11038         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
11040 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
11042         PR target/86324
11043         * target.def (translate_mode_attribute): New hook.
11044         * targhooks.h (default_translate_mode_attribute): Declare.
11045         * targhooks.c (default_translate_mode_attribute): New function.
11046         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
11047         * doc/tm.texi: Regenerate.
11048         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
11049         (rs6000_translate_mode_attribute): New function.
11051 2018-07-06  Paul Koning  <ni1d@arrl.net>
11053         * doc/md.texi (define_split): Document DONE and FAIL.
11054         (define_peephole2): Ditto.
11056 2018-07-05  Jeff Law  <law@redhat.com>
11058         PR tree-optimization/86010
11059         * tree-ssa-dse.c (compute_trims): More aggressively trim at
11060         both the head and tail of mem* and str* calls.
11062 2018-07-05  Jim Wilson  <jimw@sifive.com>
11064         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
11066 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
11068         * config/aarch64/aarch64-simd.md: correct flags text for
11069         MIN_EXPR replacement.
11071 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
11073         * configure: Regenerated.
11075 2018-07-05  Carl Love  <cel@us.ibm.com>
11077         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
11078         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
11079         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
11080         VSX_BUILTIN_DOUBLEL_V4SF.
11082 2018-07-05  Martin Sebor  <msebor@redhat.com>
11084         PR c++/86400
11085         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
11086         than its domain to compute its the upper bound of a char array.
11088 2018-07-05  Nathan Sidwell  <nathan@acm.org>
11090         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
11091         * doc/cpp.texi: Update comment.
11092         * doc/tm.texi: Rebuilt.
11093         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
11094         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
11095         * doc/extend.texi (Backwards Compatibility): Clarify it is system
11096         headers affected by extern "C".
11097         * system.h: Poison NO_IMPLICIT_EXTERN_C.
11098         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
11099         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
11100         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
11101         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
11102         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
11103         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
11104         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
11105         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
11106         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
11107         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
11108         config/sparc/sp64-elf.h, config/spu/spu.h,
11109         config/stormy16/stormy16.h, config/v850/v850.h,
11110         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
11111         define NO_IMPLICIT_EXTERN_C.
11112         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
11114 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11116         PR target/84711
11117         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11118         instead of GET_MODE_SIZE when comparing Units.
11120 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11122         PR target/84711
11123         * rtlanal.c (set_noop_p): Constrain on mode change,
11124         include hard-reg-set.h
11126 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11128         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
11130 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
11132         Revert
11133         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11135         PR sanitizer/84250
11136         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11137         libasan.
11138         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11140 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11142         PR sanitizer/84250
11143         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11144         libasan.
11145         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11147 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
11149         PR middle-end/86380
11150         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
11152 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
11154         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
11155         neg_*_op* variables.
11157 2018-07-04  Martin Liska  <mliska@suse.cz>
11159         * tree-switch-conversion.c: Define
11160         max_ratio_for_speed and max_ratio_for_size constants.
11162 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11163             Martin Liska  <mliska@suse.cz>
11165         PR middle-end/66240
11166         PR target/45996
11167         PR c/84100
11168         * common.opt: Rename align options with 'str_' prefix.
11169         * common/config/i386/i386-common.c (set_malign_value): New
11170         function.
11171         (ix86_handle_option): Use it to set -falign-* options/
11172         * config/aarch64/aarch64-protos.h (struct tune_params): Change
11173         type from int to string.
11174         * config/aarch64/aarch64.c: Update default values from int
11175         to string.
11176         * config/alpha/alpha.c (alpha_override_options_after_change):
11177         Likewise.
11178         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
11179         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11180         max skip conditionally.
11181         * config/i386/freebsd.h (SUBALIGN_LOG): New.
11182         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11183         max skip conditionally.
11184         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11185         max skip conditionally.
11186         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
11187         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11188         max skip conditionally.
11189         * config/i386/i386.c (struct ptt): Change type from int to
11190         string.
11191         (ix86_default_align): Set default values.
11192         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
11193         max skip conditionally.
11194         * config/i386/iamcu.h (SUBALIGN_LOG): New.
11195         (ASM_OUTPUT_MAX_SKIP_ALIGN):
11196         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
11197         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11198         max skip conditionally.
11199         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
11200         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
11201         * config/i386/x86-64.h (SUBALIGN_LOG): New.
11202         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11203         max skip conditionally.
11204         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11205         * config/ia64/ia64.c (ia64_option_override): Set default values
11206         for alignment options.
11207         * config/m68k/m68k.c: Handle new str_align_* options.
11208         * config/mips/mips.c (mips_set_compression_mode): Change
11209         type of constants.
11210         (mips_option_override): Set default values for options.
11211         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
11212         Likewise.
11213         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11214         Likewise.
11215         * config/rx/rx.c (rx_option_override): Likewise.
11216         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
11217         (LABEL_ALIGN): Use align_labels_log.
11218         (LOOP_ALIGN): Use align_loops_align.
11219         * config/s390/s390.c (s390_asm_output_function_label): Use new
11220         macros.
11221         * config/sh/sh.c (sh_override_options_after_change):
11222         Change type of constants.
11223         * config/spu/spu.c (spu_sched_init): Likewise.
11224         * config/sparc/sparc.c (sparc_option_override): Set default
11225         values for options.
11226         * config/visium/visium.c (visium_option_override): Likewise.
11227         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
11228         emit p2align format with last argument if it's not needed.
11229         * doc/invoke.texi: Document extended format of -falign-*.
11230         * final.c: Use align_labels alignment.
11231         * flags.h (struct target_flag_state): Change type to use
11232         align_flags.
11233         (struct align_flags_tuple): New.
11234         (struct align_flags): Likewise.
11235         (align_loops_log): Redefine macro to use new types.
11236         (align_loops_max_skip): Redefine macro to use new types.
11237         (align_jumps_log): Redefine macro to use new types.
11238         (align_jumps_max_skip): Redefine macro to use new types.
11239         (align_labels_log): Redefine macro to use new types.
11240         (align_labels_max_skip): Redefine macro to use new types.
11241         (align_functions_log): Redefine macro to use new types.
11242         (align_loops): Redefine macro to use new types.
11243         (align_jumps): Redefine macro to use new types.
11244         (align_labels): Redefine macro to use new types.
11245         (align_functions): Redefine macro to use new types.
11246         (align_functions_max_skip): Redefine macro to use new types.
11247         (align_loops_value): New macro.
11248         (align_jumps_value): New macro.
11249         (align_labels_value): New macro.
11250         (align_functions_value): New macro.
11251         * function.c (invoke_set_current_function_hook): Propagate
11252         alignment values from flags to global variables default in
11253         topleev.h.
11254         * ipa-icf.c (sem_function::equals_wpa): Use
11255         cl_optimization_option_eq instead of memcmp.
11256         * lto-streamer.h (cl_optimization_stream_out): Support streaming
11257         of string types.
11258         (cl_optimization_stream_in): Likewise.
11259         * optc-save-gen.awk: Support strings in cl_optimization.
11260         * opth-gen.awk: Likewise.
11261         * opts.c (finish_options): Remove error checking of invalid
11262         value ranges.
11263         (MAX_CODE_ALIGN): Remove.
11264         (MAX_CODE_ALIGN_VALUE): Likewise.
11265         (parse_and_check_align_values): New function.
11266         (check_alignment_argument): Likewise.
11267         (common_handle_option): Use check_alignment_argument.
11268         * opts.h (parse_and_check_align_values): Declare.
11269         * toplev.c (init_alignments): Remove.
11270         (read_log_maxskip): New.
11271         (parse_N_M): Likewise.
11272         (parse_alignment_opts): Likewise.
11273         (backend_init_target): Remove usage of init_alignments.
11274         * toplev.h (parse_alignment_opts): Declare.
11275         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
11276         argument.
11277         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
11278         * tree.c (cl_option_hasher::equal): New.
11279         * varasm.c: Use new global macros.
11281 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11283         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
11284         Use a simpler align directive also if MAXSKIP = ALIGN-1.
11285         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11286         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11287         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11288         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11289         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
11290         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
11291         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
11292         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11293         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11294         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11295         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11297 2018-07-04  Martin Liska  <mliska@suse.cz>
11298             Jonathan Wakely  <jwakely@redhat.com>
11300         * coverage.c: Use correct type.
11301         * doc/invoke.texi: Language correction.
11303 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
11305         PR target/85620
11306         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
11307         ENDBRANCH for non-tail call which may return via indirect branch.
11308         * doc/extend.texi: Document indirect_return attribute.
11310 2018-07-03  Martin Sebor  <msebor@redhat.com>
11312         PR tree-optimization/86274
11313         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
11314         precondition.
11315         (format_floating): Correct handling of infinities and NaNs.
11317 2018-07-03  Martin Sebor  <msebor@redhat.com>
11319         * print-tree.c (print_real_cst): New function.
11320         (print_node_brief): Call it.
11321         (print_node): Ditto.
11323 2018-07-03  Jeff Law  <law@redhat.com>
11325         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
11326         into a single pattern.
11328         * config/h8300/h8300.md (ors code_iterator): New.
11329         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
11330         a single pattern and single splitter.
11331         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
11332         (iorqi3_1, xorqi3_1): Likewise.
11333         (iorqi3, xorqi3 expanders): Similarly.
11335         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
11336         (movmd_internal) into a single pattern using the P mode iterator.
11337         (movmd splitters): Similarly.
11338         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
11339         (movsd splitters): Similarly.
11341         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
11342         ADDB, ADDW and ADDL into a single ADD attribute which selects the
11343         right table based on the size of the operand.
11344         * config/h8300/h8300.md (length_table): Corresponding changes. All
11345         references to "addb", "addw" and "addl" changed to "add".
11346         (btst patterns): Merge two variants into a single pattern.
11347         (tstqi, tsthi): Likewise.
11348         (addhi3_incdec, addsi3_incdec): Likewise.
11349         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
11350         (mulhi3, mulsi3): Likewise.
11351         (udivhi3, udivsi3): Likewise.
11352         (divhi3, divsi3): Likewise.
11353         (andorqi3, andorhi3, andorsi3): Likewise.
11355 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
11357         PR target/85694
11358         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
11359         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
11361 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11363         PR tree-optimization/85694
11364         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
11365         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
11366         UNSPEC_URHADD.
11367         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
11368         (<u>avg<mode>3_ceil): New patterns.
11370 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
11372         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
11373         scan-tree-dump directive.
11374         * gcc.dg/vect/slp-perm-2.c: Likewise.
11375         * gcc.dg/vect/slp-perm-3.c: Likewise.
11376         * gcc.dg/vect/slp-perm-5.c: Likewise.
11377         * gcc.dg/vect/slp-perm-6.c: Likewise.
11378         * gcc.dg/vect/slp-perm-7.c: Likewise.
11379         * gcc.dg/vect/slp-perm-8.c: Likewise.
11381 2018-07-03  Marek Polacek  <polacek@redhat.com>
11383         PR middle-end/86202
11384         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
11386 2018-07-03  Richard Biener  <rguenther@suse.de>
11388         PR ipa/86389
11389         * tree-ssa-structalias.c (find_func_clobbers): Properly
11390         handle indirect calls.
11392 2018-07-03  Jeff Law  <law@redhat.com>
11394         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
11395         (shifts): New code iterator.
11396         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11397         expander.  Fix HImode handling on H8/SX.
11398         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11399         (subqi3, subhi3, subsi3 expanders): Likewise.
11400         (andqi3, andhi3, andsi3 expanders): Likewise.
11401         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11402         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11403         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11404         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11405         (zero_extendqihi2, zero_extendqisi2): Likewise.
11406         (extendqihi2, extendqisi2): Likewise.
11407         (rotlqi3, rotlhi3, rotlsi3): Likewise.
11408         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11409         (rotlqi3_1, rotlhi3_1): Likewise.
11410         (logicalhi3_sn, logicalsi3_sn): Likewise.
11411         (logicalhi3, logicalsi3): Likewise.
11413 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11415         * tree-vect-patterns.c (vect_recog_rotate_pattern)
11416         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11417         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11418         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11419         type to append_pattern_def_seq instead of creating a stmt_vec_info
11420         directly.
11421         (build_mask_conversion): Likewise.  Remove vinfo argument.
11422         (vect_add_conversion_to_patterm): Likewise, renaming to...
11423         (vect_add_conversion_to_pattern): ...this.
11424         (vect_recog_mask_conversion_pattern): Update call to
11425         build_mask_conversion.  Pass the vector type to
11426         append_pattern_def_seq here too.
11427         (vect_recog_gather_scatter_pattern): Update call to
11428         vect_add_conversion_to_pattern.
11430 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11432         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11433         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11434         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11435         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11436         STMT_VINFO_PATTERN_DEF_SEQ to null here.
11437         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11438         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11439         append_pattern_def_seq instead of new_pattern_def_seq.
11440         (vect_recog_divmod_pattern): Do both of the above.
11441         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11442         is null.
11444 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11446         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11447         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11448         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11449         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11450         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11451         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11452         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11453         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11454         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11455         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11456         parameter with a single stmt_vec_info.
11457         (vect_recog_func_ptr): Likewise.
11458         (vect_recog_gather_scatter_pattern): Likewise, folding in...
11459         (vect_try_gather_scatter_pattern): ...this.
11460         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11461         the stmt_vec_info of the statement to be matched.  Don't clear
11462         STMT_VINFO_RELATED_STMT.
11463         (vect_pattern_recog): Update call accordingly.
11465 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11467         PR tree-optimization/85694
11468         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11469         (uavgM3_ceil): Document new optabs.
11470         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11471         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11472         functions.
11473         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11474         (savg_ceil_optab): New optabs.
11475         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11476         (vect_vect_recog_func_ptrs): Add it.
11477         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11478         constant directly from the associated lhs.
11480 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11482         * tree-vect-patterns.c (vect_split_statement): New function.
11483         (vect_convert_input): Use it to try to split an existing cast.
11485 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11487         * poly-int.h (print_hex): New function.
11488         * dumpfile.h (dump_dec, dump_hex): Declare.
11489         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11490         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11491         min_input_precision, operation_precision and operation_sign.
11492         * tree-vect-patterns.c (vect_get_range_info): New function.
11493         (vect_same_loop_or_bb_p, vect_single_imm_use)
11494         (vect_operation_fits_smaller_type): Delete.
11495         (vect_look_through_possible_promotion): Add an optional
11496         single_use_p parameter.
11497         (vect_recog_over_widening_pattern): Rewrite to use new
11498         stmt_vec_info infomration.  Handle one operation at a time.
11499         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11500         (vect_truncatable_operation_p, vect_set_operation_type)
11501         (vect_set_min_input_precision): New functions.
11502         (vect_determine_min_output_precision_1): Likewise.
11503         (vect_determine_min_output_precision): Likewise.
11504         (vect_determine_precisions_from_range): Likewise.
11505         (vect_determine_precisions_from_users): Likewise.
11506         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11507         (vect_vect_recog_func_ptrs): Put over_widening first.
11508         Add cast_forwprop.
11509         (vect_pattern_recog): Call vect_determine_precisions.
11511 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11513         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11514         statements that have been replaced by further pattern statements.
11515         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11517 2018-07-03  Richard Biener  <rguenther@suse.de>
11519         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11520         always set *dt.  Dump vectype in vectype overload.
11521         * dumpfile.h (dump_gimple_expr): New function.
11522         (dump_gimple_expr_loc): Likewise.
11523         * dumpfile.c (dump_gimple_expr): New function.
11524         (dump_gimple_expr_loc): Likewise.
11526 2018-07-02  Jeff Law  <law@redhat.com>
11528         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11529         the H8/300, H8/300H and H8/S variants into a single pattern.
11530         (movhi_h8300, movqi_h8300hs): Similarly.
11531         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11532         (QHI mode iterator): New.
11534         * config/h8300/h8300.md: Remove trailing whitespace.
11536 2018-07-02  Jim Wilson  <jimw@sifive.com>
11538         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11539         instead of emit_insn for interrupt returns.
11540         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11541         (riscv_sret, riscv_uret): Likewise.
11543 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11545         * pretty-print.c (selftest::test_pp_format): Move save and restore
11546         of quotes to class auto_fix_quotes, and add an instance.
11547         * selftest.c: Include "intl.h".
11548         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11549         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11550         * selftest.h (selftest::auto_fix_quotes): New class.
11552 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
11554         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11555         (aarch64_sve_prepare_conditional_op): Remove.
11556         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11557         Allow aarch64_simd_reg_or_zero as select operand; remove
11558         the aarch64_sve_prepare_conditional_op call.
11559         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11560         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11561         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11562         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11563         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11564         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11565         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11566         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11567         and a splitters to match all of the *_any patterns.
11568         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11570         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11571         (SVE_COND_FP_BINARY_REV): Remove.
11572         (sve_int_op_rev, sve_fp_op_rev): New.
11573         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11574         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11575         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11576         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11577         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11578         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11579         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11580         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11581         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11583         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11584         Remove match_dup 1 from the inner unspec.
11585         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11587         * config/aarch64/aarch64.md (movprfx): New attr.
11588         (length): Default movprfx to 8.
11589         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11590         (*madd<SVE_I>, *msub<SVE_I): Likewise.
11591         (*<su>mul<SVE_I>3_highpart): Likewise.
11592         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11593         (*v<ASHIFT><SVE_I>3): Likewise.
11594         (*<su><MAXMIN><SVE_I>3): Likewise.
11595         (*<su><MAXMIN><SVE_F>3): Likewise.
11596         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11597         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11598         (*div<SVE_F>4): Likewise.
11600 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
11602         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11603         in dump string.
11605 2018-07-02  Richard Biener  <rguenther@suse.de>
11607         PR tree-optimization/86363
11608         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11609         memset argument refers to a non-variable address.
11611 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11613         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11614         {PLUS,MINUS}_EXPR code to...
11615         (adjust_symbolic_bound): ...here,
11616         (combine_bound): ...here,
11617         (set_value_range_with_overflow): ...and here.
11619 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11621         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11622         code...
11623         (extract_range_from_abs_expr): ...here.
11625 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
11627         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11628         -fno-omit-frame-pointer when not optimizing.
11630 2018-07-02  Martin Liska  <mliska@suse.cz>
11632         PR ipa/86279
11633         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11634         (propagate_nothrow): Likewise.
11636 2018-07-02  Martin Liska  <mliska@suse.cz>
11638         PR ipa/86323
11639         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11641 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11643         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11644         function in r262149, changing "loc" param from source_location to
11645         const dump_location_t &.
11646         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11647         declaration, as above.
11649 2018-07-01  Paul Koning  <ni1d@arrl.net>
11651         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11652         -munit-asm, -mgnu-asm, -mdec-asm.
11653         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11654         (pdp11_output_labelref): New.
11655         (pdp11_output_def): New.
11656         (pdp11_output_addr_vec_elt): New.
11657         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
11658         %# and %@ format codes.
11659         (pdp11_option_override): New.
11660         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11661         (pdp11_output_ident): New.
11662         (pdp11_asm_named_section): New.
11663         (pdp11_asm_init_sections): New.
11664         (pdp11_file_start): New.
11665         (pdp11_file_end): New.
11666         (output_ascii): Use .ascii/.asciz for -mdec-asm.
11667         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
11668         %o, like %c but octal.
11669         (pdp11_option_override): New.
11670         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11671         -mdec-asm.
11672         (DATA_SECTION_ASM_OP): Ditto.
11673         (READONLY_DATA_SECTION_ASM_OP): New.
11674         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11675         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11676         (ASM_OUTPUT_LABELREF): Ditto.
11677         (ASM_OUTPUT_DEF): Ditto.
11678         (ASM_OUTPUT_EXTERNAL): New.
11679         (ASM_OUTPUT_SOURCE_FILENAME): New.
11680         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11681         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11682         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
11683         %# and %@ format codes.
11684         * config/pdp11/pdp11.opt (mgnu-asm): New.
11685         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11686         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11687         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11689 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
11691         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11692         dereferencing path[] beyond its length.
11693         (debug_path): New.
11694         (debug_all_paths): New.
11695         (rewire_first_differing_edge): New.
11696         (adjust_paths_after_duplication): New.
11697         (duplicate_thread_path): Call adjust_paths_after_duplication.
11698         Add new argument.
11699         (thread_through_all_blocks): Add new argument to
11700         duplicate_thread_path.
11702 2018-06-30  Jim Wilson  <jimw@sifive.com>
11704         * config/riscv/predicates.md (p2m1_shift_operand): New.
11705         (high_mask_shift_operand): New.
11706         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11707         pattern using p2m1_shift_operand.
11708         (lshsi3_zero_extend_3+2): New combiner pattern using
11709         high_mask_shift_operand.
11711 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11713         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11714         split out from...
11715         (vect_recog_rotate_pattern): ...here.
11716         (vect_convert_input): Try to insert casts of invariants in the
11717         preheader.
11718         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11719         preheader to be empty.
11721 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11723         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11724         vector type.  If given, install it in the new statement's
11725         STMT_VINFO_VECTYPE.
11726         (vect_element_precision): New function.
11727         (vect_unpromoted_value): New struct.
11728         (vect_unpromoted_value::vect_unpromoted_value): New function.
11729         (vect_unpromoted_value::set_op): Likewise.
11730         (vect_look_through_possible_promotion): Likewise.
11731         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11732         (vect_widened_op_tree, vect_convert_input): Likewise.
11733         (vect_convert_inputs, vect_convert_output): Likewise.
11734         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11735         to handle the optional cast of the multiplication result and
11736         vect_widened_op_tree to detect the widened multiplication itself.
11737         Do not require the input and output of promotion casts to have
11738         the same sign, but base the signedness of the operation on the
11739         input rather than the result.  If the pattern includes two
11740         promotions, check that those promotions have the same sign.
11741         Do not restrict the MULT_EXPR handling to a double-width result;
11742         handle quadruple-width results and wider.  Use vect_convert_inputs
11743         to convert the inputs to the common type.
11744         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
11745         to handle the optional cast of the ABS result.  Also allow a sign
11746         change or a sign extension between the ABS and MINUS.
11747         Use vect_widened_op_tree to detect the widened subtraction and use
11748         vect_convert_inputs to convert the inputs to the common type.
11749         (vect_handle_widen_op_by_const): Delete.
11750         (vect_recog_widen_op_pattern): New function.
11751         (vect_recog_widen_mult_pattern): Use it.
11752         (vect_recog_widen_shift_pattern): Likewise.
11753         (vect_recog_widen_sum_pattern): Use
11754         vect_look_through_possible_promotion to handle the promoted
11755         PLUS_EXPR operand.
11757 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11759         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11760         the containing gimple_seq *.
11761         * gimple-iterator.h (gsi_for_stmt): Declare it.
11762         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11763         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11764         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11765         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11766         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11767         checks.
11768         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11769         split out from...
11770         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
11771         statement being replaced is part of an existing pattern
11772         definition sequence, inserting the new pattern statements before
11773         the original one.
11774         (vect_pattern_recog_1): Don't return a bool.  If the statement
11775         is already part of a pattern, instead apply pattern matching
11776         to the pattern definition statements.  Don't clear the
11777         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11778         (vect_pattern_recog): Don't break after the first match;
11779         continue processing the pattern definition statements instead.
11780         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
11782 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11784         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11785         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11786         (vect_recog_widen_sum_pattern): Use it.
11788 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11790         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11791         phi is not a pattern statement and has not been replaced by
11792         a pattern statement.
11793         * tree-vect-patterns.c (type_conversion_p): Don't check
11794         STMT_VINFO_IN_PATTERN_P.
11795         (vect_recog_vector_vector_shift_pattern): Likewise.
11796         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
11797         the pattern statement rather than the original statement; check
11798         directly for a WIDEN_MULT_EXPR here.
11799         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
11800         vect_is_simple_use to return the pattern statement rather
11801         than the original statement; use is_pattern_stmt_p to check
11802         for such a pattern statement.
11803         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
11804         to return the pattern statement rather than the original statement;
11805         don't do the same transformation here.
11806         (vect_is_simple_use): If the defining statement has been replaced
11807         by a pattern statement, return the pattern statement instead.
11808         Remove the corresponding (local) transformation from the vectype
11809         overload.
11811 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11813         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
11814         end and default to null.
11815         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11816         (vectorizable_reduction): Update calls accordingly, dropping the
11817         gimple ** argument if the passed-back statement isn't needed.
11818         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
11819         (vect_recog_rotate_pattern): Likewise.
11820         (vect_recog_mask_conversion_pattern): Likewise.
11821         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11822         (vect_mask_constant_operand_p): Likewise.
11823         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
11824         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
11825         (get_group_load_store_type, get_load_store_type): Likewise.
11826         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
11827         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
11828         (vectorizable_conversion, vectorizable_assignment): Likewise.
11829         (vectorizable_shift, vectorizable_operation): Likewise.
11830         (vectorizable_store, vect_is_simple_cond): Likewise.
11831         (vectorizable_condition, vectorizable_comparison): Likewise.
11832         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
11833         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
11834         and move it to the end.  Cope with null def_stmt_outs.
11836 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11838         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
11840 2018-06-29  Jeff Law  <law@redhat.com>
11842         * config/v850/v850.c (v850_legitimate_address_p): Handle large
11843         displacements for TARGET_V850E2V3 and newer.
11844         (TARGET_LRA_P): Remove.  Defaults to LRA now.
11845         * config/v850/v850.md (sign23byte_load): Remove.
11846         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
11847         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
11849 2018-06-29  Martin Liska  <mliska@suse.cz>
11851         PR lto/85759
11852         * coverage.c (coverage_init): Mangle full path name.
11853         * doc/invoke.texi: Document the change.
11854         * gcov-io.c (mangle_path): New.
11855         * gcov-io.h (mangle_path): Likewise.
11856         * gcov.c (mangle_name): Use mangle_path for path mangling.
11858 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11860         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
11861         if starting source register is not even.
11863 2018-06-29  Martin Liska  <mliska@suse.cz>
11865         PR tree-optimization/86263
11866         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
11867         Make edge redirection.
11869 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
11871         * dumpfile.c (dump_loc): Add indentation based on scope depth.
11872         (dump_scope_depth): New variable.
11873         (get_dump_scope_depth): New function.
11874         (dump_begin_scope): New function.
11875         (dump_end_scope): New function.
11876         * dumpfile.h (get_dump_scope_depth): New declaration.
11877         (dump_begin_scope): New declaration.
11878         (dump_end_scope): New declaration.
11879         (class auto_dump_scope): New class.
11880         (AUTO_DUMP_SCOPE): New macro.
11881         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
11882         AUTO_DUMP_SCOPE.
11884 2018-06-29  Richard Biener  <rguenther@suse.de>
11886         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
11887         compute_all_dependences succeeds.
11888         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
11889         exceed --param loop-max-datarefs-for-datadeps.
11891 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
11893         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
11895 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
11897         PR target/86348
11898         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
11899         alternative 0 in preferred_for_speed attribute.
11901 2018-06-28  Paul Koning  <ni1d@arrl.net>
11903         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
11904         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
11905         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
11906         * config/pdp11/pdp11.md: Correct "length" attribute calculation
11907         for shift insn patterns.
11909 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
11911         * cgraph.c (cgraph_node::get_body): Replace assignments to
11912         "dump_file" with calls to set_dump_file.
11913         * dumpfile.c (alt_dump_file): Make static, and group with...
11914         (alt_flags): ...this definition.
11915         (dumps_are_enabled): New variable.
11916         (refresh_dumps_are_enabled): New function.
11917         (set_dump_file): New function.
11918         (set_alt_dump_file): New function.
11919         (gcc::dump_manager::dump_start): Replace assignments to
11920         "dump_file" and "alt_dump_file" with calls to set_dump_file and
11921         set_alt_dump_file.
11922         (gcc::dump_manager::dump_finish): Likewise.
11923         * dumpfile.h (alt_dump_file): Delete decl.
11924         (dumps_are_enabled): New variable decl.
11925         (set_dump_file): New function decl.
11926         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
11927         global.
11928         * tree-nested.c (lower_nested_functions): Replace assignments to
11929         "dump_file" with calls to set_dump_file.
11931 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
11933         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
11934         goto_locus of each outgoing edge of each basic block.
11936 2018-06-28  Richard Biener  <rguenther@suse.de>
11938         * dwarf2out.c (decl_scope_table): Remove.
11939         (push_decl_scope): Likewise.
11940         (pop_decl_scope): Likewise.
11941         (gen_type_die_for_member): Do not call push/pop_decl_scope.
11942         (gen_struct_or_union_type_die): Likewise.
11943         (gen_tagged_type_die): Likewise.
11944         (dwarf2out_init): Do not initialize decl_scope_table.
11945         (dwarf2out_c_finalize): Do not free it.
11947 2018-06-28  Richard Biener  <rguenther@suse.de>
11949         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
11950         deciding whether to not re-use a DIE.
11952 2018-06-28  Richard Biener  <rguenther@suse.de>
11954         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
11955         DW_AT_abstract_origin attribute.
11957 2018-06-28  Martin Liska  <mliska@suse.cz>
11959         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11960         Use newly introduced constants.
11961         * tree-switch-conversion.h (struct jump_table_cluster):
11962         Define max_ratio_for_size and max_ratio_for_speed.
11964 2018-06-28  Martin Liska  <mliska@suse.cz>
11966         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11967         Add new checking assert to catch invalid state.
11968         (jump_table_cluster::can_be_handled): Handle single case
11969         clusters.
11970         (jump_table_cluster::is_beneficial): Bail out for such case.
11971         (bit_test_cluster::find_bit_tests):
11972         Add new checking assert to catch invalid state.
11973         (bit_test_cluster::can_be_handled): Handle single case
11974         clusters.
11975         (bit_test_cluster::is_beneficial): Bail out for such case.
11976         (switch_decision_tree::analyze_switch_statement):
11977         Fix comment.
11979 2018-06-28  Martin Liska  <mliska@suse.cz>
11981         * common.opt: Introduce -completion option.
11982         * gcc.c (driver_handle_option): Handle it.
11983         (driver::main): Print completions if completion
11984         is set.
11985         * opt-suggestions.c (option_proposer::get_completions):
11986         New function.
11987         (option_proposer::suggest_completion): Likewise.
11988         (option_proposer::find_param_completions): Likewise.
11989         (verify_autocompletions): Likewise.
11990         (test_completion_valid_options): Likewise.
11991         (test_completion_valid_params): Likewise.
11992         (in_completion_p): Likewise.
11993         (empty_completion_p): Likewise.
11994         (test_completion_partial_match): Likewise.
11995         (test_completion_garbage): Likewise.
11996         (opt_proposer_c_tests): Likewise.
11997         * opt-suggestions.h: Declare new functions.
11998         * opts.c (common_handle_option): Handle OPT__completion_.
11999         * selftest-run-tests.c (selftest::run_tests): Add
12000         opt_proposer_c_tests.
12001         * selftest.c (assert_str_startswith): New.
12002         * selftest.h (assert_str_startswith): Likewise.
12003         (opt_proposer_c_tests): New.
12004         (ASSERT_STR_STARTSWITH): Likewise.
12006 2018-06-28  Martin Liska  <mliska@suse.cz>
12008         * Makefile.in: Add opt-suggestions.o.
12009         * gcc-main.c: Include opt-suggestions.h.
12010         * gcc.c (driver::driver): Likewise.
12011         (driver::~driver): Remove m_option_suggestions.
12012         (driver::build_option_suggestions): Moved to option_proposer.
12013         (driver::suggest_option): Likewise.
12014         (driver::handle_unrecognized_options): Use option_proposer.
12015         * gcc.h (class driver): Add new memver m_option_proposer.
12016         * opt-suggestions.c: New file.
12017         * opt-suggestions.h: New file.
12019 2018-06-28  Martin Liska  <mliska@suse.cz>
12021         * vec.h (class auto_string_vec): New (moved from auto_argvec).
12022         (auto_string_vec::~auto_string_vec): Likewise.
12024 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
12026         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
12027         prevent_decl_creation_for_types fields up and add reset_location field.
12028         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
12029         statement if id->reset_location is true.
12030         (copy_edges_for_bb): Do not set goto_locus on the new edges if
12031         id->reset_location is true.
12032         (copy_phis_for_bb): Force input_location on the arguments if
12033         id->reset_location is true.
12034         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
12035         is set on the function to be inlined.
12037 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
12039         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
12041 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
12043         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
12044         registers for Pmode.
12045         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
12046         hard registers for the clobbered pseudo.
12048 2018-06-27  Paul Koning  <ni1d@arrl.net>
12050         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12051         mutually exclusive options.
12052         * config/pdp11/constraints.md (h): New constraint.
12053         (O): Update definition to match shift code generation.
12054         (D): New constraint.
12055         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
12056         (CCFP): Remove.
12057         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
12058         function.
12059         (output_jump): Change arguments.
12060         (pdp11_fixed_cc_regs): New function.
12061         (pdp11_cc_mode): Ditto.
12062         (pdp11_expand_shift): Ditto.
12063         (pdp11_assemble_shift): Ditto.
12064         (pdp11_small_shift): Ditto.
12065         (pdp11_branch_cost): Remove.
12066         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
12067         from output.
12068         (pdp11_register_move_cost): Update for CC registers.
12069         (pdp11_rtx_costs): Add case for LSHIFTRT.
12070         (pdp11_output_jump): Add CCNZ mode conditional branches.
12071         (notice_update_cc_on_set): Remove.
12072         (pdp11_cc_mode): New function.
12073         (simple_memory_operand): Correct pre/post decrement case.
12074         (no_side_effect_operand): New function.
12075         (pdp11_regno_reg_class): Add CC_REGS class.
12076         (pdp11_fixed_cc_regs): New function.
12077         (pdp11_small_shift): New function.
12078         (pdp11_expand_shift): New function to expand shift insns.
12079         (pdp11_assemble_shift): New function to output shifts.
12080         (pdp11_branch_cost): Remove.
12081         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
12082         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
12083         (WCHAR_TYPE): Ditto.
12084         (PTRDIFF_TYPE): Ditto.
12085         (ADJUST_INSN_LENGTH): New macro.
12086         (FIXED_REGISTERS): Add CC registers.
12087         (CALL_USED_REGISTERS): Ditto.
12088         (reg_class): Ditto.
12089         (REG_CLASS_NAMES): Ditto.
12090         (REG_CLASS_CONTENTS): Ditto.
12091         (SELECT_CC_MODE): Use new function.
12092         (TARGET_FLAGS_REGNUM): New macro.
12093         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
12094         (cc0_reg_rtx): Remove.
12095         (CC_STATUS_MDEP): Remove.
12096         (CC_STATUS_MDEFP_INIT): Remove.
12097         (CC_IN_FPU): Remove.
12098         (NOTICE_UPDATE_CC): Remove.
12099         (REGISTER_NAMES): Add CC registers.
12100         (BRANCH_COST): Change to constant 1.
12101         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
12102         handling.
12103         * config/pdp11/pdp11.opt (mbcopy): Remove.
12104         (mbcopy-builtin): Remove.
12105         (mbranch-cheap): Remove.
12106         (mbranch-expensive): Remove.
12107         * config/pdp11/predicates.md (expand_shift_operand): Update to
12108         match shift code generation.
12109         (ccnz_operator): New predicate.
12110         * doc/invoke.texi (PDP-11 Options): Remove deleted options
12111         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
12112         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
12113         exclusive options.
12114         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
12115         description of O constraint.
12117 2018-06-27  Jeff Law  <law@redhat.com>
12118             Austin Law  <austinklaw@gmail.com>
12120         * config/v850/v850.md (addsi3_set_flags): New pattern.
12121         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
12122         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
12123         (zero_extendhisi2_v850_set_flags): Likewise.
12124         (zero_extendqisi2_v850_set_flags): Likewise.
12125         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
12126         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
12127         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
12129         * config/v850/v850-protos.h (notice_update_cc): Remove.
12130         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
12131         (v850_print_operand): Handle 'D' and "d".
12132         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
12133         Add handling of arithmetic/logical operations compared against zero.
12134         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
12135         Do not look at v850_compare_op, instead get mode from last argument.
12136         (v850_gen_compare): Remove
12137         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
12138         after reload for prologue insns.
12139         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
12140         patterns.
12141         (construct_save_jarl): Likewise.
12142         (TARGET_FLAGS_REGNUM): Define.
12143         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
12144         (NOTICE_UPDATE_CC): Remove.
12145         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
12146         than cc0.  Conditionalize on reload_completed.
12147         (cmpsi_insn, setfcc_insn): Likewise.
12148         (tst1 splitter): Turn into define_and_split which sets the flags
12149         after reload.
12150         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
12151         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
12152         (cstoresf4, cstoredf4): Clobber the flags.
12153         (cmpsi, cmpsf, cmpdf): Remove expanders.
12154         (setf_insn): Remove pattern.
12155         (addsi3): Turn into define_and_split which clobbers the flags after
12156         reload and a suitable pattern (addsi3_clobber_flags) for use after
12157         reload.
12158         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
12159         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
12160         (ashrsi3, ashrsi3_v850e2): Likewise.
12161         (bins): Clobber the flags.
12162         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
12163         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
12164         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
12165         (call_value_internal_short, call_value_internal_long): Likewise.
12166         (callt_save_interrupt, callt_return_interrupt): Likewise.
12167         (save_interrupt, return_interrupt): Likewise.
12168         (callt_save_all_interrupt, save_all_interrupt): Likewise.
12169         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
12170         (restore_all_interrupt, _restore_all_interrupt): Likewise.
12171         (All FP comparisons): Only allow after reload has completed.
12172         (trfsr): Likewise.
12173         (divh, divhu): Tweak output template.
12174         (branch_z_normal, branch_z_invert): Remove
12175         (branch_nz_normal, branch_nz_invert): Likewise.
12176         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
12178         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
12179         * config/v850/v850.c (notice_update_cc): Remove.
12180         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
12181         (CC_NO_CARRY): Likewise.
12182         (NOTICE_UPDATE_CC): Define to nothing.
12183         * config/v850/v850.md: Remove block comment on cc0 handling
12184         Remove "cc" attribute from all patterns.  Remove cc_status handling
12185         from all patterns.  Minor formatting fixes.
12187 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12189         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
12190         (cortex-a76.cortex-a55): Likewise.
12191         * config/aarch64/aarch64-tune.md: Regenerate.
12192         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
12193         cortex-a76.cortex-a55.
12195 2018-06-27  Jeff Law  <law@redhat.com>
12197         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
12198         (MULTILIB_DIRNAMES): Similarly.
12200 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
12202         * gimple.h (gimple_return_retbnd): Delete.
12203         (gimple_return_set_retbnd): Likewise.
12204         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
12205         gimple_return_set_retbnd.
12206         * gimple-pretty-print.c (dump_gimple_return): Remove call to
12207         gimple_return_retbnd and adjust.
12208         * tree-inline.h (struct copy_body_data): Remove retbnd field.
12209         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
12210         Explicitly return NULL in a couple more cases.  Move assertion
12211         on debug statements and remove unreachable code.
12212         (reset_debug_binding): Do not test id->retbnd.
12213         (expand_call_inline): Do not set it.
12215 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
12217         * configure.ac: Add --disable-gcov option.
12218         * configure: Regenerate.
12219         * Makefile.in: Honour @enable_gcov@.
12220         * doc/install.texi: Document --disable-gcov.
12222 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12224         * config/arm/arm-cpus.in (cortex-a76): New entry.
12225         (cortex-a76.cortex-a55): Likewise.
12226         * config/arm/arm-tables.opt: Regenerate.
12227         * config/arm/arm-tune.md: Likewise.
12228         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
12229         * doc/invoke.texi (ARM Options): Document cortex-a76 and
12230         cortex-a76.cortex-a55.
12232 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
12234         PR target/85769
12235         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
12237 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
12239         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
12240         comment.
12241         (EPILOGUE_USES): Likewise.
12243 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
12245         * tree-inline.c (remap_location): New function extracted from...
12246         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
12247         (copy_phis_for_bb): ...here.  Call remap_location.
12248         (copy_cfg_body): Adjust call to copy_edges_for_bb.
12250 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
12252         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
12253         unaligned vsx for 16B memset.
12255 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
12257         PR target/86285
12258         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
12259         ieee128_float_type_node to long_double_type_node unless
12260         TARGET_LONG_DOUBLE_128 is set.
12262 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
12264         * cfgloop.c (get_loop_location): Convert return type from
12265         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
12266         by implicit construction from rtx_insn *, and using
12267         dump_user_location_t::from_function_decl for the fallback case.
12268         * cfgloop.h (get_loop_location): Convert return type from
12269         location_t to dump_user_location_t.
12270         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
12271         dump_printf_loc to pass in a dump_location_t rather than a
12272         location_t, via the gimple stmt.
12273         * coverage.c (get_coverage_counts): Update calls to
12274         dump_printf_loc to pass in dump_location_t rather than a
12275         location_t.
12276         * doc/optinfo.texi (Dump types): Convert example of
12277         dump_printf_loc from taking "locus" to taking "insn".  Update
12278         description of the "_loc" calls to cover dump_location_t.
12279         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
12280         "selftest.h".
12281         (dump_user_location_t::dump_user_location_t): New constructors,
12282         from gimple *stmt and rtx_insn *.
12283         (dump_user_location_t::from_function_decl): New function.
12284         (dump_loc): Make static.
12285         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
12286         const dump_location_t &.
12287         (dump_generic_expr_loc): Delete.
12288         (dump_printf_loc): Convert param "loc" from location_t to
12289         const dump_location_t &.
12290         (selftest::test_impl_location): New function.
12291         (selftest::dumpfile_c_tests): New function.
12292         * dumpfile.h: Include "profile-count.h".
12293         (class dump_user_location_t): New class.
12294         (struct dump_impl_location_t): New struct.
12295         (class dump_location_t): New class.
12296         (dump_printf_loc): Convert 2nd param from source_location to
12297         const dump_location_t &.
12298         (dump_generic_expr_loc): Delete.
12299         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
12300         const dump_location_t &.
12301         * gimple-fold.c (fold_gimple_assign): Update call to
12302         dump_printf_loc to pass in a dump_location_t rather than a
12303         location_t, via the gimple stmt.
12304         (gimple_fold_call): Likewise.
12305         * gimple-loop-interchange.cc
12306         (loop_cand::analyze_iloop_reduction_var): Update for change to
12307         check_reduction_path.
12308         (tree_loop_interchange::interchange): Update for change to
12309         find_loop_location.
12310         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
12311         change in return-type of find_loop_location.
12312         (graphite_regenerate_ast_isl): Likewise.
12313         * graphite-optimize-isl.c (optimize_isl): Likewise.
12314         * graphite.c (graphite_transform_loops): Likewise.
12315         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
12316         pass in a dump_location_t rather than a location_t, via the
12317         gimple stmt.
12318         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12319         * ipa.c (walk_polymorphic_call_targets): Likewise.
12320         * loop-unroll.c (report_unroll): Convert "locus" param from
12321         location_t to dump_location_t.
12322         (decide_unrolling): Update for change to get_loop_location's
12323         return type.
12324         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
12325         location_t to dump_user_location_t.
12326         (grid_find_single_omp_among_assignments_1): Updates calls to
12327         dump_printf_loc to pass in a dump_location_t rather than a
12328         location_t, via the gimple stmt.
12329         (grid_parallel_clauses_gridifiable): Convert "tloc" from
12330         location_t to dump_location_t.  Updates calls to dump_printf_loc
12331         to pass in a dump_location_t rather than a location_t, via the
12332         gimple stmt.
12333         (grid_inner_loop_gridifiable_p): Likewise.
12334         (grid_dist_follows_simple_pattern): Likewise.
12335         (grid_gfor_follows_tiling_pattern): Likewise.
12336         (grid_target_follows_gridifiable_pattern): Likewise.
12337         (grid_attempt_target_gridification): Convert initialization
12338         of local "grid" from memset to zero-initialization; FIXME: does
12339         this require C++11?  Update call to dump_printf_loc to pass in a
12340         optinfo_location rather than a location_t, via the gimple stmt.
12341         * profile.c (read_profile_edge_counts): Updates call to
12342         dump_printf_loc to pass in a dump_location_t rather than a
12343         location_t
12344         (compute_branch_probabilities): Likewise.
12345         * selftest-run-tests.c (selftest::run_tests): Call
12346         dumpfile_c_tests.
12347         * selftest.h (dumpfile_c_tests): New decl.
12348         * tree-loop-distribution.c (pass_loop_distribution::execute):
12349         Update for change in return type of find_loop_location.
12350         * tree-parloops.c (parallelize_loops): Likewise.
12351         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
12352         "locus" from location_t to dump_user_location_t.
12353         (canonicalize_loop_induction_variables): Likewise.
12354         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
12355         for change in return type of find_loop_location.
12356         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
12357         to dump_printf_loc to pass in a dump_location_t rather than a
12358         location_t, via the stmt.
12359         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
12360         Likewise.
12361         * tree-vect-loop-manip.c (find_loop_location): Convert return
12362         type from source_location to dump_user_location_t.
12363         (vect_do_peeling): Update for above change.
12364         (vect_loop_versioning): Update for change in type of
12365         vect_location.
12366         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
12367         from location_t to dump_user_location_t.
12368         (vect_estimate_min_profitable_iters): Update for change in type
12369         of vect_location.
12370         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
12371         location_t to dump_location_t.
12372         (vect_slp_bb): Update for change in type of vect_location.
12373         * tree-vectorizer.c (vect_location): Convert from source_location
12374         to dump_user_location_t.
12375         (try_vectorize_loop_1): Update for change in vect_location's type.
12376         (vectorize_loops): Likewise.
12377         (increase_alignment): Likewise.
12378         * tree-vectorizer.h (vect_location): Convert from source_location
12379         to dump_user_location_t.
12380         (find_loop_location): Convert return type from source_location to
12381         dump_user_location_t.
12382         (check_reduction_path): Convert 1st param from location_t to
12383         dump_user_location_t.
12384         * value-prof.c (check_counter): Update call to dump_printf_loc to
12385         pass in a dump_user_location_t rather than a location_t; update
12386         call to error_at for change in type of "locus".
12387         (check_ic_target): Update call to dump_printf_loc to
12388         pass in a dump_user_location_t rather than a location_t, via the
12389         call_stmt.
12391 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12393         * config/s390/s390.h (enum processor_flags): Do not use
12394         default tune parameter when -march was specified.
12396 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
12398         PR target/86314
12399         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12400         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12401         operands.
12403 2018-06-26  Richard Biener  <rguenther@suse.de>
12405         PR tree-optimization/86287
12406         PR bootstrap/86316
12407         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12408         (vect_analyze_loop): Initialize n_stmts.
12410 2018-06-26  Richard Biener  <rguenther@suse.de>
12412         PR middle-end/86271
12413         * fold-const.c (fold_convertible_p): Pointer extension
12414         isn't valid.
12416 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
12418         PR debug/86064
12419         * dwarf2out.c (loc_list_has_views): Adjust comments.
12420         (dw_loc_list): Split single cross-partition range with
12421         nonzero locview.
12423 2018-06-25  Jeff Law  <law@redhat.com>
12425         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12426         on -mbig-switch by default.
12428         * config/v850/predicates.md (const_float_1_operand): Fix match_code
12429         test.
12430         (const_float_0_operand): Remove unused predicate.
12431         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12432         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12433         (recipsf2): New expander.  Original pattern now called
12434         (recipsf2_insn).
12435         (recipdf2, recipdf2_insn): Similarly.
12436         (rsqrtsf2, rsqrtsf2_insn): Similarly
12437         (rsqrtdf2, rsqrtdf2_insn): Similarly
12439 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
12441         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12442         Simplify logic for FreeBSD (twice).
12444 2018-06-25  Martin Sebor  <msebor@redhat.com>
12446         PR tree-optimization/86204
12447         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12448         a strnlen result if it's less than the length of the string.
12450 2018-06-25  Martin Sebor  <msebor@redhat.com>
12452         PR tree-optimization/85700
12453         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12454         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12455         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12457 2018-06-25  Martin Sebor  <msebor@redhat.com>
12459         * doc/extend.texi (Zero-length arrays): Update and clarify.
12461 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
12463         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12464         added IEEE/IBM long double multilib support on PowerPC little
12465         endian Linux systems.
12466         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12467         (MULTILIB_DEFAULTS): Likewise.
12468         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12469         Likewise.
12470         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12471         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12472         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12474 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
12476         PR middle-end/86311
12477         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12478         (REORDER_45): Likewise.
12480 2018-06-25  Jeff Law  <law@redhat.com>
12482         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12483         dividend to 32 bits.  Adjust length.
12484         (udivmodhi4): Cleanup output template.  Fix length.
12486 2018-06-25  Carl Love  <cel@us.ibm.com>
12488         * config/rs6000/vsx.md: Change word selector to prefered location.
12490 2018-06-25  Richard Biener  <rguenther@suse.de>
12492         PR tree-optimization/86304
12493         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12494         epilogue-if-converted loops as well.
12496 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
12498         * lto-section-out.c (lto_begin_section): Do not print section
12499         name for noaddr and unnumbered dumps.
12501 2018-06-25  Richard Biener  <rguenther@suse.de>
12503         * tree-vectorizer.h (struct vec_info_shared): New structure
12504         with parts split out from struct vec_info and loop_nest from
12505         struct _loop_vec_info.
12506         (struct vec_info): Adjust accordingly.
12507         (struct _loop_vec_info): Likewise.
12508         (LOOP_VINFO_LOOP_NEST): Adjust.
12509         (LOOP_VINFO_DATAREFS): Likewise.
12510         (LOOP_VINFO_DDRS): Likewise.
12511         (struct _bb_vec_info): Likewise.
12512         (BB_VINFO_DATAREFS): Likewise.
12513         (BB_VINFO_DDRS): Likewise.
12514         (struct _stmt_vec_info): Add dr_aux member.
12515         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12516         (DR_MISALIGNMENT_UNINITIALIZED): New.
12517         (set_dr_misalignment): Adjust.
12518         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12519         (vect_analyze_loop): Adjust prototype.
12520         (vect_analyze_loop_form): Likewise.
12521         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12522         Compute dependences lazily.
12523         (vect_record_base_alignments): Use shared datarefs/ddrs.
12524         (vect_verify_datarefs_alignment): Likewise.
12525         (vect_analyze_data_refs_alignment): Likewise.
12526         (vect_analyze_data_ref_accesses): Likewise.
12527         (vect_analyze_data_refs): Likewise.
12528         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12529         constructor parameter for shared part.
12530         (vect_analyze_loop_form): Pass in shared part and adjust.
12531         (vect_analyze_loop_2): Pass in storage for the number of
12532         stmts.  Move loop nest finding to the caller.  Compute
12533         datarefs lazily.
12534         (vect_analyze_loop): Pass in shared part.
12535         (vect_transform_loop): Verify shared datarefs are unchanged.
12536         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12537         constructor parameter for shared part.
12538         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12539         (vect_slp_bb): Verify shared datarefs are unchanged before
12540         transform.
12541         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12542         change.
12543         (new_stmt_vec_info): Initialize DR_AUX misalignment to
12544         DR_MISALIGNMENT_UNINITIALIZED.
12545         * tree-vectorizer.c (vec_info::vec_info): Add constructor
12546         parameter for shared part.
12547         (vec_info::~vec_info): Adjust.
12548         (vec_info_shared::vec_info_shared): New.
12549         (vec_info_shared::~vec_info_shared): Likewise.
12550         (vec_info_shared::save_datarefs): Likewise.
12551         (vec_info_shared::check_datarefs): Likewise.
12552         (try_vectorize_loop_1): Construct shared part live for analyses
12553         of a single loop for multiple vector sizes.
12554         * tree-parloops.c (gather_scalar_reductions): Adjust.
12556 2018-06-25  Richard Biener  <rguenther@suse.de>
12558         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12559         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12560         (vect_analyze_data_refs): Remove similar code from here and
12561         simplify accordingly.
12563 2018-06-25  Richard Biener  <rguenther@suse.de>
12565         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12566         for reverse storage order accesses rather than asserting
12567         they cannot happen here.
12569 2018-06-25  Tom de Vries  <tdevries@suse.de>
12571         PR debug/86257
12572         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12573         Use data16 instead of .byte for insn prefix.
12575 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
12577         PR C++/86082
12578         * parser.c (make_char_string_pack): Pass this literal chars
12579         through cpp_interpret_string.
12580         (cp_parser_userdef_numeric_literal): Check the result of
12581         make_char_string_pack.
12583 2018-06-24  Maya Rashish  <coypu@sdf.org>
12585         * ginclude/stddef.h: Simplify conditions around avoiding
12586         re-definition of __size_t.
12588 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12590         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12591         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12593 2018-06-22  Maya Rashish  <coypu@sdf.org>
12595         * doc/invoke.texi (mno-fancy-math-387): Update for changes
12596         made to OpenBSD and NetBSD through the years.
12598 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12600         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12601         behavior of vec_pack (vector double, vector double) to match
12602         behavior of vec_float2 (vector double, vector double).
12604 2018-06-22  Olivier Hainque  <hainque@adacore.com>
12606         * gimplify.c (gimplify_function_tree): Prevent creation
12607         of a trampoline for the address of the current function
12608         passed to entry/exit instrumentation hooks.
12610 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
12612         PR target/86222
12613         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12614         correctly.
12616 2018-06-22  Martin Liska  <mliska@suse.cz>
12618         PR tree-optimization/86263
12619         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12620         Bail out if is_enabled is false.
12621         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12622         New declaration.
12623         (jump_table_cluster::is_enabled): New function.
12625 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12627         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12628         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12629         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12630         (lto_input_ts_binfo_tree_pointers): Likewise.
12631         * tree-streamer-out.c (streamer_write_tree_bitfields,
12632         write_ts_binfo_tree_pointers): Likewise.
12633         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12635 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12637         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12639 2018-06-22  Martin Liska  <mliska@suse.cz>
12641         * symbol-summary.h (get): Make it pure and inline move
12642         functionality from ::get function.
12643         (get): Remove and inline into ::get and ::get_create.
12644         (get_create): Move code from ::get function.
12646 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12648         PR target/85994
12649         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12650         -x assembler-with-cpp.
12652 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12654         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12655         _FILE_OFFSET_BITS=64 for C++.
12657 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
12659         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12660         conversion insn that shows up when pr85657-3.c is compiled using
12661         IEEE 128-bit long double.
12662         (neg<mode>2_internal): Use the correct mode to check whether the
12663         mode is IBM extended.
12664         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12665         multiply and divide external functions from being created more
12666         than once.
12668 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12670         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12671         functions.
12672         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
12673         the edge can be forwarded.
12674         (cfg_layout_merge_blocks): Likewise.
12676 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12678         * except.c (finish_eh_generation): Commit edge insertions only after
12679         the EH edges have been redirected from post-landing to landing pads.
12681 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12683         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12684         create_tmp_var_for to create the FRAME decl.
12685         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12687 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12689         * tree-inline.c (copy_edges_for_bb): Minor tweak.
12690         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12691         debug statement when resetting its value.
12692         (expand_call_inline): Copy the locus of the call onto the assignment
12693         of the return value, if any.  Use local variable in more cases.
12695 2018-06-21  Martin Liska  <mliska@suse.cz>
12697         * ipa-pure-const.c (propagate_nothrow): Use
12698         funct_state_summaries->get.
12699         (dump_malloc_lattice): Likewise.
12700         (propagate_malloc): Likewise.
12702 2018-06-21  Richard Biener  <rguenther@suse.de>
12704         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12705         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12706         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12707         comment.
12708         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12709         BLOCK_ABSTRACT_ORIGIN unconditionally.
12711 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
12713         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12714         deleting it.
12715         * ipa-reference.c (ipa_reference_c_finalize): Delete
12716         ipa_ref_opt_sum_summaries and set it to NULL.
12718 2018-06-21  Tom de Vries  <tdevries@suse.de>
12720         PR tree-optimization/85859
12721         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12722         test with comment from bb_no_side_effects_p.
12724 2018-06-21  Richard Biener  <rguenther@suse.de>
12726         PR tree-optimization/86232
12727         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12728         max for constant niter.
12730 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12732         * config/aarch64/aarch64-simd.md
12733         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12735 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12737         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12738         Make opernads of the unspec commutative.
12740 2018-06-21  Richard Biener  <rguenther@suse.de>
12742         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12743         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12744         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12745         (vect_analyze_data_ref_dependence): Re-order checks to deal with
12746         NULL DR_STEP.
12747         (vect_record_base_alignments): Do not record base alignment
12748         for gathers or scatters.
12749         (vect_compute_data_ref_alignment): Drop return value that is always
12750         true.  Bail out early for gathers or scatters.
12751         (vect_enhance_data_refs_alignment): Bail out early for gathers
12752         or scatters.
12753         (vect_find_same_alignment_drs): Likewise.
12754         (vect_analyze_data_refs_alignment): Remove dead code.
12755         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12756         (vect_analyze_data_refs): For possible gathers or scatters do
12757         not create an alternate DR, just check their possible validity
12758         and mark them.  Adjust DECL_NONALIASED handling to not rely
12759         on DR_BASE_ADDRESS.
12760         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12761         update inits of gathers or scatters.
12762         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12763         Also copy gather/scatter flag to pattern vinfo.
12765 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12767         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12768         behavior of vec_packsu (vector unsigned long long, vector unsigned
12769         long long) to match behavior of vec_packs with same signature.
12771 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
12772             Thomas Schwinge <thomas@codesourcery.com>
12773             Cesar Philippidis  <cesar@codesourcery.com>
12775         * gimplify.c (gimplify_scan_omp_clauses): Add support for
12776         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12777         (gimplify_adjust_omp_clauses): Likewise.
12778         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
12779         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
12780         (gimplify_omp_target_update): Update handling of acc update and
12781         enter/exit data.
12782         * omp-low.c (install_var_field): Remove unused parameter
12783         base_pointers_restrict.
12784         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
12785         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
12786         FINALIZE}
12787         (omp_target_base_pointers_restrict_p): Delete.
12788         (scan_omp_target): Update call to scan_sharing_clauses.
12789         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
12790         FINALIZE}.
12791         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
12792         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12793         (convert_local_omp_clauses): Likewise.
12794         * tree-pretty-print.c (dump_omp_clause): Likewise.
12795         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
12796         FINALIZE}.
12797         (omp_clause_code_name): Likewise.
12799 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12801         PR debug/86194
12802         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
12803         be narrowed.
12805         PR tree-optimization/86231
12806         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
12807         anti-range don't overwrite *vr0min before using it to compute *vr0max.
12809 2018-06-20  Tom de Vries  <tdevries@suse.de>
12811         PR tree-optimization/86097
12812         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
12813         iv type if signedness of iv type is not the same as that of *nit.
12815 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12817         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
12818         EDGE_EH edges, verify they are all EDGE_EH.
12820 2018-06-20  Maya Rashish  <coypu@sdf.org>
12822         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
12824 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12826         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
12827         * config/aarch64/aarch64.c (xgene1_tunings): Add
12828         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
12829         (aarch64_mode_valid_for_sched_fusion_p):
12830         Allow 16-byte modes.
12831         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
12832         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
12833         128-bit modes.
12834         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
12835         New pattern.
12836         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
12837         * config/aarch64/iterators.md (VQ2): New mode iterator.
12839 2018-06-20  Martin Liska  <mliska@suse.cz>
12841         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12842         Change default ratio from 10 to 8.
12844 2018-06-20  Martin Liska  <mliska@suse.cz>
12846         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12847         New.
12848         (bit_test_cluster::find_bit_tests): Likewise.
12849         (switch_decision_tree::analyze_switch_statement): Find clusters.
12850         * tree-switch-conversion.h (struct jump_table_cluster): Document
12851         hierarchy.
12853 2018-06-20  Martin Liska  <mliska@suse.cz>
12855         * tree-switch-conversion.c (switch_conversion::collect):
12856         Record m_uniq property.
12857         (switch_conversion::expand): Bail out for special conditions.
12858         (group_cluster::~group_cluster): New.
12859         (group_cluster::group_cluster): Likewise.
12860         (group_cluster::dump): Likewise.
12861         (jump_table_cluster::emit): New.
12862         (switch_decision_tree::fix_phi_operands_for_edges): New.
12863         (struct case_node): Remove struct.
12864         (jump_table_cluster::can_be_handled): New.
12865         (case_values_threshold): Moved to header.
12866         (reset_out_edges_aux): Likewise.
12867         (jump_table_cluster::is_beneficial): New.
12868         (bit_test_cluster::can_be_handled): Likewise.
12869         (add_case_node): Remove.
12870         (bit_test_cluster::is_beneficial): New.
12871         (case_bit_test::cmp): New.
12872         (bit_test_cluster::emit): New.
12873         (expand_switch_as_decision_tree_p): Remove.
12874         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
12875         (fix_phi_operands_for_edge): Likewise.
12876         (switch_decision_tree::analyze_switch_statement): New.
12877         (compute_cases_per_edge): Move ...
12878         (switch_decision_tree::compute_cases_per_edge): ... here.
12879         (try_switch_expansion): Likewise.
12880         (switch_decision_tree::try_switch_expansion): Likewise.
12881         (record_phi_operand_mapping): Likewise.
12882         (switch_decision_tree::record_phi_operand_mapping): Likewise.
12883         (emit_case_decision_tree): Likewise.
12884         (switch_decision_tree::emit): Likewise.
12885         (balance_case_nodes): Likewise.
12886         (switch_decision_tree::balance_case_nodes): Likewise.
12887         (dump_case_nodes): Likewise.
12888         (switch_decision_tree::dump_case_nodes): Likewise.
12889         (emit_jump): Likewise.
12890         (switch_decision_tree::emit_jump): Likewise.
12891         (emit_cmp_and_jump_insns): Likewise.
12892         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
12893         (emit_case_nodes): Likewise.
12894         (switch_decision_tree::emit_case_nodes): Likewise.
12895         (conditional_probability): Remove.
12896         * tree-switch-conversion.h (enum cluster_type): New.
12897         (PRINT_CASE): New.
12898         (struct cluster): Likewise.
12899         (cluster::cluster): Likewise.
12900         (struct simple_cluster): Likewise.
12901         (simple_cluster::simple_cluster): Likewise.
12902         (struct group_cluster): Likewise.
12903         (struct jump_table_cluster): Likewise.
12904         (struct bit_test_cluster): Likewise.
12905         (struct min_cluster_item): Likewise.
12906         (struct case_tree_node): Likewise.
12907         (case_tree_node::case_tree_node): Likewise.
12908         (jump_table_cluster::case_values_threshold): Likewise.
12909         (struct case_bit_test): Likewise.
12910         (struct switch_decision_tree): Likewise.
12911         (struct switch_conversion): Likewise.
12912         (switch_decision_tree::reset_out_edges_aux): Likewise.
12914 2018-06-20  Martin Liska  <mliska@suse.cz>
12916         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
12917         (hoist_edge_and_branch_if_true): Likewise.
12918         (expand_switch_using_bit_tests_p): Likewise.
12919         (struct case_bit_test): Likewise.
12920         (case_bit_test_cmp): Likewise.
12921         (emit_case_bit_tests): Likewise.
12922         (switch_conversion::switch_conversion): New class.
12923         (struct switch_conv_info): Remove old struct.
12924         (collect_switch_conv_info): More to ...
12925         (switch_conversion::collect): ... this.
12926         (check_range): Likewise.
12927         (switch_conversion::check_range): Likewise.
12928         (check_all_empty_except_final): Likewise.
12929         (switch_conversion::check_all_empty_except_final): Likewise.
12930         (check_final_bb): Likewise.
12931         (switch_conversion::check_final_bb): Likewise.
12932         (create_temp_arrays): Likewise.
12933         (switch_conversion::create_temp_arrays): Likewise.
12934         (free_temp_arrays): Likewise.
12935         (gather_default_values): Likewise.
12936         (switch_conversion::gather_default_values): Likewise.
12937         (build_constructors): Likewise.
12938         (switch_conversion::build_constructors): Likewise.
12939         (constructor_contains_same_values_p): Likewise.
12940         (switch_conversion::contains_same_values_p): Likewise.
12941         (array_value_type): Likewise.
12942         (switch_conversion::array_value_type): Likewise.
12943         (build_one_array): Likewise.
12944         (switch_conversion::build_one_array): Likewise.
12945         (build_arrays): Likewise.
12946         (switch_conversion::build_arrays): Likewise.
12947         (gen_def_assigns): Likewise.
12948         (switch_conversion::gen_def_assigns): Likewise.
12949         (prune_bbs): Likewise.
12950         (switch_conversion::prune_bbs): Likewise.
12951         (fix_phi_nodes): Likewise.
12952         (switch_conversion::fix_phi_nodes): Likewise.
12953         (gen_inbound_check): Likewise.
12954         (switch_conversion::gen_inbound_check): Likewise.
12955         (process_switch): Use the newly created class.
12956         (switch_conversion::expand): New.
12957         (switch_conversion::~switch_conversion): New.
12958         * tree-switch-conversion.h: New file.
12960 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12962         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
12963         tree-vect-patterns.c.
12964         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
12965         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
12966         (vect_recog_sad_pattern): Likewise.
12967         (vect_recog_widen_sum_pattern): Likewise.
12968         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
12969         (vect_recog_widen_shift_pattern): Remove the type_in argument.
12970         (vect_recog_rotate_pattern): Likewise.
12971         (vect_recog_mult_pattern): Likewise.
12972         (vect_recog_vector_vector_shift_pattern): Likewise.
12973         (vect_recog_divmod_pattern): Likewise.
12974         (vect_recog_mixed_size_cond_pattern): Likewise.
12975         (vect_recog_bool_pattern): Likewise.
12976         (vect_recog_mask_conversion_pattern): Likewise.
12977         (vect_try_gather_scatter_pattern): Likewise.
12978         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
12979         (vect_recog_over_widening_pattern): Likewise.
12980         (vect_recog_gather_scatter_pattern): Likewise.
12981         (vect_recog_func_ptr): Move from tree-vectorizer.h
12982         (vect_vect_recog_func_ptrs): Move further down the file.
12983         (vect_recog_func): Likewise.  Remove the third argument.
12984         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
12985         (vect_pattern_recog_1): Expect the pattern function to do any
12986         necessary target tests.  Also expect it to provide a vector type.
12987         Remove the type_in handling.
12989 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12991         * tree-vect-patterns.c (vect_pattern_detected): New function.
12992         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
12993         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
12994         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
12995         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
12996         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12997         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12998         (vect_recog_mask_conversion_pattern)
12999         (vect_try_gather_scatter_pattern): Likewise.
13001 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13003         * tree-vect-patterns.c (vect_get_internal_def): New function.
13004         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13005         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
13006         (search_type_for_mask_1): Use it.
13008 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13010         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13011         redundant WIDEN_SUM_EXPR handling.
13012         (vect_recog_sad_pattern): Likewise.
13014 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13016         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13017         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
13018         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
13019         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
13020         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
13022 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13024         * tree-vect-stmts.c (vectorizable_call): Make sure that we
13025         use the stmt_vec_info of the original bb statement for the
13026         new zero assignment, even if the call is part of a pattern.
13028 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13030         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
13031         that the sequence is attached to the original statement rather
13032         than the pattern statement.
13033         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
13034         PATTERN_DEF_SEQ from the original statement rather than
13035         the main pattern statement.
13036         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
13037         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
13038         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
13040 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13042         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
13043         definition statements before the early exit for statements that aren't
13044         live or relevant.
13045         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
13046         split out from...
13047         (vect_transform_loop): ...here.  Process pattern definition
13048         statements without first checking whether the main pattern
13049         statement is live or relevant.
13051 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13053         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
13054         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
13056 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
13058         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
13059         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
13060         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
13061         (expand_block_compare): Change select_block_compare_mode call.
13062         (expand_strncmp_align_check): Use new functions, fix comment.
13063         (emit_final_str_compare_gpr): New function.
13064         (expand_strn_compare): Refactor and clean up code.
13065         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
13067 2018-06-19  Tony Reix  <tony.reix@atos.com>
13068             Damien Bergamini  <damien.bergamini@atos.com>
13069             David Edelsohn  <dje.gcc@gmail.com>
13071         * collect2.c (static_obj): New variable.
13072         (static_libs): New variable.
13073         (is_in_list): Uncomment declaration.
13074         (main): Track AIX libraries linked statically.
13075         (is_in_list): Uncomment definition.
13076         (scan_prog_file): Don't add AIX shared libraries initializer
13077         to constructor list if linking statically.
13079 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
13081         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
13082         constant.
13083         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
13085 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
13087         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
13088         blocks.
13090 2018-06-19  Martin Liska  <mliska@suse.cz>
13092         * config/i386/i386.c (ix86_can_inline_p): Do not use
13093         ipa_fn_summaries::get_create.
13094         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
13095         get.
13096         (devirtualization_time_bonus): Likewise.
13097         (ipcp_propagate_stage): Likewise.
13098         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13099         (edge_set_predicate): Likewise.
13100         (evaluate_conditions_for_known_args): Likewise.
13101         (evaluate_properties_for_edge): Likewise.
13102         (ipa_call_summary::reset): Tranform to ...
13103         (ipa_call_summary::~ipa_call_summary): ... this.
13104         (ipa_fn_summary::reset): Transform to ...
13105         (ipa_fn_summary::~ipa_fn_summary): ... this.
13106         (ipa_fn_summary_t::remove): Rename to ...
13107         (ipa_fn_summary_t::remove_callees): ... this.
13108         (ipa_fn_summary_t::duplicate): Use placement new
13109         instead of memory copy.
13110         (ipa_call_summary_t::duplicate): Likewise.
13111         (ipa_call_summary_t::remove): Remove.
13112         (dump_ipa_call_summary): Change get_create to get.
13113         (ipa_dump_fn_summary): Dump only when summary exists.
13114         (analyze_function_body): Use symbol_summary::get instead
13115         of get_create.
13116         (compute_fn_summary): Likewise.
13117         (estimate_edge_devirt_benefit): Likewise.
13118         (estimate_edge_size_and_time): Likewise.
13119         (inline_update_callee_summaries): Likewise.
13120         (remap_edge_change_prob): Likewise.
13121         (remap_edge_summaries): Likewise.
13122         (ipa_merge_fn_summary_after_inlining): Likewise.
13123         (write_ipa_call_summary): Likewise.
13124         (ipa_fn_summary_write): Likewise.
13125         (ipa_free_fn_summary): Likewise.
13126         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
13127         (struct ipa_call_summary): Likewise.
13128         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
13129         of get_create.
13130         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
13131         (estimate_size_after_inlining): Likewise.
13132         (estimate_growth): Likewise.
13133         (growth_likely_positive): Likewise.
13134         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13135         (inline_call): Likewise.
13136         * ipa-inline.c (caller_growth_limits): Likewise.
13137         (can_inline_edge_p): Likewise.
13138         (can_inline_edge_by_limits_p): Likewise.
13139         (compute_uninlined_call_time): Likewise.
13140         (compute_inlined_call_time): Likewise.
13141         (want_inline_small_function_p): Likewise.
13142         (edge_badness): Likewise.
13143         (update_caller_keys): Likewise.
13144         (update_callee_keys): Likewise.
13145         (inline_small_functions): Likewise.
13146         (inline_to_all_callers_1): Likewise.
13147         (dump_overall_stats): Likewise.
13148         (early_inline_small_functions): Likewise.
13149         (early_inliner): Likewise.
13150         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13151         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13152         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13153         * ipa-split.c (execute_split_functions): Likewise.
13154         * symbol-summary.h: Likewise.
13155         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13157 2018-06-19  Richard Biener  <rguenther@suse.de>
13159         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
13160         (vectorize_loops): ... here.  Fix dbgcnt handling.
13161         (try_vectorize_loop): Wrap try_vectorize_loop_1.
13163 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
13165         PR target/86197
13166         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
13167         ieee128 argument takes up only one (vector) register, not two (floating
13168         point) registers.
13170 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13172         * gimplify.c (gimplify_init_constructor): Really never clear for an
13173         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
13175 2018-06-19  Richard Biener  <rguenther@suse.de>
13177         PR tree-optimization/86179
13178         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
13179         after failed recognition.
13181 2018-06-18  Martin Sebor  <msebor@redhat.com>
13183         PR middle-end/85602
13184         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
13185         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
13186         Handle integer subtraction.
13187         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
13188         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
13190 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13192         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
13193         param from rtx to rtx_insn *.
13194         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
13195         param.
13196         (frv_ifcvt_modify_insn): Likwise.
13197         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
13198         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
13199         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
13200         as_a <rtx_insn *> cast to local "unprotected_region" once
13201         it's been established that it's not NULL or pc_rtx.
13202         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
13203         param "sethi" from rtx to rtx_insn *.
13204         (nds32_group_float_insns): Likewise for param "insn".
13205         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
13206         param.
13207         (vax_output_int_subtract): Likewise.
13208         * config/vax/vax.c (vax_output_int_add): Likewise for param
13209         "insn".
13210         (vax_output_int_subtract): Likewise.
13211         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
13212         (emit_pattern_after): Likewise for param "after".
13213         (emit_insn_after): Likewise.
13214         (emit_jump_insn_after): Likewise.
13215         (emit_call_insn_after): Likewise.
13216         (emit_debug_insn_after): Likewise.
13217         (emit_pattern_before): Likewise for param "before".
13218         (emit_insn_before): Likewise.
13219         (emit_jump_insn_before): Likewise.
13220         * final.c (get_insn_template): Likewise for param "insn", removing
13221         a cast.
13222         * output.h (get_insn_template): Likewise for 2nd param.
13223         * rtl.h (emit_insn_before): Likewise.
13224         (emit_jump_insn_before): Likewise.
13225         (emit_debug_insn_before_noloc): Likewise.
13226         (emit_insn_after): Likewise.
13227         (emit_jump_insn_after): Likewise.
13228         (emit_call_insn_after): Likewise.
13229         (emit_debug_insn_after): Likewise.
13230         (set_insn_deleted): Likewise for param.
13232 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
13234         PR target/85358
13235         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
13236         floating point modes, so that IFmode is numerically greater than
13237         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
13238         to declare the ordering.  This prevents IFmode from being
13239         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
13240         machine.  Include rs6000-modes.h to share the fractional values
13241         between genmodes* and the rest of the compiler.
13242         (IFmode): Likewise.
13243         (KFmode): Likewise.
13244         (TFmode): Likewise.
13245         * config/rs6000/rs6000-modes.h: New file.
13246         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
13247         meaning of rs6000_long_double_size so that 126..128 selects an
13248         appropriate 128-bit floating point type.
13249         (rs6000_option_override_internal): Likewise.
13250         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
13251         (TARGET_LONG_DOUBLE_128): Change the meaning of
13252         rs6000_long_double_size so that 126..128 selects an appropriate
13253         128-bit floating point type.
13254         (LONG_DOUBLE_TYPE_SIZE): Update comment.
13255         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
13256         source and destination to match the standard usage.
13257         (truncifkf2): Likewise.
13258         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
13259         ISA 2.07 to use an explicit clobber, instead of passing in a
13260         temporary.
13261         (copysign<mode>3_soft): Likewise.
13263 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13265         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13266         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
13267         (vect_slp_analyze_instance_dependence): Likewise.
13268         (vect_enhance_data_refs_alignment): Likewise.
13269         (vect_analyze_data_refs_alignment): Likewise.
13270         (vect_slp_analyze_and_verify_instance_alignment
13271         (vect_analyze_data_ref_accesses): Likewise.
13272         (vect_prune_runtime_alias_test_list): Likewise.
13273         (vect_analyze_data_refs): Likewise.
13274         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
13275         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13276         (vect_analyze_scalar_cycles_1): Likewise.
13277         (vect_get_loop_niters): Likewise.
13278         (vect_analyze_loop_form_1): Likewise.
13279         (vect_update_vf_for_slp): Likewise.
13280         (vect_analyze_loop_operations): Likewise.
13281         (vect_analyze_loop): Likewise.
13282         (vectorizable_induction): Likewise.
13283         (vect_transform_loop): Likewise.
13284         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
13285         * tree-vect-slp.c (vect_analyze_slp): Likewise.
13286         (vect_make_slp_decision): Likewise.
13287         (vect_detect_hybrid_slp): Likewise.
13288         (vect_slp_analyze_operations): Likewise.
13289         (vect_slp_bb): Likewise.
13290         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
13291         (vectorizable_bswap): Likewise.
13292         (vectorizable_call): Likewise.
13293         (vectorizable_simd_clone_call): Likewise.
13294         (vectorizable_conversion): Likewise.
13295         (vectorizable_assignment): Likewise.
13296         (vectorizable_shift): Likewise.
13297         (vectorizable_operation): Likewise.
13298         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
13300 2018-06-18  Martin Sebor  <msebor@redhat.com>
13302         PR tree-optimization/81384
13303         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
13304         * builtins.c (expand_builtin_strnlen): New function.
13305         (expand_builtin): Call it.
13306         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
13307         * builtins.def (BUILT_IN_STRNLEN): New.
13308         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
13309         Warn for bounds in excess of maximum object size.
13310         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
13311         single-value ranges.  Handle strnlen.
13312         (handle_builtin_strlen): Handle strnlen.
13313         (strlen_check_and_optimize_stmt): Same.
13314         * doc/extend.texi (Other Builtins): Document strnlen.
13316 2018-06-18  Maya Rashish  <coypu@sdf.org>
13318         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
13319         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
13320         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
13322         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
13323         here to ...
13324         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
13326 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13328         * tree.c (escaped_string::escape): Replace cast to char * by
13329         const_cast<char *> (unescaped).
13331 2018-06-18  Nick Clifton  <nickc@redhat.com>
13333         PR 84195
13334         * tree.c (escaped_string): New class.  Converts an unescaped
13335         string into its escaped equivalent.
13336         (warn_deprecated_use): Use the new class to convert the
13337         deprecation message, if present.
13338         (test_escaped_strings): New self test.
13339         (test_c_tests): Add test_escaped_strings.
13340         * doc/extend.texi (deprecated): Add a note that the
13341         deprecation message is affected by the -fmessage-length
13342         option, and that control characters will be escaped.
13343         (#pragma GCC error): Document this pragma.
13344         (#pragma GCC warning): Likewise.
13345         * doc/invoke.texi (-fmessage-length): Document this option's
13346         effect on the #warning and #error preprocessor directives and
13347         the deprecated attribute.
13349 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
13351         * tree.c (decl_value_expr_lookup): Revert latest change.
13352         (decl_value_expr_insert): Likewise.
13354 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
13356         * gimplify.c (nonlocal_vlas): Delete.
13357         (nonlocal_vla_vars): Likewise.
13358         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
13359         referenced VLAs.
13360         (gimplify_body): Do not create and destroy nonlocal_vlas.
13361         * tree-nested.c: Include diagnostic.h.
13362         (use_pointer_in_frame): Tweak.
13363         (lookup_field_for_decl): Add assertion and declare the transformation.
13364         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
13365         internal error when the reference is in a wrong context.  Do not
13366         create a debug decl by default.
13367         (note_nonlocal_block_vlas): Delete.
13368         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
13369         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
13370         create a debug decl by default.
13371         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
13372         call is in a wrong context.
13373         (fixup_vla_decls): New function.
13374         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
13375         debug variables were created.
13376         * tree.c (decl_value_expr_lookup): Add checking assertion.
13377         (decl_value_expr_insert): Likewise.
13379 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13381         PR middle-end/82479
13382         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
13383         * tree-scalar-evolution.c (interpret_expr): Likewise.
13384         (expression_expensive_p): Likewise.
13385         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
13386         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
13387         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
13388         (ssa_defined_by_minus_one_stmt_p): New.
13390 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13392         PR middle-end/64946
13393         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
13394         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
13395         * dojump.c (do_jump): Likewise.
13396         * expr.c (expand_expr_real_2): Check operand type's sign.
13397         * fold-const.c (const_unop): Handle ABSU_EXPR.
13398         (fold_abs_const): Likewise.
13399         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13400         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13401         (strip_sign_op_1): Likesise.
13402         * match.pd: Add new pattern to generate ABSU_EXPR.
13403         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13404         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13405         * tree-eh.c (operation_could_trap_helper_p): Likewise.
13406         * tree-inline.c (estimate_operator_cost): Likewise.
13407         * tree-pretty-print.c (dump_generic_node): Likewise.
13408         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13409         * tree.def (ABSU_EXPR): New.
13411 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
13413         PR middle-end/86095
13414         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13415         documented as preserved for backward compatibility only.
13416         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13418         PR rtl-optimization/86108
13419         * bb-reorder.c (create_forwarder_block): Renamed to ...
13420         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
13421         jump from new landing pad to the second part.
13422         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13423         Adjust callers.
13425 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
13427         PR middle-end/85878
13428         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13429         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13430         Only call store_expr for halves if the mode is the same.
13432         PR middle-end/86123
13433         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13434         Fix up comment formatting.
13436 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13438         * typed-splay-tree.h (typed_splay_tree::remove): New function.
13439         (typed_splay_tree::closure,
13440         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13441         (typed_splay_tree::typed_splay_tree,
13442         typed_splay_tree::operator =): Declared private.
13443         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13444         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13445         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13446         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13447         typed_splay_tree::splay_tree_splay,
13448         typed_splay_tree::splay_tree_foreach_helper,
13449         typed_splay_tree::splay_tree_insert,
13450         typed_splay_tree::splay_tree_remove,
13451         typed_splay_tree::splay_tree_lookup,
13452         typed_splay_tree::splay_tree_predecessor,
13453         typed_splay_tree::splay_tree_successor,
13454         typed_splay_tree::splay_tree_min,
13455         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13456         (typed_splay_tree::root, typed_splay_tree::comp,
13457         typed_splay_tree::delete_key,
13458         typed_splay_tree::delete_value): New data members.
13459         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13460         typed_splay_tree::remove.
13462 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
13464         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13465         -mginv and -mno-ginv to the assembler.
13466         * config/mips/mips.opt (-mcrc): New option.
13467         (-mginv): Likewise.
13468         * doc/invoke.text (-mcrc): Document.
13469         (-mginv): Likewise.
13471 2018-06-15  Nick Clifton  <nickc@redhat.com>
13473         PR 84195
13474         * tree.c (escaped_string): New class.  Converts an unescaped
13475         string into its escaped equivalent.
13476         (warn_deprecated_use): Use the new class to convert the
13477         deprecation message, if present.
13478         (test_escaped_strings): New self test.
13479         (test_c_tests): Add test_escaped_strings.
13480         * doc/extend.texi (deprecated): Add a note that the
13481         deprecation message is affected by the -fmessage-length
13482         option, and that control characters will be escaped.
13483         (#pragma GCC error): Document this pragma.
13484         (#pragma GCC warning): Likewise.
13485         * doc/invoke.texi (-fmessage-length): Document this option's
13486         effect on the #warning and #error preprocessor directives and
13487         the deprecated attribute.
13489 2018-06-15  Richard Biener  <rguenther@suse.de>
13491         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13492         here, also noting vector size used.
13493         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
13494         size used in MSG_OPTIMIZED_LOCATIONS dump.
13495         (pass_slp_vectorize::execute): Adjust.
13497 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
13499         PR target/85968
13500         * config/arc/arc.c (arc_return_address_register): Fix
13501         if-condition.
13503 2018-06-15  Richard Biener  <rguenther@suse.de>
13505         PR middle-end/86159
13506         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13507         leave useless conversion stripping to force_gimple_operand_gsi.
13508         (gimplify_build2): Likewise.
13509         (gimplify_build1): Likewise.
13511 2018-06-15  Richard Biener  <rguenther@suse.de>
13513         PR middle-end/86076
13514         * tree-cfg.c (move_stmt_op): unshare invariant addresses
13515         before adjusting their block.
13517 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13519         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13520         multilibs for *-*-rtems*.
13521         * config/riscv/t-rtems: New file.
13523 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
13525         PR middle-end/86122
13526         * match.pd ((A +- CST1) +- CST2): Punt if last resort
13527         unsigned_type_for returns NULL.
13529         PR target/85945
13530         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13531         subregs of multi-word pseudos unless the float mode has word size.
13533 2018-06-14  Richard Biener  <rguenther@suse.de>
13535         PR middle-end/86139
13536         * tree-vect-generic.c (build_word_mode_vector_type): Remove
13537         duplicate and harmful type_hash_canon.
13538         * tree.c (type_hash_canon): Assert we didn't find ourselves.
13540 2018-06-14  Richard Biener  <rguenther@suse.de>
13542         PR ipa/86124
13543         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13544         NULL cgraph_node.
13546 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13548         * config/rtems.h (STDINT_LONG32): Define.
13550 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
13551             Prachi Godbole  <prachi.godbole@imgtec.com>
13553         * config/mips/mips-cpus.def: Define P6600.
13554         * config/mips/mips-tables.opt: Regenerate.
13555         * config/mips/mips.c (mips_ucbranch_type): New enum.
13556         (mips_rtx_cost_data): Add support for P6600.
13557         (mips_issue_rate): Likewise.
13558         (mips_multipass_dfa_lookahead): Likewise.
13559         (mips_avoid_hazard): Likewise.
13560         (mips_reorg_process_insns): Likewise.
13561         (mips_classify_branch_p6600): New function.
13562         * config/mips/mips.h (TUNE_P6600): New define.
13563         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13564         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13565         * config/mips/mips.md: Include p6600.md.
13566         (processor): Add p6600.
13567         * config/mips/p6600.md: New file.
13568         * doc/invoke.texi: Add p6600 to supported architectures.
13570 2018-06-13  Martin Sebor  <msebor@redhat.com>
13572         PR tree-optimization/86114
13573         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13574         of integer types.
13575         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13577 2018-06-13  Richard Biener  <rguenther@suse.de>
13579         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13580         Properly set vector type of the intermediate stmt.
13581         * tree-vect-stmts.c (vectorizable_operation): The destination
13582         var always has vectype_out type.
13584 2018-06-13  Jeff Law  <law@redhat.com>
13586         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13587         integer 0 for argument to print_rtl_with_bb.
13588         (rl78_reorg): Likewise.
13590 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
13592         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13593         from rtx to rtx_insn *.
13594         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13595         "label".
13596         (add_sched_insns_for_speculation): Likewise for local "target",
13597         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13598         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13599         from rtx_insn ** to rtx_code_label **.
13600         (reorg_emit_nops): Likewise.
13601         (c6x_reorg): Likewise for local "call_labels".
13602         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13603         rtx to rtx_insn *.
13604         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13605         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13606         the loops over LABEL_REFS.
13607         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13608         braf_label.
13609         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13610         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13611         (split_branches): Strengthen local "olabel" from rtx to
13612         rtx_insn *, adding a safe_as_a cast.
13613         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13614         to "rtx_insn *".
13615         (add_insn_after): Likewise for first two params.
13616         (add_insn_before): Likewise.
13617         (remove_insn): Likewise for param.
13618         (emit_pattern_before_noloc): Likewise for second and third params.
13619         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13620         (emit_call_insn_before_noloc): Likewise.
13621         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13622         to "rtx_insn *".
13623         (emit_barrier_before): Likewise.
13624         (emit_label_before): Strengthen "label" param from "rtx" to
13625         "rtx_code_label *".  Strengthen "before" param from "rtx" to
13626         "rtx_insn *".
13627         (emit_insn_after_1): Strengthen "after" param from "rtx" to
13628         "rtx_insn *".
13629         (emit_pattern_after_noloc): Likewise.
13630         (emit_insn_after_noloc): Likewise.
13631         (emit_jump_insn_after_noloc): Likewise.
13632         (emit_call_insn_after_noloc): Likewise.
13633         (emit_debug_insn_after_noloc): Likewise.
13634         (emit_barrier_after): Likewise.
13635         (emit_label_after): Likewise for both params.
13636         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
13637         "loc" param from "int" to "location_t".
13638         (emit_insn_after_setloc): Likewise.
13639         (emit_jump_insn_after_setloc): Likewise.
13640         (emit_call_insn_after_setloc): Likewise.
13641         (emit_debug_insn_after_setloc): Likewise.
13642         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
13643         "loc" param from "int" to "location_t".
13644         (emit_pattern_before): Convert NULL_RTX to NULL.
13645         (emit_insn_before_setloc): Convert "loc" param from "int" to
13646         "location_t".
13647         (emit_jump_insn_before_setloc): Likewise.
13648         (emit_call_insn_before_setloc): Likewise.
13649         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13650         rtx_insn *.  Convert "loc" param from "int" to "location_t".
13651         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13652         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13653         Convert 3rd param from "int" to "location_t".
13654         (emit_barrier_before, emit_barrier_after, next_real_insn):
13655         Strengthen param from rtx to rtx_insn *.
13656         (emit_label_before): Strengthen 1st param from "rtx" to
13657         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
13658         "rtx_insn *".
13659         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13660         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13661         Strengthen 2nd param from "rtx" to "rtx_insn *".
13662         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13663         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13664         Likewise. Convert 3rd param from "int" to "location_t".
13665         (emit_label_after): Strengthen 1st param from "rtx" to
13666         "rtx_code_label *".
13667         (next_real_insn, remove_insn): Strengthen param from "rtx" to
13668         "rtx_insn *".
13669         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13670         from "rtx" to "rtx_insn *".
13672 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
13674         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13675         bodies streamed in with -Q.
13676         * dumpfile.c (dump_files): Add lto-stream-out dump file.
13677         * dumpfile.h (tree_dump_index): Add lto_stream_out.
13678         * gimple-streamer-out.c: Include gimple-pretty-print.h
13679         (output_bb): Dump stmts streamed.
13680         * lto-section-out.c: Include print-tree.h
13681         (lto_begin_section): Dump sections created.
13682         (lto_output_decl_index): Dump decl encoded.
13683         * lto-streamer-out.c: Include print-tree.h
13684         (create_output_block): Dump output block created.
13685         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13686         (output_function): Dump function output.
13687         (output_constructor): Dump constructor streamed.
13688         (write_global_stream): Output indexes encoded.
13689         (produce_asm_for_decls): Dump streams encoded.
13690         * lto-streamer.c (streamer_dump_file): New global var.
13691         * lto-streamer.h (streamer_dump_file): Declare.
13692         * passes.c (ipa_write_summaries): Initialize streamer dump.
13693         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13694         in.
13696 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
13698         PR target/86048
13699         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13700         offsets for register save directives.  Emit a second batch of save
13701         directives, if need be, when the function accesses prior frames.
13703 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13705         * config/arc/fpu.md (fmasf4): Force operand to register.
13706         (fnmasf4): Likewise.
13708 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13710         * config/arc/arc-protos.h (arc_pad_return): Remove.
13711         * config/arc/arc.c (machine_function): Remove force_short_suffix
13712         and size_reason.
13713         (arc_print_operand): Adjust printing of '&'.
13714         (arc_verify_short): Remove conditional printing of short suffix.
13715         (arc_final_prescan_insn): Remove reference to size_reason.
13716         (pad_return): New function.
13717         (arc_reorg): Call pad_return.
13718         (arc_pad_return): Remove.
13719         (arc_init_machine_status): Remove reference to force_short_suffix.
13720         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13721         (attr length): When attribute iscompact is true force to 2
13722         regardless; in the case of maybe check if we want to force the
13723         instruction to have 4 bytes length.
13724         (nopv): Change it to generate 4 byte long nop as well.
13725         (blockage): New pattern.
13726         (simple_return): Remove call to arc_pad_return.
13727         (p_return_i): Likewise.
13729 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13731         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13733 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13735         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13736         ARC cores.
13738 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13740         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13741         for ARC700 and ARCv2.
13743 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
13745         PR target/86076
13746         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13747         operands[2] instead of operands[1].
13750 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13752         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13753         case, check whether the outer register overlaps an unallocatable
13754         register, not just whether it fits the required class.
13756 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13758         * poly-int.h (can_div_trunc_p): Add new overload in which all values
13759         are poly_ints.
13760         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13761         (memrefs_conflict_p): Likewise.
13762         (init_alias_analysis): Likewise.
13763         * cfgexpand.c (expand_debug_expr): Likewise.
13764         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13765         * cse.c (fold_rtx): Likewise.
13766         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13767         * expr.c (emit_block_move_hints): Likewise.
13768         (clear_storage_hints, push_block, emit_push_insn): Likewise.
13769         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13770         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13771         (emit_group_store): Likewise.
13772         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
13773         to read the PRE/POST_MODIFY increment.
13774         * calls.c (store_one_arg): Use strip_offset.
13775         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13776         poly_int_rtx_p.
13777         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13778         by a VEC_SELECT.
13779         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
13780         (simplify_binary_operation_1): Extend CONST_INT handling to
13781         poly_int_rtx_p.
13782         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
13783         than a HOST_WIDE_INT.
13784         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
13785         poly_int64.
13786         (adjust_mems, add_stores): Update accodingly.
13787         (vt_canonicalize_addr): Track polynomial offsets.
13788         (emit_note_insn_var_location): Likewise.
13789         (vt_add_function_parameter): Likewise.
13790         (vt_initialize): Likewise.
13792 2018-06-12  Jeff Law  <law@redhat.com>
13794         * config.gcc (alpha*-*-freebsd*): Remove.
13795         * config/alpha/freebsd.h: Remove.
13797 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
13799         PR other/69968
13800         * spellcheck-tree.c (levenshtein_distance): Rename to...
13801         (get_edit_distance): ...this, and update for underlying renaming.
13802         * spellcheck-tree.h (levenshtein_distance): Rename to...
13803         (get_edit_distance): ...this.
13804         * spellcheck.c (levenshtein_distance): Rename to...
13805         (get_edit_distance): ...this.  Convert from Levenshtein distance
13806         to Damerau-Levenshtein distance by supporting transpositions of
13807         adjacent characters.  Rename "v1" to "v_next" and "v0" to
13808         "v_one_ago".
13809         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
13810         (selftest::test_edit_distance_unit_test_oneway): ...this, and
13811         update for underlying renaming.
13812         (selftest::levenshtein_distance_unit_test): Rename to...
13813         (selftest::test_get_edit_distance_unit): ...this, and update for
13814         underlying renaming.
13815         (selftest::test_find_closest_string): Add example from PR 69968
13816         where transposition helps
13817         (selftest::test_metric_conditions): Update for renaming.
13818         (selftest::test_metric_conditions): Likewise.
13819         (selftest::spellcheck_c_tests): Likewise.
13820         * spellcheck.h (levenshtein_distance): Rename both overloads to...
13821         (get_edit_distance): ...this.
13822         (best_match::consider): Update for renaming.
13824 2018-06-12  Martin Sebor  <msebor@redhat.com>
13826         PR tree-optimization/85259
13827         * builtins.c (compute_objsize): Handle constant offsets.
13828         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
13829         true iff a warning has been issued.
13830         * gimple.h (gimple_nonartificial_location): New function.
13831         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
13832         gimple_nonartificial_location and handle -Wno-system-headers.
13833         (handle_builtin_stxncpy): Same.
13835 2018-06-12  Martin Sebor  <msebor@redhat.com>
13837         PR c/85931
13838         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
13840 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
13842         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
13843         BUILTIN_VEC_XST entries for pointer to double and long long.
13845 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
13847         PR target/85990
13848         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13849         Update comments.
13850         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13851         Likewise.
13853 2018-06-12  Martin Liska  <mliska@suse.cz>
13855         * doc/options.texi: Document IntegerRange.
13857 2018-06-12  Martin Liska  <mliska@suse.cz>
13859         * config/i386/i386.opt: Make MPX-related options as Deprecated.
13860         * opt-functions.awk: Handle Deprecated flag.
13861         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
13862         and report error.
13863         (read_cmdline_option): Report warning for a deprecated option.
13864         * opts.h (struct cl_option): Add new field cl_deprecated.
13865         (CL_ERR_DEPRECATED): New.
13867 2018-06-12  Martin Liska  <mliska@suse.cz>
13869         * doc/options.texi: Document Deprecated option flag.
13871 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13873         * config/arc/arc-arch.h (arc_extras): New enum.
13874         (arc_cpu_t):Add field extra.
13875         (arc_cpu_types): Consider the extras.
13876         * config/arc/arc-cpus.def: Add extras info.
13877         * config/arc/arc-opts.h (processor_type): Consider extra field.
13878         * config/arc/arc.c (arc_override_options): Handle extra field.
13880 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13882         * config/arc/arc-arch.h: Update ARC_OPTX macro.
13883         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
13884         field.
13885         * config/arc/arc.c (arc_init): Update pic warning.
13886         (irq_range): Update irq range parsing warnings.
13887         (arc_override_options): Update various warning messages.
13888         (arc_handle_aux_attribute): Likewise.
13890 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
13892         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
13894 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13896         * doc/sourcebuild.texi: Document usage of line number 0 in verify
13897         compiler messages directives.
13899 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
13901         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
13902         * config/mips/mips-tables.opt: Regenerate.
13903         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
13904         mips64r6.
13905         * doc/invoke.texi: Document -march=i6500.
13907 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
13909         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
13910         (i6400_gpmul): Add cpu_unit.
13911         (i6400_gpdiv): Likewise.
13912         (i6400_msa_add_d): Update reservations.
13913         (i6400_msa_int_add) Likewise.
13914         (i6400_msa_short_logic3) Likewise.
13915         (i6400_msa_short_logic2) Likewise.
13916         (i6400_msa_short_logic) Likewise.
13917         (i6400_msa_move) Likewise.
13918         (i6400_msa_cmp) Likewise.
13919         (i6400_msa_short_float2) Likewise.
13920         (i6400_msa_div_d) Likewise.
13921         (i6400_msa_long_logic1) Likewise.
13922         (i6400_msa_long_logic2) Likewise.
13923         (i6400_msa_mult) Likewise.
13924         (i6400_msa_long_float2) Likewise.
13925         (i6400_msa_long_float4) Likewise.
13926         (i6400_msa_long_float5) Likewise.
13927         (i6400_msa_long_float8) Likewise.
13928         (i6400_fpu_fadd): Include frint type.
13929         (i6400_fpu_store): New define_insn_reservation.
13930         (i6400_fpu_load): Likewise.
13931         (i6400_fpu_move): Likewise.
13932         (i6400_fpu_fcmp): Likewise.
13933         (i6400_fpu_fmadd): Likewise.
13934         (i6400_int_mult): Include imul3nc type and update reservation.
13935         (i6400_int_div): Include idiv3 type and update reservation.
13936         (i6400_int_load): Update to check type not move_type.
13937         (i6400_int_store): Likewise.
13938         (i6400_int_prefetch): Set zero latency.
13940 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
13942         * gcc.c: Document new %@{...} sequence.
13943         (LINK_COMMAND_SPEC): Use it for the -L switches.
13944         (cpp_unique_options): Use it for the -I switches.
13945         (at_file_argbuf): New global variable.
13946         (in_at_file): Likewise.
13947         (alloc_args): Create at_file_argbuf.
13948         (clear_args): Truncate at_file_argbuf.
13949         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
13950         (open_at_file): New function.
13951         (close_at_file): Likewise.
13952         (create_at_file): Delete.
13953         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
13954         <'o'>: Likewise.
13955         <'@'>: New case.
13956         (validate_switches_from_spec): Deal with %@{...} sequence.
13957         (validate_switches): Likewise.
13958         (driver::finalize): Call clear_args.
13960 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
13962         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
13964 2018-06-11  Martin Sebor  <msebor@redhat.com>
13966         * doc/invoke.texi (-Wall): List -Wc++17-compat.
13967         (Wno-class-memaccess): Add @opindex.
13968         (Wno-templates, Wno-multiple-inheritance): Same.
13969         (Wno-virtual-inheritance, Wno-namespaces): Same.
13970         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
13971         (Wno-format-overflow, Wno-format-truncation): Same.
13972         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
13973         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
13974         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
13975         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
13976         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
13977         (Wno-misspelled-isr): Same.
13979 2018-06-11  Martin Sebor  <msebor@redhat.com>
13981         * PR tree-optimization/86083
13982         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
13984 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
13986         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
13988 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
13990         PR target/85755
13991         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
13992         on the correct operand.
13993         (*movdi_internal64): Ditto.
13995 2018-06-11  Martin Liska  <mliska@suse.cz>
13997         PR tree-optimization/86089
13998         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
14000 2018-06-11  Julia Koval  <julia.koval@intel.com>
14002         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
14003         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
14004         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
14006 2018-06-11  Olivier Hainque  <hainque@adacore.com>
14008         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
14009         for Ada with strict dwarf2.
14011 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
14013         PR target/85755
14014         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
14015         addresses.
14017 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
14019         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
14021 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
14023         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
14024         TARGET_ELF.
14026 2018-06-08  Martin Liska  <mliska@suse.cz>
14028         * tree-cfg.h (debug_function): Fix argument type to match
14029         implementation.
14031 2018-06-08  Martin Liska  <mliska@suse.cz>
14033         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
14034         Remove usage of MPX-related (and removed) fields.
14035         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
14037 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
14039         * cfg.c (debug): Use TDF_NONE rather than 0.
14040         * cfghooks.c (debug): Likewise.
14041         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
14042         (struct dump_option_value_info): Convert to...
14043         (struct kv_pair): ...this template type.
14044         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
14045         rather than 0.
14046         (optinfo_verbosity_options): Likewise.
14047         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
14048         OPTGROUP_NONE.
14049         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
14050         than int for "optgroup_flags" param.
14051         (dump_generic_expr_loc): Use dump_flags_t rather than int for
14052         "dump_kind" param.
14053         (dump_dec): Likewise.
14054         (dump_finish): Use TDF_NONE rather than 0.
14055         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
14056         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
14057         than 0.  Update for change to option_ptr.
14058         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
14059         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
14060         0.  Update for changes to optinfo_verbosity_options and
14061         optgroup_options.
14062         (opt_info_switch_p): Convert optgroup_flags from int to
14063         optgroup_flags_t.
14064         (dump_basic_block): Use dump_flags_t rather than int
14065         for "dump_kind" param.
14066         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
14067         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
14068         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
14069         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
14070         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
14071         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
14072         TDF_NONE): Convert from macros to...
14073         (enum dump_flag): ...this new enum.
14074         (dump_flags_t): Update to use enum.
14075         (operator|, operator&, operator~, operator|=, operator&=):
14076         Implement for dump_flags_t.
14077         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
14078         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
14079         Convert from macros to...
14080         (enum optgroup_flag): ...this new enum.
14081         (optgroup_flags_t): New typedef.
14082         (operator|, operator|=): Implement for optgroup_flags_t.
14083         (struct dump_file_info): Convert field "alt_flags" to
14084         dump_flags_t.  Convert field "optgroup_flags" to
14085         optgroup_flags_t.
14086         (dump_basic_block): Use dump_flags_t rather than int for param.
14087         (dump_generic_expr_loc): Likewise.
14088         (dump_dec): Likewise.
14089         (dump_register): Convert param "optgroup_flags" to
14090         optgroup_flags_t.
14091         (opt_info_enable_passes): Likewise.
14092         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
14093         than 0.
14094         * gimple-pretty-print.c (debug): Likewise.
14095         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
14096         (merged_store_group::apply_stores): Likewise.
14097         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
14098         * gimple.c (verify_gimple_pp): Likewise.
14099         * graphite-poly.c (print_pbb_body): Likewise.
14100         * passes.c (pass_manager::register_one_dump_file): Convert
14101         local "optgroup_flags" to optgroup_flags_t.
14102         * print-tree.c (print_node): Use TDF_NONE rather than 0.
14103         (debug): Likewise.
14104         (debug_body): Likewise.
14105         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
14106         to optgroup_flags_t.
14107         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
14108         than 0.
14109         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
14110         (convert_mult_to_fma): Likewise.
14111         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
14112         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
14113         * tree-vect-data-refs.c (dump_lower_bound): Convert param
14114         "dump_kind" to dump_flags_t.
14116 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14118         * config/rs6000/rs6000.c (min, max): Delete.
14120 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14122         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
14123         -mabi=spe and -mabi=no-spe.
14125 2018-06-08  Martin Liska  <mliska@suse.cz>
14127         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
14128         where we expect an existing summary.
14130 2018-06-08  Martin Liska  <mliska@suse.cz>
14132         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
14133         * ipa-inline.h (estimate_edge_growth): Likewise.
14135 2018-06-08  Martin Liska  <mliska@suse.cz>
14137         * cgraph.c (function_version_hasher::hash): Use
14138         cgraph_node::get_uid ().
14139         (function_version_hasher::equal):
14140         * cgraph.h (cgraph_node::get_uid): New method.
14141         * ipa-inline.c (update_caller_keys): Use
14142         cgraph_node::get_uid ().
14143         (update_callee_keys): Likewise.
14144         * ipa-utils.c (searchc): Likewise.
14145         (ipa_reduced_postorder): Likewise.
14146         * lto-cgraph.c (input_node): Likewise.
14147         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
14148         * symbol-summary.h (symtab_insertion): Likewise.
14149         (symtab_removal): Likewise.
14150         (symtab_duplication): Likewise.
14151         * tree-pretty-print.c (dump_function_header): Likewise.
14152         * tree-sra.c (convert_callers_for_node): Likewise.
14154 2018-06-08  Martin Liska  <mliska@suse.cz>
14156         * cgraph.c (symbol_table::create_edge): Always assign a new
14157         unique number.
14158         (symbol_table::free_edge): Do not recycle numbers.
14159         * cgraph.h (cgraph_edge::get): New method.
14160         * symbol-summary.h (symtab_removal): Use it.
14161         (symtab_duplication): Likewise.
14162         (call_summary::hashable_uid): Remove.
14164 2018-06-08  Martin Liska  <mliska@suse.cz>
14166         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
14167         (initialize_growth_caches): Remove.
14168         (free_growth_caches): Likewise.
14169         (do_estimate_edge_time): Use edge_growth_cache.
14170         (do_estimate_edge_size): Likewise.
14171         (do_estimate_edge_hints): Likewise.
14172         * ipa-inline.c (reset_edge_caches): Likewise.
14173         (recursive_inlining): Likewise.
14174         (inline_small_functions): Likewise.
14175         * ipa-inline.h (initialize_growth_caches): Remove.
14176         (estimate_edge_size): Likewise.
14177         (estimate_edge_time): Likewise.
14178         (estimate_edge_hints): Likewise.
14179         (reset_edge_growth_cache): Likewise.
14180         * symbol-summary.h (call_summary::remove): New method.
14182 2018-06-08  Martin Liska  <mliska@suse.cz>
14184         * ipa-cp.c (class edge_clone_summary): New summary.
14185         (grow_edge_clone_vectors): Remove.
14186         (ipcp_edge_duplication_hook): Remove.
14187         (class edge_clone_summary_t): New call_summary class.
14188         (ipcp_edge_removal_hook): Remove.
14189         (edge_clone_summary_t::duplicate): New function.
14190         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
14191         (create_specialized_node): Likewise.
14192         (ipcp_driver): Initialize edge_clone_summaries and do not
14193         register hooks.
14195 2018-06-08  Martin Liska  <mliska@suse.cz>
14197         * symbol-summary.h (get): New function.
14198         (call_summary::m_initialize_when_cloning): New class member.
14200 2018-06-08  Martin Liska  <mliska@suse.cz>
14202         * cgraph.c (cgraph_node::remove): Do not recycle uid.
14203         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
14204         (symbol_table::allocate_cgraph_symbol): Do not set uid.
14205         * passes.c (uid_hash_t): Record removed_nodes by their uids.
14206         (remove_cgraph_node_from_order): Use the removed_nodes set.
14207         (do_per_function_toporder): Likwise.
14208         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
14209         instead of summary_uid.
14210         (symtab_removal): Likewise.
14211         (symtab_duplication): Likewise.
14213 2018-06-08  Martin Liska  <mliska@suse.cz>
14215         * ipa-cp.c (ipcp_store_bits_results): Use
14216         ipcp_transformation_sum.
14217         (ipcp_store_vr_results): Likewise.
14218         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
14219         to ...
14220         (ipcp_transformation_initialize): ... this.
14221         (ipa_set_node_agg_value_chain):
14222         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
14223         (write_ipcp_transformation_info): Likewise.
14224         (read_ipcp_transformation_info): Likewise.
14225         (ipcp_update_bits): Likewise.
14226         (ipcp_update_vr): Likewise.
14227         (ipcp_transform_function): Likewise.
14228         * ipa-prop.h: Rename ipcp_transformation_summary to
14229         ipcp_transformation.
14230         (class ipcp_transformation_t): New function summary.
14231         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
14232         (ipa_get_agg_replacements_for_node): Likewise.
14234 2018-06-08  Martin Liska  <mliska@suse.cz>
14236         * ipa-pure-const.c (struct funct_state_d): Do it class instead
14237         of struct.
14238         (class funct_state_summary_t): New function_summary class.
14239         (has_function_state): Remove.
14240         (get_function_state): Likewise.
14241         (set_function_state): Likewise.
14242         (add_new_function): Likewise.
14243         (funct_state_summary_t::insert): New function.
14244         (duplicate_node_data): Remove.
14245         (remove_node_data): Remove.
14246         (funct_state_summary_t::duplicate): New function.
14247         (register_hooks): Create new funct_state_summaries.
14248         (pure_const_generate_summary): Use it.
14249         (pure_const_write_summary): Likewise.
14250         (pure_const_read_summary): Likewise.
14251         (propagate_pure_const): Likewise.
14252         (propagate_nothrow): Likewise.
14253         (dump_malloc_lattice): Likewise.
14254         (propagate_malloc): Likewise.
14255         (execute): Do not register hooks, just remove summary
14256         instead.
14257         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
14258         constructor.
14260 2018-06-08  Martin Liska  <mliska@suse.cz>
14262         * ipa-reference.c (remove_node_data): Remove.
14263         (duplicate_node_data): Likewise.
14264         (class ipa_ref_var_info_summary_t): New class.
14265         (class ipa_ref_opt_summary_t): Likewise.
14266         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
14267         (get_reference_optimization_summary): Use
14268         ipa_ref_opt_sum_summaries.
14269         (set_reference_vars_info): Remove.
14270         (set_reference_optimization_summary): Likewise.
14271         (ipa_init): Create summaries.
14272         (init_function_info): Use function summary.
14273         (ipa_ref_opt_summary_t::duplicate): New function.
14274         (ipa_ref_opt_summary_t::remove): New function.
14275         (get_read_write_all_from_node): Fix GNU coding style.
14276         (propagate): Use function summary.
14277         (write_node_summary_p): Fix GNU coding style.
14278         (stream_out_bitmap): Likewise.
14279         (ipa_reference_read_optimization_summary): Use function summary.
14280         (ipa_reference_c_finalize): Do not release hooks.
14282 2018-06-08  Martin Liska  <mliska@suse.cz>
14284         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
14285         (analyze_function_body): Extract multiple calls of get_create.
14286         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14287         * ipa-inline.c (recursive_inlining): Use ::get method.
14288         * ipa-inline.h (estimate_edge_growth): Likewise.
14290 2018-06-08  Martin Liska  <mliska@suse.cz>
14292         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
14293         HSA_INVALID.
14294         (hsa_function_summary::hsa_function_summary): Use the new enum
14295         value.
14296         (hsa_gpu_implementation_p): Use hsa_summaries::get.
14297         * hsa-gen.c (hsa_get_host_function): Likewise.
14298         (get_brig_function_name): Likewise.
14299         * ipa-hsa.c (process_hsa_functions): Likewise.
14300         (ipa_hsa_write_summary): Likewise.
14301         * symbol-summary.h (symtab_duplication): Use ::get function/
14302         (get): New function.
14304 2018-06-08  Martin Liska  <mliska@suse.cz>
14306         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
14307         of get.
14308         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
14309         (hsa_register_kernel): Likewise.
14310         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
14311         * hsa-gen.c (hsa_get_host_function): Likewise.
14312         (get_brig_function_name): Likewise.
14313         (generate_hsa): Likewise.
14314         (pass_gen_hsail::execute): Likewise.
14315         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
14316         (devirtualization_time_bonus): Likewise.
14317         (ipcp_propagate_stage): Likewise.
14318         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14319         (edge_set_predicate): Likewise.
14320         (evaluate_conditions_for_known_args): Likewise.
14321         (evaluate_properties_for_edge): Likewise.
14322         (ipa_fn_summary::reset): Likewise.
14323         (ipa_fn_summary_t::duplicate): Likewise.
14324         (dump_ipa_call_summary): Likewise.
14325         (ipa_dump_fn_summary): Likewise.
14326         (analyze_function_body): Likewise.
14327         (compute_fn_summary): Likewise.
14328         (estimate_edge_devirt_benefit): Likewise.
14329         (estimate_edge_size_and_time): Likewise.
14330         (estimate_calls_size_and_time): Likewise.
14331         (estimate_node_size_and_time): Likewise.
14332         (inline_update_callee_summaries): Likewise.
14333         (remap_edge_change_prob): Likewise.
14334         (remap_edge_summaries): Likewise.
14335         (ipa_merge_fn_summary_after_inlining): Likewise.
14336         (ipa_update_overall_fn_summary): Likewise.
14337         (read_ipa_call_summary): Likewise.
14338         (inline_read_section): Likewise.
14339         (write_ipa_call_summary): Likewise.
14340         (ipa_fn_summary_write): Likewise.
14341         (ipa_free_fn_summary): Likewise.
14342         * ipa-hsa.c (process_hsa_functions): Likewise.
14343         (ipa_hsa_write_summary): Likewise.
14344         (ipa_hsa_read_section): Likewise.
14345         * ipa-icf.c (sem_function::merge): Likewise.
14346         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14347         (do_estimate_edge_time): Likewise.
14348         (estimate_size_after_inlining): Likewise.
14349         (estimate_growth): Likewise.
14350         (growth_likely_positive): Likewise.
14351         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14352         (inline_call): Likewise.
14353         * ipa-inline.c (caller_growth_limits): Likewise.
14354         (can_inline_edge_p): Likewise.
14355         (can_inline_edge_by_limits_p): Likewise.
14356         (compute_uninlined_call_time): Likewise.
14357         (compute_inlined_call_time): Likewise.
14358         (want_inline_small_function_p): Likewise.
14359         (edge_badness): Likewise.
14360         (update_caller_keys): Likewise.
14361         (update_callee_keys): Likewise.
14362         (recursive_inlining): Likewise.
14363         (inline_small_functions): Likewise.
14364         (inline_to_all_callers_1): Likewise.
14365         (dump_overall_stats): Likewise.
14366         (early_inline_small_functions): Likewise.
14367         (early_inliner): Likewise.
14368         * ipa-inline.h (estimate_edge_growth): Likewise.
14369         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14370         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14371         * ipa-prop.h (IPA_NODE_REF): Likewise.
14372         (IPA_EDGE_REF): Likewise.
14373         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14374         (propagate_malloc): Likewise.
14375         * ipa-split.c (execute_split_functions): Likewise.
14376         * symbol-summary.h: Rename get to get_create.
14377         (get): Likewise.
14378         (get_create): Likewise.
14379         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14381 2018-06-08  Martin Liska  <mliska@suse.cz>
14383         * symbol-summary.h (release): Move definition out of class
14384         declaration.
14385         (symtab_removal): Likewise.
14386         (symtab_duplication): Likewise.
14388 2018-06-08  Martin Liska  <mliska@suse.cz>
14390         * symbol-summary.h (function_summary): Move constructor
14391         implementation out of class declaration.
14392         (release): Likewise.
14393         (symtab_insertion): Likewise.
14394         (symtab_removal): Likewise.
14395         (symtab_duplication): Likewise.
14396         (get): Likewise.
14398 2018-06-08  Martin Liska  <mliska@suse.cz>
14400         * Makefile.in: Remove support for MPX (macros, related functions,
14401         fields in cgraph_node, ...).
14402         * builtin-types.def (BT_BND): Likewise.
14403         (BT_FN_BND_CONST_PTR): Likewise.
14404         (BT_FN_CONST_PTR_BND): Likewise.
14405         (BT_FN_VOID_PTR_BND): Likewise.
14406         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14407         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14408         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14409         (expand_builtin_mempcpy_with_bounds): Likewise.
14410         (expand_builtin_memset_with_bounds): Likewise.
14411         (expand_builtin_memset_args): Likewise.
14412         (std_expand_builtin_va_start): Likewise.
14413         (expand_builtin): Likewise.
14414         (expand_builtin_with_bounds): Likewise.
14415         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14416         (DEF_LIB_BUILTIN_CHKP): Likewise.
14417         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14418         (DEF_CHKP_BUILTIN): Likewise.
14419         (BUILT_IN_MEMCPY): Likewise.
14420         (BUILT_IN_MEMMOVE): Likewise.
14421         (BUILT_IN_MEMPCPY): Likewise.
14422         (BUILT_IN_MEMSET): Likewise.
14423         (BUILT_IN_STPCPY): Likewise.
14424         (BUILT_IN_STRCAT): Likewise.
14425         (BUILT_IN_STRCHR): Likewise.
14426         (BUILT_IN_STRCPY): Likewise.
14427         (BUILT_IN_STRLEN): Likewise.
14428         (BUILT_IN_MEMCPY_CHK): Likewise.
14429         (BUILT_IN_MEMMOVE_CHK): Likewise.
14430         (BUILT_IN_MEMPCPY_CHK): Likewise.
14431         (BUILT_IN_MEMSET_CHK): Likewise.
14432         (BUILT_IN_STPCPY_CHK): Likewise.
14433         (BUILT_IN_STRCAT_CHK): Likewise.
14434         (BUILT_IN_STRCPY_CHK): Likewise.
14435         * calls.c (store_bounds): Likewise.
14436         (emit_call_1): Likewise.
14437         (special_function_p): Likewise.
14438         (maybe_warn_nonstring_arg): Likewise.
14439         (initialize_argument_information): Likewise.
14440         (finalize_must_preallocate): Likewise.
14441         (compute_argument_addresses): Likewise.
14442         (expand_call): Likewise.
14443         * cfgexpand.c (expand_call_stmt): Likewise.
14444         (expand_return): Likewise.
14445         (expand_gimple_stmt_1): Likewise.
14446         (pass_expand::execute): Likewise.
14447         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14448         (cgraph_node::remove): Likewise.
14449         (cgraph_node::dump): Likewise.
14450         (cgraph_node::verify_node): Likewise.
14451         * cgraph.h (chkp_function_instrumented_p): Likewise.
14452         (symtab_node::get_alias_target): Likewise.
14453         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14454         (cgraph_local_p): Likewise.
14455         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14456         (cgraph_edge::rebuild_references): Likewise.
14457         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14458         (walk_polymorphic_call_targets): Likewise.
14459         (cgraph_node::expand_thunk): Likewise.
14460         (symbol_table::output_weakrefs): Likewise.
14461         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14462         (ix86_handle_option): Likewise.
14463         * config/i386/constraints.md: Likewise.
14464         * config/i386/i386-builtin-types.def (BND): Likewise.
14465         (VOID): Likewise.
14466         (PVOID): Likewise.
14467         (ULONG): Likewise.
14468         * config/i386/i386-builtin.def (BDESC_END): Likewise.
14469         (BDESC_FIRST): Likewise.
14470         (BDESC): Likewise.
14471         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14472         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14473         * config/i386/i386.c (enum reg_class): Likewise.
14474         (ix86_target_string): Likewise.
14475         (ix86_option_override_internal): Likewise.
14476         (ix86_conditional_register_usage): Likewise.
14477         (ix86_valid_target_attribute_inner_p): Likewise.
14478         (ix86_set_indirect_branch_type): Likewise.
14479         (ix86_set_current_function): Likewise.
14480         (ix86_function_arg_regno_p): Likewise.
14481         (init_cumulative_args): Likewise.
14482         (ix86_function_arg_advance): Likewise.
14483         (ix86_function_arg): Likewise.
14484         (ix86_pass_by_reference): Likewise.
14485         (ix86_function_value_regno_p): Likewise.
14486         (ix86_function_value_1): Likewise.
14487         (ix86_function_value_bounds): Likewise.
14488         (ix86_return_in_memory): Likewise.
14489         (ix86_setup_incoming_vararg_bounds): Likewise.
14490         (ix86_va_start): Likewise.
14491         (indirect_thunk_need_prefix): Likewise.
14492         (print_reg): Likewise.
14493         (ix86_print_operand): Likewise.
14494         (ix86_expand_call): Likewise.
14495         (ix86_output_function_return): Likewise.
14496         (reg_encoded_number): Likewise.
14497         (BDESC_VERIFYS): Likewise.
14498         (ix86_init_mpx_builtins): Likewise.
14499         (ix86_init_builtins): Likewise.
14500         (ix86_emit_cmove): Likewise.
14501         (ix86_emit_move_max): Likewise.
14502         (ix86_expand_builtin): Likewise.
14503         (ix86_builtin_mpx_function): Likewise.
14504         (ix86_get_arg_address_for_bt): Likewise.
14505         (ix86_load_bounds): Likewise.
14506         (ix86_store_bounds): Likewise.
14507         (ix86_load_returned_bounds): Likewise.
14508         (ix86_store_returned_bounds): Likewise.
14509         (ix86_class_likely_spilled_p): Likewise.
14510         (ix86_hard_regno_mode_ok): Likewise.
14511         (x86_order_regs_for_local_alloc): Likewise.
14512         (ix86_mitigate_rop): Likewise.
14513         (ix86_bnd_prefixed_insn_p): Likewise.
14514         (ix86_mpx_bound_mode): Likewise.
14515         (ix86_make_bounds_constant): Likewise.
14516         (ix86_initialize_bounds): Likewise.
14517         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14518         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14519         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14520         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14521         (TARGET_CHKP_BOUND_MODE): Likewise.
14522         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14523         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14524         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14525         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14526         * config/i386/i386.h (TARGET_MPX): Likewise.
14527         (TARGET_MPX_P): Likewise.
14528         (VALID_BND_REG_MODE): Likewise.
14529         (FIRST_BND_REG): Likewise.
14530         (LAST_BND_REG): Likewise.
14531         (enum reg_class): Likewise.
14532         (BND_REG_P): Likewise.
14533         (BND_REGNO_P): Likewise.
14534         (BNDmode): Likewise.
14535         (ADJUST_INSN_LENGTH): Likewise.
14536         * config/i386/i386.md: Likewise.
14537         * config/i386/i386.opt: Likewise.
14538         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14539         (defined): Likewise.
14540         (LINK_MPX): Likewise.
14541         (MPX_SPEC): Likewise.
14542         (LIBMPX_SPEC): Likewise.
14543         (LIBMPXWRAPPERS_SPEC): Likewise.
14544         (CHKP_SPEC): Likewise.
14545         * config/i386/predicates.md: Likewise.
14546         * dbxout.c (dbxout_type): Likewise.
14547         * doc/extend.texi: Likewise.
14548         * doc/invoke.texi: Likewise.
14549         * doc/md.texi: Likewise.
14550         * doc/tm.texi: Likewise.
14551         * doc/tm.texi.in: Likewise.
14552         * dwarf2out.c (is_base_type): Likewise.
14553         (gen_formal_types_die): Likewise.
14554         (gen_subprogram_die): Likewise.
14555         (gen_type_die_with_usage): Likewise.
14556         (gen_decl_die): Likewise.
14557         (dwarf2out_late_global_decl): Likewise.
14558         * expr.c (expand_assignment): Likewise.
14559         (emit_storent_insn): Likewise.
14560         (store_expr_with_bounds): Likewise.
14561         (store_expr): Likewise.
14562         (expand_expr_real_1): Likewise.
14563         * expr.h (store_expr_with_bounds): Likewise.
14564         * function.c (use_register_for_decl): Likewise.
14565         (struct bounds_parm_data): Likewise.
14566         (assign_parms_augmented_arg_list): Likewise.
14567         (assign_parm_find_entry_rtl): Likewise.
14568         (assign_parm_is_stack_parm): Likewise.
14569         (assign_parm_load_bounds): Likewise.
14570         (assign_bounds): Likewise.
14571         (assign_parms): Likewise.
14572         (expand_function_start): Likewise.
14573         * gcc.c (CHKP_SPEC): Likewise.
14574         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14575         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14576         (wrestrict_dom_walker::check_call): Likewise.
14577         * gimple.c (gimple_build_call_from_tree): Likewise.
14578         * gimple.h (enum gf_mask): Likewise.
14579         (gimple_call_with_bounds_p): Likewise.
14580         (gimple_call_set_with_bounds): Likewise.
14581         * gimplify.c (gimplify_init_constructor): Likewise.
14582         * ipa-cp.c (initialize_node_lattices): Likewise.
14583         (propagate_constants_across_call): Likewise.
14584         (find_more_scalar_values_for_callers_subset): Likewise.
14585         * ipa-hsa.c (process_hsa_functions): Likewise.
14586         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14587         * ipa-icf.c (sem_function::merge): Likewise.
14588         * ipa-inline.c (early_inliner): Likewise.
14589         * ipa-pure-const.c (warn_function_noreturn): Likewise.
14590         (warn_function_cold): Likewise.
14591         (propagate_pure_const): Likewise.
14592         * ipa-ref.h (enum GTY): Likewise.
14593         * ipa-split.c (find_retbnd): Likewise.
14594         (consider_split): Likewise.
14595         (split_function): Likewise.
14596         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14597         * ipa.c (walk_polymorphic_call_targets): Likewise.
14598         (symbol_table::remove_unreachable_nodes): Likewise.
14599         (process_references): Likewise.
14600         (cgraph_build_static_cdtor_1): Likewise.
14601         * lto-cgraph.c (lto_output_node): Likewise.
14602         (output_refs): Likewise.
14603         (compute_ltrans_boundary): Likewise.
14604         (input_overwrite_node): Likewise.
14605         (input_node): Likewise.
14606         (input_cgraph_1): Likewise.
14607         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14608         * passes.c (pass_manager::execute_early_local_passes): Likewise.
14609         (class pass_chkp_instrumentation_passes): Likewise.
14610         (make_pass_chkp_instrumentation_passes): Likewise.
14611         * passes.def: Likewise.
14612         * rtl.h (struct GTY): Likewise.
14613         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14614         * stor-layout.c (layout_type): Likewise.
14615         * symtab.c: Likewise.
14616         * target.def: Likewise.
14617         * targhooks.c (default_chkp_bound_type): Likewise.
14618         (default_chkp_bound_mode): Likewise.
14619         (default_builtin_chkp_function): Likewise.
14620         (default_chkp_function_value_bounds): Likewise.
14621         (default_chkp_make_bounds_constant): Likewise.
14622         (default_chkp_initialize_bounds): Likewise.
14623         * targhooks.h (default_chkp_bound_type): Likewise.
14624         (default_chkp_bound_mode): Likewise.
14625         (default_builtin_chkp_function): Likewise.
14626         (default_chkp_function_value_bounds): Likewise.
14627         (default_chkp_make_bounds_constant): Likewise.
14628         (default_chkp_initialize_bounds): Likewise.
14629         * toplev.c (compile_file): Likewise.
14630         (process_options): Likewise.
14631         * tree-core.h (DEF_BUILTIN): Likewise.
14632         (DEF_BUILTIN_CHKP): Likewise.
14633         * tree-inline.c (declare_return_variable): Likewise.
14634         (remap_gimple_stmt): Likewise.
14635         (copy_bb): Likewise.
14636         (initialize_inlined_parameters): Likewise.
14637         (expand_call_inline): Likewise.
14638         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14639         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14640         (make_pass_ipa_chkp_produce_thunks): Likewise.
14641         (make_pass_chkp): Likewise.
14642         (make_pass_chkp_opt): Likewise.
14643         (make_pass_chkp_instrumentation_passes): Likewise.
14644         * tree-pretty-print.c (dump_generic_node): Likewise.
14645         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14646         * tree-ssa-dce.c (propagate_necessity): Likewise.
14647         (eliminate_unnecessary_stmts): Likewise.
14648         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14649         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14650         * tree-ssa-sccvn.h: Likewise.
14651         * tree-ssa-strlen.c (get_string_length): Likewise.
14652         (valid_builtin_call): Likewise.
14653         (adjust_last_stmt): Likewise.
14654         (handle_builtin_strchr): Likewise.
14655         (handle_builtin_strcpy): Likewise.
14656         (handle_builtin_stxncpy): Likewise.
14657         (handle_builtin_memcpy): Likewise.
14658         (handle_builtin_strcat): Likewise.
14659         (strlen_check_and_optimize_stmt): Likewise.
14660         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14661         * tree-streamer-in.c: Likewise.
14662         * tree-streamer.c (record_common_node): Likewise.
14663         * tree.c (tree_code_size): Likewise.
14664         (wide_int_to_tree_1): Likewise.
14665         (type_contains_placeholder_1): Likewise.
14666         (build_common_tree_nodes): Likewise.
14667         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14668         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14669         (POINTER_BOUNDS_P): Likewise.
14670         (BOUNDED_TYPE_P): Likewise.
14671         (BOUNDED_P): Likewise.
14672         (CALL_WITH_BOUNDS_P): Likewise.
14673         (pointer_bounds_type_node): Likewise.
14674         * value-prof.c (gimple_ic): Likewise.
14675         * var-tracking.c (vt_add_function_parameters): Likewise.
14676         * varasm.c (make_decl_rtl): Likewise.
14677         (assemble_start_function): Likewise.
14678         (output_constant): Likewise.
14679         (maybe_assemble_visibility): Likewise.
14680         * varpool.c (ctor_for_folding): Likewise.
14681         * chkp-builtins.def: Remove.
14682         * ipa-chkp.c: Remove.
14683         * ipa-chkp.h: Remove.
14684         * rtl-chkp.c: Remove.
14685         * rtl-chkp.h: Remove.
14686         * tree-chkp-opt.c: Remove.
14687         * tree-chkp.c: Remove.
14688         * tree-chkp.h: Remove.
14690 2018-06-07  Carl Love  <cel@us.ibm.com>
14692         * config/rs6000/vsx.md (vextract_fp_from_shorth,
14693         vextract_fp_from_shortl): Add BE support.
14695 2018-06-07  Paul Koning  <ni1d@arrl.net>
14697         * compare-elim.c (try_merge_compare): Don't merge compare if
14698         address contains a side effect.
14699         (try_eliminate_compare): Likewise.
14701 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
14703         * config.gcc: Support "tremont".
14704         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14705         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14706         PROCESSOR_TREMONT.
14707         * config/i386/i386.c (m_TREMONT): Define.
14708         (processor_target_table): Add "tremont".
14709         (PTA_TREMONT): Define.
14710         (ix86_lea_outperforms): Add TARGET_TREMONT.
14711         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14712         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14713         and M_INTEL_GOLDMONT_PLUS.
14714         (fold_builtin_cpu): Add "tremont".
14715         (ix86_add_stmt_cost): Add TARGET_TREMONT.
14716         (ix86_option_override_internal): Add "tremont".
14717         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14718         (processor_type): Add PROCESSOR_TREMONT.
14719         * config/i386/x86-tune.def: Add m_TREMONT.
14720         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14722 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14724         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14725         symbol defined for msp430i* devices to be lower case.
14727 2018-06-07  Richard Biener  <rguenther@suse.de>
14729         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14730         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
14731         Properly wrap signed arithmetic if overflow wraps.
14733 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
14735         PR tree-optimization/69615
14736         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14737         of a cast from a same precision integral SSA_NAME in a bb dominated
14738         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
14739         cast to utype if rhs2 has already a compatible type.
14741 2018-06-07  Richard Biener  <rguenther@suse.de>
14743         PR tree-optimization/85935
14744         * graphite-scop-detection.c (find_params_in_bb): Analyze
14745         condition operands with respect to the correct loop.  Assert
14746         the analysis doesn't fail.
14748 2018-06-04  Carl Love  <cel@us.ibm.com>
14750         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14751         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
14752         as it is slightly cheaper.
14753         (first_match_or_eos_index_<mode>):
14754         Calculate index using natural element order.
14755         (first_match_index_<mode>):
14756         Calculate index using natural element order.
14757         (first_match_or_eos_index_<mode>):
14758         Calculate index using natural order.
14759         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14760         for BE and LE modes.
14761         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14762         P9V_BUILTIN_VCLZLSBB_V16QI.
14763         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14764         specific.
14766 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14768         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14769         indentation and line wrap for many prototypes.  Add missing
14770         @smallexample directives around block of prototypes for vec_xl and
14771         vec_xst.
14773 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
14775         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14776         track if we pass or return IEEE 128-bit floating point.
14777         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14778         C++ mangling that is compatible with GCC 8.1.
14779         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
14780         (init_cumulative_args): Note if we pass or return IEEE 128-bit
14781         floating point types.
14782         (rs6000_function_arg_advance_1): Likewise.
14783         (rs6000_mangle_type): Optionally generate mangled names that match
14784         what GCC 8.1 generated for IEEE 128-bit floating point types.
14785         (rs6000_globalize_decl_name): If we have an external function that
14786         passes or returns IEEE 128-bit types, generate a weak reference
14787         from the mangled name used in GCC 8.1 to the current mangled
14788         name.
14789         (rs6000_init_builtins): Make __ibm128 use the long double type if
14790         long double is IBM extended double.  Make __float128 use the long
14791         double type if long double is IEEE 128-bit.
14793         PR target/85657
14794         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
14795         macro for __ibm128 built-in functions.
14796         (PACK_IF): Add __ibm128 pack/unpack functions.
14797         (UNPACK_IF): Likewise.
14798         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
14799         enable long double built-in functions if long double is IEEE
14800         128-bit floating point.
14801         (rs6000_invalid_builtin): Update long double built-in function
14802         error message.
14803         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
14804         functions, adjust the built-in function to use the long double
14805         built-in function if __ibm128 and long double are the same type.
14806         * doc/extend.texi (PowerPC builtins): Update documention for
14807         __builtin_{,un}pack_longdouble.  Add documentation for
14808         __builtin_{,un}pack_ibm128.
14810 2018-06-06  Jim Wilson  <jimw@sifive.com>
14812         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
14813         (struct machine_function): New field interrupt_mode.
14814         (riscv_handle_type_attribute): New function.  Add forward declaration.
14815         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
14816         (riscv_expand_epilogue): Check interrupt_mode field.
14817         (riscv_set_current_function): Check interrupt attribute args and
14818         set interrupt_mode field.
14819         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
14820         (riscv_sret, riscv_uret): New.
14821         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
14822         new arguments to interrupt attribute.
14824 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
14826         PR target/63177
14827         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
14828         Don't handle -mcpu=power8 if -mpower9-vector is also used.
14830 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14832         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
14833         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
14834         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14835         several redundant entries.
14837 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
14839         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
14840         type from "rtx" to "rtx_insn *".
14841         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
14842         for local "call_insn", removing cast.
14843         (ix86_expand_call): Likewise, introducing a "call_insn" local.
14845 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
14847         PR tree-optimization/86066
14848         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
14849         for BIT_INSERT_EXPR stores.
14851 2018-06-06  Richard Biener  <rguenther@suse.de>
14853         PR tree-optimization/86062
14854         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14855         component refs ontop
14856         of to be offsetted base.
14858 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14860         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
14861         to be static and remove check on interrupt attribute name.
14863 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14865         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14866         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
14868 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
14870         PR target/79924
14871         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
14872         second argument.
14873         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
14874         Remove second argument, change how error is called.
14875         (aarch64_layout_arg): Remove second argument from
14876         aarch64_err_no_fpadvsimd call.
14877         (aarch64_init_cumulative_args): Ditto.
14878         (aarch64_gimplify_va_arg_expr): Ditto.
14879         * config/aarch64/aarch64.md (mov<mode>): Ditto.
14881 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
14883         * config/i386/i386.md (simple_return_indirect_internal): New expander.
14884         (*simple_return_indirect_internal<mode>): Rename from
14885         simple_return_indirect_internal.  Use W mode iterator.
14886         (rstorssp): New expander.
14887         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
14888         (clrssbsy): New expander.
14889         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
14891 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14893         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
14894         __typeof__.
14895         (cmse_check_pointed_object): Likewise.
14897 2018-06-05  Martin Liska  <mliska@suse.cz>
14899         PR gcov-profile/47618
14900         * doc/invoke.texi: Document how -fprofile-dir format
14901         is extended.
14903 2018-06-05  Richard Biener  <rguenther@suse.de>
14905         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
14906         removal pretend DOM info isn't available so we do not update
14907         it and only remove edges, not dominated blocks.  Actually free
14908         DOM info in case we removed something.  Remove unreachable blocks.
14909         (mfb_keep_latches): Work with either DOM info or marked backedges.
14910         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
14911         first.  Mark backedges if DOM info isn't available.
14912         (Re-)compute DOM info after cleanup_control_flow_pre.
14914 2018-06-05  Richard Biener  <rguenther@suse.de>
14916         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
14917         (locus_discrim_hasher::hash): Adjust.
14918         (locus_discrim_hasher::equal): Likewise.
14919         (next_discriminator_for_locus): Work on line directly.
14920         (same_line_p): Pass in expanded locus1 as well.
14921         (assign_discriminators): Avoid redundant location expansions.
14923 2018-06-05  Richard Biener  <rguenther@suse.de>
14925         PR tree-optimization/86046
14926         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
14927         if required after clearing TREE_ADDRESSABLE.
14929 2018-06-05  Richard Biener  <rguenther@suse.de>
14931         PR tree-optimization/86047
14932         * tree-ssa-loop.c (for_each_index): Glob handling of all
14933         decls and constants and really handle all of them.
14935 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14937         PR target/81497
14938         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
14939         qualifier_void_pointer and qualifier_const_void_pointer.
14940         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
14941         (arm_init_builtins): Handle the above.
14942         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
14943         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
14944         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
14945         void intrinsics.
14947 2018-06-05  Martin Liska  <mliska@suse.cz>
14949         * auto-profile.c (read_autofdo_file): Do not use
14950         gcov_ctr_summary struct.
14951         (afdo_callsite_hot_enough_for_early_inline): Likewise.
14952         * coverage.c (struct counts_entry): Likewise.
14953         (read_counts_file): Read just single summary entry.
14954         (get_coverage_counts): Use gcov_summary struct.
14955         * coverage.h (get_coverage_counts): Likewise.
14956         * gcov-dump.c (dump_working_sets): Likewise.
14957         (tag_summary): Dump just single summary.
14958         * gcov-io.c (gcov_write_summary): Write just histogram
14959         summary.
14960         (gcov_read_summary): Read just single summary.
14961         (compute_working_sets): Use gcov_summary struct.
14962         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
14963         of GCOV_COUNTERS_SUMMABLE.
14964         (GCOV_COUNTERS_SUMMABLE): Remove.
14965         (GCOV_FIRST_VALUE_COUNTER): Replace with
14966         GCOV_COUNTER_V_INTERVAL.
14967         (struct gcov_ctr_summary): Remove.
14968         (struct gcov_summary): Directly use fields of former
14969         gcov_ctr_summary.
14970         (compute_working_sets): Use gcov_summary struct.
14971         * gcov.c (read_count_file): Do not use ctrs fields.
14972         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
14973         struct.
14974         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
14975         struct.
14976         * profile.c: Likewise.
14977         * profile.h: Likewise.
14979 2018-06-05  Martin Liska  <mliska@suse.cz>
14981         PR gcov-profile/84846
14982         * gcov.c (output_lines): Print working directory only
14983         in intermediate format.
14985 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
14987         * config/s390/s390-builtin-types.def: Add void function type.
14988         * config/s390/s390-builtins.def: Use the function type for the
14989         tbeginc builtin.
14991 2018-06-04  Jim Wilson  <jimw@sifive.com>
14993         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
14994         to int.
14995         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
14996         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
14997         handle EH_RETURN_DATA_REGNO registers properly.
14998         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
14999         (riscv_expand_epilogue): Update comment.  Change argument name and
15000         type.  Update code to use new name and type.  Pass new args to
15001         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
15002         EXCEPTION_RETURN.
15003         * config/riscv/riscv.md (NORMAL_RETURN): New.
15004         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
15005         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
15006         (eh_return): Call gen_eh_return_internal and emit barrier.
15007         (eh_return_internal): Call riscv_expand_epilogue.
15009 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
15011         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
15012         bit_insertion field and declare can_be_merged_into method.
15013         (merged_store_group::can_be_merged_into): New method.
15014         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
15015         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
15016         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
15018 2018-06-04  Richard Biener  <rguenther@suse.de>
15020         PR tree-optimization/85955
15021         * builtins.c (fold_builtin_sincos): Convert pointers to
15022         destination to appropriate type before dereferencing.
15024 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
15026         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
15028 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
15030         * expr.c (expand_expr_real_1): Force the operand into memory if
15031         its TYPE_MODE is BLKmode and if there is no integer mode for
15032         the number of bits being extracted.
15034 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15036         PR target/85832
15037         PR target/86036
15038         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15039         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
15041 2018-06-04  Richard Biener  <rguenther@suse.de>
15043         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
15044         (cleanup_tree_cfg_noloop): ... single caller.  Do
15045         start_recording_case_labels later.
15047 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
15049         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
15050         to _IMMINTRIN_H_INCLUDED.
15051         * config/i386/pconfigintrin.h: Ditto.
15052         * config/i386/waitpkgintrin.h: Ditto.
15053         * config/i386/immintrin.h: Add includes for sgxintrin.h,
15054         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
15055         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
15056         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
15057         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
15058         waitpkgintrin.h and cldemoteintrin.h.
15060 2018-06-04  Richard Biener  <rguenther@suse.de>
15062         PR tree-optimization/86038
15063         * tracer.c (find_best_successor): Check probability for
15064         being initialized, bail out if not.
15066 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
15068         PR target/86003
15069         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
15070         of bits to ignore when comparing architectures.
15072 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15074         PR tree-optimization/69615
15075         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
15076         maximum or minimum of the type, try to merge it also as if
15077         range1 is + [-, x - 1] or + [x + 1, -].
15079         PR c++/86025
15080         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
15082 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
15084         PR tree-optimization/86034
15085         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
15086         the unsigned bitfield type in a bit insertion sequence if it does not
15087         have a larger precision than the bitfield size.
15088         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
15090 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
15092         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
15094 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15096         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
15097         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
15098         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
15099         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
15101 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15103         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
15104         Disable -fdelete-null-pointer-checks for ELF toolchain.
15106 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15107             Kito Cheng  <kito.cheng@gmail.com>
15109         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
15110         (nds32le-*-*, nds32be-*-*): Integrate checking process.
15111         (nds32*-*-*): Add glibc and uclibc conditions.
15112         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
15113         (TARGET_EXCEPT_UNWIND_INFO): Define.
15114         * config/nds32/elf.h: New file.
15115         * config/nds32/linux.h: New file.
15116         * config/nds32/nds32-elf.opt: New file.
15117         * config/nds32/nds32-linux.opt: New file.
15118         * config/nds32/nds32-fp-as-gp.c
15119         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
15120         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
15121         TARGET_LINUX_ABI.
15122         (nds32_asm_file_end): Ditto.
15123         (nds32_print_operand): Ditto.
15124         (nds32_insert_attributes): Ditto.
15125         (nds32_init_libfuncs): New function.
15126         (TARGET_HAVE_TLS): Define.
15127         (TARGET_INIT_LIBFUNCS): Define.
15128         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
15129         spec content.
15130         (TARGET_ELF): Apply different mcmodel setting.
15131         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
15132         been migrated into elf.h and linux.h files.
15133         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
15134         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
15135         (mcmodel): The content has been migrated into nds32-elf.opt and
15136         nds32-linux.opt files.
15137         * config/nds32/t-elf: New file.
15138         * config/nds32/t-linux: New file.
15140 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15141             Shiva Chen  <shiva0217@gmail.com>
15143         * config/nds32/constants.md (unspec_volatile_element): Add
15144         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
15145         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
15146         optimization.
15147         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
15148         (make_pass_nds32_fp_as_gp): Declare.
15149         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
15150         optmization pass.
15151         (nds32_asm_function_end_prologue): Remove unused asm output.
15152         (nds32_asm_function_begin_epilogue): Remove unused asm output.
15153         (nds32_asm_file_start): Output necessary fp_as_gp information.
15154         (nds32_option_override): Adjust register usage.
15155         (nds32_expand_prologue): Consider fp_as_gp situation.
15156         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
15157         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
15158         (epilogue): Ditto.
15159         (return): Ditto.
15160         (simple_return): Ditto.
15161         (omit_fp_begin): Output special directive for fp_as_gp.
15162         (omit_fp_end): Output special directive for fp_as_gp.
15163         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
15164         mforbid-fp-as-gp): New options.
15166 2018-06-01  Mark Wielaard  <mark@klomp.org>
15168         * dwarf2out.c (dwarf2out_finish): Remove generation of
15169         DW_AT_loclists_base.
15171 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
15173         * gimple-ssa-store-merging.c: Include gimple-fold.h.
15174         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
15175         (struct merged_store_group): Add bit_insertion field.
15176         (dump_char_array): Use standard hexadecimal format.
15177         (merged_store_group::merged_store_group): Set bit_insertion to false.
15178         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
15179         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
15180         also print the mask in the dump file.
15181         (pass_store_merging::gate): Minor tweak.
15182         (imm_store_chain_info::coalesce_immediate): Fix wrong association
15183         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
15184         stores with INTEGER_CST stores.
15185         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
15186         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
15187         and use it throughout.  Generate bit insertion sequences if need be.
15188         (pass_store_merging::process_store): Remove redundant condition.
15189         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
15191 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
15193         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
15194         the 128-bit floating point types.  Fix function comment.
15196 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15198         * config/aarch64/aarch64-simd.md
15199         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
15200         mnemonics.
15201         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
15202         mnemonics.
15204 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15206         PR tree-optimization/85989
15207         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
15208         variable.
15209         (backprop::intersect_uses): Check it when deciding whether this
15210         is a backedge reference.
15211         (backprop::process_block): Add each phi to m_visited_phis
15212         after visiting it, then clear it at the end.
15214 2018-06-01  Richard Biener  <rguenther@suse.de>
15216         * tree-vectorizer.h (vect_dr_stmt): New function.
15217         (vect_get_load_cost): Adjust.
15218         (vect_get_store_cost): Likewise.
15219         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
15220         Use vect_dr_stmt instead of DR_SMTT.
15221         (vect_record_base_alignments): Likewise.
15222         (vect_calculate_target_alignment): Likewise.
15223         (vect_compute_data_ref_alignment): Likewise and make static.
15224         (vect_update_misalignment_for_peel): Likewise.
15225         (vect_verify_datarefs_alignment): Likewise.
15226         (vector_alignment_reachable_p): Likewise.
15227         (vect_get_data_access_cost): Likewise.  Pass down
15228         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
15229         (vect_get_peeling_costs_all_drs): Likewise.
15230         (vect_peeling_hash_get_lowest_cost): Likewise.
15231         (vect_enhance_data_refs_alignment): Likewise.
15232         (vect_find_same_alignment_drs): Likewise.
15233         (vect_analyze_data_refs_alignment): Likewise.
15234         (vect_analyze_group_access_1): Likewise.
15235         (vect_analyze_group_access): Likewise.
15236         (vect_analyze_data_ref_access): Likewise.
15237         (vect_analyze_data_ref_accesses): Likewise.
15238         (vect_vfa_segment_size): Likewise.
15239         (vect_small_gap_p): Likewise.
15240         (vectorizable_with_step_bound_p): Likewise.
15241         (vect_prune_runtime_alias_test_list): Likewise.
15242         (vect_analyze_data_refs): Likewise.
15243         (vect_supportable_dr_alignment): Likewise.
15244         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
15245         (vect_gen_prolog_loop_niters): Likewise.
15246         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
15247         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
15248         modify DR_STMT.
15249         (vect_recog_mask_conversion_pattern): Likewise.
15250         (vect_try_gather_scatter_pattern): Likewise.
15251         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
15252         to vect_get_store_cost.
15253         (vect_get_store_cost): Get stmt_info instead of DR.
15254         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
15255         (vect_get_load_cost): Get stmt_info instead of DR.
15257 2018-06-01  Richard Biener  <rguenther@suse.de>
15259         PR middle-end/86017
15260         * gimple-fold.c (var_decl_component_p): Also allow offsetted
15261         vars wrapped in MEM_REFs.
15263 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15265         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15266         Fix subreg tests so that we only return a choice between
15267         GENERAL_REGS and FP_REGS if the original classes included both.
15269 2018-06-01  Richard Biener  <rguenther@suse.de>
15271         PR ipa/85960
15272         * tree-ssa-structalias.c (get_function_part_constraint):
15273         Handle NULL fi->decl.
15274         (find_func_aliases_for_call): Properly handle indirect
15275         fi from direct call.
15276         (find_func_clobbers): Likewise.
15277         (ipa_pta_execute): Likewise.
15278         (create_variable_info_for): For functions that are ifunc_resolver
15279         resolve to a varinfo that contains the result of the resolver call.
15280         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
15281         aliases.
15283 2018-05-31  Michael Collison  <michael.collison@arm.com>
15285         * config/aarch64/aarch64.md:
15286         (*fix_to_zero_extenddfdi2): New pattern.
15287         * gcc.target/aarch64/fix_extend1.c: New testcase.
15289 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
15291         PR middle-end/78809
15292         PR middle-end/83026
15293         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
15294         and BUILT_IN_STRNCMP_EQ.
15295         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
15296         BUILT_IN_STRNCMP_EQ.
15297         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
15298         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15299         (gimple_fold_builtin): Likewise.
15300         * tree-ssa-strlen.c (compute_string_length): New function.
15301         (determine_min_obsize): New function.
15302         (handle_builtin_string_cmp): New function to handle calls to
15303         string compare functions.
15304         (strlen_optimize_stmt): Add handling to builtin string compare
15305         calls.
15306         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
15307         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15308         * tree.c (build_common_builtin_nodes): Add new defines of
15309         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
15311 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
15313         PR target/85984
15314         * bb-reorder.c (pass_partition_blocks::gate): Return false for
15315         functions with naked attribute.
15317 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
15319         * config/i386/sse.md (avx_vec_concat<mode>):
15320         Substitute concat_tg_mode mode attribute with xtg_mode.
15321         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
15322         (concat_tg_mode): Remove mode attribute.
15324 2018-05-31  Martin Sebor  <msebor@redhat.com>
15326         PR c/82063
15327         * calls.c (alloc_max_size): Correct a logic error/typo.
15328         Treat excessive arguments as infinite.  Warn for invalid arguments.
15329         * doc/invoke.texi (-Walloc-size-larger-than): Update.
15331 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
15333         PR target/85829
15334         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
15335         and movx for Haswell.
15337 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
15338             Cesar Philippidis  <cesar@codesourcery.com>
15340         PR middle-end/85879
15341         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
15342         when emitting error on private/firstprivate reductions.
15343         * omp-low.c (lower_omp_target): Avoid reference-type processing
15344         on pointers for firstprivate clause.
15346 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15348         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15349         (st1x2): Likewise.
15350         (st1x3): Likewise.
15351         * config/aarch64/aarch64-simd.md
15352         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
15353         (aarch64_ld1_x3_<mode>): Likewise
15354         (aarch64_st1x2<VALLDIF:mode>): Likewise
15355         (aarch64_st1_x2_<mode>): Likewise
15356         (aarch64_st1x3<VALLDIF:mode>): Likewise
15357         (aarch64_st1_x3_<mode>): Likewise
15358         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
15359         (vld1_s8_x3): Likewise.
15360         (vld1_u16_x3): Likewise.
15361         (vld1_s16_x3): Likewise.
15362         (vld1_u32_x3): Likewise.
15363         (vld1_s32_x3): Likewise.
15364         (vld1_u64_x3): Likewise.
15365         (vld1_s64_x3): Likewise.
15366         (vld1_f16_x3): Likewise.
15367         (vld1_f32_x3): Likewise.
15368         (vld1_f64_x3): Likewise.
15369         (vld1_p8_x3): Likewise.
15370         (vld1_p16_x3): Likewise.
15371         (vld1_p64_x3): Likewise.
15372         (vld1q_u8_x3): Likewise.
15373         (vld1q_s8_x3): Likewise.
15374         (vld1q_u16_x3): Likewise.
15375         (vld1q_s16_x3): Likewise.
15376         (vld1q_u32_x3): Likewise.
15377         (vld1q_s32_x3): Likewise.
15378         (vld1q_u64_x3): Likewise.
15379         (vld1q_s64_x3): Likewise.
15380         (vld1q_f16_x3): Likewise.
15381         (vld1q_f32_x3): Likewise.
15382         (vld1q_f64_x3): Likewise.
15383         (vld1q_p8_x3): Likewise.
15384         (vld1q_p16_x3): Likewise.
15385         (vld1q_p64_x3): Likewise.
15386         (vst1_s64_x2): Likewise.
15387         (vst1_u64_x2): Likewise.
15388         (vst1_f64_x2): Likewise.
15389         (vst1_s8_x2): Likewise.
15390         (vst1_p8_x2): Likewise.
15391         (vst1_s16_x2): Likewise.
15392         (vst1_p16_x2): Likewise.
15393         (vst1_s32_x2): Likewise.
15394         (vst1_u8_x2): Likewise.
15395         (vst1_u16_x2): Likewise.
15396         (vst1_u32_x2): Likewise.
15397         (vst1_f16_x2): Likewise.
15398         (vst1_f32_x2): Likewise.
15399         (vst1_p64_x2): Likewise.
15400         (vst1q_s8_x2): Likewise.
15401         (vst1q_p8_x2): Likewise.
15402         (vst1q_s16_x2): Likewise.
15403         (vst1q_p16_x2): Likewise.
15404         (vst1q_s32_x2): Likewise.
15405         (vst1q_s64_x2): Likewise.
15406         (vst1q_u8_x2): Likewise.
15407         (vst1q_u16_x2): Likewise.
15408         (vst1q_u32_x2): Likewise.
15409         (vst1q_u64_x2): Likewise.
15410         (vst1q_f16_x2): Likewise.
15411         (vst1q_f32_x2): Likewise.
15412         (vst1q_f64_x2): Likewise.
15413         (vst1q_p64_x2): Likewise.
15414         (vst1_s64_x3): Likewise.
15415         (vst1_u64_x3): Likewise.
15416         (vst1_f64_x3): Likewise.
15417         (vst1_s8_x3): Likewise.
15418         (vst1_p8_x3): Likewise.
15419         (vst1_s16_x3): Likewise.
15420         (vst1_p16_x3): Likewise.
15421         (vst1_s32_x3): Likewise.
15422         (vst1_u8_x3): Likewise.
15423         (vst1_u16_x3): Likewise.
15424         (vst1_u32_x3): Likewise.
15425         (vst1_f16_x3): Likewise.
15426         (vst1_f32_x3): Likewise.
15427         (vst1_p64_x3): Likewise.
15428         (vst1q_s8_x3): Likewise.
15429         (vst1q_p8_x3): Likewise.
15430         (vst1q_s16_x3): Likewise.
15431         (vst1q_p16_x3): Likewise.
15432         (vst1q_s32_x3): Likewise.
15433         (vst1q_s64_x3): Likewise.
15434         (vst1q_u8_x3): Likewise.
15435         (vst1q_u16_x3): Likewise.
15436         (vst1q_u32_x3): Likewise.
15437         (vst1q_u64_x3): Likewise.
15438         (vst1q_f16_x3): Likewise.
15439         (vst1q_f32_x3): Likewise.
15440         (vst1q_f64_x3): Likewise.
15441         (vst1q_p64_x3): Likewise.
15443 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15445         * config/msp430/msp430.c (msp430_output_labelref): Prepend
15446         user_label_prefix to name.
15448         * tree-core.h: Update comment about the format of NAME string
15449         passed to handler in attribute_spec.
15451         * config/msp430/msp430.md: Remove erroneous subreg expression from
15452         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15453         zero_extend{q,h}isi2.
15455 2018-05-30  Borislav Petkov  <bp@suse.de>
15457         * doc/extend.texi: Document some architecture specific
15458         constraints and sort entries.
15460 2018-05-30  Martin Sebor  <msebor@redhat.com>
15462         PR middle-end/85369
15463         * builtins.c (expand_builtin_stpcpy_1): New function.
15464         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15465         only if the former succeeds.
15467 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15469         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15470         in saphira.
15472 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15474         * doc/invoke.texi (-flinker-output): Document
15476 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15478         * passes.c (ipa_write_summaries): Only modify statements if body
15479         is in memory.
15480         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15481         incrementally linking.
15482         (ipa_passes): Likewise.
15483         * lto-cgraph.c (lto_output_node): When incrementally linking do not
15484         pass down resolution info.
15485         * common.opt (flag_incremental_link): Update info.
15486         * gcc.c (plugin specs): Turn flinker-output=* to
15487         -plugin-opt=-linker-output-known
15488         * toplev.c (compile_file): Also cut compilation when doing incremental
15489         link.
15490         * flag-types. (enum lto_partition_model): Add
15491         LTO_LINKER_OUTPUT_NOLTOREL.
15492         (invoke.texi): Add -flinker-output docs.
15493         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15494         link same way as WPA; do not stream in dead initializers.
15496         * dwarf2out.c (dwarf2out_die_ref_for_decl,
15497         darf2out_register_external_decl): Support incremental link.
15499 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15501         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15503 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15505         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15506         it down to simple_object_copy_lto_debug_sections.
15507         (run_gcc): Determine incremental LTO link time and configure
15508         lto1 into non-wpa mode, disable renaming of debug sections.
15510 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15512         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15513         descriptions of various incorrectly documented functions.
15515 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15517         Revert:
15518         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15519         address check not strict.
15521 2018-05-30  Richard Biener  <rguenther@suse.de>
15523         PR tree-optimization/85964
15524         * tracer.c (better_p): Drop initialized count check, we only
15525         call the function with initialized counts now.
15526         (find_best_successor): Do find a best edge if one
15527         has uninitialized count.
15528         (find_best_predecessor): Likewise.  Do BB frequency check only
15529         if count is initialized.
15531 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
15533         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15534         (aarch64_ldrstr_offset_compare): New.
15535         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15536         load/store orderings.
15537         (aarch64_gen_adjusted_ldpstp): Likewise.
15539 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
15541         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15542         Check for subset of GENERAL_REGS and FP_REGS.
15543         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15544         r=w alternative.
15546 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
15548         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15549         and wi::to_poly_offset.  Add the current offset and then check
15550         whether the sum fits, rather than using an unchecked addition of
15551         a checked term.  Check for a shwi rather than a uhwi.
15552         * expr.c (get_bit_range): Use tree_to_poly_uint64.
15553         (store_constructor): Use poly_int_tree_p.
15554         (expand_expr_real_1): Likewise.
15555         * function.c (assign_temp): Likewise.
15556         * fold-const.c (const_binop): Use poly_int_tree_p and
15557         wi::to_poly_offset.
15558         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
15559         division.
15560         * ipa-icf-gimple.c (func_checker::compare_operand): Use
15561         to_poly_offset for MEM offsets.
15562         * ipa-icf.c (sem_variable::equals): Likewise.
15563         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15564         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15565         wi::to_poly_offset for BIT_FIELD_REF offsets.
15566         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15567         wi::to_poly_offset.
15568         * var-tracking.c (emit_note_insn_var_location): Use
15569         tree_to_poly_uint64.
15571 2018-05-29  Jim Wilson  <jimw@sifive.com>
15573         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15575 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
15577         PR target/85950
15578         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15579         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15580         sequence.
15581         (sse4_1_round<mode>2): Use nonimmediate_operand
15582         for operand 1 predicate.
15584 2018-05-29  Martin Sebor  <msebor@redhat.com>
15585             Richard Biener  <rguenther@suse.de>
15587         PR testsuite/85888
15588         * calls.c (get_size_range): Call determine_value_range instead
15589         of get_value_range..
15590         * tree-vrp.h (determine_value_range): Declared new function.
15591         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15593 2018-05-29  Richard Biener  <rguenther@suse.de>
15595         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15596         sure to use non-pattern stmts for get_earlier_stmt arguments.
15597         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15598         called on pattern stmts.
15599         (get_later_stmt): Likewise.
15601 2018-05-29  Martin Liska  <mliska@suse.cz>
15603         PR gcov-profile/85759
15604         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15605         env variables.
15607 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
15609         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15610         VEC_UNPACK_*_EXPR.
15611         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15612         VEC_PACK_*_EXPR.
15614         PR target/85918
15615         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15616         VEC_PACK_FLOAT_EXPR): New tree codes.
15617         * tree-pretty-print.c (op_code_prio): Handle
15618         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15619         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15620         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15621         * tree-inline.c (estimate_operator_cost): Likewise.
15622         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15623         * fold-const.c (const_binop): Likewise.
15624         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15625         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15626         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15627         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15628         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15629         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15630         * expr.c (expand_expr_real_2): Likewise.
15631         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15632         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15633         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15634         optabs.
15635         * optabs.c (expand_widen_pattern_expr): For
15636         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15637         sign from result type rather than operand's type.
15638         (expand_binop_directly): For vec_packu_float_optab and
15639         vec_packs_float_optab allow result type to be different from operand's
15640         type.
15641         * optabs-tree.c (optab_for_tree_code): Handle
15642         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15643         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
15644         * tree-vect-generic.c (expand_vector_operations_1):  Handle
15645         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15646         VEC_PACK_FLOAT_EXPR.
15647         * tree-vect-stmts.c (supportable_widening_operation): Handle
15648         FIX_TRUNC_EXPR.
15649         (supportable_narrowing_operation): Handle FLOAT_EXPR.
15650         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15651         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15652         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
15653         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15654         mode attributes.
15655         (vec_pack<floatprefix>_float_<mode>): New expander.
15656         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15657         attributes.
15658         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15659         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15660         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15661         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15662         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15663         Document.
15664         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15665         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15666         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15667         VEC_PACK_FLOAT_EXPR): Document.
15669 2018-05-29  Richard Biener  <rguenther@suse.de>
15671         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15672         member.
15673         (stmt_vec_info_vec): Make pointer.
15674         (init_stmt_vec_info_vec): Remove.
15675         (free_stmt_vec_info_vec): Likewise.
15676         (set_stmt_vec_info_vec): New function.
15677         (free_stmt_vec_infos): Likewise.
15678         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15679         (set_vinfo_for_stmt): Likewise.
15680         (get_earlier_stmt): Likewise.
15681         (get_later_stmt): Likewise.
15682         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15683         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15684         (vec_info::~vec_info): Free stmt_vec_infos.
15685         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15686         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15687         (pass_slp_vectorize::execute): Likewise.
15688         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15689         (free_stmt_vec_info_vec): Likewise.
15690         (set_stmt_vec_info_vec): New function.
15691         (free_stmt_vec_infos): Likewise.
15692         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15693         the global stmt_vec_info_vec.
15694         * tree-parloops.c (gather_scalar_reductions): Use
15695         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15696         vector.
15698 2018-05-29  Richard Biener  <rguenther@suse.de>
15700         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15702 2018-05-29  Martin Liska  <mliska@suse.cz>
15703             David Malcolm  <dmalcolm@redhat.com>
15705         * vec.c (test_reverse): New.
15706         (vec_c_tests): Add new test.
15707         * vec.h (vl_ptr>::reverse): New function.
15709 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
15711         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15713         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15714         and later.
15716 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15718         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15720 2018-05-28  Richard Biener  <rguenther@suse.de>
15722         PR tree-optimization/85933
15723         * tree-vect-data-refs.c (vect_record_base_alignments): Only
15724         look at stmts marked as vectorizable.
15726 2018-05-28  Richard Biener  <rguenther@suse.de>
15728         PR tree-optimization/85934
15729         * tree-vect-generic.c (expand_vector_operations_1): Hoist
15730         vector boolean check before scalar optimization.
15732 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
15734         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15735         for armv5te.
15737 2018-05-28  Mark Wielaard  <mark@klomp.org>
15739         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15740         if it is an expression containing a minus sign.
15742 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
15744         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15746 2018-05-27  Paul Koning  <ni1d@arrl.net>
15748         * config/pdp11/pdp11.md (truncsihi2): Remove.
15750 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
15751             Chung-Ju Wu  <jasonwucj@gmail.com>
15753         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15754         implementation.
15755         (unaligned_store_dw): Ditto.
15756         * config/nds32/nds32-memory-manipulation.c
15757         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15758         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15759         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15760         (emit_setmem_word_loop): Rename to ...
15761         (emit_setmem_doubleword_loop): ... this.
15762         (nds32_gen_dup_4_byte_to_word_value): New function.
15763         (nds32_gen_dup_8_byte_to_double_word_value): New function.
15764         (nds32_expand_setmem_loop): Refine implementation.
15765         (nds32_expand_setmem_loop_v3m): Ditto.
15766         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15767         pattern.
15769 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15771         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15773 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15775         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15776         (nds32_init_machine_status): Initialize machine->attr_naked_p and
15777         machine->attr_no_prologue_p.
15778         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
15779         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
15780         (nds32_expand_epilogue): Consider attr_naked_p.
15781         (nds32_expand_epilogue_v3pop): Likewise.
15782         (nds32_can_use_return_insn): Likewise.
15783         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
15784         attr_no_prologue_p fields.
15785         * config/nds32/nds32.opt (mret-in-naked-func): New option.
15787 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
15789         PR target/85918
15790         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
15791         attributes.
15792         * config/i386/sse.md
15793         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
15794         Rename to ...
15795         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
15796         ... this.
15797         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
15798         Rename to ...
15799         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
15800         ... this.
15801         (*<floatsuffix>floatv2div2sf2): Rename to ...
15802         (*float<floatunssuffix>v2div2sf2): ... this.
15803         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
15804         (float<floatunssuffix>v2div2sf2_mask): ... this.
15805         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
15806         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
15807         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
15808         to ...
15809         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
15810         ... this.
15811         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15812         Rename to ...
15813         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15814         ... this.
15815         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15816         Rename to ...
15817         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15818         ... this.
15819         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
15820         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
15821         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
15822         gen_ufix_truncv8dfv8si2.
15823         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
15824         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
15825         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
15826         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
15827         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
15828         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
15829         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
15830         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
15832 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15834         PR target/85900
15835         PR target/85345
15836         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
15838 2018-05-25  Jim Wilson  <jimw@sifive.com>
15840         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
15841         * config/riscv/riscv.c (struct machine_function): Add
15842         interrupt_handler_p and attribute_checked_p fields.
15843         (riscv_attribute_table): Add interrupt.
15844         (riscv_interrupt_type_p): New.
15845         (riscv_save_reg_p): Save extra regs for interrupt handler.
15846         (riscv_use_save_libcall): Return false  for interrupt handler.
15847         (riscv_first_stack_step): Add forward declaration.
15848         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
15849         for interrupt handler with large frame.  Use it for saved reg list.
15850         (riscv_expand_prologue): Move flag_stack_usage_info support to
15851         eliminate duplication.
15852         (riscv_expand_epilogue): Generate mret for interrupt handler.
15853         (riscv_epilogue_uses): New.
15854         (riscv_can_use_return_insn): Return false for interrupt handler.
15855         (riscv_function_ok_for_sibcall): Likewise.
15856         (riscv_set_current_function): Add interrupt handler support.
15857         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
15858         * config/riscv/riscv.md (UNSPECV_MRET): New.
15859         (GP_REGNUM): New.
15860         (riscv_frflags, riscv_fsflags): Use tab after opcode.
15861         (riscv_mret): New.
15862         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
15864 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
15866         PR tree-optimization/85712
15867         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
15868         this candidate has already been replaced in-situ by a copy.
15870 2018-05-25  Jason Merrill  <jason@redhat.com>
15872         PR c++/80485 - inline function non-zero address.
15873         * symtab.c (nonzero_address): Check DECL_COMDAT.
15875 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
15877         PR target/83628
15878         * config/alpha/alpha.md (ashlsi3): New insn pattern.
15879         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
15880         extension of SImode operation.  Use const123_operand predicate.
15881         (*saddsi_1): Remove.
15882         (*saddl_se_1): Ditto.
15883         (*ssubsi_1): Ditto.
15884         (*ssubl_se_1): Ditto.
15885         * config/alpha/predicates.md (const123_operand): New predicate.
15886         * config/alpha/constraints.md (P): Use IN_RANGE.
15888 2018-05-25  Richard Biener  <rguenther@suse.de>
15890         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
15891         defaulted to true.
15892         (ref_maybe_used_by_stmt_p): Likewise.
15893         (stmt_may_clobber_ref_p): Likewise.
15894         (stmt_may_clobber_ref_p_1): Likewise.
15895         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
15896         and pass it along.
15897         (ref_maybe_used_by_stmt_p): Likewise.
15898         (stmt_may_clobber_ref_p): Likewise.
15899         (stmt_may_clobber_ref_p_1): Likewise.
15900         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
15901         the alias oracle to disambiguate DRs with stmts DR analysis
15902         couldn't handle.
15903         (vect_analyze_data_refs): Do not give up on not analyzable
15904         DRs for BB vectorization.  Remove code truncating the dataref
15905         vector.
15907 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
15909         PR target/85832
15910         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15911         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
15912         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
15914 2018-05-25  Richard Biener  <rguenther@suse.de>
15916         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
15917         function, combining stmt data ref gathering and fatal analysis
15918         parts.
15919         (vect_analyze_data_refs): Remove now redudnant code and simplify.
15920         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
15921         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
15922         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
15923         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
15925 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15927         PR tree-optimization/85720
15928         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
15929         SCC if all partitions are builtins.
15930         (version_loop_by_alias_check): New parameter.  Generate cancelable
15931         runtime alias check if all partitions are builtins.
15932         (distribute_loop): Update call to above function.
15934 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15936         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
15937         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
15938         (parm_default_def_partition_arg): Ditto.
15939         (set_parm_default_def_partition): Ditto.
15940         (get_parm_default_def_partitions): Ditto and make it static.
15941         (get_undefined_value_partitions): Ditto and make it static.
15942         (remove_ssa_form): Refactor call to init_var_map here.
15943         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
15944         computation for loop region.
15945         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
15946         (register_default_def): Delete.
15947         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
15948         (parm_default_def_partition_arg): Ditto.
15949         (set_parm_default_def_partition): Ditto.
15950         (get_parm_default_def_partitions): Ditto and make it static.
15951         (get_undefined_value_partitions): Ditto and make it static.
15952         (coalesce_with_default, coalesce_with_default): Update comment.
15953         (create_coalesce_list_for_region): New func factored out from
15954         create_outofssa_var_map.
15955         (populate_coalesce_list_for_outofssa): New func factored out from
15956         create_outofssa_var_map and coalesce_ssa_name.
15957         (create_outofssa_var_map): Delete.
15958         (coalesce_ssa_name): Refactor to support live range computation.
15959         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
15960         (get_parm_default_def_partitions): Delete.
15961         (get_undefined_value_partitions): Ditto.
15962         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
15963         computation for loop region.
15964         (new_tree_live_info, loe_visit_block): Ditto.
15965         (live_worklist, set_var_live_on_entry): Ditto.
15966         (calculate_live_on_exit, verify_live_on_entry): Ditto.
15967         * tree-ssa-live.h (struct _var_map): New fields.
15968         (init_var_map): Change decl.
15969         (region_contains_p): New.
15971 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15973         * tree-ssa-live.h (live_merge_and_clear): Delete.
15975 2018-05-25  Richard Biener  <rguenther@suse.de>
15977         PR c++/85912
15978         * tree-dump.c (dequeue_and_dump): Remove access to removed
15979         operand 2 of a SWITCH_EXPR.
15981 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15983         * doc/sourcebuild.texi (vect_double_cond_arith): Include
15984         multiplication and division.
15985         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
15986         (cond_udiv@var{m}, cond_umod@var{m}): Document.
15987         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
15988         (cond_udiv_optab, cond_umod_optab): New optabs.
15989         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
15990         (IFN_COND_RDIV): New internal functions.
15991         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
15992         TRUNC_MOD_EXPR and RDIV_EXPR.
15993         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
15994         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
15995         New unspecs.
15996         (SVE_INT_BINARY): Include mult.
15997         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
15998         (optab, sve_int_op): Handle mult.
15999         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
16000         UNSPEC_COND_DIV.
16001         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
16002         for SVE_INT_BINARY_SD.
16004 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16006         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
16007         (optab, sve_int_op): Handle div and udiv.
16008         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
16009         for SVE_INT_BINARY_SD.
16010         (*<optab><mode>3): New insn for the same.
16012 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16014         * tree-vect-patterns.c: Include predict.h.
16015         (vect_recog_divmod_pattern): Restrict check for division support
16016         to when optimizing for size.
16018 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16020         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
16021         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
16022         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
16023         (gimple_match_op::set_op): Likewise.
16024         (gimple_resimplify4): Declare.
16025         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
16026         (expr::gen_transform): Likewise.
16027         (decision_tree::gen): Generate a simplification routine for 4 operands.
16028         * gimple-match-head.c (gimple_simplify): Add an overload for
16029         4 operands.  In the top-level function, handle up to 4 call
16030         arguments and call gimple_resimplify4.
16031         (gimple_resimplify4): New function.
16032         (build_call_internal): Pass a fourth operand.
16033         (maybe_push_to_seq): Likewise.
16034         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
16035         Fold VEC_COND_EXPRs of an operation and a default value into
16036         an IFN_COND_* function if possible.
16037         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
16038         New unspecs.
16039         (SVE_COND_FP_BINARY): Include them.
16040         (optab, sve_fp_op): Handle them.
16041         (SVE_INT_BINARY_REV): New code iterator.
16042         (SVE_COND_FP_BINARY_REV): New int iterator.
16043         (commutative): New int attribute.
16044         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
16045         Declare.
16046         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
16047         function.
16048         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
16049         (*cond_<optab><mode>): New patterns for reversed operands.
16051 2018-05-25  Richard Biener  <rguenther@suse.de>
16053         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
16054         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
16055         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
16056         (STMT_VINFO_GROUPED_ACCESS): Adjust.
16057         * tree-vect-data-refs.c (everywhere): Adjust users.
16058         * tree-vect-loop.c (everywhere): Likewise.
16059         * tree-vect-slp.c (everywhere): Likewise.
16060         * tree-vect-stmts.c (everywhere): Likewise.
16061         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
16063 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16065         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
16066         Rename to...
16067         (gcc_cv_as_section_exclude): ... this.
16068         Try Solaris as #exclude syntax.
16069         * configure: Regenerate.
16070         * config.in: Regenerate.
16071         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
16072         SECTION_EXCLUDE.
16073         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
16074         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
16076         * varasm.c (default_elf_asm_named_section): Don't check if
16077         HAVE_GAS_SECTION_EXCLUDE is defined.
16079 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16081         * doc/md.texi: Update the documentation of the cond_* optabs
16082         to mention the new final operand.  Fix GET_MODE_NUNITS call.
16083         Describe the scalar case too.
16084         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
16085         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
16086         instead of 2.
16087         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
16088         (get_conditional_internal_fn): Update comment.
16089         * tree-vect-loop.c (vectorizable_reduction): Pass the original
16090         accumulator value as a final argument to conditional functions.
16091         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
16092         a define_expand and add an "else" operand.  Assert for now that
16093         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
16094         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
16095         (*cond_<optab><mode>): New patterns.
16096         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
16097         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16098         (UNSPEC_COND_EOR): Delete.
16099         (optab): Remove associated mappings.
16100         (SVE_INT_BINARY): New code iterator.
16101         (sve_int_op): Remove int attribute and add "minus" to the code
16102         attribute.
16103         (SVE_COND_INT_OP): Delete.
16104         (SVE_COND_FP_OP): Rename to...
16105         (SVE_COND_FP_BINARY): ...this.
16107 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16109         * optabs.c (can_reuse_operands_p): New function.
16110         (maybe_legitimize_operands): Try to reuse the results for
16111         earlier operands.
16113 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16115         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
16116         Add {q} suffix to insn mnemonic.
16118 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16120         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
16121         (msp430_warn_func_return): New.
16123 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
16125         * fold-const.c (tree_nonzero_bits): New function.
16126         * fold-const.h (tree_nonzero_bits): Likewise.
16127         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
16128         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
16130 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
16132         PR target/85900
16133         PR target/85345
16134         * varasm.c (assemble_alias): Check ifunc_resolver only on
16135         FUNCTION_DECL.
16137 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16139         PR target/85903
16140         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
16141         when memory input operand is handled.
16143 2018-05-24  Luis Machado  <luis.machado@linaro.org>
16145         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
16146         global.
16147         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
16149 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16151         * match.pd: Delay FMA folds until after vectorization.
16153 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16155         PR target/83009
16156         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16157         address check not strict.
16159 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16161         * gimple-match.h (gimple_match_op): New class.
16162         (mprts_hook): Replace parameters with a gimple_match_op *.
16163         (maybe_build_generic_op): Likewise.
16164         (gimple_simplified_result_is_gimple_val): Replace parameters with
16165         a const gimple_match_op *.
16166         (gimple_simplify): Replace code_helper * and tree * parameters with
16167         a gimple_match_op * parameter.
16168         (gimple_resimplify1): Replace code_helper *, tree and tree *
16169         parameters with a gimple_match_op * parameter.
16170         (gimple_resimplify2): Likewise.
16171         (gimple_resimplify3): Likewise.
16172         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
16173         parameters with a gimple_match_op * parameter.
16174         * gimple-match-head.c (gimple_simplify): Change prototypes of
16175         auto-generated functions to take a gimple_match_op * instead of
16176         separate code_helper * and tree * parameters.  Make the same
16177         change in the top-level overload and update calls to the
16178         gimple_resimplify routines.  Update calls to the auto-generated
16179         functions and to maybe_push_res_to_seq in the publicly-facing
16180         operation-specific gimple_simplify overloads.
16181         (gimple_match_op::MAX_NUM_OPS): Define.
16182         (gimple_resimplify1): Replace rcode and ops with a single res_op
16183         parameter.  Update call to gimple_simplify.
16184         (gimple_resimplify2): Likewise.
16185         (gimple_resimplify3): Likewise.
16186         (mprts_hook): Replace parameters with a gimple_match_op *.
16187         (maybe_build_generic_op): Likewise.
16188         (build_call_internal): Replace type, nargs and ops with
16189         a gimple_match_op *.
16190         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
16191         with a single gimple_match_op *.  Update calls to mprts_hook,
16192         build_call_internal and gimple_simplified_result_is_gimple_val.
16193         Factor out code that is common to the tree_code and combined_fn cases.
16194         * genmatch.c (expr::gen_transform): Replace tem_code and
16195         tem_ops with a gimple_match_op called tem_op.  Update calls
16196         to the gimple_resimplify functions and maybe_push_res_to_seq.
16197         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
16198         res_ops.  Update call to the gimple_resimplify functions.
16199         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
16200         (decision_tree::gen): Make the functions take a gimple_match_op *
16201         called res_op instead of separate res_code and res_ops parameters.
16202         Update call accordingly.
16203         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
16204         and ops with a single res_op parameter.  Update calls to
16205         maybe_build_generic_op and maybe_push_res_to_seq.
16206         (fold_stmt_1): Update calls to gimple_simplify and
16207         replace_stmt_with_simplification.
16208         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
16209         and gimple_simplified_result_is_gimple_val.
16210         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
16211         gimple_simplify.
16212         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
16213         with a gimple_match_op *.
16214         (vn_nary_build_or_lookup): Likewise.  Update call to
16215         vn_nary_build_or_lookup_1.
16216         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
16217         gimple_match_op *.  Update calls to the gimple_resimplify routines
16218         and to gimple_simplified_result_is_gimple_val.
16219         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
16220         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
16221         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
16222         (visit_nary_op): Likewise.
16223         (visit_reference_op_load): Likewise.
16225 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16227         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
16228         modifier for printing the step amount.
16230 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
16232         PR target/78849
16233         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
16234         types.
16236 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
16238         * doc/sourcebuild.texi (Endianness): New subsubsection.
16240 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16242         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16243         <prefetch_dynamic_strides>: New const bool field.
16244         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16245         prefetch_dynamic_strides.
16246         (exynosm1_prefetch_tune): Likewise.
16247         (thunderxt88_prefetch_tune): Likewise.
16248         (thunderx_prefetch_tune): Likewise.
16249         (thunderx2t99_prefetch_tune): Likewise.
16250         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
16251         false.
16252         (aarch64_override_options_internal): Update to set
16253         PARAM_PREFETCH_DYNAMIC_STRIDES.
16254         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16255         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16256         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16257         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16258         prefetch-dynamic-strides setting.
16260 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16262         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16263         <minimum_stride>: New const int field.
16264         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16265         minimum_stride field defaulting to -1.
16266         (exynosm1_prefetch_tune): Likewise.
16267         (thunderxt88_prefetch_tune): Likewise.
16268         (thunderx_prefetch_tune): Likewise.
16269         (thunderx2t99_prefetch_tune): Likewise.
16270         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
16271         <default_opt_level>: Set to 3.
16272         (aarch64_override_options_internal): Update to set
16273         PARAM_PREFETCH_MINIMUM_STRIDE.
16274         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16275         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16276         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16277         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16278         stride is constant and is below the minimum stride threshold.
16280 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16282         * config/arm/arm-cpus.in (mode26): Delete.
16283         (armv4): Delete mode26 reference.
16284         * config/arm/arm.c (arm_configure_build_target): Delete use of
16285         isa_bit_mode26.
16287 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16289         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
16290         New insn pattern.
16291         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
16292         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
16293         for non-SSE modes.
16294         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
16295         (floatunsdidf2): Ditto.
16297 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16299         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
16300         (fixuns_trunc<mode>si2_avx512f): Ditto.
16301         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
16302         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
16303         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
16305 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16307         PR rtl-optimization/79985
16308         * df-scan.c (df_insn_refs_collect): Remove special case for
16309         global registers and asm statements.
16311 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16313         * extend.texi (Global Register Variables): Rewrite the bullet list.
16314         Note that the register is available for allocation. Note that access
16315         via inline asm must use constraints. Add note about async-signal
16316         handlers. Remove paragraph about automagic register selection.
16318 2018-05-23  Richard Biener  <rguenther@suse.de>
16320         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
16321         of fixed offset from memset VN.
16323 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
16325         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
16326         first_interp field.
16327         (alloc_cand_and_find_basis): Initialize first_interp field.
16328         (slsr_process_mul): Modify first_interp field.
16329         (slsr_process_add): Likewise.
16330         (slsr_process_cast): Modify first_interp field for each new
16331         interpretation.
16332         (slsr_process_copy): Likewise.
16333         (dump_candidate): Dump first_interp field.
16334         (replace_mult_candidate): Process all interpretations, not just
16335         subsequent ones.
16336         (replace_rhs_if_not_dup): Likewise.
16337         (replace_one_candidate): Likewise.
16339 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
16341         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
16342         Add new boolean.
16343         (aarch64_needs_frame_chain): New function.
16344         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
16346 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
16348         PR target/84882
16349         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
16350         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
16351         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
16352         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
16353         as true for strict-align.
16354         (aarch64_can_inline_p): Perform checks even when callee has no
16355         attributes to check for strict alignment.
16356         * doc/extend.texi (AArch64 Function Attributes): Document
16357         no-strict-align.
16358         * doc/invoke.texi: (AArch64 Options): Likewise.
16360 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
16362         PR tree-optimization/85853
16363         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
16364         the handling of the root of the node to...
16365         (vect_slp_analyze_node_operations_1): ...this new function,
16366         and run the whole thing with the child nodes' def types
16367         set according to their SLP node's def type.
16369 2018-05-23  Richard Biener  <rguenther@suse.de>
16371         PR middle-end/85874
16372         * tree-data-ref.c (create_runtime_alias_checks): Defer
16373         and ignore overflow warnings.
16375 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
16377         PR tree-optimization/85822
16378         * tree-vrp.c (is_masked_range_test): Fix handling of negative
16379         constants.
16381 2018-05-23  Richard Biener  <rguenther@suse.de>
16383         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16384         memset constants via native_interpret_expr.
16386 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
16388         PR target/85345
16389         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
16390         attribute.
16391         (cgraph_node::create_alias): Likewise.
16392         (cgraph_node::get_availability): Check ifunc_resolver instead
16393         of looking up ifunc attribute.
16394         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
16395         * varasm.c (do_assemble_alias): Likewise.
16396         (assemble_alias): Likewise.
16397         (default_binds_local_p_3): Likewise.
16398         * cgraph.h (cgraph_node): Add ifunc_resolver.
16399         (cgraph_node::only_called_directly_or_aliased_p): Return false
16400         for IFUNC resolver.
16401         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16402         attribute.
16403         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16404         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16405         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16406         instead of looking up ifunc attribute.
16408 2018-05-22  Luis Machado  <luis.machado@linaro.org>
16410         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16412 2018-05-22  Martin Sebor  <msebor@redhat.com>
16414         PR middle-end/85359
16415         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16416         only when expasion succeeds.
16417         (expand_builtin_strcmp): Same.
16418         (expand_builtin_strncmp): Same.
16420 2018-05-22  Martin Sebor  <msebor@redhat.com>
16422         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16424 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16425             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16427         * config/aarch64/aarch64-ldpstp.md: Replace uses of
16428         aarch64_mem_pair_operand with memory_operand and delete operand swapping
16429         code.
16430         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16431         Add check for legitimate_address.
16432         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16433         (aarch64_swap_ldrstr_operands): New.
16434         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16435         Define prototype.
16437 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16438             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16440         * config/aarch64/aarch64.md: New patterns to generate stp
16441         and ldp.
16442         (store_pair_sw, store_pair_dw): New patterns to generate stp for
16443         single words and double words.
16444         (load_pair_sw, load_pair_dw): Likewise.
16445         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16446         Delete.
16447         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16448         Delete.
16449         * config/aarch64/aarch64-ldpstp.md: Modify peephole
16450         for different mode ldpstp and add peephole for merged zero stores.
16451         Likewise for loads.
16452         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16453         Add size check.
16454         (aarch64_gen_store_pair): Rename calls to match new patterns.
16455         (aarch64_gen_load_pair): Rename calls to match new patterns.
16456         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16457         (load_pair<DREG:mode><DREG2:mode>): ... This.
16458         (store_pair<mode>): Rename to...
16459         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16460         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16461         New mode iterators.
16462         (V_INT_EQUIV): Handle SImode.
16463         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16464         New predicate.
16466 2018-05-22  Martin Sebor  <msebor@redhat.com>
16468         PR c/85623
16469         * calls.c (maybe_warn_nonstring_arg): Use string length to set
16470         or ajust the presumed bound on an operation to avoid unnecessary
16471         warnings.
16473 2018-05-22  Martin Sebor  <msebor@redhat.com>
16475         PR tree-optimization/85826
16476         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16477         assuming that a DECL necesarily has a constant size.
16479 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
16481         PR middle-end/85862
16482         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16484 2018-05-22  Richard Biener  <rguenther@suse.de>
16486         PR tree-optimization/85834
16487         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16488         non-constant and non-zero memset arguments.
16490 2018-05-22  Martin Liska  <mliska@suse.cz>
16492         PR ipa/85607
16493         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16495 2018-05-22  Richard Biener  <rguenther@suse.de>
16497         PR tree-optimization/85863
16498         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16499         comparisons when vectype is specified.
16500         (vectorizable_condition): Do not specify vectype for
16501         vect_is_simple_cond when SLP vectorizing.
16503 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
16505         PR target/85657
16506         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16507         define __ibm128 as long double.
16508         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16509         as a distinct type when IEEE 128-bit support is enabled.
16510         (init_float128_ieee): Fix up conversions between IFmode and IEEE
16511         128-bit types to use the correct functions.
16512         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16513         convert between 128-bit floating point types that have different
16514         modes but the same representation, instead of using gen_lowpart to
16515         makean alias.
16516         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16517         KFmode.
16518         (IFKF_reg): New attributes to give the register constraints for
16519         IFmode and KFmode.
16520         (extend<mode>tf2_internal): New insns to mark an explicit
16521         conversion between 128-bit floating point types that have a
16522         different mode but share the same representation.
16524 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
16526         PR tree-optimization/85814
16527         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16528         a null return from get_strinfo when unsharing the next
16529         strinfo in the chain.
16531 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
16533         PR gcc/84923
16534         * varasm.c (weak_finish): Clean up weak_decls.
16536 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16538         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16539         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16540         UNSPEC_UADALP values.
16541         * config/aarch64/iterators.md (ABAL): New int iterator.
16542         (ABDL2): Likewise.
16543         (ADALP): Likewise.
16544         (sur): Add mappings for the above.
16545         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16546         New define_insn.
16547         (aarch64_<sur>abal<mode>_4): Likewise.
16548         (aarch64_<sur>adalp<mode>_3): Likewise.
16549         (<sur>sadv16qi): New define_expand.
16551 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
16553         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16554         (*movdf_internal): Ditto.
16555         (*rcpsf2_sse): Ditto.
16556         (*rsqrtsf2_sse): Ditto.
16557         (*sqrt<mode>2_sse): Ditto.
16559 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
16561         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16562         eor3q<mode>4.
16563         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16564         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16565         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16566         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16567         vbcaxq_s64): New.
16568         * config/aarch64/arm_neon.h: Likewise.
16569         * config/aarch64/iterators.md (VQ_I): New.
16571 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
16573         * config.gcc: Add arc/t-multilib-linux to tmake_file for
16574         arc*-*-linux*.
16575         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16576         MULTILIB_DIRNAMES
16578 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16580         * config/nds32/constraints.md (S): New constraint.
16581         * config/nds32/nds32.md (call_internal): Use constraint S.
16582         (call_value_internal): Likewise.
16583         (sibcall_internal): Likewise.
16584         (sibcall_value_internal): Likewise.
16586 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16587             Chung-Ju Wu  <jasonwucj@gmail.com>
16589         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16590         into consideration.
16592 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16593             Chung-Ju Wu  <jasonwucj@gmail.com>
16595         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16596         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16597         (nds32_rtx_costs_impl): Simplify.
16598         (nds32_address_cost_impl): Simplify.
16599         (nds32_init_rtx_costs): New function.
16600         (nds32_rtx_costs_speed_prefer): Likewise.
16601         (nds32_rtx_costs_size_prefer): Likewise.
16602         (nds32_address_cost_speed_prefer): Likewise.
16603         (nds32_address_cost_speed_fwprop): Likewise.
16604         (nds32_address_cost_size_prefer): Likewise.
16605         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16606         * config/nds32/nds32.c (nds32_option_override): Use
16607         nds32_init_rtx_costs function.
16609 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16611         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16612         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16613         (TARGET_PIPELINE_N8): Likewise.
16614         (TARGET_PIPELINE_N10): Likewise.
16615         (TARGET_PIPELINE_N13): Likewise.
16616         (TARGET_PIPELINE_GRAYWOLF): Likewise.
16618 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
16620         * config/nds32/nds32-fpu.md: Update copyright year.
16622 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16624         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16626 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16628         * config/nds32/nds32.c
16629         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16630         * config/nds32/nds32.opt (minline-asm-r15): New option.
16632 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16634         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16635         MASK_HW_ABS.
16636         * config/nds32/nds32.md (abssi2): New pattern.
16638 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
16640         * config/i386/i386.md (rex64namesuffix): New mode attribute.
16641         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16642         Merge insn pattern from sse_cvtsi2ss<round_name> and
16643         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16644         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16645         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16646         using SWI48 mode iterator.
16647         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16648         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16649         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16650         pattern from sse_cvttss2si<round_saeonly_name>
16651         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16652         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16653         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16654         using SWI48 mode iterator.
16655         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16656         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16657         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16658         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16659         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16660         using SWI48 mode iterator.
16661         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16662         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16663         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16664         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16665         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16666         SWI48 mode iterator.
16667         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16668         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16669         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16670         pattern from sse_cvttsd2si<round_saeonly_name>
16671         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16673 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16675         * config/nds32/nds32-md-auxiliary.c
16676         (nds32_valid_smw_lwm_base_p): Refine.
16677         (nds32_output_smw_single_word): Refine.
16678         (nds32_output_smw_double_word): New.
16679         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16681 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16683         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16684         (nds32_output_stack_pop): Refine.
16685         (nds32_expand_unaligned_load): Refine.
16686         (nds32_expand_unaligned_store): Refine.
16688 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16689             Chung-Ju Wu  <jasonwucj@gmail.com>
16691         * config/nds32/constants.md: Add TP_REGNUM constant.
16692         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16693         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16694         UNSPEC_ADD32.
16695         * config/nds32/nds32-doubleword.md: Consider flag_pic.
16696         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16697         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16698         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16699         and PIC code generation.
16700         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16701         code generation.
16702         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16703         optimization.
16704         * config/nds32/nds32.md: Support TLS and PIC.
16705         * config/nds32/nds32.c: Support TLS and PIC.
16706         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16707         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16708         predicate.
16710 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16712         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16713         mode with E_ prefix.
16715 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16716             Chung-Ju Wu  <jasonwucj@gmail.com>
16718         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16719         * config/nds32/nds32-md-auxiliary.c
16720         (symbolic_reference_mentioned_p): New.
16721         (nds32_legitimize_ict_address): New.
16722         (nds32_expand_ict_move): New.
16723         (nds32_indirect_call_referenced_p): New.
16724         (nds32_symbol_binds_local_p): Delete.
16725         (nds32_long_call_p): Modify.
16726         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16727         * config/nds32/nds32-protos.h
16728         (symbolic_reference_mentioned_p): Declare.
16729         (nds32_legitimize_ict_address): Declare.
16730         (nds32_expand_ict_move): Declare.
16731         (nds32_indirect_call_referenced_p): Declare.
16732         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16733         (nds32_relax_group): Use nds32_ict_const_p as condition.
16734         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16735         (nds32_asm_file_start): Output ict_model directive in asm code.
16736         (nds32_legitimate_address_p): Consider indirect call.
16737         (nds32_print_operand): Consider indirect call.
16738         (nds32_print_operand_address): Consider indirect call.
16739         (nds32_insert_attributes): Handle "indirect_call" attribute.
16740         (TARGET_LEGITIMATE_ADDRESS_P): Define.
16741         (TARGET_LEGITIMATE_CONSTANT_P): Define.
16742         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16743         (TARGET_DELEGITIMIZE_ADDRESS): Define.
16744         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16745         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16746         (TARGET_ICT_MODEL_SMALL): Define.
16747         (TARGET_ICT_MODEL_LARGE): Define.
16748         * config/nds32/nds32.md (movsi): Consider ict model.
16749         (call, call_value): Consider ict model.
16750         (sibcall, sibcall_value): Consider ict model.
16751         * config/nds32/nds32.opt (mict-model): New option.
16752         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16753         model.
16755 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
16756             Monk Chiang  <sh.chiang04@gmail.com>
16757             Jim Wilson <jimw@sifive.com>
16759         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16760         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
16761         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16762         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16763         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
16764         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16765         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16766         compute save_libcall_adjustment properly.
16767         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16768         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16769         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16770         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16771         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16772         (ABI_SPEC): Handle mabi=ilp32e.
16773         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16774         (RVE): Add RVE mask.
16775         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16776         <-march>: Add rv32e as an example.
16778 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
16780         PR c++/82899
16781         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
16782         (intra_create_variable_infos): Handle C++ constructors.
16784 2018-05-18  Martin Liska  <mliska@suse.cz>
16786         * passes.def: Remove a redundant pass.
16788 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
16790         PR bootstrap/85838
16791         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
16793 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16795         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
16796         (ARMv4): Update.
16797         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16798         (ARMv6m): Update.
16799         (armv2, armv2a, armv3, armv3m): Delete architectures.
16800         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
16801         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
16802         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
16803         Delete cpus.
16804         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
16805         (*mulsidi3adddi): Likewise.
16806         (mulsidi3): Likewise.
16807         (*mulsidi3_nov6): Likewise.
16808         (umulsidi3): Likewise.
16809         (umulsidi3_nov6): Likewise.
16810         (umaddsidi4): Likewise.
16811         (*umulsidi3adddi): Likewise.
16812         (smulsi3_highpart): Likewise.
16813         (*smulsi3_highpart_nov6): Likewise.
16814         (umulsi3_highpart): Likewise.
16815         (*umulsi3_highpart_nov6): Likewise.
16816         * config/arm/arm.h (arm_arch3m): Delete.
16817         * config/arm/arm.c (arm_arch3m): Delete.
16818         (arm_option_override_internal): Update armv3-related comment.
16819         (arm_configure_build_target): Delete use of isa_bit_mode32.
16820         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
16821         (arm_rtx_costs_internal): Delete check of arm_arch3m.
16822         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
16823         (mulsa3): Likewise.
16824         (mulusa3): Likewise.
16825         * config/arm/arm-protos.h (arm_arch3m): Delete.
16826         * config/arm/arm-tables.opt: Regenerate.
16827         * config/arm/arm-tune.md: Likewise.
16828         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
16829         deleted architectures.
16831 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16833         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
16834         (armv5t, armv5te): New features.
16835         (ARMv5, ARMv5e): Delete fgroups.
16836         (ARMv5t, ARMv5te): Adjust for above changes.
16837         (ARMv6m): Likewise.
16838         (armv5, armv5e): Delete arches.
16839         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
16840         arm_arch5.
16841         (*call_reg_arm): Likewise.
16842         (*call_value_reg_armv5): Likewise.
16843         (*call_value_reg_arm): Likewise.
16844         (*call_symbol): Likewise.
16845         (*call_value_symbol): Likewise.
16846         (*sibcall_insn): Likewise.
16847         (*sibcall_value_insn): Likewise.
16848         (clzsi2): Likewise.
16849         (prefetch): Likewise.
16850         (define_split and define_peephole2 dependent on arm_arch5):
16851         Likewise.
16852         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
16853         arm_arch5e.
16854         (TARGET_ARM_QBIT): Likewise.
16855         (TARGET_DSP_MULTIPLY): Likewise.
16856         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
16857         (arm_arch5, arm_arch5e): Delete.
16858         (arm_arch5t, arm_arch5te): Declare.
16859         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
16860         (arm_arch5t): Declare.
16861         (arm_option_reconfigure_globals): Update for the above.
16862         (arm_options_perform_arch_sanity_checks): Update comment, replace
16863         use of arm_arch5 with arm_arch5t.
16864         (use_return_insn): Likewise.
16865         (arm_emit_call_insn): Likewise.
16866         (output_return_instruction): Likewise.
16867         (arm_final_prescan_insn): Likewise.
16868         (arm_coproc_builtin_available): Likewise.
16869         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
16870         arm_arch5e with arm_arch5t and arm_arch5te.
16871         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
16872         (arm_arch5t, arm_arch5te): Declare.
16873         * config/arm/arm-tables.opt: Regenerate.
16874         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
16875         * config/arm/t-multilib: Likewise.
16876         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
16877         instead of arm_arch5.
16878         (*call_reg_thumb1): Likewise.
16879         (*call_value_reg_thumb1_v5): Likewise.
16880         (*call_value_reg_thumb1): Likewise.
16881         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
16882         unreachable path.
16883         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
16885 2018-05-18  Martin Liska  <mliska@suse.cz>
16887         PR gcov-profile/84846
16888         * doc/gcov.texi: Document -t option of gcov tool.
16890 2018-05-18  Martin Liska  <mliska@suse.cz>
16892         PR gcov-profile/84846
16893         * gcov.c (print_usage): Add new -t option.
16894         (process_args): Handle the option.
16895         (generate_results): Use stdout as output when requested by
16896         the option.
16898 2018-05-18  Martin Liska  <mliska@suse.cz>
16900         PR gcov-profile/84846
16901         * coverage.c (coverage_init): Write PWD to .gcno file.
16902         * doc/gcov.texi: Document how working directory is printed.
16903         * gcov-dump.c (dump_gcov_file): Print PWD.
16904         * gcov.c (output_intermediate_file): Likewise.
16905         (read_graph_file): Read PWD string.
16906         (output_lines): Print PWD.
16908 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16910         PR middle-end/85817
16911         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
16912         for retval and return false if all args to phi are zero.
16914 2018-05-18  Richard Biener  <rguenther@suse.de>
16916         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
16917         method.
16918         (evrp_dom_walker::before_dom_children): Call it.
16920 2018-05-18  Richard Biener  <rguenther@suse.de>
16922         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
16923         results when processing array refs with variable index.
16925 2018-05-18  Toon Moene  <toon@moene.org>
16927         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
16928         directly after that of -floop-interchange. Indicate that both
16929         options are enabled by default when specifying -O3.
16931 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16933         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
16934         iterator.  Delete separate integer-mode vec_set<mode> expander.
16935         (aarch64_simd_vec_setv2di): Delete.
16936         (vec_setv2di): Delete.
16937         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
16938         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
16939         the "w, r" alternative.
16941 2018-05-18  Martin Liska  <mliska@suse.cz>
16943         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
16944         * tree-pass.h (make_pass_lower_switch_O0): New function.
16945         * tree-switch-conversion.c (node_has_low_bound): Remove.
16946         (node_has_high_bound): Likewise.
16947         (node_is_bounded): Likewise.
16948         (class pass_lower_switch): Make it a template type and create
16949         two instances.
16950         (pass_lower_switch::execute): Add template argument.
16951         (make_pass_lower_switch): New function.
16952         (make_pass_lower_switch_O0): New function.
16953         (do_jump_if_equal): Remove.
16954         (emit_case_nodes): Simplify to just handle all 3 cases and leave
16955         all the hard work to tree optimization passes.
16957 2018-05-18  Martin Liska  <mliska@suse.cz>
16959         * dbgcnt.c (limit_low): Renamed from limit.
16960         (limit_high): New variable.
16961         (dbg_cnt_is_enabled): Check for upper limit.
16962         (dbg_cnt): Adjust dumping.
16963         (dbg_cnt_set_limit_by_index): Add new argument for high
16964         value.
16965         (dbg_cnt_set_limit_by_name): Likewise.
16966         (dbg_cnt_process_single_pair): Parse new format.
16967         (dbg_cnt_process_opt): Use strtok.
16968         (dbg_cnt_list_all_counters): Remove 'value' and add
16969         'limit_high'.
16970         * doc/invoke.texi: Document changes.
16972 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
16974         * doc/sourcebuild.texi (scalar_all_fma): Document.
16975         * tree.def (FMA_EXPR): Delete.
16976         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
16977         * internal-fn.c (ternary_direct): New macro.
16978         (expand_ternary_optab_fn): Likewise.
16979         (direct_ternary_optab_supported_p): Likewise.
16980         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
16981         * builtins.c (fold_builtin_fma): Delete.
16982         (fold_builtin_3): Don't call it.
16983         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
16984         * expr.c (expand_expr_real_2): Likewise.
16985         * fold-const.c (operand_equal_p): Likewise.
16986         (fold_ternary_loc): Likewise.
16987         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
16988         * gimple.c (DEFTREECODE): Likewise.
16989         * gimplify.c (gimplify_expr): Likewise.
16990         * optabs-tree.c (optab_for_tree_code): Likewise.
16991         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
16992         * tree-eh.c (operation_could_trap_p): Likewise.
16993         (stmt_could_throw_1_p): Likewise.
16994         * tree-inline.c (estimate_operator_cost): Likewise.
16995         * tree-pretty-print.c (dump_generic_node): Likewise.
16996         (op_code_prio): Likewise.
16997         * tree-ssa-loop-im.c (stmt_cost): Likewise.
16998         * tree-ssa-operands.c (get_expr_operands): Likewise.
16999         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
17000         * fold-const-call.h (fold_fma): Delete.
17001         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
17002         CFN_FNMA and CFN_FNMS.
17003         (fold_fma): Delete.
17004         * genmatch.c (combined_fn): New enum.
17005         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
17006         (commutative_op): New function.
17007         (commutate): Use it.  Handle more than 2 operands.
17008         (dt_operand::gen_gimple_expr): Use commutative_op.
17009         (parser::parse_expr): Allow :c to be used with non-binary
17010         operators if the commutative operand is known.
17011         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
17012         CFN_FMS, CFN_FNMA and CFN_FNMS.
17013         (backprop::process_assign_use): Remove FMA_EXPR handling.
17014         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
17015         (gen_hsa_fma): New function.
17016         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
17017         IFN_FNMA and IFN_FNMS.
17018         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
17019         * gimple-fold.h (follow_all_ssa_edges): Declare.
17020         * gimple-fold.c (follow_all_ssa_edges): New function.
17021         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
17022         gimple_build interface and use follow_all_ssa_edges to fold the result.
17023         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
17024         instead of checking for optabs directly.
17025         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
17026         rather than FMA_EXPRs.
17027         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
17028         call to IFN_FMA instead of an FMA_EXPR.
17030 2018-05-17  Jim Wilson  <jimw@sifive.com>
17032         * expr.c (do_tablejump): When converting index to Pmode, if we have a
17033         sign extended promoted subreg, and the range does not have the sign bit
17034         set, then do a sign extend.
17036         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
17037         test, check for sign extended subreg and/or constant operands, and
17038         do a sign extend in that case.
17040 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
17042         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
17043         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
17044         Add untyped.
17045         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
17046         Change logics_shift_reg to logics_shift_imm.
17047         (thunderx2t99_fp_loadpair_basic): Delete.
17048         (thunderx2t99_fp_storepair_basic): Delete.
17049         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
17050         (thunderx2t99_asimd_polynomial): Delete.
17051         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
17052         and neon_fp_mul_d_scalar_q.
17053         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
17054         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
17055         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
17056         (thunderx2t99_asimd_lut): Add missing tbl types.
17057         (thunderx2t99_asimd_ext): Delete.
17058         (thunderx2t99_asimd_load1_1_mult): Delete.
17059         (thunderx2t99_asimd_load1_2_mult): Delete.
17060         (thunderx2t99_asimd_load1_ldp): New.
17061         (thunderx2t99_asimd_load1): New.
17062         (thunderx2t99_asimd_load2): Add missing *load2* types.
17063         (thunderx2t99_asimd_load3): New.
17064         (thunderx2t99_asimd_load4): New.
17065         (thunderx2t99_asimd_store1_1_mult): Delete.
17066         (thunderx2t99_asimd_store1_2_mult): Delete.
17067         (thunderx2t99_asimd_store2_mult): Delete.
17068         (thunderx2t99_asimd_store2_onelane): Delete.
17069         (thunderx2t99_asimd_store_stp): New.
17070         (thunderx2t99_asimd_store1): New.
17071         (thunderx2t99_asimd_store2): New.
17072         (thunderx2t99_asimd_store3): New.
17073         (thunderx2t99_asimd_store4): New.
17075 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
17077         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
17078         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
17080 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
17081             Segher Boessenkool  <segher@kernel.crashing.org>
17083         PR target/85698
17084         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
17085         operand.
17087 2018-05-17  Richard Biener  <rguenther@suse.de>
17089         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
17090         for pruning loop and prune defs feeding only already visited PHIs.
17092 2018-05-17  Richard Biener  <rguenther@suse.de>
17094         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
17096 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
17097             Richard Biener  <rguenther@suse.de>
17099         PR tree-optimization/85793
17100         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
17101         for VMAT_ELEMENTWISE.
17103 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17105         * internal-fn.h (lookup_internal_fn): Declare
17106         * internal-fn.c (lookup_internal_fn): New function.
17107         * gimple.c (gimple_build_call_from_tree): Handle calls to
17108         internal functions.
17109         * gimple-pretty-print.c (dump_gimple_call): Print "." before
17110         internal function names.
17111         * tree-pretty-print.c (dump_generic_node): Likewise.
17112         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
17114 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17116         * gimple-fold.h (gimple_build): Make the function forms take
17117         combined_fn rather than built_in_function.
17118         (gimple_simplify): Likewise.
17119         * gimple-match-head.c (gimple_simplify): Likewise.
17120         * gimple-fold.c (gimple_build): Likewise.
17121         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
17122         rather than gimple_build_call_internal.
17123         (get_initial_defs_for_reduction): Likewise.
17124         (vect_create_epilog_for_reduction): Likewise.
17125         (vectorizable_live_operation): Likewise.
17127 2018-05-17  Martin Liska  <mliska@suse.cz>
17129         * gimple-ssa-sprintf.c (format_directive): Do not use
17130         space in between 'G_' and '('.
17132 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
17134         PR target/85323
17135         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
17136         even if the mask is not all ones.
17138         PR target/85323
17139         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
17140         vector.
17141         (ix86_gimple_fold_builtin): Likewise.
17143         PR target/85323
17144         * config/i386/i386.c: Include tree-vector-builder.h.
17145         (ix86_vector_shift_count): New function.
17146         (ix86_fold_builtin): Fold shift builtins by scalar count.
17147         (ix86_gimple_fold_builtin): Likewise.
17149         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
17150         _mm512_setzero): New intrinsics.
17152 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
17153             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17155         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
17156         code generation for cases where splatting a value is not useful.
17157         * simplify-rtx.c (simplify_ternary_operation): Simplify
17158         vec_merge across a vec_duplicate and a paradoxical subreg forming
17159         a vector mode to a vec_concat.
17161 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
17163         * config.gcc: Support "goldmont-plus".
17164         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17165         "goldmont-plus".
17166         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17167         PROCESSOR_GOLDMONT_PLUS.
17168         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
17169         (processor_target_table): Add "goldmont-plus".
17170         (PTA_GOLDMONT_PLUS): Define.
17171         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
17172         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
17173         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
17174         (fold_builtin_cpu): Add "goldmont-plus".
17175         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
17176         (ix86_option_override_internal): Add "goldmont-plus".
17177         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
17178         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
17179         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
17180         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
17182 2018-05-17  Richard Biener  <rguenther@suse.de>
17184         PR tree-optimization/85757
17185         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
17186         remove defs that only feed that PHI from further processing.
17188 2018-05-16  Jim Wilson  <jimw@sifive.com>
17190         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
17191         asterisk to name.
17192         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
17194 2018-05-16  Mark Wielaard  <mark@klomp.org>
17196         * dwarf2out.c (count_index_strings): New function.
17197         (output_indirect_strings): Call count_index_strings and generate
17198         header for dwarf_version >= 5.
17200 2018-05-16  Mark Wielaard  <mark@klomp.org>
17202         * dwarf2out.c (dwarf_FORM): New function.
17203         (set_indirect_string): Use dwarf_FORM.
17204         (reset_indirect_string): Likewise.
17205         (size_of_die): Likewise.
17206         (value_format): Likewise.
17207         (output_die): Likewise.
17208         (add_skeleton_AT_string): Likewise.
17209         (output_macinfo_op): Likewise.
17210         (index_string): Likewise.
17211         (output_index_string_offset): Likewise.
17212         (output_index_string): Likewise.
17213         (count_index_strings): Likewise.
17215 2018-05-16  Carl Love  <cel@us.ibm.com>
17217         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17218         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
17220 2018-05-16  Martin Jambor  <mjambor@suse.cz>
17222         * ipa-prop.c (ipa_free_all_edge_args): Remove.
17223         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
17225 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
17227         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
17228         (fnma<mode>4): Likewise.
17229         (fms<mode>4): Likewise.
17230         (fnms<mode>4): Likewise.
17231         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
17232         (aarch64_fnma<mode>4): Likewise.
17233         (aarch64_fms<mode>4): Likewise.
17234         (aarch64_fnms<mode>4): Likewise.
17235         (aarch64_fnmadd<mode>4): Likewise.
17237 2018-05-16  Jason Merrill  <jason@redhat.com>
17239         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
17241 2018-05-16  Richard Biener  <rguenther@suse.de>
17243         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
17244         (dump_stmt_cost): Declare.
17245         (add_stmt_cost): Dump cost we add.
17246         (add_stmt_costs): New function.
17247         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
17248         No longer exported.
17249         (vect_analyze_stmt): Adjust prototype.
17250         (vectorizable_condition): Likewise.
17251         (vectorizable_live_operation): Likewise.
17252         (vectorizable_reduction): Likewise.
17253         (vectorizable_induction): Likewise.
17254         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
17255         cost vector to pass to vectorizable_ and record afterwards.
17256         (vect_model_reduction_cost): Take cost vector argument and adjust.
17257         (vect_model_induction_cost): Likewise.
17258         (vectorizable_reduction): Likewise.
17259         (vectorizable_induction): Likewise.
17260         (vectorizable_live_operation): Likewise.
17261         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
17262         SLP_TREE_NUMBER_OF_VEC_STMTS.
17263         (vect_analyze_slp_cost_1): Remove.
17264         (vect_analyze_slp_cost): Likewise.
17265         (vect_slp_analyze_node_operations): Take visited args and
17266         a target cost vector.  Avoid processing already visited stmt sets.
17267         (vect_slp_analyze_operations): Use a local cost vector to gather
17268         costs and register those of non-discarded instances.
17269         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
17270         (vect_schedule_slp_instance): Remove copying of
17271         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
17272         zero.
17273         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
17274         adding cost.  Record cost entry location.
17275         (vect_prologue_cost_for_slp_op): Function to compute cost of
17276         a constant or invariant generated for SLP vect in the prologue,
17277         split out from vect_analyze_slp_cost_1.
17278         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
17279         (vect_model_promotion_demotion_cost): Likewise.
17280         (vect_model_store_cost): Likewise, make static.
17281         (vect_model_load_cost): Likewise.
17282         (vectorizable_bswap): Add cost vector arg and adjust.
17283         (vectorizable_call): Likewise.
17284         (vectorizable_simd_clone_call): Likewise.
17285         (vectorizable_conversion): Likewise.
17286         (vectorizable_assignment): Likewise.
17287         (vectorizable_shift): Likewise.
17288         (vectorizable_operation): Likewise.
17289         (vectorizable_store): Likewise.
17290         (vectorizable_load): Likewise.
17291         (vectorizable_condition): Likewise.
17292         (vectorizable_comparison): Likewise.
17293         (can_vectorize_live_stmts): Likewise.
17294         (vect_analyze_stmt): Likewise.
17295         (vect_transform_stmt): Adjust calls to vectorizable_*.
17296         * tree-vectorizer.c: Include gimple-pretty-print.h.
17297         (dump_stmt_cost): New function.
17299 2018-05-16  Richard Biener  <rguenther@suse.de>
17301         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
17302         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
17303         * tree-ssa-dse.c: Include tree-ssa-loop.h.
17304         (check_name): New callback.
17305         (dse_classify_store): Track cycles via a visited bitmap of PHI
17306         defs and simplify handling of in-loop and across loop dead stores
17307         and properly fail for loop-variant refs.  Handle byte-tracking with
17308         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
17309         limiting the walk.
17311 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
17313         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
17314         (vect_get_mask_type_for_stmt): Likewise.
17315         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
17316         split out from...
17317         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
17318         to determine the statement's vector type and the vector type that
17319         should be used for calculating nunits.  Deal with cases in which
17320         the type has to be deferred.
17321         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
17322         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
17323         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
17324         (vect_determine_vf_for_stmt): New functions, split out from...
17325         (vect_determine_vectorization_factor): ...here.
17326         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
17327         (vect_get_mask_type_for_stmt): New functions, split out from
17328         vect_determine_vectorization_factor.
17330 2018-05-16  Richard Biener  <rguenther@suse.de>
17332         * tree-cfg.c (verify_gimple_assign_ternary): Properly
17333         verify the [VEC_]COND_EXPR embedded comparison.
17335 2018-05-15  Martin Sebor  <msebor@redhat.com>
17337         PR tree-optimization/85753
17338         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
17339         RECORD_TYPE in addition to ARRAY_TYPE.
17341 2018-05-15  Martin Sebor  <msebor@redhat.com>
17343         PR middle-end/85643
17344         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17346 2018-05-15  Richard Biener  <rguenther@suse.de>
17348         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
17349         add by_clobber_p one.  Change algorithm to collect all defs
17350         representing uses we need to walk and try reducing them to
17351         a single one before failing.
17352         (dse_dom_walker::dse_optimize_stmt): Adjust.
17354 2018-05-13  Mark Wielaard  <mark@klomp.org>
17356         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
17357         (size_of_loc_descr): Likewise.
17358         (output_loc_operands): Likewise.
17359         (output_loc_operands_raw): Likewise.
17360         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
17361         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17362         (hash_loc_operands): Likewise.
17363         (compare_loc_operands): Likewise.
17365 2018-05-14  Mark Wielaard  <mark@klomp.org>
17367         * dwarf2out.c (count_index_addrs): New function.
17368         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
17370 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17372         PR tree-optimization/83648
17373         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
17374         return value as malloc candidate.
17376 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17378         PR ipa/85734
17379         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
17380         param as true in call to suggest_attribute.
17382 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
17384         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
17385         -mreadonly-in-sdata.
17387 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17389         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
17390         New pattern.
17391         (aarch64_crypto_aesd_fused): Likewise.
17393 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
17395         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17396         (movsi_aarch64): Likewise.
17397         (load_pairsi): Likewise.
17398         (load_pairdi): Likewise.
17399         (store_pairsi): Likewise.
17400         (store_pairdi): Likewise.
17401         (load_pairsf): Likewise.
17402         (load_pairdf): Likewise.
17403         (store_pairsf): Likewise.
17404         (store_pairdf): Likewise.
17405         (zero_extend): Likewise.
17406         (trunc): Swap alternatives.
17407         (fcvt_target): Add '?' to prefer w over r.
17409 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
17411         PR target/85756
17412         * config/i386/i386.md: Disallow non-commutative arithmetics in
17413         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17414         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
17415         in the peephole2 before it.
17417 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
17419         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17420         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17421         (ix86_handle_option): Handle -mcldemote.
17422         * config.gcc: New header.
17423         * config/i386/cldemoteintrin.h: New file.
17424         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17425         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17426         -mcldemote.
17427         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17428         OPTION_MASK_ISA_CLDEMOTE.
17429         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17430         (ix86_valid_target_attribute_inner_p): Ditto.
17431         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17432         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17433         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17434         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17435         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17436         (cldemote): New.
17437         * config/i386/i386.opt: Add -mcldemote.
17438         * config/i386/x86intrin.h: New header.
17439         * doc/invoke.texi: Add -mcldemote.
17441 2018-05-14  Richard Biener  <rguenther@suse.de>
17443         * doc/match-and-simplify.texi: Adjust :s documentation.
17445 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
17447         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17448         intended memcpy size.
17449         (REORDER_45): Likewise.
17451 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
17453         * sort.cc: New file.
17454         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17455         * vec.c (qsort_chk): Use gcc_qsort.
17456         * Makefile.in (OBJS-libcommon): Add sort.o.
17457         (build/sort.o): New target.  Use it...
17458         (BUILD_RTL): ... here, and...
17459         (build/gencfn-macros): ... here, and...
17460         (build/genmatch): ... here.
17462 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17463             Chung-Ju Wu  <jasonwucj@gmail.com>
17465         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17466         * config/nds32/nds32-graywolf.md: New file.
17467         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17468         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17469         pipeline.
17470         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17471         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17472         * config/nds32/nds32.md (pipeline_model): Add graywolf.
17473         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17474         * config/nds32/pipelines.md: Include n15 settings.
17476 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17477             Chung-Ju Wu  <jasonwucj@gmail.com>
17479         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17480         * config/nds32/nds32-n13.md: New file.
17481         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17482         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17483         pipeline.
17484         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17485         * config/nds32/nds32.md (pipeline_model): Add n13.
17486         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17487         * config/nds32/pipelines.md: Include n13 settings.
17489 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17490             Chung-Ju Wu  <jasonwucj@gmail.com>
17492         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17493         * config/nds32/nds32-n10.md: New file.
17494         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17495         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17496         pipeline.
17497         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17498         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17499         * config/nds32/nds32.md (pipeline_model): Add n10.
17500         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17501         * config/nds32/pipelines.md: Include n10 settings.
17503 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
17504             Kito Cheng  <kito.cheng@gmail.com>
17505             Chung-Ju Wu  <jasonwucj@gmail.com>
17507         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17508         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17509         Add enum values for DSP extension instructions.
17510         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17511         New constraints.
17512         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17513         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17514         New code iterators.
17515         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17516         * config/nds32/nds32-dspext.md: New file for DSP implementation.
17517         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17518         * config/nds32/nds32-intrinsic.md: Likewise.
17519         * config/nds32/nds32_intrinsic.h: Likewise.
17520         * config/nds32/nds32-md-auxiliary.c: Likewise.
17521         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17522         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17523         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17524         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17525         * config/nds32/nds32-protos.h: New declarations for DSP extension.
17526         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17527         TYPE_DMAC in switch statement.
17528         * config/nds32/nds32.c: New checking and implementation for DSP
17529         extension instructions.
17530         * config/nds32/nds32.h: Likewise.
17531         * config/nds32/nds32.md: Likewise.
17532         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17533         * config/nds32/predicates.md: Implement new predicates for DSP
17534         extension.
17536 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
17538         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17539         Reformat alternatives and attributes so it is easier to identify
17540         which constraints/attributes go with which instruction.
17541         (mov<mode>_hardfloat32, FMOVE64): Likewise.
17542         (mov<mode>_softfloat32, FMOVE64): Likewise.
17543         (mov<mode>_hardfloat64, FMOVE64): Likewise.
17544         (mov<mode>_softfloat64, FMOVE64): Likewise.
17546 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17548         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17549         subsection.
17550         (Basic PowerPC Built-in Functions): The new name of the
17551         subsection previously known as "PowerPC Built-in Functions".
17552         (Basic PowerPC Built-in Functions Available on all Configurations):
17553         New subsubsection.
17554         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17555         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17556         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17557         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17559 2018-05-11  Martin Jambor  <mjambor@suse.cz>
17561         PR ipa/85655
17562         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17563         single const.
17565 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17567         PR target/85733
17568         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17570 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
17572         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17573         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17574         (ix86_handle_option): Handle -mwaitpkg.
17575         * config.gcc: New header.
17576         * config/i386/cpuid.h (bit_WAITPKG): New bit.
17577         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17578         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17579         function type.
17580         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17581         OPTION_MASK_ISA_WAITPKG.
17582         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17583         (ix86_option_override_internal): Add PTA_WAITPKG.
17584         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17585         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17586         IX86_BUILTIN_TPAUSE.
17587         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17588         __builtin_ia32_umwait and __builtin_ia32_tpause.
17589         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17590         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17591         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17592         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17593         UNSPECV_TPAUSE): New.
17594         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17595         * config/i386/i386.opt: Add -mwaitpkg.
17596         * config/i386/waitpkgintrin.h: New file.
17597         * config/i386/x86intrin.h: New header.
17598         * doc/invoke.texi: Add -mwaitpkg.
17600 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17602         PR target/85606
17603         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17604         equivalent.
17605         (cortex-m0): Use armv6s-m isa.
17606         (cortex-m0plus): Likewise.
17607         (cortex-m1): Likewise.
17608         (cortex-m0.small-multiply): Likewise.
17609         (cortex-m0plus.small-multiply): Likewise.
17610         (cortex-m1.small-multiply): Likewise.
17612 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
17613             Jakub Jelinek  <jakub@redhat.com>
17615         PR tree-optimization/85692
17616         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17617         source permute as well.
17619 2018-05-11  Martin Liska  <mliska@suse.cz>
17621         PR sanitizer/85556
17622         * doc/extend.texi: Document LLVM style format for no_sanitize
17623         attribute.
17625 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17627         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17628         mode_supports_vsx_dform_quad to mode_supports_dq_form.
17629         (mode_supports_vsx_dform_quad): Likewise.
17630         (mode_supports_vmx_dform): Move these functions to be next to the
17631         other mode_supports functions.
17632         (mode_supports_dq_form): Likewise.
17633         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17634         mode_supports_dq_form.
17635         (reg_offset_addressing_ok_p): Likewise.
17636         (offsettable_ok_by_alignment): Likewise.
17637         (rs6000_legitimate_offset_address_p): Likewise.
17638         (legitimate_lo_sum_address_p): Likewise.
17639         (rs6000_legitimize_address): Likewise.
17640         (rs6000_legitimize_reload_address): Likewise.
17641         (rs6000_secondary_reload_inner): Likewise.
17642         (rs6000_preferred_reload_class): Likewise.
17643         (rs6000_output_move_128bit): Likewise.
17645 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17647         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17648         Generate SImode target register for null target.
17649         <case IX86_BUILTIN_XGETBV>: Ditto.
17650         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17651         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17653 2018-05-10  Carl Love  <cel@us.ibm.com>
17655         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17656         dcbtt and dcbtstt if operands[2] is 0.
17658 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17660         PR target/85693
17661         * config/i386/sse.md (usadv64qi): New expander.
17663 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
17665         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17666         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17667         -maltivec=be support.
17668         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17669         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17670         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17671         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17672         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17673         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17674         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17675         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17676         altivec_vsumsws): Adjust.
17677         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17678         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17679         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17680         support.
17681         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17682         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17683         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17684         (altivec_lve<VI_char>x): Delete expand.
17685         (*altivec_lve<VI_char>x_internal): Rename to...
17686         (altivec_lve<VI_char>x): ... this.
17687         (altivec_lvxl_<mode>): Delete expand.
17688         (*altivec_lvxl_<mode>_internal): Rename to ...
17689         (altivec_lvxl_<mode>): ... this.
17690         (altivec_stvxl_<mode>): Delete expand.
17691         (*altivec_stvxl_<mode>_internal): Rename to ...
17692         (altivec_stvxl_<mode>): ... this.
17693         (altivec_stve<VI_char>x): Delete expand.
17694         (*altivec_stve<VI_char>x_internal): Rename to ...
17695         (altivec_stve<VI_char>x): ... this.
17696         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17697         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17698         reduc_plus_scal_<mode>): Adjust.
17699         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17700         comment.
17701         (rs6000_cpu_cpp_builtins): Adjust.
17702         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17703         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17704         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17705         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17706         -maltivec=be support.
17707         (rs6000_split_vec_extract_var): Adjust.
17708         (rs6000_split_v4si_init): Adjust.
17709         (swap_selector_for_mode): Delete.
17710         (altivec_expand_lvx_be, altivec_expand_stvx_be,
17711         altivec_expand_stvex_be): Delete.
17712         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17713         -maltivec=be support.
17714         (rs6000_gimple_fold_builtin): Ditto.
17715         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17716         Adjust.
17717         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17718         (TARGET_DIRECT_MOVE_64BIT): Adjust.
17719         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17720         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17721         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17722         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17723         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17724         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17725         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17726         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17727         anonymous split): Adjust.
17728         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17729         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17731 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
17733         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17734         when --with-gxx-include-dir is also specified.
17735         * configure: Regenerate.
17737 2018-05-09  Jim Wilson  <jimw@sifive.com>
17739         PR target/84797
17740         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17741         * config/riscv/t-withmultilib: New.
17742         * config/riscv/withmultilib.h: New.
17743         * doc/install.texi: Document RISC-V --with-multilib-list support.
17745 2018-05-09  Richard Biener  <rguenther@suse.de>
17747         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17748         vector.
17749         (vect_bb_vectorization_profitable_p): Adjust.  Compute
17750         actual scalar cost using the cost vector and the add_stmt_cost
17751         machinery.
17753 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17755         PR rtl-optimization/85645
17756         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17757         in the REG_CFA_REGISTER note for LR, don't leave it empty.
17759 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17761         PR rtl-optimization/85645
17762         * shrink-wrap.c (spread_components): Return a boolean saying if
17763         anything was changed.
17764         (try_shrink_wrapping_separate): Iterate spread_components until
17765         nothing changes anymore.
17767 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17769         PR rtl-optimization/85645
17770         * regrename.c (build_def_use): Also kill the chains that include the
17771         destination of a REG_CFA_REGISTER note.
17773 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17775         PR rtl-optimization/85645
17776         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17777         insn that has a REG_CFA_REGISTER note.
17779 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
17781         * cfgexpand.c (expand_clobber): New function.
17782         (expand_gimple_stmt_1): Use it.
17783         * tree-vect-stmts.c (vect_clobber_variable): New function,
17784         split out from...
17785         (vectorizable_simd_clone_call): ...here.
17786         (vectorizable_store): Emit a clobber either side of an
17787         IFN_STORE_LANES sequence.
17788         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
17790 2018-05-09  Tom de Vries  <tom@codesourcery.com>
17792         PR target/85626
17793         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
17794         (define_insn "trap_if_false"): Add exit after trap.
17796 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
17798         PR rtl-optimization/85638
17799         * bb-reorder.c: Include common/common-target.h.
17800         (create_forwarder_block): New function extracted from...
17801         (fix_up_crossing_landing_pad): ...here.  Rename into...
17802         (dw2_fix_up_crossing_landing_pad): ...this.
17803         (sjlj_fix_up_crossing_landing_pad): New function.
17804         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
17805         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
17806         from both partitions and exit the loop after one iteration.
17808 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17810         Revert:
17811         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17812         subsection.
17813         (Basic PowerPC Built-in Functions): The new name of the
17814         subsection previously known as "PowerPC Built-in Functions".
17815         (Basic PowerPC Built-in Functions Available on all Configurations):
17816         New subsubsection.
17817         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17818         subsubsection.
17819         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17820         subsubsection.
17821         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17822         subsubsection.
17823         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17824         subsubsection.
17826 2018-05-08  Jim Wilson  <jimw@sifive.com>
17828         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
17829         (LD_EMUL_SUFFIX): New.
17830         (LINK_SPEC): Use it.
17832 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17834         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17835         subsection.
17836         (Basic PowerPC Built-in Functions): The new name of the
17837         subsection previously known as "PowerPC Built-in Functions".
17838         (Basic PowerPC Built-in Functions Available on all Configurations):
17839         New subsubsection.
17840         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17841         subsubsection.
17842         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17843         subsubsection.
17844         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17845         subsubsection.
17846         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17847         subsubsection.
17849 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17851         PR target/85683
17852         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
17853         after cmpelim optimization.
17855 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
17857         * config.gcc: Support "goldmont".
17858         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
17859         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17860         PROCESSOR_GOLDMONT.
17861         * config/i386/i386.c (m_GOLDMONT): Define.
17862         (processor_target_table): Add "goldmont".
17863         (PTA_GOLDMONT): Define.
17864         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
17865         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
17866         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
17867         (fold_builtin_cpu): Add "goldmont".
17868         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
17869         (ix86_option_override_internal): Add "goldmont".
17870         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
17871         (processor_type): Add PROCESSOR_GOLDMONT.
17872         * config/i386/i386.md: Add CPU "glm".
17873         * config/i386/glm.md: New file.
17874         * config/i386/x86-tune.def: Add m_GOLDMONT.
17875         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
17877 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17879         PR target/85572
17880         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
17881         E_V4DImode.
17882         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
17883         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
17884         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
17886         PR target/85317
17887         * config/i386/i386.c (ix86_fold_builtin): Handle
17888         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
17890         PR target/85480
17891         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
17892         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
17894 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
17896         PR target/85658
17897         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
17898         (check_arch): Likewise.
17899         (check_fpu): Return the result rather than printing it.
17900         (end arch): Fix operator precedence.
17901         (end cpu): Likewise.
17902         (END): Print the result from check_fpu.
17904 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17905             Alan Hayward  <alan.hayward@arm.com>
17906             David Sherwood  <david.sherwood@arm.com>
17908         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
17909         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
17910         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
17911         (*fcmuo<mode>_and): New patterns.
17913 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17915         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
17916         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
17917         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
17918         (cmp_op, sve_imm_con): New code attributes.
17919         (SVE_COND_INT_CMP, imm_con): Delete.
17920         (cmp_op): Remove above unspecs from int attribute.
17921         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
17922         to...
17923         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
17924         comparison-specific unspecs.
17925         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
17926         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
17927         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
17928         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
17929         (*vec_fcm<cmp_op><mode>): Rename to...
17930         (*fcm<cmp_op><mode>): ...this and adjust likewise.
17931         (*vec_fcmuo<mode>): Rename to...
17932         (*fcmuo<mode>): ...this and adjust likewise.
17933         (*pred_fcm<cmp_op><mode>): New pattern.
17934         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
17935         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
17936         functions.
17937         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
17938         and UNORDERED.
17939         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
17940         (aarch64_emit_sve_predicated_cond): New function.
17941         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
17942         (aarch64_emit_unspec_cond_or): Replace with...
17943         (aarch64_emit_sve_or_conds): ...this new function.  Use
17944         aarch64_emit_sve_ptrue_op for the individual comparisons and
17945         aarch64_emit_binop to OR them together.
17946         (aarch64_emit_inverted_unspec_cond): Replace with...
17947         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
17948         aarch64_emit_sve_ptrue_op for the comparison and
17949         aarch64_emit_unop to invert the result.
17950         (aarch64_expand_sve_vec_cmp_float): Update after the above
17951         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
17953 2018-05-07  Nathan Sidwell  <nathan@acm.org>
17955         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
17956         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
17957         (Backwards Compatibility): Likewise.
17959 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17961         PR bootstrap/85681
17962         Revert:
17963         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17965         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17966         <prefetch_dynamic_strides>: New const bool field.
17967         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17968         prefetch_dynamic_strides.
17969         (exynosm1_prefetch_tune): Likewise.
17970         (thunderxt88_prefetch_tune): Likewise.
17971         (thunderx_prefetch_tune): Likewise.
17972         (thunderx2t99_prefetch_tune): Likewise.
17973         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17974         to false.
17975         (aarch64_override_options_internal): Update to set
17976         PARAM_PREFETCH_DYNAMIC_STRIDES.
17977         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17978         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17979         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17980         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17981         prefetch-dynamic-strides setting.
17983         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17985         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17986         <minimum_stride>: New const int field.
17987         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17988         minimum_stride field.
17989         (exynosm1_prefetch_tune): Likewise.
17990         (thunderxt88_prefetch_tune): Likewise.
17991         (thunderx_prefetch_tune): Likewise.
17992         (thunderx2t99_prefetch_tune): Likewise.
17993         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17994         (aarch64_override_options_internal): Update to set
17995         PARAM_PREFETCH_MINIMUM_STRIDE.
17996         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17997         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17998         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17999         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18000         stride is constant and is below the minimum stride threshold.
18002 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18004         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
18005         to 512.
18007 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18009         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18010         <prefetch_dynamic_strides>: New const bool field.
18011         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18012         prefetch_dynamic_strides.
18013         (exynosm1_prefetch_tune): Likewise.
18014         (thunderxt88_prefetch_tune): Likewise.
18015         (thunderx_prefetch_tune): Likewise.
18016         (thunderx2t99_prefetch_tune): Likewise.
18017         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18018         to false.
18019         (aarch64_override_options_internal): Update to set
18020         PARAM_PREFETCH_DYNAMIC_STRIDES.
18021         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18022         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18023         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18024         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18025         prefetch-dynamic-strides setting.
18027 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18029         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18030         <minimum_stride>: New const int field.
18031         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18032         minimum_stride field.
18033         (exynosm1_prefetch_tune): Likewise.
18034         (thunderxt88_prefetch_tune): Likewise.
18035         (thunderx_prefetch_tune): Likewise.
18036         (thunderx2t99_prefetch_tune): Likewise.
18037         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18038         (aarch64_override_options_internal): Update to set
18039         PARAM_PREFETCH_MINIMUM_STRIDE.
18040         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18041         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18042         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18043         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18044         stride is constant and is below the minimum stride threshold.
18046 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
18048         PR c++/85659
18049         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
18050         the type is addressable.  Don't force op into register if it has
18051         BLKmode.
18053 2018-05-05  Roland McGrath  <mcgrathr@google.com>
18055         PR other/77609
18056         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
18057         any section for which we don't know a specific type it should have,
18058         regardless of name.  Previously this was done only for the exact
18059         names ".init_array", ".fini_array", and ".preinit_array".
18060         (default_elf_asm_named_section): Add comment about
18061         relationship with default_section_type_flags and SECTION_NOTYPE.
18062         (get_section): Don't consider it a type conflict if one side has
18063         SECTION_NOTYPE and the other doesn't, as long as neither has the
18064         SECTION_BSS et al used in the default_section_type_flags logic.
18066 2018-05-05  Tom de Vries  <tom@codesourcery.com>
18068         PR target/85653
18069         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
18070         (workaround_barsyncs): New function.
18071         (nvptx_reorg): Use workaround_barsyncs.
18072         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18073         (define_expand "nvptx_membar_cta"): New define_expand.
18074         (define_insn "*nvptx_membar_cta"): New insn.
18076 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
18078         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
18079         To improve optimization opportunities.
18080         * builtin-types.def: The new needed builtin types for the above.
18082 2018-05-04  Richard Biener  <rguenther@suse.de>
18084         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
18085         * gimple-ssa-store-merging.c
18086         (imm_store_chain_info::output_merged_store): Remove redundant create,
18087         release split_store vector contents on failure.
18088         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
18089         scalar stmt vector on cache hit.
18091 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
18093         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
18094         Xilinx FP support.
18095         * config.gcc (powerpc-xilinx-eabi*): Remove.
18096         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
18097         support.
18098         (fusion_addis_mem_combo_load): Ditto.
18099         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
18100         FP support.
18101         (rs6000_cpu_cpp_builtins): Ditto.
18102         * config/rs6000/rs6000-linux.c
18103         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
18104         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
18105         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
18106         support.
18107         (rs6000_setup_reg_addr_masks): Ditto.
18108         (rs6000_init_hard_regno_mode_ok): Ditto.
18109         (rs6000_option_override_internal): Ditto.
18110         (legitimate_lo_sum_address_p): Ditto.
18111         (rs6000_legitimize_address): Ditto.
18112         (rs6000_legitimize_reload_address): Ditto.
18113         (rs6000_legitimate_address_p): Ditto.
18114         (abi_v4_pass_in_fpr): Ditto.
18115         (setup_incoming_varargs): Ditto.
18116         (rs6000_gimplify_va_arg): Ditto.
18117         (rs6000_split_multireg_move): Ditto.
18118         (rs6000_savres_strategy): Ditto.
18119         (rs6000_emit_prologue_components): Ditto.
18120         (rs6000_emit_epilogue_components): Ditto.
18121         (rs6000_emit_prologue): Ditto.
18122         (rs6000_emit_epilogue): Ditto.
18123         (rs6000_elf_file_end): Ditto.
18124         (rs6000_function_value): Ditto.
18125         (rs6000_libcall_value): Ditto.
18126         * config/rs6000/rs6000.h: Ditto.
18127         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
18128         (TARGET_MINMAX): ... this.  New.
18129         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
18130         * config/rs6000/rs6000.md: Remove Xilinx FP support.
18131         (*movsi_internal1_single): Delete.
18132         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
18133         mfpu=, mxilinx-fpu): Delete.
18134         * config/rs6000/singlefp.h: Delete.
18135         * config/rs6000/sysv4.h: Remove Xilinx FP support.
18136         * config/rs6000/t-rs6000: Ditto.
18137         * config/rs6000/t-xilinx: Delete.
18138         * config/rs6000/titan.md: Adjust for fp_type removal.
18139         * config/rs6000/vsx.md: Remove Xilinx FP support.
18140         (VStype_simple): Delete.
18141         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
18142         * config/rs6000/xfpu.h: Delete.
18143         * config/rs6000/xfpu.md: Delete.
18144         * config/rs6000/xilinx.h: Delete.
18145         * config/rs6000/xilinx.opt: Delete.
18146         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
18147         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
18149 2018-05-04  Tom de Vries  <tom@codesourcery.com>
18151         PR libgomp/85639
18152         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
18153         if ignore == 0.
18155 2018-05-04  Richard Biener  <rguenther@suse.de>
18157         PR middle-end/85627
18158         * tree-complex.c (update_complex_assignment): We are always in SSA form.
18159         (expand_complex_div_wide): Likewise.
18160         (expand_complex_operations_1): Likewise.
18161         (expand_complex_libcall): Preserve EH info of the original stmt.
18162         (tree_lower_complex): Handle removed blocks.
18163         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
18164         on complex multiplication and division libcall builtins.
18166 2018-05-04  Richard Biener  <rguenther@suse.de>
18168         PR middle-end/85574
18169         * fold-const.c (negate_expr_p): Restrict negation of operand
18170         zero of a division to when we know that can happen without
18171         overflow.
18172         (fold_negate_expr_1): Likewise.
18174 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
18176         PR libstdc++/85466
18177         * real.h (real_nextafter): Declare.
18178         * real.c (real_nextafter): New function.
18179         * fold-const-call.c (fold_const_nextafter): New function.
18180         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
18181         CASE_CFN_NEXTTOWARD.
18182         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
18183         even when arg1_mode is different from arg0_mode.
18185 2018-05-03  Nathan Sidwell  <nathan@acm.org>
18187         * doc/extend.texi (Deprecated Features): Remove
18188         -ffriend-injection.
18189         (Backwards Compatibility): Likewise.
18190         * doc/invoke.texi (C++ Language Options): Likewise.
18191         (C++ Dialect Options): Likewise.
18193 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
18195         PR target/85530
18196         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
18197         _mm512_mask_mullox_epi64): New intrinsics.
18199 2018-05-03  Tom de Vries  <tom@codesourcery.com>
18201         PR testsuite/85106
18202         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18203         dump files): Add offload-tree.
18205 2018-05-03  Richard Biener  <rguenther@suse.de>
18207         PR tree-optimization/85615
18208         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
18209         to loops not nested in BBs loop father to avoid creating multi-entry
18210         loops.
18212 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18214         PR tree-optimization/70291
18215         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
18216         arguments.  Change return type to tree.  Emit libcall as a new
18217         statement rather than replacing existing one when inplace_p is true.
18218         (expand_complex_multiplication_components): New function.
18219         (expand_complex_multiplication): Expand floating-point complex
18220         multiplication using the above.
18221         (expand_complex_division): Rename inner_type parameter to type.
18222         Update expand_complex_libcall call-site.
18223         (expand_complex_operations_1): Update expand_complex_multiplication
18224         and expand_complex_division call-sites.
18226 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18228         PR target/85582
18229         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18230         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18231         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
18232         the highest significant bit of the shift count mask is clear.  In
18233         check whether and[sq]i3 is needed verify that all significant bits
18234         of the shift count other than the highest are set.
18236 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18238         PR libgomp/82428
18239         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
18240         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
18241         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
18242         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
18243         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
18244         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
18245         __builtin_goacc_parlevel_size.
18247 2018-05-02  Richard Biener  <rguenther@suse.de>
18249         PR tree-optimization/85597
18250         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
18251         do not use split vect_get_vec_defs call but call vect_get_slp_defs
18252         directly.
18254 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18256         PR testsuite/85106
18257         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18258         dump files): Add ltrans-tree.
18260 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18262         PR testsuite/85106
18263         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18264         dump files): Add wpa-ipa.
18266 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
18268         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
18269         powerpc*-*-linux*paired* target.
18270         * config/rs6000/750cl.h: Delete.
18271         * config/rs6000/paired.h: Delete.
18272         * config/rs6000/paired.md: Delete.
18273         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
18274         float support.
18275         * config/rs6000/rs6000-builtin.def: Remove paired float support.
18276         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
18277         comment.  Remove paired float support.
18278         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
18279         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
18280         VECTOR_PAIRED.
18281         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
18282         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
18283         declarations.
18284         * config/rs6000/rs6000.c: Remove paired float support.
18285         (paired_expand_vector_init, paired_expand_vector_move,
18286         paired_emit_vector_compare, paired_emit_vector_cond_expr,
18287         (paired_expand_lv_builtin, paired_expand_stv_builtin,
18288         paired_expand_builtin, paired_expand_predicate_builtin,
18289         paired_init_builtins): Delete.
18290         * config/rs6000/rs6000.h: Remove paired float support.
18291         * config/rs6000/rs6000.md: Remove paired float support.
18292         (move_from_CR_ov_bit): Delete.
18293         * config/rs6000/rs6000.opt (mpaired): Delete.
18294         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
18295         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
18297 2018-05-02  Richard Biener  <rguenther@suse.de>
18299         PR middle-end/85567
18300         * gimplify.c (gimplify_save_expr): When in SSA form allow
18301         SAVE_EXPRs to compute to SSA vars.
18303 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18305         PR target/85582
18306         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18307         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18308         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
18309         clobber operands[2], instead use a new pseudo.  Formatting fixes.
18311 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
18313         PR tree-optimization/85586
18314         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
18315         exit early for statements in the same group if the accesses are
18316         not strided.
18318 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18320         PR lto/85451
18321         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
18322         error message.
18324 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
18326         PR tree-optimization/85143
18327         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
18329 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18331         PR lto/85451
18332         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
18333         not found" error message.
18335 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18337         PR other/83786
18338         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
18339         * vec.c (test_ordered_remove_if): New function.
18340         (vec_c_tests): Call test_ordered_remove_if.
18341         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
18342         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
18343         * tree-vect-patterns.c (vect_pattern_recog_1): Use
18344         VEC_ORDERED_REMOVE_IF.
18346 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18348         PR tree-optimization/82665
18349         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
18350         pointer subtraction where arguments come from a memchr call.
18352 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
18354         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
18355         --push-state --as-needed and --pop-state instead of --as-needed and
18356         --no-as-needed if ld supports it.
18357         * configure: Regenerated.
18359         PR web/85578
18360         * doc/install.texi2html: Replace _002d with - and _002a with * in
18361         generated html files using sed.
18363 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18365         PR c++/85523
18366         * gcc-rich-location.c (blank_line_before_p): New function.
18367         (use_new_line): New function.
18368         (gcc_rich_location::add_fixit_insert_formatted): New function.
18369         * gcc-rich-location.h
18370         (gcc_rich_location::add_fixit_insert_formatted): New function.
18372 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18374         * selftest.c (assert_streq): Rename "expected" and "actual" to
18375         "val1" and "val2".  Extend NULL-handling to cover both inputs
18376         symmetrically, while still requiring both to be non-NULL for a pass.
18377         * selftest.h (assert_streq): Rename "expected" and "actual" to
18378         "val1" and "val2".
18379         (ASSERT_EQ): Likewise.
18380         (ASSERT_EQ_AT): Likewise.
18381         (ASSERT_KNOWN_EQ): Likewise.
18382         (ASSERT_KNOWN_EQ_AT): Likewise.
18383         (ASSERT_NE): Likewise.
18384         (ASSERT_MAYBE_NE): Likewise.
18385         (ASSERT_MAYBE_NE_AT): Likewise.
18386         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
18387         the assertion to pass.
18388         (ASSERT_STREQ_AT): Likewise.
18390 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18392         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
18393         interaction with -pie.
18395 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18397         * selftest.h: Fix alphabetization of per-source-file selftest
18398         declarations.
18400 2018-04-30  Jason Merrill  <jason@redhat.com>
18402         PR c++/61982 - dead stores to destroyed objects.
18403         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18404         of clobber.
18406 2018-04-30  Jason Merrill  <jason@redhat.com>
18408         * tree.c (build_clobber): New.
18409         * tree.h: Declare it.
18410         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18412 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18414         * diagnostic-show-locus.c (layout::layout): Update for
18415         location_get_source_line returning a char_span.
18416         (struct char_span): Move to input.h.
18417         (struct correction): Update for fields in char_span becoming
18418         private.
18419         (struct source_line): Update for location_get_source_line
18420         returning a char_span.
18421         (layout::print_line): Likewise.
18422         * edit-context.c (edited_file::print_content): Likewise.
18423         (edited_file::print_diff_hunk): Likewise.
18424         (edited_file::print_run_of_changed_lines): Likewise.
18425         (edited_file::get_num_lines): Likewise.
18426         (edited_line::edited_line): Likewise.
18427         * final.c (asm_show_source): Likewise.
18428         * input.c (location_get_source_line): Convert return type
18429         from const char * to char_span, losing the final "line_len"
18430         param.
18431         (dump_location_info): Update for the above.
18432         (get_substring_ranges_for_loc): Likewise.  Use a char_span
18433         when handling the literal within the line.
18434         (test_reading_source_line): Update for location_get_source_line
18435         returning a char_span.
18436         * input.h (class char_span): Move here from
18437         diagnostic-show-locus.c, converting from a struct to a class.
18438         Make data members private.
18439         (char_span::operator bool): New.
18440         (char_span::length): New.
18441         (char_span::get_buffer): New.
18442         (char_span::operator[]): New.
18443         (char_span::subspan): Make const.
18444         (char_span::xstrdup): New.
18445         (location_get_source_line): Convert return type from const char *
18446         to char_span, losing the final "line_size" param.
18448 2018-04-30  Jan Hubicka  <jh@suse.cz>
18450         * lto-wrapper.c (ltrans_priorities): New static var.
18451         (cmp_priority): New.
18452         (run_gcc): Read priorities and if doing parallel build order
18453         the Makefile by them.
18455 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18457         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18459 2018-04-30  Richard Biener  <rguenther@suse.de>
18461         * tree-cfg.c (verify_address): Remove base argument, add
18462         flag whether to check TREE_ADDRESSABLE and do that.
18463         (verify_expr): Remove.
18464         (verify_types_in_gimple_reference): Add pieces from verify_expr.
18465         (verify_gimple_assign_single): Likewise.
18466         (verify_gimple_switch): Likewise.
18467         (verify_expr_location_1): Dereference tp once.  Add (disabled)
18468         piece from verify_expr.
18469         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18471 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18473         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18475 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18477         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18478         (small_data_pattern): Likewise.
18479         (arc_rewrite_small_data): Likewise.
18480         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18481         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18482         (get_symbol_alignment): New function.
18483         (legitimate_small_data_address_p): Likewise.
18484         (legitimate_scaled_address): Update, call
18485         legitimate_small_data_address_p.
18486         (output_sdata): New static variable.
18487         (arc_print_operand): Update how we handle small data operands.
18488         (arc_print_operand_address): Likewise.
18489         (arc_legitimate_address_p): Update, use
18490         legitimate_small_data_address_p.
18491         (arc_rewrite_small_data_p): Remove.
18492         (arc_rewrite_small_data_1): Likewise.
18493         (arc_rewrite_small_data): Likewise.
18494         (small_data_pattern): Likewise.
18495         (compact_sda_memory_operand): Update to use
18496         legitimate_small_data_address_p and get_symbol_alignment.
18497         (prepare_move_operands): Don't rewite sdata pattern.
18498         (prepare_extend_operands): Remove.
18499         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18500         pattern.
18501         (zero_extendqisi2): Likewise.
18502         (zero_extendhisi2): Likewise.
18503         (extendqihi2): Likewise.
18504         (extendqisi2): Likewise.
18505         (extendhisi2): Likewise.
18506         (addsi3): Likewise.
18507         (subsi3): Likewise.
18508         (andsi3): Likewise.
18509         * config/arc/constraints.md (Usd): Change it to memory constraint.
18511 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18513         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18514         as source of std instructions.
18515         * config/arc/arc.md (movsi_insn): Update pattern predicate to
18516         allow 6-bit constants as source for store instructions.
18517         (movdi_insn): Update instruction pattern to allow 6-bit constants
18518         as source for store instructions.
18520 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18522         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18524 2018-04-30  Nathan Sidwell  <nathan@acm.org>
18525             Sandra Loosemore <sandra@codesourcery.com>
18527         * dumpfile.c (dump_open): Allow '-' for stdout.
18528         * doc/invoke.texi (Developer Options): Document dump filename
18529         determination early.  Document stdin/stdout selection.
18531 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
18533         Microblaze Target: PIC data text relative
18535         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18536         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18537         Add declaration.
18538         * config/microblaze/microblaze.h (microblaze_constant_address_p):
18539         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18540         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18541         New addressing mode for data-text relative position indepenedent code.
18542         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18543         'ADDRESS_SYMBOLIC_TXT_REL'.
18544         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18545         (microblaze_legitimate_pic_operand): Exclude function calls from
18546         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18547         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18548         addresses cases.
18549         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18550         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18551         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18552         for 'address + offset'.
18553         (microblaze_expand_prologue): Add new function prologue call for
18554         'r20' assignation.
18555         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18556         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18557         table in case of TARGET_PIC_DATA_TEXT_REL.
18558         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18559         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18560         Add new macros 'UNSPEC_TEXT',
18561         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18562         + exclude function calls from 'UNSPEC_PLT' in case of data text
18563         relative mode.
18564         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18565         new target hook for generating address diff vector tables in case of
18566         flag_pic.
18567         * doc/tm.texi : Regenerate.
18568         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18569         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18570         of addr diff vector generation.
18571         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18572         target hook definition.
18573         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18574         Add default function for generate_pic_addr_diff_vec -> flag_pic.
18575         * doc/invoke.texi (Add new pic option): Add new microblaze pic
18576         option for data text relative.
18578 2018-04-30  Richard Biener  <rguenther@suse.de>
18580         * tree-chrec.h (evolution_function_is_constant_p): Remove
18581         redundant check.
18582         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18584 2018-04-30  Richard Biener  <rguenther@suse.de>
18586         PR bootstrap/85571
18587         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18589 2018-04-30  Richard Biener  <rguenther@suse.de>
18591         PR tree-optimization/28364
18592         PR tree-optimization/85275
18593         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18594         copying first exit test.
18596 2018-04-28  Mark Wielaard  <mark@klomp.org>
18598         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18599         dwarf_version >= 5.
18600         (dwarf_AT): Handle DW_AT_addr_base.
18601         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18603 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
18605         PR target/84431
18606         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18607         (*ashl<dwi>3_doubleword_mask_1): Ditto.
18608         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18609         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18611 2018-04-28  Richard Biener  <rguenther@suse.de>
18613         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18614         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18615         to reflect use.  Only add interesting stmts.
18617 2018-04-27  Martin Jambor  <mjambor@suse.cz>
18619         PR ipa/85549
18620         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18621         the jump function allows for passing through aggregate values.
18623 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
18625         * input.h (in_system_header_at): Convert from macro to inline
18626         function.
18627         (from_macro_expansion_at): Likewise.
18628         (from_macro_definition_at): Likewise.
18630 2018-04-27  Jeff Law  <law@redhat.com>
18632         * config.gcc: Mark tile* targets as deprecated/obsolete.
18634 2018-04-27  Richard Biener  <rguenther@suse.de>
18636         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18637         fix for ILP32.
18639 2018-04-27  Richard Biener  <rguenther@suse.de>
18641         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18643 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
18645         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18646         with Yd constraint. Set "preferred_for_speed" attribute from
18647         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18648         with Yd constraint.
18649         (*movdi_internal): Ditto.
18650         (movti_interunit splitters): Remove
18651         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18652         (movdi_interunit splitters): Ditto.
18653         * config/i386/constraints.md (Ye): Remove.
18654         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18656 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18658         PR target/85512
18659         * config/aarch64/constraints.md (Usg): Limit to 31.
18660         (Usj): Limit to 63.
18662 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
18664         PR tree-optimization/85529
18665         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18666         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18667         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18668         zero extension or masking of the MSB bit.
18669         (optimize_range_tests): Add FIRST_BB argument, pass it through
18670         to optimize_range_tests_var_bound.
18671         (maybe_optimize_range_tests, reassociate_bb): Adjust
18672         optimize_range_tests callers.
18674 2018-04-26  Richard Biener  <rguenther@suse.de>
18675             Jakub Jelinek  <jakub@redhat.com>
18677         * cgraph.h (symbol_table): Just declare debug method here.
18678         * symtab.c (symbol_table::debug): Define.
18680 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
18682         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18684 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
18686         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18687         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18688         (*movdi_internal): Substitute Yi and Yj constraint with x
18689         and Ym and Yn constraint with y constraint.  Update "isa"
18690         attribute and set "preferred_for_speed" attribute from
18691         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18692         (*movsi_internal): Ditto.
18693         (*movdf_internal): Ditto.
18694         (*movsf_internal): Ditto.
18695         (*zero_extendsidi2): Ditto.
18696         * config/i386/sse.md (vec_set<mode>_0): Ditto.
18697         (sse2_loadld): Ditto.
18698         (*vec_extract<ssevecmodelower>_0): Ditto.
18699         (*vec_extractv4si_0_zext_sse4): Ditto.
18700         (vec_concatv2di): Ditto.
18701         (*vec_dup<mode>): Ditto.
18702         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18703         * config/i386/constraints.md (Yi): Remove.
18704         (Yj): Remove.
18705         (Ym): Remove.
18706         (Yn): Remove.
18708 2018-04-26  Nathan Sidwell  <nathan@acm.org>
18710         * dumpfile.c (dump_open): New.
18711         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18712         (dump_finish): Detect stdio/stderr by value not name.
18714 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
18716         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18718 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18720         PR target/84952
18721         * config/nvptx/nvptx.c (verify_neutering_jumps)
18722         (verify_neutering_labels): New function
18723         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18725 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18727         PR target/84025
18728         * config/nvptx/nvptx.c (needs_neutering_p): New function.
18729         (nvptx_single): Use needs_neutering_p to skip over insns that do not
18730         need neutering.
18732 2018-04-26  Richard Biener <rguenther@suse.de>
18733             Tom de Vries  <tom@codesourcery.com>
18735         PR lto/85422
18736         * lto-streamer-out.c (output_function): Fixup loops if required to match
18737         discovery done in the reader.
18739 2018-04-26  Richard Biener  <rguenther@suse.de>
18741         PR tree-optimization/85116
18742         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18743         have a loop exit from the single latch predecessor.  Remove
18744         case of header with just condition.
18745         (ch_base::copy_headers): Exclude infinite loops from any
18746         processing.
18747         (pass_ch::execute): Record exits.
18749 2018-04-26  Richard Biener  <rguenther@suse.de>
18751         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18752         prologue cost vector and pass it to vect_get_load_cost.
18753         (vect_get_peeling_costs_all_drs): Likewise.
18754         (vect_peeling_hash_get_lowest_cost): Likewise.
18755         (vect_enhance_data_refs_alignment): Likewise.
18757 2018-04-26  Richard Biener  <rguenther@suse.de>
18759         PR middle-end/85450
18760         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18761         checking of integer<->pointer conversions.
18762         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18763         sign-/zero-extending pointer types.
18764         (expand_omp_for_static_chunk): Likewise.
18766 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
18767             Jean Lee  <xiaoyur347@gmail.com>
18769         * config/mips/mips.c (mips_asan_shadow_offset): New function.
18770         (TARGET_ASAN_SHADOW_OFFSET): Define.
18771         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18772         true for -fsanitize=address.
18774 2018-04-25  Mark Wielaard  <mark@klomp.org>
18776         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18777         shorter ones.
18779 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18781         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
18782         than "alu", remove explicit "memory" and "imm_disp" attributes.
18783         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
18785         PR middle-end/85414
18786         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
18787         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
18788         gen_lowpart_no_emit.
18790 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
18792         PR target/85473
18793         * config/i386/i386.c (ix86_expand_builtin): Change memory
18794         operand to XI, extend p0 to Pmode.
18795         * config/i386/i386.md: Change unspec volatile and operand
18796         1 mode to XI, change operand 0 mode to P.
18798 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18800         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
18801         GET_MODE_MASK before any checking.
18802         (nds32_can_use_bset_p): Likewise.
18803         (nds32_can_use_btgl_p): Likewise.
18805 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18807         * config/nds32/nds32-doubleword.md: New define_split pattern for
18808         illegal register number.
18810 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18812         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
18814 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18816         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
18818 2018-04-25  Richard Biener  <rguenther@suse.de>
18820         * lto-streamer.h (LTO_major_version): Bump to 8.
18822 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18824         * BASE-VER: Set to 9.0.0.
18826 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
18828         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
18829         in __abskf2 and __powikf2.
18831 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18833         PR target/85512
18834         * config/aarch64/constraints.md (Usg, Usj): New constraints.
18835         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
18836         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
18837         Use the above on operand 2.  Reindent.
18838         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
18840 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18842         PR target/85485
18843         * common/config/i386/i386-common.c (ix86_handle_option): Don't
18844         handle OPT_mcet.
18845         * config/i386/i386.opt (mcet): Removed.
18846         * doc/install.texi: Remove -mcet documentation.
18847         * doc/invoke.texi: Likewise.
18849 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18851         PR target/85485
18852         * doc/install.texi: Remove -mcet from bootstrap-cet.
18854 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18856         PR target/85511
18857         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
18858         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
18859         if TARGET_64BIT.
18861         PR target/85503
18862         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
18863         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
18864         containing a CONST_VECTOR.
18866 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
18868         * doc/install.texi: Update newlib dependency for nvptx.
18870 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18872         PR target/85508
18873         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
18874         instead of INTVAL when shifting x left.
18876 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
18878         PR tree-optimization/85478
18879         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
18880         vect_grouped_store_supported for single element vectors.
18882 2018-04-24  Richard Biener  <rguenther@suse.de>
18884         PR target/85491
18885         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
18886         load cost increase to the case of non-constant step.
18888 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18890         PR target/84828
18891         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
18892         destination if any_malformed_asm.
18894 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
18896         PR middle-end/85496
18897         * expr.c (store_field): In the bitfield case, if the value comes from
18898         a function call and is returned in registers by means of a PARALLEL,
18899         do not change the mode of the temporary unless BLKmode and VOIDmode.
18901 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
18903         PR rtl-optimization/85423
18904         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
18905         dependencies to debug insns when the previous insn is non-debug.
18907 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18909         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
18910         enums into a single definition.
18911         (fls): Fix predicates and printing.
18912         (seti): Likewise.
18914 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18916         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
18917         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
18918         and short u6 immediate.
18919         (check_if_valid_sleep_operand): Remove.
18920         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
18922 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18924         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
18925         flag_always_save_lp condition.
18926         * config/nds32/nds32.opt (malways-save-lp): New option.
18928 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18930         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
18931         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
18932         * config/nds32/nds32.h
18933         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
18934         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
18936 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18938         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
18939         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
18941 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18942             Chung-Ju Wu  <jasonwucj@gmail.com>
18944         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
18945         Declare.
18946         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
18947         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
18949 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18951         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
18953 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18955         * config/nds32/nds32-protos.h (nds32_data_alignment,
18956         nds32_local_alignment): Declare.
18957         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
18958         nds32_local_alignment): New functions.
18959         (TARGET_CONSTANT_ALIGNMENT): Define.
18960         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
18962 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18964         * config/nds32/nds32.c
18965         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
18966         (TARGET_MODES_TIEABLE_P): Likewise.
18968 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18970         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
18971         level Ofast and Og.
18973 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
18974             Chung-Ju Wu  <jasonwucj@gmail.com>
18976         * config/nds32/constants.md (unspec_volatile_element): Add enum values
18977         for unaligned access.
18978         * config/nds32/nds32-intrinsic.c: Implementation of expanding
18979         unaligned access.
18980         * config/nds32/nds32-intrinsic.md: Likewise.
18981         * config/nds32/nds32_intrinsic.h: Likewise.
18982         * config/nds32/nds32.h (nds32_builtins): Likewise.
18983         * config/nds32/nds32.opt (munaligned-access): New option.
18984         * config/nds32/nds32.c (nds32_asm_file_start): Display
18985         flag_unaligned_access status.
18987 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
18989         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
18990         -mno-relax is present.
18991         * config/riscv/linux.h (LINK_SPEC): Ditto.
18993 2018-04-20  Martin Sebor  <msebor@redhat.com>
18995         PR c/85365
18996         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
18997         for null pointers.
18998         (gimple_fold_builtin_stxcpy_chk): Same.
18999         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
19001 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
19003         PR target/85456
19004         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
19005         __powikf2 when long double is IEEE 128-bit.
19007 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
19009         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
19010         step to make sure stack always aligned.
19012 2018-04-20  Carl Love  <cel@us.ibm.com>
19014         PR target/83402
19015         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
19016         size check for arg0.
19018 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
19019             Tom de Vries  <tom@codesourcery.com>
19021         PR target/85445
19022         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
19023         Emit insns for calls too.
19024         (nvptx_find_par): Always look for worker-level predecessor insn.
19025         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
19026         calls.
19027         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
19028         (nvptx_process_pars): Propagate frames for calls.
19030 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
19032         PR target/85469
19033         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
19034         Removed.
19035         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
19036         (ix86_handle_option): Don't handle OPT_mibt.
19037         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
19038         __SHSTK__.
19039         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
19040         has_ibt and ibt.
19041         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
19042         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
19043         (ix86_target_macros): Define __CET__ with flag_cf_protection
19044         for -fcf-protection.
19045         * config/i386/i386.c (isa2_opts): Remove -mibt.
19046         * config/i386/i386.h (TARGET_IBT): Removed.
19047         (TARGET_IBT_P): Likewise.
19048         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
19049         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
19050         * config/i386/i386.opt (mcet): Update help message.
19051         (mshstk): Likewise.
19052         (mibt): Removed.
19053         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
19054         -mcet as an alias for -mshstk.
19056 2018-04-20  Richard Biener <rguenther@suse.de>
19058         PR middle-end/85475
19059         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
19060         complexity by forcing a single use of the multiply operand.
19062 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19064         ipa/85449
19065         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
19066         recursion dependency to only apply to non-clones.
19068 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19070         ipa/85447
19071         * ipa-cp.c (create_specialized_node): Check that clones of
19072         self-recursive edges exist during IPA-CP.
19074 2018-04-19  Toon Moene  <toon@moene.org>
19076         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
19077         by -O3.
19079 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19081         PR tree-optimization/85467
19082         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
19083         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
19084         VECTOR_CST element to type.
19086 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19088         PR target/85397
19089         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
19090         * config/i386/i386.md (builtin_setjmp_setup): Removed.
19091         (builtin_longjmp): Likewise.
19092         (save_stack_nonlocal): New pattern.
19093         (restore_stack_nonlocal): Likewise.
19095 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19097         PR target/85404
19098         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19099         Replace ASM_OUTPUT_LABEL with fprintf.
19101 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19103         PR target/85417
19104         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19105         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
19106         * config/i386/i386-c.c (ix86_target_macros_internal): Also
19107         define __IBT__ and __SHSTK__ for -fcf-protection.
19108         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
19109         TARGET_IBT.
19110         (ix86_trampoline_init): Likewise.
19111         (x86_output_mi_thunk): Likewise.
19112         (ix86_notrack_prefixed_insn_p): Likewise.
19113         (ix86_option_override_internal): Don't disallow -fcf-protection.
19114         * config/i386/i386.md (rdssp<mode>): Also enable for
19115         -fcf-protection.
19116         (incssp<mode>): Likewise.
19117         (nop_endbr): Likewise.
19118         * config/i386/i386.opt (mcet): Change help message to built-in
19119         functions only.
19120         (mibt): Likewise.
19121         (mshstk): Likewise.
19122         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
19123         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
19124         enable CET built-in functions.
19126 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
19128         * common/config/i386/i386-common.c
19129         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
19130         OPTION_MASK_ISA_MOVDIRI_UNSET,
19131         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
19132         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
19133         * config.gcc (movdirintrin.h): New header.
19134         * config/i386/cpuid.h (bit_MOVDIRI,
19135         bit_MOVDIR64B): New bits.
19136         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
19137         and -mmvodir64b.
19138         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
19139         (VOID, PVOID, PCVOID)): New function types.
19140         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
19141         __builtin_ia32_directstoreu_u64,
19142         __builtin_ia32_movdir64b): New builtins.
19143         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
19144         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
19145         and -mmovdiri.
19146         (ix86_valid_target_attribute_inner_p): Ditto.
19147         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
19148         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
19149         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
19150         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
19151         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
19152         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
19153         (movdiri<mode>, movdir64b_<mode>): New.
19154         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
19155         * config/i386/immintrin.h: Include movdirintrin.h.
19156         * config/i386/movdirintrin.h: New file.
19157         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
19159 2018-04-19  Richard Biener  <rguenther@suse.de>
19161         PR middle-end/85455
19162         * cfg.c (clear_bb_flags): When loop state says we have
19163         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
19165 2018-04-19  Richard Biener  <rguenther@suse.de>
19167         PR tree-optimization/84737
19168         * tree-vect-data-refs.c (vect_copy_ref_info): New function
19169         copying restrict info.
19170         (vect_setup_realignment): Use it.
19171         * tree-vectorizer.h (vect_copy_ref_info): Declare.
19172         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
19173         the first DR to all generated stores.
19174         (vectorizable_load): Likewise for loads.
19176 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19178         PR tree-optimization/85446
19179         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
19180         the integral and pointer types to have the same precision.
19182         * doc/install.texi: Document --disable-cet being the default and
19183         --enable-cet=auto.
19185 2018-04-18  Martin Liska  <mliska@suse.cz>
19187         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
19188         style.
19190 2018-04-18  Martin Liska  <mliska@suse.cz>
19192         Revert
19193         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
19195         PR ipa/83983
19196         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19197         arguments if they are comparable.
19199 2018-04-18  Martin Liska  <mliska@suse.cz>
19201         Revert
19202         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19204         PR lto/84805
19205         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19206         incomplete types.
19208 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
19210         PR target/85388
19211         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
19212         ENDBR after calling __morestack.
19214 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
19216         PR jit/85384
19217         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
19218         by using gcc_base_ver to generate a gcc_driver_version, and use
19219         it when generating GCC_DRIVER_NAME.
19220         * configure: Regenerate.
19222 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
19224         PR target/81084
19225         * config.gcc: Obsolete powerpc*-*-*spe*.
19227 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19229         PR debug/84637
19230         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
19231         (stabstr_D): Change type of unum from unsigned int to
19232         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
19233         type.
19235 2018-04-17  Jim Wilson  <jimw@sifive.com>
19237         PR 84856
19238         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
19239         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
19240         Set arg_pointer_offset after using pretend_args_size.
19242 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19244         PR rtl-optimization/85431
19245         * dse.c (record_store): Ignore zero width stores.
19247         PR sanitizer/85230
19248         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
19249         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
19250         __builtin_stack_restore rather than after it.
19251         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
19252         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
19253         argument instead of virtual_dynamic_stack_rtx.
19255 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19257         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
19258         New prototype.
19259         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19260         Add note to error message to explain internal mapping of overloaded
19261         built-in function name to non-overloaded built-in function name.
19262         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
19263         function.
19265 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
19267         PR target/85424
19268         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
19269         where the inputs overlap with the output.
19271 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19273         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
19274         (=v, v) alternative and explicit "memory" attribute.
19275         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
19276         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19277         attributes.
19278         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
19279         "sselog1" type instead of "sselog".
19280         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
19281         "sselog".  Remove explicit "memory" attribute.
19282         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
19283         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19284         attributes.
19285         (vec_extract_hi_v32hi): Merge all alternatives into one, use
19286         "sselog1" type instead of "sselog".  Remove explicit "memory"
19287         attribute.
19288         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
19289         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19290         attribute.
19291         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
19292         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19293         attributes.
19294         (vec_extract_hi_v64qi): Merge all alternatives into one, use
19295         "sselog1" type instead of "sselog".  Remove explicit "memory"
19296         attribute.
19297         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
19298         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19299         attribute.
19301         PR target/85430
19302         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
19304         PR middle-end/85414
19305         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
19306         on a SUBREG.
19308 2018-04-17  Martin Jambor  <mjambor@suse.cz>
19310         PR ipa/85421
19311         * ipa-cp.c (create_specialized_node): Call
19312         expand_all_artificial_thunks if necessary.
19314 2018-04-17  Martin Liska  <mliska@suse.cz>
19316         PR lto/85405
19317         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
19318         in message, remote space in between '_G' and '('.
19320 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19322         PR target/85281
19323         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
19324         avx512f_vmcmp<mode>3<round_saeonly_name>,
19325         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
19326         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
19327         avx512f_rndscale<mode><round_saeonly_name>,
19328         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
19329         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
19330         Use %<iptr>2 instead of %2 for -masm=intel.
19331         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
19332         avx512f_vcvttss2usi<round_saeonly_name>,
19333         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
19334         -masm=intel.
19335         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
19336         avx512f_vcvttsd2usi<round_saeonly_name>,
19337         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
19338         Use %q1 instead of %1 for -masm=intel.
19339         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
19340         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
19341         of %3 for -masm=intel.
19342         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
19343         -masm=intel.
19344         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
19345         -masm=intel.
19346         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
19347         -masm=intel.
19348         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
19349         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
19350         %g1.
19351         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
19352         -masm=intel.
19353         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
19354         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
19355         %g1 and one with %0 and %1.
19356         (avx512er_vmrcp28<mode><round_saeonly_name>,
19357         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
19358         %1 for -masm=intel.
19359         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
19360         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
19361         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
19362         of %0 and %{%4%} for -masm=intel.
19363         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
19364         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
19365         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
19366         order of %0 and %{%5%}%{z%} for -masm=intel.
19368 2018-04-17  Jan Hubicka  <jh@suse.cz>
19370         PR lto/85405
19371         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
19373 2018-04-17  Martin Liska  <mliska@suse.cz>
19375         PR ipa/85329
19376         * multiple_target.c (create_dispatcher_calls): Set apostrophes
19377         for target_clone error message.  Make default implementation
19378         clone to be a local declaration.
19379         (separate_attrs): Add new argument and check for an empty
19380         string.
19381         (expand_target_clones): Handle it.
19382         (ipa_target_clone): Make redirection just for target_clones
19383         functions.
19385 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
19386             Tom de Vries  <tom@codesourcery.com>
19388         PR middle-end/84955
19389         * omp-expand.c (expand_oacc_for): Add dummy false branch for
19390         tiled basic blocks without omp continue statements.
19392 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
19394         PR target/83660
19395         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19396         vec_extract expression as having side effects to make sure it gets
19397         a cleanup point.
19399 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
19401         PR target/85403
19402         * config/i386/i386.c (get_builtin_code_for_version): Check
19403         error_mark_node.
19405 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
19407         PR target/84331
19408         * config.gcc: Support "skylake".
19409         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19410         PROCESSOR_SKYLAKE.
19411         * config/i386/i386.c (m_SKYLAKE): Define.
19412         (processor_target_table): Add "skylake".
19413         (ix86_option_override_internal): Add "skylake".
19414         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19415         PROCESSOR_CANNONLAKE.
19416         (get_builtin_code_for_version): Fix priority for
19417         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19418         PROCESSOR_SKYLAKE-AVX512.
19419         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19420         (processor_type): Add PROCESSOR_SKYLAKE.
19422 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
19423             Jason Merrill  <jason@redhat.com>
19425         PR c++/85112
19426         * convert.c (convert_to_integer_1): Use direct recursion for
19427         enumeral types and types with a precision less than the number
19428         of bits in their mode.
19430 2018-04-16  Julia Koval  <julia.koval@intel.com>
19432         PR target/84413
19433         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19434         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19436 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
19438         PR target/85293
19439         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19440         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19441         and -mno-direct-move.
19443 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
19445         PR target/83402
19446         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19447         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19448         Ensure negative shifts result in {0}.
19450 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
19452         PR rtl-optimization/79916
19453         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19454         regs (if any) to define how to gnerate SD moves when LRA is in
19455         progress.
19457 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
19459         PR rtl-optimization/85393
19460         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19461         * except.c (expand_dw2_landing_pad_for_region): Make static.
19462         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19463         a label and unconditional jump to old_bb, rather than
19464         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19465         basic block.
19467         PR rtl-optimization/85376
19468         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19469         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19470         instead of a specific value.
19472 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19473             Bin Cheng  <bin.cheng@arm.com>
19475         PR tree-optimization/82965
19476         PR tree-optimization/83991
19477         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19478         by_profile_only parameter.
19479         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19480         information if the loop was predicted to iterate too many times.
19481         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19483 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19485         PR lto/71991
19486         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19487         always inline.
19489 2018-04-13  Martin Liska  <mliska@suse.cz>
19490             Jakub Jelinek  <jakub@redhat.com>
19492         PR middle-end/81657
19493         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19494         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19495         * builtins.c (expand_builtin_memory_copy_args): Use
19496         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19497         handle dest_addr == pc_rtx.
19499 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
19501         PR target/85291
19502         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19503         asked to not generate direct moves.
19504         (fix_trunc<mode>si2_stfiwx): Similar.
19505         (fix_trunc<mode>si2_internal): Similar.
19507 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19509         PR debug/83157
19510         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19511         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19512         lookup if dest in some wider mode is known to be const0_rtx and
19513         if so, record permanent equivalence for it to be ZERO_EXTEND of
19514         the narrower mode destination.
19516 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19518         * lto-streamer-out.c (output_function): Revert 259346.
19519         * omp-expand.c (expand_oacc_for): Likewise.
19521 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
19523         PR rtl-optimization/85354
19524         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19525         * sel-sched.c (sel_global_init): ... here.
19527 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
19529         PR target/85238
19530         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19531         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19532         mode for PE-COFF targets.
19533         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19534         (i386_pe_asm_lto_end): Likewise.
19535         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19536         (TARGET_ASM_LTO_END): Likewise.
19537         * config/i386/winnt.c (saved_debug_info_level): New static variable.
19538         (i386_pe_asm_lto_start): New function.
19539         (i386_pe_asm_lto_end): Likewise.
19541 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19542             Richard Biener  <rguenther@suse.de>
19544         PR middle-end/84955
19545         * lto-streamer-out.c (output_function): Fix CFG loop state before
19546         streaming out.
19547         * omp-expand.c (expand_oacc_for): Handle calls to internal
19548         functions like regular functions.
19550 2018-04-12  Richard Biener  <rguenther@suse.de>
19552         PR lto/85371
19553         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19554         for the early LTO debug to properly generate references to it
19555         during DIE emission.  Do not re-use that for the skeleton for
19556         split-dwarf.
19557         (dwarf2out_early_finish): Likewise.
19559 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19561         PR target/85328
19562         * config/i386/sse.md
19563         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19564         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19565         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19566         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19567         and output is a reg, avoid creating invalid lowpart subreg, but
19568         instead split into a 512-bit move.  Don't split if not AVX512VL,
19569         input is xmm16+ reg and output is a mem.
19570         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19571         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19572         xmm16+ reg and output is a mem.
19574 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19576         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19577         also for flag_dwarf2_cfi_asm.
19579 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19581         PR rtl-optimization/85342
19582         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19583         a bool scalar var inside of the loop instead.  Don't try to update
19584         recog_data.operand after failed apply_change_group.
19586 2018-04-12  Tom de Vries  <tom@codesourcery.com>
19588         PR target/85296
19589         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19590         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
19591         array with flexible array member as array without given dimension.
19592         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19593         argument for undefined param to true.
19595 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
19597         PR target/85321
19598         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19599         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19600         from PowerPC section.
19601         * config/rs6000/sysv4.opt (mcall-): Improve help text.
19602         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19603         help text that is too long.
19604         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19605         help text that is too long.
19606         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19607         help text that is too long.
19609 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
19611         * config/alpha/alpha.md (stack_probe_internal): Rename
19612         from "probe_stack".  Update all callers.
19614 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19616         PR rtl-optimization/84566
19617         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19618         sched_macro_fuse_insns.
19620 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19622         PR target/84301
19623         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19624         (compute_block_dependences): ... from here.
19626 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19628         PR tree-optimization/85331
19629         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19630         from int to HOST_WIDE_INT.
19632 2018-04-11  Martin Jambor  <mjambor@suse.cz>
19634         PR ipa/84149
19635         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19636         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19637         not the same as the source val.
19638         (cgraph_edge_brings_value_p): New parameter.
19639         (gather_edges_for_value): Pass destination value to
19640         cgraph_edge_brings_value_p.
19641         (perhaps_add_new_callers): Likewise.
19642         (get_info_about_necessary_edges): Likewise and exclude values brought
19643         only by self-recursive edges.
19644         (create_specialized_node): Redirect only clones of self-calling edges.
19645         (+self_recursive_pass_through_p): New function.
19646         (find_more_scalar_values_for_callers_subset): Use it.
19647         (find_aggregate_values_for_callers_subset): Likewise.
19648         (known_aggs_to_agg_replacement_list): Removed.
19649         (decide_whether_version_node): Re-calculate known constants for all
19650         remaining context clones.
19652 2018-04-11  Richard Biener  <rguenther@suse.de>
19654         PR lto/85339
19655         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19656         from early DWARF output.
19657         (dwarf2out_early_finish): Output line info unconditionally into
19658         early DWARF and add reference to it.
19660 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19662         PR target/85281
19663         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19664         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19665         other than V2DFmode using iptr mode attribute.
19666         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19668 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19670         PR rtl-optimization/84659
19671         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19673 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19675         PR debug/85302
19676         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19677         SIZEP is NULL.
19678         (output_loc_list): Pass address of a dummy size variable even in the
19679         locview handling loop.
19680         (index_location_lists): Add comment on why skip_loc_list_entry can't
19681         call size_of_locs.
19683 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19685         PR target/85261
19686         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19687         into register.
19689 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
19691         PR target/85321
19692         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19693         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19694         and -mstring-compare-inline-limit.
19696 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19698         PR target/85287
19699         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19700         for stack clash protection in a register whenever we need it to be in
19701         a register.
19703 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19705         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19706         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19708 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19710         PR target/85321
19711         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19712         the help text.
19713         (mlong-double-): Ditto.
19714         * config/rs6000/sysv4.opt (msdata=): Ditto.
19715         (mtls-size=): Ditto.
19717 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19719         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19720         erroneous entries for
19721         "vector int vec_ldl (int, long int *)", and
19722         "vector unsigned int vec_ldl (int, unsigned long int *)".
19723         Add comments and entries for
19724         "vector bool char vec_ldl (int, bool char *)",
19725         "vector bool short vec_ldl (int, bool short *)",
19726         "vector bool int vec_ldl (int, bool int *)",
19727         "vector bool long long vec_ldl (int, bool long long *)",
19728         "vector pixel vec_ldl (int, pixel *)",
19729         "vector long long vec_ldl (int, long long *)",
19730         "vector unsigned long long vec_ldl (int, unsigned long long *)".
19731         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19732         type tree bool_long_long_type_node and correct definition of
19733         bool_V2DI_type_node to make reference to this new type tree.
19734         (rs6000_mangle_type): Replace erroneous reference to
19735         bool_long_type_node with bool_long_long_type_node.
19736         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19737         comments to emphasize sign distinctions for char and int types and
19738         replace RS6000_BTI_bool_long constant with
19739         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
19740         use of RS6000_BTI_pixel.
19741         (bool_long_type_node): Remove this macro definition.
19742         (bool_long_long_type_node): New macro definition
19744 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19746         PR rtl-optimization/85300
19747         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19748         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19749         simplify_unary_operation fails.
19751 2018-04-10  Martin Liska  <mliska@suse.cz>
19753         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19754         cgraph_edge and ipa_ref.
19756 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19758         PR target/85177
19759         PR target/85255
19760         * config/i386/sse.md
19761         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19762         computation of the VEC_MERGE selector from mask.
19763         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19764         Fix decoding of the VEC_MERGE selector into mask.
19766 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19768         PR tree-optimization/85286
19769         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19771 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19773         * final.c (final_1): Set insn_last_address as well as
19774         insn_current_address.
19776 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19778         PR target/85173
19779         * explow.c (emit_stack_probe): Call validize_mem on memory location
19780         before passing it to gen_probe_stack.  Create address operand and
19781         legitimize it for the probe_stack_address case.
19783 2018-04-09  Jan Hubicka  <jh@suse.cz>
19785         PR lto/85078
19786         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
19787         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
19788         * tree.c (free_lang_data_in_type): Fix handling of binfos;
19789         walk basetypes.
19790         (free_lang_data): Rebuild type inheritance graph.
19792 2018-04-09  Martin Sebor  <msebor@redhat.com>
19794         * invoke.texi (-finline-small-functions): Mention other optimization
19795         options.
19796         (-findirect-inlining, -fpartial-inlining): Same.
19797         (-finline-functions-called-once): Same.
19798         (-freorder-blocks-and-partition): Same.
19800 2018-04-09  Jan Hubicka  <jh@suse.cz>
19802         PR rtl/84058
19803         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
19804         jumps; choose last target that matches the criteria (i.e.
19805         no partition changes for non-crossing jumps).
19806         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
19807         support for redirecting crossing jumps to non-crossing.
19809 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
19811         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
19812         also for naked functions.
19814 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
19816         * config/arc/arc.md (add_shift): New pattern.
19817         (add_shift2): Likewise.
19818         (sub_shift): Likewise.
19819         (sub_shift_cmp0_noout): Likewise.
19820         (compare_si_ashiftsi): Likewise.
19821         (xbfu_cmp0_noout): New combine pattern.
19822         (xbfu_cmp0"): Likewise.
19823         (movsi_set_cc_insn): Place the predicable variant first.
19824         (commutative_binary_cmp0_noout): Remove clobber.
19825         (commutative_binary_cmp0): New pattern.
19826         (noncommutative_binary_cmp0): Likewise.
19827         (noncommutative_binary_cmp0_noout): Likewise.
19828         (noncommutative_binary_comparison_result_used): Removed.
19829         (rsub_cmp0): New pattern.
19830         (rsub_cmp0_noout): Likewise.
19831         (extzvsi): Changed, keep only meaningful variants.
19832         (SQH, SEZ): New iterators.
19833         (SQH_postfix): New mode attribute.
19834         (SEZ_prefix): New code attribute.
19835         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
19836         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
19837         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
19838         of numerical value.
19839         (noncommutative_operator): Check the availability of barrel
19840         shifter option.
19842 2018-04-09  Richard Biener  <rguenther@suse.de>
19844         PR tree-optimization/85284
19845         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
19846         Only use the niter constraining form of simple_iv when the exit
19847         is always executed.
19849 2018-04-09  Tom de Vries  <tom@codesourcery.com>
19851         PR target/84041
19852         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19853         (define_expand "*memory_barrier"): New define_expand.
19854         (define_insn "memory_barrier"): New insn.
19856 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19858         PR rtl-optimization/80463
19859         PR rtl-optimization/83972
19860         PR rtl-optimization/83480
19862         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
19863         correct producer for the insn.
19864         (tidy_control_flow): Fixup seqnos in case of debug insns.
19866 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19868         PR rtl-optimization/83913
19870         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
19871         different sched-times when merging exprs.
19873 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19875         PR rtl-optimization/83962
19877         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
19878         tidy_fallthru_edge and tidy_control_flow.
19880 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19882         PR rtl-optimization/83530
19884         * sel-sched.c (force_next_insn): New global variable.
19885         (remove_insn_for_debug): When force_next_insn is true, also leave only
19886         next insn in the ready list.
19887         (sel_sched_region): When the region wasn't scheduled, make another pass
19888         over it with force_next_insn set to 1.
19890 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19892         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
19893         into tm_file.
19894         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19895         for interrupt control.
19896         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19897         functions for interrupt control.
19898         * config/nds32/nds32-intrinsic.md: Likewise.
19899         * config/nds32/nds32_intrinsic.h: Likewise.
19900         * config/nds32/nds32.h (nds32_builtins): Likewise.
19902 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19904         * config/nds32/nds32.c (nds32_init_machine_status,
19905         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
19906         strict_aligned_p field.
19907         (nds32_expand_to_rtl_hook): New function.
19908         (TARGET_EXPAND_TO_RTL_HOOK): Define.
19909         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
19911 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19912             Chung-Ju Wu  <jasonwucj@gmail.com>
19914         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
19915         * config/nds32/nds32-n7.md: New file.
19916         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
19917         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
19918         pipeline.
19919         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
19920         * config/nds32/nds32.md (pipeline_model): Add n7.
19921         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
19922         * config/nds32/pipelines.md: Include n7 settings.
19924 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19925             Chung-Ju Wu  <jasonwucj@gmail.com>
19927         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
19928         * config/nds32/nds32-e8.md: New file.
19929         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
19930         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
19931         pipeline.
19932         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
19933         * config/nds32/nds32.md (pipeline_model): Add e8.
19934         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
19935         * config/nds32/pipelines.md: Include e8 settings.
19937 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19938             Chung-Ju Wu  <jasonwucj@gmail.com>
19940         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
19941         * config/nds32/nds32-n8.md: New file.
19942         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
19943         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
19944         pipeline.
19945         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
19946         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
19947         * config/nds32/nds32.md (pipeline_model): Add n8.
19948         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
19949         * config/nds32/pipelines.md: Include n8 settings.
19951 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19952             Chung-Ju Wu  <jasonwucj@gmail.com>
19954         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
19955         * config/nds32/nds32-n9-2r1w.md: New file.
19956         * config/nds32/nds32-n9-3r2w.md: New file.
19957         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
19958         nds32_register_ports): New or modify for cpu n9.
19959         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
19960         pipeline.
19961         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
19962         * config/nds32/nds32-utils.c: New file.
19963         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
19964         TARGET_MUL_SLOW): Define.
19965         * config/nds32/nds32.md (pipeline_model): New attribute.
19966         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
19967         New options that support cpu n9.
19968         * config/nds32/pipelines.md: Include n9 settings.
19969         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
19971 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19973         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
19974         information if necessary.
19975         (output_cond_branch_compare_zero): Likewise.
19976         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
19977         (nds32_target_alignment): Refine for alignment.
19978         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
19979         (FUNCTION_BOUNDARY): Modify.
19980         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
19981         align case.
19982         * config/nds32/nds32.opt (malways-align, malign-functions): New.
19984 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19986         * config/nds32/constants.md (unspec_volatile_element): Add values for
19987         TLB operation and data prefetch.
19988         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19989         functions for TLB operation and data prefetch.
19990         * config/nds32/nds32-intrinsic.md: Likewise.
19991         * config/nds32/nds32_intrinsic.h: Likewise.
19992         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
19993         (nds32_print_operand): Likewise.
19994         * config/nds32/nds32.h (nds32_builtins): Likewise.
19996 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
19997         Andrew Pinski <pinsika@gcc.gnu.org>
19999         PR middle-end/82976
20000         * match.pd: Use constant_boolean_node of correct type instead of
20001         boolean_true_node or boolean_false_node for simplifying
20002         pointer comparisons to zero.
20004 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20006         PR tree-optimization/80021
20007         * tree.c (verify_type_variant): Make error call in verify_variant_match
20008         translatable and remove final full stop.
20010 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20012         * config/nds32/constants.md (unspec_volatile_element): Add
20013         UNSPEC_VOLATILE_EH_RETURN.
20014         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
20015         nds32_output_stack_pop): Support dwarf exception handling process.
20016         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
20017         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
20018         exception handling process.
20019         (nds32_compute_stack_frame): Likewise.
20020         (nds32_return_addr_rtx): Likewise.
20021         (nds32_initial_elimination_offset): Likewise.
20022         (nds32_expand_prologue): Likewise.
20023         (nds32_expand_epilogue): Likewise.
20024         (nds32_dynamic_chain_address): New function.
20025         * config/nds32/nds32.h (machine_function): Add fields for dwarf
20026         exception handling.
20027         (DYNAMIC_CHAIN_ADDRESS): Define.
20028         (EH_RETURN_DATA_REGNO): Define.
20029         (EH_RETURN_STACKADJ_RTX): Define.
20030         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
20031         patterns for dwarf exception handling.
20033 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20035         * config/nds32/nds32.h: Clean up obsolete macros.
20037 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20039         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20040         Add enum values for particular instructions.
20041         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20042         particular intrinsic functions.
20043         * config/nds32/nds32-intrinsic.md: Likewise.
20044         * config/nds32/nds32_intrinsic.h: Likewise.
20045         * config/nds32/nds32.h (nds32_builtins): Likewise.
20046         * config/nds32/nds32.md (type): Add pbsad and pbsada.
20047         (btst, ave): New patterns for particular instructions.
20049 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20051         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20052         Add enum values for atomic load/store and memory sync.
20053         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
20054         and memory sync.
20055         * config/nds32/nds32-intrinsic.md: Likewise.
20056         * config/nds32/nds32_intrinsic.h: Likewise.
20057         * config/nds32/nds32.h (nds32_builtins): Likewise.
20059 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20061         PR tree-optimization/85257
20062         * fold-const.c (native_encode_vector): If not all elts could fit
20063         and off is -1, return 0 rather than offset.
20064         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
20065         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
20066         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
20067         adjust buffer in native_interpret_expr call.
20069 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20071         * config/nds32/constants.md (unspec_volatile_element): Add cache
20072         control enum values.
20073         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
20074         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
20075         * config/nds32/nds32.c (nds32_cctl_names): New.
20076         (nds32_print_operand): Handle cache control register names.
20077         * config/nds32/nds32.h (nds32_builtins): New enum values.
20078         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
20079         macros.
20080         * config/nds32/nds32.md (type): Add mmu.
20081         * config/nds32/pipelines.md (simple_insn): Add mmu.
20083 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20085         * config/nds32/nds32.md (type): Remove call.
20086         * config/nds32/pipelines.md (simple_insn): Likewise.
20088 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20090         * config/nds32/constants.md (unspec_volatile_element): Add
20091         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
20092         UNSPEC_VOLATILE_FMFCFG.
20093         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
20094         description for fmfcfg and fmfcsr.
20095         (bdesc_1arg): Add fmtcsr.
20096         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
20097         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
20098         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
20099         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
20100         unspec_fmfcfg): New patterns.
20101         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
20102         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
20103         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
20104         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
20105         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
20106         __nds32__fmfcfg): Define.
20108 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20110         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
20111         intrinsic register names.
20112         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
20113         intrinsic register enum values and macros.
20115 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20117         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
20118         for load/store addressing form.
20119         (nds32_print_operand_address): Likewise.
20121 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
20123         PR target/85196
20124         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
20125         based on LABEL_REF.  Remove useless assertion.
20126         (pic_address_needs_scratch): Fix formatting.
20127         (sparc_legitimize_pic_address): Minor tweaks.
20128         (sparc_delegitimize_address): Adjust assertion accordingly.
20129         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
20130         into symbolic_operand.
20131         (movsi_high_pic_label_ref): Likewise.
20132         (movsi_lo_sum_pic_label_ref): Likewise.
20133         (movdi_pic_label_ref): Likewise.
20134         (movdi_high_pic_label_ref): Likewise.
20135         (movdi_lo_sum_pic_label_ref): Likewise.
20137 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
20139         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
20140         custom LIB_SPEC setup.
20142 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
20143             Kito Cheng  <kito.cheng@gmail.com>
20145         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
20146         * config/riscv/freebsd.h: New.
20148 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20150         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
20151         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
20152         file.
20154 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20155             Kito Cheng  <kito.cheng@gmail.com>
20157         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
20158         nds32_output_call, nds32_symbol_binds_local_p): New functions.
20159         * config/nds32/nds32-protos.h (nds32_output_call,
20160         nds32_output_return): Declare.
20161         * config/nds32/nds32.md: Refine all the call and return patterns.
20163 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20165         PR debug/85252
20166         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
20167         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
20169         PR rtl-optimization/84872
20170         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
20171         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
20172         EDGE_CROSSING edge.
20174 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
20176         * expr.c (copy_blkmode_to_reg): Revert 254862.
20177         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
20179 2018-04-06  Richard Biener  <rguenther@suse.de>
20181         PR middle-end/85244
20182         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
20183         after seeing a component reference with an adjacent field.  Treat
20184         refs to arrays at struct end of external decls similar to
20185         refs to unconstrained commons.
20187 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20189         PR sanitizer/85213
20190         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
20191         look through SAVE_EXPRs with non-side-effects argument.  Adjust
20192         recursive calls.
20193         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
20194         save_p here.
20196 2018-04-06  Richard Biener  <rguenther@suse.de>
20198         PR middle-end/85180
20199         * alias.c (find_base_term): New wrapper around find_base_term
20200         unwinding CSELIB_VAL_PTR changes.
20201         (find_base_term): Do not restore CSELIB_VAL_PTR during the
20202         recursion.
20204 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20206         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
20207         instructions.
20208         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
20209         constant definitions.
20210         ("nop"): lr 0,0 -> nopr r0
20211         ("nop_lr0", "nop_lr1"): New insn definitions.
20213 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20215         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
20216         NDS32_V3PUSH_AVAILABLE_P macro.
20218 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
20219             Chung-Ju Wu  <jasonwucj@gmail.com>
20221         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
20222         (nds32*-*-*): Add float and fpu_config into supported_defaults.
20223         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
20224         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
20225         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
20226         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
20227         * config/nds32/constraints.md: New constraints and checking for hard
20228         float configuration.
20229         * config/nds32/iterators.md: New mode iterator and attribute for hard
20230         float configuration.
20231         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
20232         patterns.
20233         * config/nds32/nds32-fpu.md: New file.
20234         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
20235         deal with hard float code generation.
20236         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
20237         ARCH_V3S.
20238         (abi_type, float_reg_number): New enum type.
20239         * config/nds32/nds32-predicates.c: New predicates for hard float.
20240         * config/nds32/nds32-protos.h: Declare functions for hard float.
20241         * config/nds32/nds32.c: Implementation for hard float configuration.
20242         * config/nds32/nds32.h: Definitions for hard float configuration.
20243         * config/nds32/nds32.md: Include hard float machine description and
20244         modify patterns for hard float configuration.
20245         * config/nds32/nds32.opt: New options for hard float configuration.
20246         * config/nds32/predicates.md: New predicates for hard float
20247         configuration.
20249 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20251         * common/config/nds32/nds32-common.c
20252         (nds32_option_optimization_table): Enable -mreleax-hint by default.
20254 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
20256         PR middle-end/85195
20257         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
20258         CONSTRUCTOR_ELT (ctor, ...)->value.
20260 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
20262         PR target/85193
20263         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
20265 2018-04-05  Tom de Vries  <tom@codesourcery.com>
20267         PR target/85204
20268         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
20269         cond jump.
20271 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20272             Kito Cheng  <kito.cheng@gmail.com>
20274         * config/nds32/constraints.md (U33): Fine-tune checking condition.
20275         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
20276         * config/nds32/nds32.h (nds32_16bit_address_type): Add
20277         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
20279 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20280             Kito Cheng  <kito.cheng@gmail.com>
20282         * config/nds32/constraints.md (Ufe): New memory constraint.
20283         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
20284         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
20285         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
20286         operands.
20287         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
20288         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
20290 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20292         * config/nds32/nds32.md: Use optimize_size in the condition for
20293         alu-shift instructions.
20295 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20297         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
20299 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20301         * config/nds32/nds32.md (negsi2): Refine pattern.
20303 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20304             Chung-Ju Wu  <jasonwucj@gmail.com>
20306         * config/nds32/iterators.md (shift_rotate): New code iterator.
20307         (shift): New code attribute.
20308         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
20309         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
20310         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
20311         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
20312         bit-wise operations.
20313         (andsi3, *andsi3): Ditto.
20314         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
20315         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
20316         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
20317         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
20318         nds32_ior_operand, nds32_xor_operand): New predicates.
20320 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20322         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
20323         (addsi3, subsi3): ... this.
20325 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20327         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
20329 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20331         * config/nds32/nds32.md: Adjust indention.
20333 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20335         * config/nds32/nds32.md (feature): New attribute.
20337 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20339         * config/nds32/nds32.md (subtype): New attribute.
20341 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20343         PR target/85203
20344         * config/arm/arm-builtins.c (arm_expand_builtin): Change
20345         expansion to perform a bitwise AND of the argument followed by a
20346         boolean negation of the result.
20348 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
20350         PR rtl-optimization/84878
20351         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
20352         the basic block.  Assert the use reference is not artificial and that
20353         it has an associated insn.
20355 2018-04-04  Michael Matz  <matz@suse.de>
20357         * builtins.c (compute_objsize): Pass correct operand
20358         to array_at_struct_end_p.
20360 2018-04-04  Richard Biener  <rguenther@suse.de>
20362         PR lto/85176
20363         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
20364         from contexts for DINFO_LEVEL_TERSE and below.
20366 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20368         * config/nds32/nds32-doubleword.md (move_<mode>): Require
20369         resiter_operand condition.
20370         * config/nds32/nds32.md (*move<mode>): Ditto.
20372 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20373             Monk Chiang  <sh.chiang04@gmail.com>
20375         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
20377 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20379         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
20381 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20382             Kito Cheng  <kito.cheng@gmail.com>
20384         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
20385         nds32_cond_code_str, output_cond_branch,
20386         output_cond_branch_compare_zero, nds32_expand_cbranch,
20387         nds32_expand_cstore, nds32_expand_movcc,
20388         nds32_output_cbranchsi4_equality_zero,
20389         nds32_output_cbranchsi4_equality_reg,
20390         nds32_output_cbranchsi4_equality_reg_or_const_int,
20391         nds32_output_cbranchsi4_greater_less_zero: New functions.
20392         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
20393         nds32_expand_cstore, nds32_expand_movcc,
20394         nds32_output_cbranchsi4_equality_zero,
20395         nds32_output_cbranchsi4_equality_reg,
20396         nds32_output_cbranchsi4_equality_reg_or_const_int,
20397         nds32_output_cbranchsi4_greater_less_zero): Declare.
20398         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20399         nds32_rimm11s_operand): New predicates.
20400         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20401         * config/nds32/nds32.md: Rewrite all the branch and conditional move
20402         patterns.
20404 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20406         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20407         * config/nds32/nds32.md: Ditto.
20408         * config/nds32/pipelines.md: Ditto.
20410 2018-04-04  Richard Biener  <rguenther@suse.de>
20412         PR tree-optimization/85168
20413         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20414         propagating abnormals.
20416 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20418         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20420 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20421             Kito Cheng  <kito.cheng@gmail.com>
20423         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20424         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20425         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20426         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20427         * config/nds32/nds32.md (sibcall_internal): New.
20428         (sibcall_register): Remove.
20429         (sibcall_immediate): Remove.
20430         (sibcall_value_internal): New.
20431         (sibcall_value_register): Remove.
20432         (sibcall_value_immediate): Remove.
20433         * config/nds32/predicates.md (nds32_general_register_operand): New.
20434         (nds32_call_address_operand): New.
20436 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20438         PR rtl-optimization/85167
20439         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20440         bb_defs if *split_p, instead preinitialize it to NULL.
20442         PR tree-optimization/85156
20443         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20444         evaluating the argument multiple times.
20446 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
20448         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20449         than vector.
20450         (_mm_cvtpd_ps): Likewise.
20451         (_mm_cvttpd_epi32): Likewise.
20452         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20453         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20454         vector, pixel, and bool following altivec.h include.
20456 2018-04-03  Martin Sebor  <msebor@redhat.com>
20458         * doc/extend.texi (Common Function Attributes): Clarify.
20459         (const attribute): Likewise.
20460         (pure attribute): Likewise.
20462 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20464         PR target/85169
20465         * config/i386/i386.c (ix86_expand_vector_set): Use
20466         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
20468 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
20470         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20471         instructions when changing rounding bits to preserve precision bits
20472         in the x87 control word.
20474 2018-04-03  Martin Liska  <mliska@suse.cz>
20476         PR tree-optimization/82491
20477         * rtl.h (strip_offset_and_add): Replace += suboffset with
20478         poly_uint64 () + suboffset.
20480 2018-03-29  Martin Liska  <mliska@suse.cz>
20481             Martin Jambor  <mjambor@suse.cz>
20483         PR ipa/84947
20484         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20485         param_type is not an integral or pointer type.
20487 2018-04-03  Richard Biener  <rguenther@suse.de>
20489         * sese.h (recompute_all_dominators): Remove.
20491 2018-04-02  Martin Sebor  <msebor@redhat.com>
20493         * doc/invoke.texi (-Wrestrict): Fix typos.
20495 2018-04-02  Jim Wilson  <jimw@sifive.com>
20497         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20498         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20499         (<optab>di3, <optab>si3_extend): Likewise.
20500         (<optab>si3_mask, <optab>si3_mask_1): New.
20501         (<optab>di3_mask, <optab>di3_mask_1): New.
20502         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20503         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20504         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20506 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
20508         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20509         example.
20511 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20513         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20514         (nds32_canonicalize_comparison): New function.
20516 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20517             Kito Cheng  <kito.cheng@gmail.com>
20518             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20520         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20521         * config/nds32/constants.md (unspec_volatile_element): Add
20522         UNSPEC_VOLATILE_RELAX_GROUP.
20523         * config/nds32/nds32-relax-opt.c: New file.
20524         * config/nds32/nds32-predicates.c
20525         (nds32_symbol_load_store_p): New function.
20526         * config/nds32/nds32-protos.h
20527         (nds32_symbol_load_store_p): Declare function.
20528         (make_pass_nds32_relax_opt): Declare new rtl pass function.
20529         * config/nds32/nds32.c
20530         (nds32_register_pass): New function to register pass.
20531         (nds32_register_passes): New function to register passes.
20532         * config/nds32/nds32.md (relax_group): New pattern.
20533         * config/nds32/nds32.opt (mrelax-hint): New option.
20534         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20536 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
20538         * config/nds32/t-nds32: Modify files dependency.
20540 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20542         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20543         (PROFILE_HOOK): Define its implementation.
20545 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20547         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20548         type and 32-bit size.
20550 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
20552         PR middle-end/85090
20553         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20554         (V_128_256): New mode iterator.
20555         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20556         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20557         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20558         of V.
20559         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20560         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20562 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
20564         PR target/83315
20565         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20566         NaN inputs correctly.
20568 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
20570         PR target/80546
20571         * config/rs6000/vsx.md (??r): New mode attribute.
20572         (*vsx_mov<mode>_64bit): Use it.
20573         (*vsx_mov<mode>_32bit): Likewise.
20575 2018-03-30  Martin Sebor  <msebor@redhat.com>
20577         PR tree-optimization/84818
20578         * builtins.c (check_access): Use warning_n.
20580 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20582         PR target/83822
20583         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20584         condition.
20585         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20586         condition.
20588 2018-03-30  Julia Koval  <julia.koval@intel.com>
20590         PR target/84413
20591         * x86-tune.def (movx, partial_reg_dependency): Enable for
20592         m_SKYLAKE_AVX512.
20594 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
20596         PR inline-asm/84985
20597         * lra-constraints.c (process_alt_operands): Move setting
20598         this_alternative_matches below.
20600 2018-03-29  Martin Liska  <mliska@suse.cz>
20602         PR lto/84995.
20603         * doc/invoke.texi: Document how LTO works with debug info.
20604         Describe auto-load support of binutils.  Mention 'x86-64'
20605         as valid option value of -march option.
20607 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
20609         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20611         PR c/85094
20612         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20613         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20614         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20615         checking.
20617 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20619         PR target/84912
20620         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20621         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20622         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20623         for RS6000_BTM_POWERPC64.
20624         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20625         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20626         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20627         definition.
20628         (DIVDE): Use it.
20629         (DIVDEU): Likewise.
20631 2018-03-28  Carl Love  <cel@us.ibm.com>
20633         Revert
20634         2017-09-27  Carl Love  <cel@us.ibm.com>
20636         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20637         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20638         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20639         fctiw instruction.
20641 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20643         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20644         instead of __vector bool.
20645         (_mm_max_pu8): Likewise.
20646         (_mm_min_pi16): Likewise.
20648 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20650         PR target/84912
20651         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20652         (DIVWEUO): Likewise.
20653         (DIVDEO): Likewise.
20654         (DIVDEUO): Likewise.
20655         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20656         DIVWEUO and DIVDEUO.
20657         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20658         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20659         (div_extend): Likewise.
20660         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20661         builtin function.
20662         (__builtin_divweuo): Likewise.
20663         (__builtin_divdeo): Likewise.
20664         (__builtin_divdeuo): Likewise.
20666 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20668         PR target/85095
20669         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20670         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20672         PR tree-optimization/82004
20673         * gimple-match-head.c (optimize_pow_to_exp): New function.
20674         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20675         Don't fold to exp if optimize_pow_to_exp is false.
20677 2018-03-28  Martin Liska  <mliska@suse.cz>
20679         PR other/84819
20680         * calls.c (initialize_argument_information): Fix trailing space.
20681         * common.opt: Fix typo and provide better explanation for
20682         -fsanitize-coverage option.
20683         * config/i386/i386.opt: Fix typo.
20685 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20686             Martin Liska  <mliska@suse.cz>
20688         PR sanitizer/85081
20689         * gimplify.c (asan_poison_variable): Don't do the check for
20690         gimplify_omp_ctxp here.
20691         (gimplify_decl_expr): Do it here.
20692         (gimplify_target_expr): Likewise.
20694 2018-03-28  Martin Liska  <mliska@suse.cz>
20696         PR target/84988
20697         * config/i386/i386.c (ix86_function_arg_advance): Do not call
20698         chkp_type_bounds_count if MPX is not enabled.
20700 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20702         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20704 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
20706         PR target/84914
20707         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20708         function to create the function decl for complex long double
20709         multiply and divide for -mabi=ieeelongdouble.
20710         (init_float128_ieee): Call it.
20712 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
20714         PR target/85044
20715         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20716         -fcf-protection=branch -mibt.
20717         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20719 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20721         PR target/81863
20722         * config/arm/arm.c (arm_valid_symbolic_address): Handle
20723         arm_word_relocations.
20725 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
20727         PR target/85056
20728         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20729         extern array declarations.
20731 2018-03-27  Richard Biener  <rguenther@suse.de>
20733         PR middle-end/84067
20734         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20735         explicit single_use checks.
20737 2018-03-27  Richard Biener  <rguenther@suse.de>
20739         PR tree-optimization/85082
20740         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20741         Valueize the VUSE.
20743 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20745         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20746         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20747         Turn on fasynchronous-unwind-tables and funwind-tables.
20749 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
20751         PR target/85073
20752         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20753         (*bmi_blsr_<mode>_ccz): Ditto.
20755 2018-03-26  Tom de Vries  <tom@codesourcery.com>
20757         PR tree-optimization/85063
20758         * omp-general.c (offloading_function_p): New function.  Factor out
20759         of ...
20760         * omp-offload.c (pass_omp_target_link::gate): ... here.
20761         * omp-general.h (offloading_function_p): Declare.
20762         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20763         with attribute omp declare target for offloading functions.
20765 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
20767         PR tree-optimization/84005
20768         * tree-data-ref.h (get_base_for_alignment): Declare.
20769         * tree-data-ref.c (get_base_for_alignment_1): New function.
20770         (get_base_for_alignment): Likewise.
20771         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20772         get_base_for_alignment to find a suitable base object, instead
20773         of always using drb->base_address.
20775 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20777         PR inline-asm/85022
20778         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
20779         known size by default.
20781 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
20783         PR inline-asm/85030
20784         * lra-constraints.c (process_alt_operands): Don't match BLKmode
20785         and non BLKmode operands.
20787 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20789         PR target/85026
20790         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
20791         Clean up attributes.
20793 2018-03-23  Richard Biener  <rguenther@suse.de>
20795         PR debug/85020
20796         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
20797         we are going to emit early debug for LTO.
20799 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20801         PR inline-asm/85034
20802         * function.c (match_asm_constraints_1): Don't optimize if input
20803         doesn't satisfy general_operand predicate for output's mode.
20805         PR inline-asm/85022
20806         * alias.c (write_dependence_p): Don't require for x_canonicalized
20807         non-VOIDmode if x has VOIDmode.
20809         PR sanitizer/85029
20810         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
20811         just don't try to optimize it rather than assert it never happens.
20813 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20815         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
20816         macro expansions for definition of ST_INTERNAL_<mode> and
20817         LD_INTERNAL_<mode> builtins.
20818         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
20819         Remove prototype.
20820         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
20821         function.
20822         (altivec_expand_st_builtin): Likewise.
20823         (altivec_expand_builtin): Remove calls to deleted functions.
20824         (rs6000_address_for_altivec): Delete this function.
20825         * config/rs6000/vector.md: Remove expands for
20826         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
20828 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
20830         PR target/84826
20831         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
20832         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
20833         re-computing once computed.
20834         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
20835         (arm_init_machine_status): Initialize
20836         machine->static_chain_stack_bytes.
20838 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20840         PR target/84760
20841         * doc/extend.texi: Add four new prototypes for vec_ld.
20842         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
20843         definitions for more logical presentation.
20844         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
20845         entries for V1TI variants of __builtin_altivec_ld builtin.
20846         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
20847         handling of V1TI variant of LVX icode pattern.
20848         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
20849         (rs6000_gimple_fold_builtin): Likewise.
20850         (altivec_init_builtins): Add code to define
20851         __builtin_altivec_lvx_v1ti function.
20853 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20855         PR inline-asm/84941
20856         * function.c (match_asm_constraints_1): Don't do the optimization
20857         if input isn't a REG, SUBREG, MEM or constant.
20859 2018-03-22  Tom de Vries  <tom@codesourcery.com>
20861         PR tree-optimization/84956
20862         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
20863         bb_has_abnormal_pred.
20865 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20867         PR sanitizer/85018
20868         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
20869         DECL_INITIAL (decl) to decl at the end.
20870         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
20871         adjust the comment.
20873 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
20875         * doc/extend.texi (__builtin_tgmath): Document when complex
20876         integer types are treated as _Complex _Float64.
20878 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20880         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
20882 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20884         PR tree-optimization/84960
20885         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
20886         if it is ENTRY block, move them into single succ of ENTRY in that case.
20888 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
20890         PR tree-optimization/84811
20891         * poly-int.h (poly_span_traits): Remove the T3 parameter and
20892         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
20893         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
20894         (known_subrange_p): Update accordingly.  Cast each value involved
20895         in the size comparison, rather than casting the result of the
20896         subtraction.
20898 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20900         PR tree-optimization/84982
20901         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
20902         by flipping the least significant bit rather than all bits from
20903         bitpos to bitpos + bitsize - 1.
20905 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20907         * doc/extend.texi (Deprecated Features): Remove mention of
20908         long-deleted deprecations.
20910 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20912         PR jit/84288
20913         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
20914         * configure: Regenerate.
20916 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20918         PR tree-optimization/83126
20919         * tree-parloops.c (num_phis): New function.
20920         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
20922 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20924         * doc/extend.texi (Deprecated Features): Update deprecated flags,
20925         mention anon-struct/union members and trailing attributes.
20927 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
20929         PR tree-optimization/84969
20930         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
20931         builtin memset partitions if they set different rhs values.
20933 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20935         PR rtl-optimization/84989
20936         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
20937         VEC_DUPLICATE with scalar result mode.
20939 2018-03-21  Martin Liska  <mliska@suse.cz>
20941         PR ipa/84963
20942         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
20943         not intended return statement.
20945 2018-03-21  Martin Liska  <mliska@suse.cz>
20947         PR target/84988
20948         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
20949         (chkp_find_bound_slots_1): Limit number of iterations.
20951 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
20953         PR target/84838
20954         * Minor grammar fixes for x86 options.
20956 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20958         PR debug/84875
20959         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
20960         holding REG_CFA_RESTORE notes, instead turn them into a USE.
20962 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
20964         PR target/83789
20965         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
20966         (altivec_lvx_<mode>_1op): Likewise.
20967         (altivec_stvx_<mode>_2op): Likewise.
20968         (altivec_stvx_<mode>_1op): Likewise.
20969         (altivec_lvx_<VM2:mode>): New define_expand.
20970         (altivec_stvx_<VM2:mode>): Likewise.
20971         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
20972         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20973         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
20974         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20975         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
20976         (rs6000_gen_lvx): Likewise.
20977         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
20978         (altivec_expand_stv_builtin): Likewise.
20979         (altivec_expand_builtin): Likewise.
20980         * config/rs6000/vector.md: Likewise.
20982 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20984         PR target/82518
20985         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
20986         BYTES_BIG_ENDIAN.
20988 2018-03-20  Richard Biener  <rguenther@suse.de>
20990         PR target/84986
20991         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
20992         sign-conversions as zero, fall back to standard scalar_stmt
20993         cost for the rest.
20995 2018-03-20  Martin Liska  <mliska@suse.cz>
20997         PR ipa/84825
20998         * predict.c (rebuild_frequencies): Handle case when we have
20999         PROFILE_ABSENT, but flag_guess_branch_prob is false.
21001 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21003         PR target/84990
21004         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
21005         flag_section_anchors.
21006         * varasm.c (use_blocks_for_decl_p): Remove hack for
21007         dw2_force_const_mem.
21009         PR target/84845
21010         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
21011         to ...
21012         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
21013         be created, use lowpart_subreg of operands[0] rather than operands[0]
21014         itself.
21015         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
21016         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
21017         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
21018         and n constraint instead of aarch64_shift_imm_di and Usd.
21019         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
21020         (*aarch64_<optab>_reg_minus<mode>3): ... this.
21022 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
21024         PR target/82989
21025         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
21026         to favor GPR over NEON registers.
21027         (<shift>di3_neon): Likewise.
21029 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21031         PR target/84952
21032         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
21033         (nvptx_process_pars): Emit bar.sync asap and alap.
21035 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21037         PR target/84954
21038         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
21039         seen_label if seen_label is already set.
21041 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21043         PR target/84945
21044         * config/i386/i386.c (fold_builtin_cpu): For features above 31
21045         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
21046         Use 1U instead of 1.  Formatting fixes.
21048         PR c/84953
21049         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
21050         instead of TREE_TYPE (s1) for the return value.
21052 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
21054         PR tree-optimization/84946
21055         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
21056         bitsize + bitsize in poly_uint64 rather than poly_int64.
21058         PR sanitizer/78651
21059         * dwarf2asm.c: Include fold-const.c.
21060         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
21061         of decl rather than decl itself.
21063         PR rtl-optimization/84643
21064         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
21066 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
21068         PR sanitizer/78651
21069         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
21070         calling assemble_variable.
21072 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
21074         PR target/81647
21075         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
21076         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
21078 2018-03-19  Jim Wilson  <jimw@sifive.com>
21080         PR bootstrap/84856
21081         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
21082         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
21083         (riscv_first_stack_step): Likewise.
21084         (riscv_option_override): Use STACK_BOUNDARY instead of
21085         MIN_STACK_BOUNDARY.
21086         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
21087         MIN_STACK_BOUNDARY.
21088         (BIGGEST_ALIGNMENT): Set to 128.
21089         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
21090         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
21091         STACK_BOUNDARY.
21093 2018-03-19  Richard Biener  <rguenther@suse.de>
21095         PR tree-optimization/84933
21096         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
21097         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
21099 2018-03-19  Richard Biener  <rguenther@suse.de>
21101         PR tree-optimization/84859
21102         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
21103         (cond_if_else_store_replacement): Perform sinking operation on
21104         single-store BBs regardless of MAX_STORES_TO_SINK setting.
21105         Generalize what a BB with a single eligible store is.
21107 2018-03-19  Richard Biener  <rguenther@suse.de>
21109         PR tree-optimization/84929
21110         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
21111         chrec_is_positive against non-chrec arg.
21113 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
21115         PR target/84711
21116         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
21118 2018-03-18  Martin Liska  <mliska@suse.cz>
21120         PR rtl-optimization/84635
21121         * regrename.c (build_def_use): Use matches_mode only when
21122         matches >= 0.
21124 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
21126         PR tree-optimization/84913
21127         * tree-vect-loop.c (vectorizable_reduction): Don't try to
21128         vectorize chains of COND_EXPRs.
21130 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21132         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
21134 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21136         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
21138 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21140         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21142 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21143             Kito Cheng  <kito.cheng@gmail.com>
21145         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
21146         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
21147         (nds32_adjust_reg_alloc_order): New function.
21148         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
21150 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
21152         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
21153         nds32_print_operand, nds32_print_operand_address): Use
21154         HOST_WIDE_INT_PRINT_DEC instead.
21156 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21158         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
21160 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
21162         PR target/84902
21163         * config/i386/i386.c (initial_ix86_tune_features,
21164         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
21165         unsigned long long.
21166         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
21167         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
21168         rather than 1u << ix86_tune.  Formatting fix.
21169         (ix86_option_override_internal): Change ix86_arch_mask from
21170         unsigned int to unsigned HOST_WIDE_INT, initialize to
21171         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
21172         (ix86_function_specific_restore): Likewise.
21174 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21176         PR target/84899
21177         * postreload.c (reload_combine_recognize_pattern): Perform
21178         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
21179         truncate_int_for_mode the result for the destination's mode.
21181         PR c/84909
21182         * hsa-gen.c (mem_type_for_type): Fix comment typo.
21183         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
21184         Likewise.
21185         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21186         Likewise.
21188 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
21190         PR target/84876
21191         * lra-assigns.c (lra_split_hard_reg_for): Don't use
21192         regno_allocno_class_array and sorted_pseudos.
21193         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
21194         insns where regno is used.
21196 2018-03-16  Martin Liska  <mliska@suse.cz>
21198         PR ipa/84833
21199         * multiple_target.c (create_dispatcher_calls): Redirect
21200         reference in the symbol table.
21202 2018-03-16  Martin Liska  <mliska@suse.cz>
21204         PR ipa/84722
21205         * multiple_target.c (create_dispatcher_calls): Redirect also
21206         an alias.
21208 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21210         PR c++/79937
21211         PR c++/82410
21212         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
21213         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
21214         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
21216 2018-03-16  Julia Koval  <julia.koval@intel.com>
21218         * doc/invoke.texi (Skylake Server): Add CLWB.
21219         Cannonlake): Remove CLWB.
21221 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21223         PR tree-optimization/84841
21224         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
21225         1 << 3.
21226         (FLOAT_ONE_CONST_TYPE): Define.
21227         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
21228         (sort_by_operand_rank): Put entries with higher constant_type last
21229         rather than first to match comments.
21231 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
21233         * config/nios2/nios2.md (movsi_internal): Fix thinko in
21234         split predicate.
21236 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21238         PR c++/79085
21239         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
21240         check and use address of target always.
21242 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
21244         PR target/84574
21245         * config/i386/i386.c (indirect_thunk_needed): Update comments.
21246         (indirect_thunk_bnd_needed): Likewise.
21247         (indirect_thunks_used): Likewise.
21248         (indirect_thunks_bnd_used): Likewise.
21249         (indirect_return_needed): New.
21250         (indirect_return_bnd_needed): Likewise.
21251         (output_indirect_thunk_function): Add a bool argument for
21252         function return.
21253         (output_indirect_thunk_function): Don't generate alias for
21254         function return thunk.
21255         (ix86_code_end): Call output_indirect_thunk_function to generate
21256         function return thunks.
21257         (ix86_output_function_return): Set indirect_return_bnd_needed
21258         and indirect_return_needed instead of indirect_thunk_bnd_needed
21259         and indirect_thunk_needed.
21261 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
21263         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
21264         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
21265         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
21267 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
21268             Paul Hua <paul.hua.gm@gmail.com>
21270         PR c/84852
21271         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
21273 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
21275         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
21276         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
21277         resp. SFmode cases.
21279 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
21281         PR target/84711
21282         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
21283         instead of GET_MODE_SIZE when comparing Units.
21285 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
21287         PR target/68256
21288         * varasm.c (hash_section): Return an unchangeble hash value
21289         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
21290         Return !aarch64_can_use_per_function_literal_pools_p ().
21292 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21294         PR target/84860
21295         * optabs.c (emit_conditional_move): Pass address of cmode's copy
21296         rather than address of cmode as last argument to prepare_cmp_insn.
21298 2018-03-15  Julia Koval  <julia.koval@intel.com>
21300         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
21301         F_AVX512VNNI, F_AVX512BITALG): New.
21303 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
21305         PR target/83451
21306         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
21307         insn for floating-point loads and stores.
21309 2018-03-14  Carl Love  <cel@us.ibm.com>
21311         * config/rs6000/rs6000-c.c: Add macro definitions for
21312         ALTIVEC_BUILTIN_VEC_PERMXOR.
21313         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
21314         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
21315         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
21316         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
21317         UNSPEC_VPERMXOR.
21318         * config/doc/extend.texi: Add prototypes for vec_permxor.
21320 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
21322         PR c/84852
21323         * diagnostic-show-locus.c (class layout_point): Convert m_line
21324         from int to linenum_type.
21325         (line_span::comparator): Use linenum "compare" function when
21326         comparing line numbers.
21327         (test_line_span): New function.
21328         (layout_range::contains_point): Convert param "row" from int to
21329         linenum_type.
21330         (layout_range::intersects_line_p): Likewise.
21331         (layout::will_show_line_p): Likewise.
21332         (layout::print_source_line): Likewise.
21333         (layout::should_print_annotation_line_p): Likewise.
21334         (layout::print_annotation_line): Likewise.
21335         (layout::print_leading_fixits): Likewise.
21336         (layout::annotation_line_showed_range_p): Likewise.
21337         (struct line_corrections): Likewise for field m_row.
21338         (line_corrections::line_corrections): Likewise for param "row".
21339         (layout::print_trailing_fixits): Likewise.
21340         (layout::get_state_at_point): Likewise.
21341         (layout::get_x_bound_for_row): Likewise.
21342         (layout::print_line): Likewise.
21343         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
21344         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
21345         * input.c (selftest::test_linenum_comparisons): New function.
21346         (selftest::input_c_tests): Call it.
21347         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
21348         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
21349         * selftest.h (ASSERT_GT): New macro.
21350         (ASSERT_GT_AT): New macro.
21351         (ASSERT_LT): New macro.
21352         (ASSERT_LT_AT): New macro.
21354 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
21356         PR rtl-optimization/84780
21357         * combine.c (distribute_links): Don't make a link based on pc_rtx.
21359 2018-03-14  Martin Liska  <mliska@suse.cz>
21361         * tree.c (record_node_allocation_statistics): Use
21362         get_stats_node_kind.
21363         (get_stats_node_kind): New function extracted from
21364         record_node_allocation_statistics.
21365         (free_node): Use get_stats_node_kind.
21367 2018-03-14  Richard Biener  <rguenther@suse.de>
21369         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
21370         that the value-set of ANTIC_IN doesn't grow.
21372         Revert
21373         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21374         member.
21375         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21376         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21378 2018-03-14  Julia Koval  <julia.koval@intel.com>
21380         * config.gcc (icelake-client, icelake-server): New.
21381         (icelake): Remove.
21382         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
21383         (initial_ix86_arch_features): Ditto.
21384         (PTA_SKYLAKE): Add SGX.
21385         (PTA_ICELAKE): Remove.
21386         (PTA_ICELAKE_CLIENT): New.
21387         (PTA_ICELAKE_SERVER): New.
21388         (ix86_option_override_internal): Split up icelake on icelake client and
21389         icelake server.
21390         (get_builtin_code_for_version): Ditto.
21391         (fold_builtin_cpu): Ditto.
21392         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
21393         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
21394         * config/i386/i386.h (processor_type): Ditto.
21395         * doc/invoke.texi: Ditto.
21397 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
21399         PR sanitizer/83392
21400         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21401         INTEGER_CST offset, add it together with bitpos / 8 and
21402         sign extend based on POINTER_SIZE.
21404         PR target/84844
21405         Revert
21406         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
21408         PR target/78090
21409         * config/i386/constraints.md (Yc): New register constraint.
21410         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21411         Use Yc constraint for alternative 2 of operand 0.  Remove
21412         preferred_for_speed attribute.
21414 2018-03-14  Richard Biener  <rguenther@suse.de>
21416         PR tree-optimization/84830
21417         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21418         with the old one to avoid oscillations.
21420 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
21422         PR target/83712
21423         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21424         pseudos.
21425         (assign_by_spills): Return a flag of reload assignment failure.
21426         Do not process the reload assignment failures.  Do not spill other
21427         reload pseudos if they has the same reg class.  Update n if
21428         necessary.
21429         (lra_assign): Add a return arg.  Set up from the result of
21430         assign_by_spills call.
21431         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21432         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21433         usage_insns if it is not NULL.
21434         (spill_hard_reg_in_range): New function.
21435         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21436         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21437         function prototypes.
21438         (lra_assign): Change prototype.
21439         * lra.c (lra): Add code to deal with fails by splitting hard reg
21440         live ranges.
21442 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
21444         * config/riscv/riscv.opt (mrelax): New option.
21445         * config/riscv/riscv.c (riscv_file_start): Emit ".option
21446         "norelax" when riscv_mrelax is disabled.
21447         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21449 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21451         PR target/84743
21452         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21453         reassociation for int modes.
21455 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21457         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21458         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21459         for big-endian.
21460         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21461         * config/aarch64/aarch64-sve.md
21462         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21463         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21464         (*extend<mode><Vwide>2): Rename to...
21465         (aarch64_sve_extend<mode><Vwide>2): ...this.
21466         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21467         renaming the old pattern to...
21468         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
21469         unsigned packs.
21470         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21471         define_expand, renaming the old pattern to...
21472         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21473         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21474         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21475         account when deciding which SVE instruction the optab should use.
21476         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21478 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21480         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21481         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21482         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21483         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21484         (tlsdesc_small_<mode>): Turn a define_expand and use
21485         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
21486         (tlsdesc_small_advsimd_<mode>): ...this.
21487         (tlsdesc_small_sve_<mode>): New pattern.
21489 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21491         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21492         (UNSPEC_UMUL_HIGHPART): New constants.
21493         (MUL_HIGHPART): New int iteraor.
21494         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21495         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21496         define_expand.
21497         (*<su>mul<mode>3_highpart): New define_insn.
21499 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21501         PR lto/84805
21502         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21503         incomplete types.
21505 2018-03-13  Martin Liska  <mliska@suse.cz>
21507         PR ipa/84658.
21508         * (sem_item_optimizer::sem_item_optimizer): Initialize new
21509         vector.
21510         (sem_item_optimizer::~sem_item_optimizer): Release it.
21511         (sem_item_optimizer::merge_classes): Register variable aliases.
21512         (sem_item_optimizer::fixup_pt_set): New function.
21513         (sem_item_optimizer::fixup_points_to_sets): Likewise.
21514         * ipa-icf.h: Declare new variables and functions.
21516 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
21518         PR middle-end/84834
21519         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21520         integer_pow2p@2 and test integer_pow2p in condition.
21521         (A < 0 ? C : 0): Similarly for @1.
21523         PR middle-end/84831
21524         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21525         characters starting at p contain '\0' character, don't look beyond
21526         that.
21528         PR target/84827
21529         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21530         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21532         PR target/84828
21533         * reg-stack.c (change_stack): Change update_end var from int to
21534         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21535         also call set_block_for_insn on the newly added insns and rescan.
21537         PR target/84786
21538         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21539         on the last operand.
21541         PR c++/84704
21542         * tree.c (stabilize_reference_1): Return save_expr (e) for
21543         STATEMENT_LIST even if it doesn't have side-effects.
21545 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
21547         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21549 2018-03-12  Renlin Li  <renlin.li@arm.com>
21551         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21552         aarch64_output_scalar_simd_mov_immediate.
21554 2018-03-12  Martin Sebor  <msebor@redhat.com>
21556         PR tree-optimization/83456
21557         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21558         for perfectly overlapping calls to memcpy.
21559         (gimple_fold_builtin_memory_chk): Same.
21560         (gimple_fold_builtin_strcpy): Handle no-warning.
21561         (gimple_fold_builtin_stxcpy_chk): Same.
21562         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21564 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21566         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21567         parameter.  Use it for SFmode.
21568         (rs6000_function_arg_advance_1): Adjust.
21569         (rs6000_function_arg): Adjust.
21570         (rs6000_gimplify_va_arg): Pass false for that new parameter.
21572 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21574         PR rtl-optimization/84169
21575         PR rtl-optimization/84780
21576         * combine.c (can_combine_p): Check for a 2-insn combination whether
21577         the destination register is used between the two insns, too.
21579 2018-03-12  Richard Biener  <rguenther@suse.de>
21581         PR tree-optimization/84803
21582         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21583         for refs DR analysis didn't process.
21585 2018-03-12  Richard Biener  <rguenther@suse.de>
21587         PR tree-optimization/84777
21588         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21589         force-vectorize loops ignore whether we are optimizing for size.
21591 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
21593         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21594         (TARGET_MD_ASM_ADJUST): Define.
21596 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
21597             Kito Cheng  <kito.cheng@gmail.com>
21598             Chung-Ju Wu  <jasonwucj@gmail.com>
21600         * config/nds32/nds32.c (nds32_compute_stack_frame,
21601         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21602         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21603         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21604         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21605         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21606         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21607         * config/nds32/nds32.md (prologue, epilogue): Use macro
21608         NDS32_V3PUSH_AVAILABLE_P to do checking.
21610 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
21612         PR debug/58150
21613         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21614         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21615         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
21616         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
21617         addition of most attributes on !orig_type_die or the attribute not
21618         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21620 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21621             Chung-Ju Wu  <jasonwucj@gmail.com>
21623         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21624         __NDS32_VH__ macro.
21625         * config/nds32/nds32.opt (mvh): New option.
21627 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21628             Chung-Ju Wu  <jasonwucj@gmail.com>
21630         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21631         function.
21632         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21633         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21634         definition.
21636 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21637             Chung-Ju Wu  <jasonwucj@gmail.com>
21639         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21640         function.
21641         * config/nds32/nds32-multiple.md (strlensi): New pattern.
21642         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21644 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
21645             Kito Cheng  <kito.cheng@gmail.com>
21646             Chung-Ju Wu  <jasonwucj@gmail.com>
21648         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21649         UNSPEC_FFMISM and UNSPEC_FLMISM.
21650         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21651         for ffb, ffmism and flmism.
21652         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21653         (unspec_ffmism): Ditto.
21654         (unspec_flmism): Ditto.
21655         (nds32_expand_builtin_impl): Check if string extension is available.
21656         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21657         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21659 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
21661         Reverting patch:
21662         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21664         PR target/83712
21665         * lra-assigns.c (assign_by_spills): Return a flag of reload
21666         assignment failure.  Do not process the reload assignment
21667         failures.  Do not spill other reload pseudos if they has the same
21668         reg class.
21669         (lra_assign): Add a return arg.  Set up from the result of
21670         assign_by_spills call.
21671         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21672         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21673         usage_insns if it is not NULL.
21674         (spill_hard_reg_in_range): New function.
21675         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21676         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21677         function prototypes.
21678         (lra_assign): Change prototype.
21679         * lra.c (lra): Add code to deal with fails by splitting hard reg
21680         live ranges.
21682 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
21684         PR target/84807
21685         * config/i386/i386.opt: Replace Enforcment with Enforcement.
21687 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
21689         PR debug/84620
21690         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21691         (dw_val_node): Add val_symbolic_view.
21692         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21693         (symview_upper_bound): New.
21694         (new_line_info_table): Initialize symviews_since_reset.
21695         (dwarf2out_source_line): Count symviews_since_reset and set
21696         symview_upper_bound.
21697         (dw_val_equal_p): Handle symview.
21698         (add_AT_symview): New.
21699         (print_dw_val): Handle symview.
21700         (attr_checksum, attr_checksum_ordered): Likewise.
21701         (same_dw_val_p, size_of_die): Likewise.
21702         (value_format, output_die): Likewise.
21703         (add_high_low_attributes): Use add_AT_symview for entry_view.
21704         (dwarf2out_finish): Reset symview_upper_bound, clear
21705         zero_view_p.
21707 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
21709         PR target/83969
21710         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21711         Add strict argument and use it.
21712         (rs6000_split_multireg_move): Update for new strict argument.
21713         (mem_operand_gpr): Disallow all non-offsettable addresses.
21714         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21716 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
21718         PR target/84772
21719         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21720         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21721         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21723         PR c++/84767
21724         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21725         decl, use remap_type if we want to use the type.
21727 2018-03-09  Martin Sebor  <msebor@redhat.com>
21729         PR tree-optimization/84526
21730         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21731         Remove dead code.
21732         (builtin_access::generic_overlap): Be prepared to handle non-array
21733         base objects.
21735 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
21737         PR rtl-optimization/84682
21738         * lra-constraints.c (process_address_1): Check is_address flag
21739         for address constraints.
21740         (process_alt_operands): Likewise.
21741         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21742         preprocess_constraints.
21743         * recog.h (preprocess_constraints): Add oploc parameter.
21744         Adjust callers.
21745         * recog.c (preprocess_constraints): Test address_operand for
21746         CT_ADDRESS constraints.
21748 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21750         PR target/83712
21751         * lra-assigns.c (assign_by_spills): Return a flag of reload
21752         assignment failure.  Do not process the reload assignment
21753         failures.  Do not spill other reload pseudos if they has the same
21754         reg class.
21755         (lra_assign): Add a return arg.  Set up from the result of
21756         assign_by_spills call.
21757         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21758         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21759         usage_insns if it is not NULL.
21760         (spill_hard_reg_in_range): New function.
21761         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21762         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21763         function prototypes.
21764         (lra_assign): Change prototype.
21765         * lra.c (lra): Add code to deal with fails by splitting hard reg
21766         live ranges.
21768 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21770         PR target/83193
21771         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21772         Accept complain bool parameter.  Only emit errors if it is true.
21773         (arm_parse_cpu_option_name): Likewise.
21774         (arm_target_thumb_only): Adjust callers of the above.
21775         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21776         prototype to take a default true bool parameter.
21777         (arm_parse_arch_option_name): Likewise.
21779 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
21780             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
21782         PR jit/64089
21783         PR jit/84288
21784         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
21785         * configure: Regenerate.
21786         * configure.ac ("linker --version-script option"): New.
21787         ("linker soname option"): New.
21789 2018-03-09  Richard Biener  <rguenther@suse.de>
21791         PR tree-optimization/84775
21792         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
21793         immediate uses of predicate stmts and mark them modified.
21795         Revert
21796         PR tree-optimization/84178
21797         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21798         to caller.
21799         (version_loop_for_if_conversion): Delay update_ssa call.
21800         (tree_if_conversion): Delay update_ssa until after predicate
21801         insertion.
21803 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
21805         PR target/84763
21806         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
21807         when the function accesses prior frames.
21809 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21811         PR debug/84456
21812         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
21813         gen_llsym, otherwise call maybe_gen_llsym.
21815         PR inline-asm/84742
21816         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
21817         has ',' character inside of it.
21819 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21821         PR target/84748
21822         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21823         as clobbering CC_REGNUM.
21825 2018-03-08  Richard Biener  <rguenther@suse.de>
21827         PR middle-end/84552
21828         * tree-scalar-evolution.c: Include tree-into-ssa.h.
21829         (follow_copies_to_constant): Do not follow SSA names registered
21830         for update.
21832 2018-03-08  Richard Biener  <rguenther@suse.de>
21834         PR tree-optimization/84178
21835         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21836         to caller.
21837         (version_loop_for_if_conversion): Delay update_ssa call.
21838         (tree_if_conversion): Delay update_ssa until after predicate
21839         insertion.
21841 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
21843         PR tree-optimization/84178
21844         * tree-if-conv.c (release_bb_predicate): Remove the
21845         the assertion that the stmts have NULL use_ops.
21846         Discard the statements, asserting that they haven't
21847         yet been added to a BB.
21849 2018-03-08  Richard Biener  <rguenther@suse.de>
21851         PR tree-optimization/84746
21852         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
21853         (phi_translate): Pass in destination ANTIC_OUT set.
21854         (phi_translate_1): Likewise.  For a simplified result lookup
21855         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
21856         (phi_translate_set): Adjust.
21857         (do_pre_regular_insertion): Likewise.
21858         (do_pre_partial_partial_insertion): Likewise.
21860 2018-03-08  Martin Liska  <mliska@suse.cz>
21862         PR gcov-profile/84735
21863         * doc/gcov.texi: Document usage of profile files.
21864         * gcov-io.h: Document changes in the format.
21866 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
21868         PR debug/84404
21869         PR debug/84408
21870         * dwarf2out.c (struct dw_line_info_table): Update comments for
21871         view == -1.
21872         (FORCE_RESET_NEXT_VIEW): New.
21873         (FORCE_RESETTING_VIEW_P): New.
21874         (RESETTING_VIEW_P): Check for -1 too.
21875         (ZERO_VIEW_P): Likewise.
21876         (new_line_info_table): Force-reset next view.
21877         (dwarf2out_begin_function): Likewise.
21878         (dwarf2out_source_line): Simplify zero_view_p initialization.
21879         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
21880         view directly.  Omit view when omitting .loc at line 0.
21882 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21884         PR tree-optimization/84740
21885         * tree-switch-conversion.c (process_switch): Call build_constructors
21886         only if info.phi_count is non-zero.
21888         PR tree-optimization/84739
21889         * tree-tailcall.c (find_tail_calls): Check call arguments against
21890         DECL_ARGUMENTS (current_function_decl) rather than
21891         DECL_ARGUMENTS (func) when checking for tail recursion.
21893 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21895         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
21896         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
21897         Volker Reichelt's entry and add entries for people that perform
21898         GCC fuzzy testing and report numerous bugs.
21900 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
21902         PR target/82411
21903         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
21904         readonly data in sdata, if that is disabled.
21905         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
21906         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
21907         -mreadonly-in-sdata option.
21909 2018-03-07  Martin Sebor  <msebor@redhat.com>
21911         PR tree-optimization/84468
21912         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
21913         basic block when looking for nul assignment.
21915 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
21917         PR target/84277
21918         * except.h (output_function_exception_table): Adjust prototype.
21919         * except.c (output_function_exception_table): Remove FNNAME parameter
21920         and add SECTION parameter.  Ouput one part of the table at a time.
21921         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
21922         the first part of the exception table and emit unwind directives.
21923         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
21924         (i386_pe_seh_cold_init): Likewise.
21925         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
21926         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
21927         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
21928         (ix86_output_call_insn): Emit a nop in one more case for SEH.
21929         * config/i386/winnt.c: Include except.h.
21930         (struct seh_frame_state): Add reg_offset, after_prologue and
21931         in_cold_section fields.
21932         (i386_pe_seh_end_prologue): Set seh->after_prologue.
21933         (i386_pe_seh_cold_init): New function.
21934         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
21935         to seh->in_cold_section.
21936         (seh_emit_push): Record the offset of the push.
21937         (seh_emit_save): Record the offet of the save.
21938         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
21939         Test seh->after_prologue to disregard the epilogue.
21940         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
21941         (i386_pe_end_cold_function): New function.
21943 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21945         PR fortran/84565
21946         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
21947         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
21949         PR c++/84704
21950         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
21951         on tmp_var.
21952         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
21953         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
21955         PR middle-end/84723
21956         * multiple_target.c: Include tree-inline.h and intl.h.
21957         (expand_target_clones): Diagnose and fail if node->definition and
21958         !tree_versionable_function_p (node->decl).
21960 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
21962         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
21963         sprint_ul.
21964         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
21965         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
21966         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
21968 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
21970         PR target/84710
21971         * combine.c (try_combine): Use reg_or_subregno instead of handling
21972         just paradoxical SUBREGs and REGs.
21974 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
21976         * config/arc/arc.c (arc_finalize_pic): Remove function.
21977         (arc_must_save_register): We use single base PIC register, remove
21978         checks to save/restore the PIC register.
21979         (arc_expand_prologue): Likewise.
21980         * config/arc/arc-protos.h (arc_set_default_type_attributes):
21981         Remove.
21982         (arc_verify_short): Likewise.
21983         (arc_attr_type): Likewise.
21984         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
21985         (walk_stores): Likewise.
21986         (arc_address_cost): Make it static.
21987         (arc_verify_short): Likewise.
21988         (branch_dest): Likewise.
21989         (arc_attr_type): Likewise.
21990         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
21991         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
21992         (arc_final_prescan_insn): Remove inserting the nops due to
21993         hardware hazards.  It is done in reorg step.
21994         (insn_length_variant_t): Remove.
21995         (insn_length_parameters_t): Likewise.
21996         (arc_insn_length_parameters): Likewise.
21997         (arc_get_insn_variants): Likewise.
21998         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
22000 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
22002         PR inline-asm/84683
22003         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
22004         assertion failure.
22006         PR tree-optimization/84687
22007         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
22008         on new_node->decl.
22009         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
22011 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22013         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
22014         Rename to ppc_speculation_barrier.
22015         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
22016         __builtin_ppc_speculation_barrier.
22018 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22020         PR target/84700
22021         * combine.c (combine_simplify_rtx): Don't try to simplify if
22022         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
22023         are equal to x.
22025 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
22027         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
22028         to 32 bytes when compiling for POWER9.
22030 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22032         PR target/84564
22033         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
22034         regparm >= 3 with no arg reg available also for calls with
22035         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
22037         PR target/84524
22038         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
22039         orig,vex.
22040         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
22042 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
22044         PR target/84264
22045         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
22047 2018-03-05  Richard Biener  <rguenther@suse.de>
22049         PR tree-optimization/84486
22050         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
22051         When inserting a __builtin_assume_aligned call set the LHS
22052         SSA name alignment info accordingly.
22054 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
22056         PR tree-optimization/84114
22057         * config/aarch64/aarch64.c (aarch64_reassociation_width)
22058         Avoid reassociation of FLOAT_MODE addition.
22060 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
22062         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
22063         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
22064         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
22065         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
22066         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
22067         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
22068         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
22069         and -mwbnoinvd.
22070         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
22071         __builtin_ia32_wbinvd): New builtins.
22072         (SPECIAL_ARGS2): New.
22073         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
22074         (SPECIAL_ARGS2): New.
22075         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
22076         (ix86_valid_target_attribute_inner_p): Ditto.
22077         (ix86_init_mmx_sse_builtins): Add special_args2.
22078         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
22079         TARGET_WBNOINVD_P): New.
22080         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
22081         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
22082         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
22083         * config/i386/immintrin.h (_wbinvd): New intrinsic.
22084         * config/i386/pconfigintrin.h: New file.
22085         * config/i386/wbnoinvdintrin.h: Ditto.
22086         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
22087         wbnoinvdintrin.h.
22088         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
22090 2018-03-05  Richard Biener  <rguenther@suse.de>
22092         PR tree-optimization/84670
22093         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22094         member.
22095         (BB_VISITED_WITH_VISITED_SUCCS): New define.
22096         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22097         (compute_antic_aux): Only assert the number of values in ANTIC_IN
22098         doesn't grow if all successors (recursively) were visited at least
22099         once.
22101 2018-03-05  Richard Biener  <rguenther@suse.de>
22103         PR tree-optimization/84650
22104         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
22105         if executed in the loop pipeline.
22107 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
22109         * doc/configfiles.texi (Configuration Files): Move info about
22110         conditionalizing $target-protos.h to...
22111         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
22112         differs from $target-protos.h.
22114 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
22115             Chung-Ju Wu  <jasonwucj@gmail.com>
22117         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
22118         * config/nds32/nds32-multiple.md (setmemsi): Define.
22119         * config/nds32/nds32-memory-manipulation.c
22120         (nds32_gen_dup_4_byte_to_word_value): New.
22121         (emit_setmem_word_loop): New.
22122         (emit_setmem_byte_loop): New.
22123         (nds32_expand_setmem_loop): New.
22124         (nds32_expand_setmem_loop_v3m): New.
22125         (nds32_expand_setmem_unroll): New.
22126         (nds32_expand_setmem): New.
22128 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22129             Chung-Ju Wu  <jasonwucj@gmail.com>
22131         * config/nds32/nds32-memory-manipulation.c
22132         (nds32_emit_load_store): New.
22133         (nds32_emit_post_inc_load_store): New.
22134         (nds32_emit_mem_move): New.
22135         (nds32_emit_mem_move_block): New.
22136         (nds32_expand_movmemsi_loop_unknown_size): New.
22137         (nds32_expand_movmemsi_loop_known_size): New.
22138         (nds32_expand_movmemsi_loop): New.
22139         (nds32_expand_movmemsi_unroll): New.
22140         (nds32_expand_movmemqi): Rename ...
22141         (nds32_expand_movmemsi): ... to this.
22142         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
22143         (movmemsi): ... to this.
22144         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
22145         (nds32_expand_movmemsi): ... to this.
22147 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22148             Monk Chiang  <sh.chiang04@gmail.com>
22149             Chung-Ju Wu  <jasonwucj@gmail.com>
22151         * config/nds32/nds32-protos.h
22152         (nds32_expand_load_multiple): New arguments.
22153         (nds32_expand_store_multiple): Ditto.
22154         (nds32_valid_multiple_load_store): Rename ...
22155         (nds32_valid_multiple_load_store_p): ... to this.
22156         * config/nds32/nds32-memory-manipulation.c
22157         (nds32_expand_load_multiple): Refine implementation.
22158         (nds32_expand_store_multiple): Ditto.
22159         * config/nds32/nds32-multiple.md
22160         (load_multiple): Update nds32_expand_load_multiple interface.
22161         (store_multiple): Update nds32_expand_store_multiple interface.
22162         * config/nds32/nds32-predicates.c
22163         (nds32_valid_multiple_load_store): Rename ...
22164         (nds32_valid_multiple_load_store_p): ... to this and refine
22165         implementation.
22166         * config/nds32/predicates.md
22167         (nds32_load_multiple_and_update_address_operation): New predicate.
22168         (nds32_store_multiple_and_update_address_operation): New predicate.
22170 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22171             Chung-Ju Wu  <jasonwucj@gmail.com>
22173         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
22174         (combo): New attribute.
22175         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
22177 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
22179         * config/nds32/nds32.opt: Change -mcmodel= default value.
22181 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
22182             Monk Chiang  <sh.chiang04@gmail.com>
22183             Chung-Ju Wu  <jasonwucj@gmail.com>
22185         * config/nds32/constants.md (unspec_element): New enum.
22186         * config/nds32/constraints.md (Umw): New constraint.
22187         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
22188         * config/nds32/nds32-intrinsic.md: Likewise.
22189         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
22190         (nds32_valid_smw_lwm_base_p): New.
22191         (nds32_output_smw_single_word): New.
22192         (nds32_output_lmw_single_word): New.
22193         (nds32_expand_unaligned_load): New.
22194         (nds32_expand_unaligned_store): New.
22195         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
22196         (nds32_output_smw_single_word): Declare.
22197         (nds32_output_lmw_single_word): Declare.
22198         (nds32_expand_unaligned_load): Declare.
22199         (nds32_expand_unaligned_store): Declare.
22200         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
22201         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
22202         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
22203         NDS32_BUILTIN_UASTORE_DW.
22204         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
22205         predicate.
22207 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
22208             Kito Cheng  <kito.cheng@gmail.com>
22209             Chung-Ju Wu  <jasonwucj@gmail.com>
22211         * config/nds32/nds32-intrinsic.c
22212         (nds32_expand_builtin_null_ftype_reg): Delete.
22213         (nds32_expand_builtin_reg_ftype_imm): Ditto.
22214         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
22215         (nds32_read_argument): New.
22216         (nds32_legitimize_target): Ditto.
22217         (nds32_legitimize_argument): Ditto.
22218         (nds32_check_constant_argument): Ditto.
22219         (nds32_expand_unop_builtin): Ditto.
22220         (nds32_expand_unopimm_builtin): Ditto.
22221         (nds32_expand_binop_builtin): Ditto.
22222         (nds32_builtin_decl_impl): Ditto.
22223         (builtin_description): Ditto.
22224         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
22225         (nds32_init_builtins_impl): Ditto.
22226         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
22227         (nds32_builtin_decl): New.
22228         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
22229         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
22231 2018-03-02  Jeff Law  <law@redhat.com>
22233         * reorg.c (stop_search_p): Handle DEBUG_INSN.
22234         (redundant_insn, fill_simple_delay_slots): Likewise.
22235         (fill_slots_from_thread): Likewise.
22236         * resource.c (mark_referenced_resources): Likewise.
22237         (mark_set_resources, find_dead_or_set_registers): Likewise.
22239 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22241         * substring-locations.h (format_warning_va): Formatting fix for
22242         ATTRIBUTE_GCC_DIAG.
22243         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
22244         argument.
22245         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
22246         * substring-locations.c: Include intl.h.
22247         (format_warning_va): Turned into small wrapper around
22248         format_warning_n_va, renamed to ...
22249         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
22250         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
22251         use ngettext.
22252         (format_warning_at_substring_n): New function.
22253         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
22254         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
22255         format_warning_at_substring with just a shorter name instead of
22256         const function pointer.
22257         (fmtwarn_n): New function.
22258         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
22259         appropriate, get rid of all the fmtstr temporaries, move conditionals
22260         with G_() wrapped string literals directly into fmtwarn arguments,
22261         cast dir.len to (int), formatting fixes.
22263 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
22265         * doc/invoke.texi: Remove "Cilk Plus" references.
22267 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22268             Richard Biener  <rguenther@suse.de>
22270         PR ipa/84628
22271         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
22272         for error or warning attributes if CALL_FROM_THUNK_P is set.
22273         Formatting fixes.
22275 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22277         PR target/56540
22278         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
22279         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
22281         PR target/56540
22282         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
22283         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
22285         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
22286         instead of -1U in last predictors element's probability member.
22288 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
22290         PR ipa/83983
22291         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22292         arguments if they are comparable.
22294 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22296         PR tree-optimization/84634
22297         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
22298         masks and masked_loop_p with a single loop_masks, making sure it's
22299         null for bb vectorization.
22301 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22303         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
22304         (vect_analyze_data_ref_access): Use loop->safe_len rather than
22305         loop->force_vectorize to check whether there is no alias.
22307 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22309         PR target/84614
22310         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
22311         prototypes.
22312         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
22313         comments.
22314         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
22315         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
22316         instead of a loop around prev_real_insn.
22317         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
22318         prev_real_insn.
22320         PR inline-asm/84625
22321         * config/i386/i386.c (ix86_print_operand): Use conditional
22322         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
22323         zero vector.
22325 2018-03-02  Richard Biener  <rguenther@suse.de>
22327         PR tree-optimization/84427
22328         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
22329         (bitmap_set_subtract_values): Rewrite to handle multiple
22330         exprs per value.
22331         (clean): Likewise.
22332         (prune_clobbered_mems): Likewise.
22333         (phi_translate): Take edge instead of pred/phiblock.
22334         (phi_translate_1): Likewise.
22335         (phi_translate_set): Likewise.  Insert all translated
22336         exprs for a value into the set, keeping possibly multiple
22337         expressions per value.
22338         (compute_antic_aux): Adjust for phi_translate changes.
22339         When intersecting union the expressions and prune those
22340         not in the final value set, keeping possibly multiple
22341         expressions per value.  Do not use value-insertion
22342         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
22343         all expressions.  Add verification that the value-sets
22344         only shrink during iteration.
22345         (compute_partial_antic_aux): Adjust for the phi_translate changes.
22346         (do_pre_regular_insertion): Likewise.
22347         (do_pre_partial_partial_insertion): Likewise.
22349 2018-03-02  Richard Biener  <rguenther@suse.de>
22351         PR target/82005
22352         * config/darwin.c (saved_debug_info_level): New static global.
22353         (darwin_asm_lto_start): Disable debug info generation for LTO out.
22354         (darwin_asm_lto_end): Restore debug info generation settings.
22356 2018-03-01  Martin Liska  <mliska@suse.cz>
22358         PR sanitizer/82484
22359         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
22360         volatile arguments.
22362 2018-03-01  Richard Biener  <rguenther@suse.de>
22364         PR debug/84645
22365         * dwarf2out.c (gen_variable_die): Properly handle late VLA
22366         type annotation with LTO when debug was disabled at compile-time.
22368 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
22370         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
22371         XINT with INTVAL.
22372         (mips_final_postscan_insn): Likewise.
22374 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
22376         PR rtl-optimization/84528
22377         * alias.c (init_alias_target): Add commentary.
22378         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
22379         a unique base value if the frame pointer is not eliminated
22380         to the stack pointer.
22382 2018-03-01  Tom de Vries  <tom@codesourcery.com>
22384         PR rtl-optimization/83327
22385         * lra-int.h (hard_regs_spilled_into): Declare.
22386         * lra.c (hard_regs_spilled_into): Define.
22387         (init_reg_info): Init hard_regs_spilled_into.
22388         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22389         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
22390         (process_bb_lives): Handle hard_regs_spilled_into.
22391         (lra_create_live_ranges_1): Before doing liveness propagation, clear
22392         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
22394 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
22396         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22397         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22398         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22399         * config/rs6000/aix72.h: New file.
22401 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
22403         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22404         instead of warning_at with conditional singular and plural messages
22405         where possible.
22407         PR target/52991
22408         * stor-layout.c (update_alignment_for_field): For
22409         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22410         && !DECL_PACKED (field), do the alignment update, just use
22411         only desired_align instead of MAX (type_align, desired_align)
22412         as the alignment.
22413         (place_field): Don't do known_align < desired_align handling
22414         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22415         is non-NULL, instead do it after rli->prev_field handling and
22416         only if not within a bitfield word.  For DECL_PACKED (field)
22417         use type_align of BITS_PER_UNIT.
22419 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
22421         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22422         superfluous parentheses and trailing spaces.
22424 2018-02-28  Richard Biener  <rguenther@suse.de>
22426         PR tree-optimization/84584
22427         * graphite-scop-detection.c (scop_detection::add_scop): Discard
22428         SCoPs with fake exit edge.
22430 2018-02-28  Martin Liska  <mliska@suse.cz>
22432         PR testsuite/84597
22433         * timevar.c (timer::print): Fix format to properly print 100%
22434         values.
22436 2018-02-28  Richard Biener  <rguenther@suse.de>
22438         PR middle-end/84607
22439         * genmatch.c (capture_info::walk_match): Do not mark
22440         captured expressions without operands as expr_p given
22441         they act more like predicates and should be subject to
22442         "lost tail" side-effect preserving.
22444 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
22446         PR rtl-optimization/81611
22447         * auto-inc-dec.c (attempt_change): Move dead note from
22448         mem_insn if it's the next use of regno
22449         (find_address): Take address use of reg holding
22450         non-incremented value.  Add parm to limit search to the named
22451         reg only.
22452         (merge_in_block): Attempt to use a mem insn that is the next
22453         use of the original regno.
22455 2018-02-27  Martin Sebor  <msebor@redhat.com>
22457         PR c++/83871
22458         * doc/invoke.texi (-Wmissing-attributes): New option.
22459         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22461 2018-02-27  Martin Sebor  <msebor@redhat.com>
22463         PR translation/84207
22464         * diagnostic-core.h (warning_n, error_n, inform_n): Change
22465         n argument to unsigned HOST_WIDE_INT.
22466         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22467         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
22468         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22469         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22471 2018-02-27  Richard Biener  <rguenther@suse.de>
22473         PR tree-optimization/84512
22474         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22475         Do not use the estimate returned from record_stmt_cost for
22476         the scalar iteration cost but sum properly using add_stmt_cost.
22478 2018-02-27  Richard Biener  <rguenther@suse.de>
22480         PR tree-optimization/84466
22481         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22482         Adjust last change to less strictly validate use operands.
22484 2018-02-27  Martin Liska  <mliska@suse.cz>
22486         PR gcov-profile/84548
22487         * gcov.c (process_file): Allow partial overlap and consider it
22488         also as group functions.
22489         (output_lines): Properly calculate range of lines for a group.
22491 2018-02-27  Martin Liska  <mliska@suse.cz>
22493         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22494         'ggc' suffixes.  Change first column width.
22495         (timer::print): Fix formatting of the column.
22497 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
22499         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22500         preserve inline entry blocks for the sake of debug inline
22501         entry point markers alone.
22502         (remove_unused_locals): Suggest in comments a better place to
22503         force the preservation of inline entry blocks that are
22504         otherwise unused, but do not preserve them.
22506 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22508         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22510 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22512         PR target/84039
22513         * config/i386/constraints.md (Bs): Replace
22514         ix86_indirect_branch_register with
22515         TARGET_INDIRECT_BRANCH_REGISTER.
22516         (Bw): Likewise.
22517         * config/i386/i386.md (indirect_jump): Likewise.
22518         (tablejump): Likewise.
22519         (*sibcall_memory): Likewise.
22520         (*sibcall_value_memory): Likewise.
22521         Peepholes of indirect call and jump via memory: Likewise.
22522         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22523         (*sibcall_value_GOT_32): Likewise.
22524         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22525         (GOT_memory_operand): Likewise.
22526         (call_insn_operand): Likewise.
22527         (sibcall_insn_operand): Likewise.
22528         (GOT32_symbol_operand): Likewise.
22529         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22531 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22533         PR rtl-optimization/83496
22534         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22535         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
22536         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22537         redundant insn, if any.
22538         (relax_delay_slots): Likewise.
22539         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22541 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
22543         PR tree-optimization/83965
22544         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22545         that grouped statements are part of a reduction chain.  Return
22546         true if the statement is not marked as a reduction itself but
22547         is part of a group.
22548         (vect_recog_dot_prod_pattern): Don't check whether the statement
22549         is part of a group here.
22550         (vect_recog_sad_pattern): Likewise.
22551         (vect_recog_widen_sum_pattern): Likewise.
22553 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22555         PR debug/84545
22556         * final.c (rest_of_clean_state): Also look for calls inside sequences.
22558 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22560         PR target/84530
22561         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22562         the bool argument.
22563         (ix86_output_indirect_function_return): New prototype.
22564         (ix86_split_simple_return_pop_internal): Likewise.
22565         * config/i386/i386.c (indirect_return_via_cx): New.
22566         (indirect_return_via_cx_bnd): Likewise.
22567         (indirect_thunk_name): Handle return va CX_REG.
22568         (output_indirect_thunk_function): Create alias for
22569         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22570         (ix86_output_indirect_jmp): Remove the bool argument.
22571         (ix86_output_indirect_function_return): New function.
22572         (ix86_split_simple_return_pop_internal): Likewise.
22573         * config/i386/i386.md (*indirect_jump): Don't pass false
22574         to ix86_output_indirect_jmp.
22575         (*tablejump_1): Likewise.
22576         (simple_return_pop_internal): Change it to define_insn_and_split.
22577         Call ix86_split_simple_return_pop_internal to split it for
22578         -mfunction-return=.
22579         (simple_return_indirect_internal): Call
22580         ix86_output_indirect_function_return instead of
22581         ix86_output_indirect_jmp.
22583 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
22585         PR bootstrap/84405
22586         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22587         memset and value initialization afterwards.
22589 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
22591         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22593 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22595         PR target/84521
22596         * common/config/aarch64/aarch64-common.c
22597         (aarch_option_optimization_table[]): Switch
22598         off fomit-frame-pointer
22600 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22601             Chung-Ju Wu  <jasonwucj@gmail.com>
22603         * config/nds32/nds32-multiple.md (load_multiple): Disallow
22604         volatile memory.
22605         (store_multiple): Ditto.
22607 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22609         * config.gcc: Add --with-cpu support for nds32 target.
22610         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22611         * config/nds32/nds32.opt: Add -mcpu= option.
22613 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
22615         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22616         isel=yes): Warn for these deprecated options.
22618 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
22620         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22621         ISA_2_5_MASKS_EMBEDDED.
22623 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
22625         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22626         p->max as pointers rather than using iterative_hash_expr.
22628 2018-02-23  Carl Love  <cel@us.ibm.com>
22630         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22631         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22632         BU_P8V_OVERLOAD_2.
22633         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22634         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22635         P8V_BUILTIN_VEC_VUNSIGNED2.
22637 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
22639         PR target/81572
22640         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22641         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22642         LRA_UNKNOWN_ALT.
22643         * lra-constraints.c (curr_insn_transform): Set up
22644         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
22645         LRA_UNKNOWN_ALT.
22646         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22647         * lra-eliminations.c (spill_pseudos): Ditto.
22648         (process_insn_for_elimination): Ditto.
22649         * lra-lives.c (reg_early_clobber_p): Use the new macros.
22650         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22651         LRA_NON_CLOBBERED_ALT.
22653 2018-02-22  Martin Sebor  <msebor@redhat.com>
22655         PR tree-optimization/84480
22656         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22657         to maybe_diag_stxncpy_trunc.  Call it.
22658         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22659         from gimple_fold_builtin_strcpy.  Print inlining stack.
22660         (handle_builtin_stxncpy): Print inlining stack.
22661         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22663 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
22665         PR target/84176
22666         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22667         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22668         and -fcheck-pointer-bounds are used together.
22669         (indirect_thunk_prefix): New enum.
22670         (indirect_thunk_need_prefix): New function.
22671         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
22672         "_nt" instead of "_bnd" for NOTRACK prefix.
22673         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22674         (output_indirect_thunk_function): Likewise.
22675         (): Likewise.
22676         (ix86_code_end): Update output_indirect_thunk_function calls.
22677         (ix86_output_indirect_branch_via_reg): Replace
22678         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22679         (ix86_output_indirect_branch_via_push): Likewise.
22680         (ix86_output_function_return): Likewise.
22681         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22682         incompatible with -fcf-protection=branch and
22683         -fcheck-pointer-bounds.
22685 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22687         PR target/83335
22688         * config/aarch64/aarch64.c (aarch64_print_address_internal):
22689         Change gcc_assert call to output_operand_lossage.
22691 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22693         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22695 2018-02-22  DJ Delorie  <dj@redhat.com>
22696             Sebastian Perta  <sebastian.perta@renesas.com>
22697             Oleg Endo  <olegendo@gcc.gnu.org>
22699         * config/rx/rx.c (rx_rtx_costs): New function.
22700         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22702 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22704         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22706 2018-02-22  Martin Liska  <mliska@suse.cz>
22708         PR driver/83193
22709         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22710         Add "native" as a possible value.
22712 2018-02-22  Martin Liska  <mliska@suse.cz>
22714         PR driver/83193
22715         * config/i386/i386.c (ix86_option_override_internal):
22716         Add "native" as a possible value for -march and -mtune.
22718 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
22720         PR target/84502
22721         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22722         to all type variants.
22724         PR tree-optimization/84503
22725         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22726         width as info->bitpos + info->bitsize - start.
22727         (merged_store_group::merge_overlapping): Simplify width computation.
22728         (check_no_overlap): New function.
22729         (imm_store_chain_info::try_coalesce_bswap): Compute expected
22730         start + width and last_order of the group, fail if check_no_overlap
22731         fails.
22732         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22733         to group if check_no_overlap fails.
22735 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22737         * config/rs6000/altivec.md: Delete contraint arguments to
22738         define_expand, define_split, and define_peephole2, and in
22739         define_insn_and_split if always unused.
22740         * config/rs6000/darwin.md: Ditto.
22741         * config/rs6000/dfp.md: Ditto.
22742         * config/rs6000/rs6000.md: Ditto.
22743         * config/rs6000/sync.md: Ditto.
22744         * config/rs6000/vector.md: Ditto.
22745         * config/rs6000/vsx.md: Ditto.
22747 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22749         * config/rs6000/altivec.md: Write output control strings as braced
22750         blocks instead of double-quoted strings.
22751         * config/rs6000/darwin.md: Ditto.
22752         * config/rs6000/rs6000.md: Ditto.
22753         * config/rs6000/vector.md: Ditto.
22754         * config/rs6000/vsx.md: Ditto.
22756 2018-02-21  Jason Merrill  <jason@redhat.com>
22758         PR c++/84314 - ICE with templates and fastcall attribute.
22759         * attribs.c (build_type_attribute_qual_variant): Remove assert.
22761 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22763         * ipa-cp.c (determine_versionability): Fix comment typos.
22765 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22767         PR c/84229
22768         * ipa-cp.c (determine_versionability): Do not version functions caling
22769         va_arg_pack.
22771 2018-02-21  Martin Liska  <mliska@suse.cz>
22773         PR driver/83193
22774         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22775         Add "native" as a possible value.
22776         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
22777         the macro when native cpu detection is available.
22779 2018-02-21  Martin Liska  <mliska@suse.cz>
22781         PR driver/83193
22782         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
22783         Add "native" as a possible value.
22784         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
22785         when native cpu detection is available.
22787 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
22788             Martin Sebor  <msebor@redhat.com>
22790         PR tree-optimization/84478
22791         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22792         false.
22793         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
22794         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
22795         support which is conservatively correct, for 2 only stay conservative
22796         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
22797         argument to the 2 argument get_range_strlen, adjust 6 arg
22798         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
22799         false.
22800         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
22801         (gimple_fold_builtin_strlen): Pass true as last argument to
22802         get_range_strlen.
22804 2018-02-20  Martin Sebor  <msebor@redhat.com>
22806         PR middle-end/84095
22807         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
22808         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
22809         (builtin_memref::builtin_memref): Factor out parts into
22810         set_base_and_offset and call it.
22812 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22814         PR middle-end/84406
22815         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
22816         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
22817         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
22818         search at the associated MODE_INT.
22820 2018-02-20  Jeff Law  <law@redhat.com>
22822         PR middle-end/82123
22823         PR tree-optimization/81592
22824         PR middle-end/79257
22825         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
22826         for range data rather than using global data.
22827         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
22828         range data rather than using global data.
22829         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
22830         pass it to children as needed.
22831         (struct directive::fmtresult): Similarly.
22832         (struct directive::set_width): Similarly.
22833         (struct directive::set_precision): Similarly.
22834         (format_integer, format_directive, parse_directive): Similarly.
22835         (format_none): Accept unnamed vr_values parameter.
22836         (format_percent, format_floating, format_character): Similarly.
22837         (format_string, format_plain): Similarly.
22838         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
22839         the EVRP range analyzer for range data rather than using global data.
22840         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
22841         gimple-ssa-evrp-analyze.h
22842         (class sprintf_dom_walker): Add after_dom_children member function.
22843         Add evrp_range_analyzer member.
22844         (sprintf_dom_walker::before_dom_children): Call into the EVRP
22845         range analyzer as needed.
22846         (sprintf_dom_walker::after_dom_children): New member function.
22847         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
22848         if not optimizing.
22849         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
22850         (evrp_range_analyzer::pop_to_marker): Likewise.
22852 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22854         PR tree-optimization/84419
22855         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
22856         with the required type if its current type is compatible but
22857         different.
22859 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
22861         PR middle-end/82004
22862         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
22863         after vectorization.
22865 2018-02-20  Martin Liska  <mliska@suse.cz>
22867         PR driver/83193
22868         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
22869         possible values if we don't have a hint.
22871 2018-02-20  Martin Liska  <mliska@suse.cz>
22873         PR c/84310
22874         PR target/79747
22875         * final.c (shorten_branches): Build align_tab array with one
22876         more element.
22877         * opts.c (finish_options): Add alignment option limit check.
22878         (MAX_CODE_ALIGN): Likewise.
22879         (MAX_CODE_ALIGN_VALUE): Likewise.
22880         * doc/invoke.texi: Document maximum allowed option value for
22881         all -falign-* options.
22883 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
22885         PR target/84146
22886         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
22887         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
22888         * var-tracking.c (emit_note_insn_var_location): Remove all references
22889         to NOTE_INSN_CALL_ARG_LOCATION.
22890         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
22891         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
22892         Use copy_rtx_if_shared.
22893         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
22894         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
22895         (dwarf2out_var_location): Remove handling of
22896         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
22897         on call_insn.
22898         * final.c (final_scan_insn): Remove all references to
22899         NOTE_INSN_CALL_ARG_LOCATION.
22900         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
22901         before dumping final insns.
22902         * except.c (emit_note_eh_region_end): Remove all references to
22903         NOTE_INSN_CALL_ARG_LOCATION.
22904         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
22905         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
22906         * config/arc/arc.c (hwloop_optimize): Likewise.
22907         * config/arm/arm.c (create_fix_barrier): Likewise.
22908         * config/s390/s390.c (s390_chunkify_start): Likewise.
22909         * config/sh/sh.c (find_barrier): Likewise.
22910         * config/i386/i386.c (rest_of_insert_endbranch,
22911         ix86_seh_fixup_eh_fallthru): Likewise.
22912         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
22913         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
22914         * config/frv/frv.c (frv_function_prologue): Likewise.
22915         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
22916         reg note.
22917         (note_outside_basic_block_p): Remove all references to
22918         NOTE_INSN_CALL_ARG_LOCATION.
22919         * gengtype.c (adjust_field_rtx_def): Likewise.
22920         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
22921         Likewise.
22922         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
22923         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
22925         PR c++/84444
22926         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
22927         is ADDR_EXPR.
22929         PR tree-optimization/84452
22930         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
22931         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
22932         is NULL.
22934 2018-02-19  Martin Liska  <mliska@suse.cz>
22936         PR sanitizer/82183
22937         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
22939 2018-02-19  Martin Liska  <mliska@suse.cz>
22940             Richard Sandiford  <richard.sandiford@linaro.org>
22942         PR tree-optimization/82491
22943         * gimple-fold.c (get_base_constructor): Make earlier bail out
22944         to prevent ubsan.
22946 2018-02-19  Carl Love  <cel@us.ibm.com>
22948         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
22949         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
22950         BU_P8V_OVERLOAD_1.
22951         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
22952         P8V_BUILTIN_VEC_NEG.
22954 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
22956         * config/rl78/rl78.md (movdf): New define expand.
22958 2018-02-19  Martin Liska  <mliska@suse.cz>
22960         PR other/80589
22961         * doc/invoke.texi: Fix typo.
22962         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
22964 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
22966         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
22967         handle rs6000_single_float and rs6000_double_float specially for
22968         e500 family CPUs.
22970 2018-02-16  Jeff Law  <law@redhat.com>
22972         * config/rx/rx.c (add_pop_cfi_notes): New function.;
22973         (pop_regs): Use it.
22975 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22977         PR ipa/84425
22978         * ipa-inline.c (inline_small_functions): Fix a typo.
22980 2018-02-16  Nathan Sidwell  <nathan@acm.org>
22982         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
22984 2018-02-16  Carl Love  <cel@us.ibm.com>
22986         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
22987         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
22988         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
22989         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
22990         expansion to P8V_BUILTIN_VEC_FLOAT2.
22992 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
22994         PR rtl-optimization/70023
22995         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
22996         src_regno into account.
22998 2018-02-16  Carl Love  <cel@us.ibm.com>
23000         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
23001         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
23002         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
23003         * config/rs6000/rs6000.c: Remove case statements for
23004         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
23005         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
23006         and P9V_BUILTIN_VEC_VINSERT4B.
23007         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
23008         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
23009         * config/rs6000/vsx.md:
23010         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
23011         vec_insert4b.
23013 2018-02-16  Carl Love  <cel@us.ibm.com>
23015         * config/rs6000/altivec.h: Add builtin names vec_extract4b
23016         vec_insert4b.
23017         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
23018         definitions.
23019         * config/rs6000/rs6000-c.c: Add the definitions for
23020         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
23021         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
23022         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
23023         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
23024         definition for insert4b and define insn *insert3b_internal.
23025         * doc/extend.texi: Add documentation for vec_extract4b.
23027 2018-02-16  Nathan Sidwell  <nathan@acm.org>
23029         * doc/extend.texi (Backwards Compatibility): Mention friend
23030         injection.  Note for-scope is deprecated.
23031         * doc/invoke.texi (-ffriend-injection): Deprecate.
23033 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
23035         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
23036         that moved to I2, also allow destinations that are a paradoxical
23037         subreg (instead of a normal reg).
23039 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
23041         PR target/83831
23042         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
23043         to QImode.
23045 2018-02-16  Richard Biener  <rguenther@suse.de>
23047         PR tree-optimization/84037
23048         PR tree-optimization/84016
23049         PR target/82862
23050         * config/i386/i386.c (ix86_builtin_vectorization_cost):
23051         Adjust vec_construct for the fact we need additional higher latency
23052         128bit inserts for AVX256 and AVX512 vector builds.
23053         (ix86_add_stmt_cost): Scale vector construction cost for
23054         elementwise loads.
23056 2018-02-16  Richard Biener  <rguenther@suse.de>
23058         PR tree-optimization/84417
23059         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
23060         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
23061         (non_rewritable_lvalue_p): Likewise, use poly-ints.
23063 2018-02-16  Martin Liska  <mliska@suse.cz>
23065         PR sanitizer/84307
23066         * internal-fn.def (ASAN_CHECK): Set proper flags.
23067         (ASAN_MARK): Likewise.
23069 2018-02-16  Julia Koval  <julia.koval@intel.com>
23071         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
23072         from PTA_CANNONLAKE.
23074 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
23076         PR target/84272
23077         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
23078         Use ++iter rather than iter++ for std::list iterators.
23079         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
23080         defer deleting them until all nodes in the forest are processed.  Do
23081         free even leaf nodes.  Change to_process into auto_vec.
23083         PR bootstrap/84405
23084         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
23085         * vec.h (vec_default_construct): Use memset instead of placement new
23086         if BROKEN_VALUE_INITIALIZATION is defined.
23087         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
23088         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
23089         is defined.
23091         PR rtl-optimization/83723
23092         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
23093         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
23094         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
23095         recursive calls.
23096         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
23097         callers.
23098         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
23100 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
23102         PR rtl-optimization/81443
23103         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
23104         from inner REGs to paradoxical SUBREGs.
23106 2018-02-16  Richard Biener  <rguenther@suse.de>
23108         PR tree-optimization/84399
23109         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23110         For operands we can analyze at their definition make sure we can
23111         analyze them at each use as well.
23113 2018-02-16  Richard Biener  <rguenther@suse.de>
23115         PR tree-optimization/84190
23116         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
23117         volatile accesses if the decl isn't volatile.
23119 2018-02-15  Jason Merrill  <jason@redhat.com>
23121         PR c++/84314 - ICE with templates and fastcall attribute.
23122         * attribs.c (build_type_attribute_qual_variant): Don't clobber
23123         TYPE_CANONICAL on an existing type.
23125 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
23127         PR tree-optimization/84383
23128         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
23129         dstoff nor call operand_equal_p if dstbase is NULL.
23131         PR tree-optimization/84334
23132         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
23133         also a CONSTANT_CLASS_P, punt.
23135 2018-02-14  Jim Wilson  <jimw@sifive.com>
23137         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
23138         first SMALL_OPERAND check.  New local min_second_step.  Move assert
23139         to where locals are set.  Add TARGET_RVC support.
23140         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
23142 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
23144         * doc/invoke.texi: Correct -Wformat-overflow code sample.
23146 2018-02-14  Martin Sebor  <msebor@redhat.com>
23148         PR tree-optimization/83698
23149         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
23150         arrays constrain the offset range to their bounds.
23151         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
23152         (builtin_access::overlap): Avoid setting the size of overlap if it's
23153         already been set.
23154         (maybe_diag_overlap): Also consider arrays when deciding what values
23155         of offsets to include in diagnostics.
23157 2018-02-14  Martin Sebor  <msebor@redhat.com>
23159         PR c/84108
23160         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
23161         that correspond to the kind of a declaration.
23163 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
23165         PR target/83984
23166         * config/pa/pa.md: Load address of PIC label using the linkage table
23167         if the label is nonlocal.
23169 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23171         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
23172         warning message if user requests -maltivec=be.
23173         * doc/invoke.texi: Document deprecation of -maltivec=be.
23175 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
23177         PR target/84220
23178         * config/rs6000/rs6000-c.c: Update definitions for
23179         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
23180         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
23182 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23184         PR target/84239
23185         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
23186         add _get_ssp intrinsics. Remove argument from
23187         __builtin_ia32_rdssp[d|q].
23188         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
23189         * config/i386/i386-builtin.def: Remove argument from
23190         __builtin_ia32_rdssp[d|q].
23191         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
23192         ix86_expand_special_args_builtin for _rdssp[d|q].
23193         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
23194         Clear register before usage.
23195         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
23196         Add documentation for new _get_ssp and _inc_ssp intrinsics.
23198 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
23200         PR tree-optimization/84357
23201         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
23202         operand 1 of an ARRAY_REF too.
23204 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
23206         PR target/83831
23207         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
23208         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23209         declarations.
23210         (set_of_reg): New struct.
23211         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
23212         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
23213         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23214         functions.
23215         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
23216         Split into bitclr, bitset, bitinvert patterns if appropriate.
23217         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
23218         use rx_fuse_in_memory_bitop.
23219         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
23220         to named insn, correct maximum insn length.
23222 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
23224         PR target/79242
23225         * machmode.def: Define a complex mode for PARTIAL_INT.
23226         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
23227         MODE_PARTIAL_INT.
23228         * doc/rtl.texi: Document CSPImode.
23229         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
23230         handling.
23231         (msp430_hard_regno_nregs_with_padding): Likewise.
23233 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
23235         PR target/84279
23236         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
23238 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
23240         PR rtl-optimization/84169
23241         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
23242         we generated a parallel as new i3 and we split that to new i2 and i3
23243         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
23244         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
23245         those to i2, not i1.  Partially rewrite this scan code.
23247 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23249         PR c/82210
23250         * stor-layout.c (place_field): For variable length fields, adjust
23251         offset_align afterwards not just based on the field's alignment,
23252         but also on the size.
23254         PR middle-end/84309
23255         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
23256         of exps and logs in the use_exp2 case.
23258 2018-02-13  Jeff Law  <law@redhat.com>
23260         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
23261         entry for "vector".
23263         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
23264         ARGS as unused.
23266 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
23268         PR debug/84342
23269         PR debug/84319
23270         * common.opt (gas-loc-support, gas-locview-support): New.
23271         (ginline-points, ginternal-reset-location-views): New.
23272         * doc/invoke.texi: Document them.  Use @itemx where intended.
23273         (gvariable-location-views): Adjust.
23274         * target.def (reset_location_view): New.
23275         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
23276         (TARGET_RESET_LOCATION_VIEW): New.
23277         * doc/tm.texi: Rebuilt.
23278         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
23279         (dwarf2out_default_as_locview_support): New.
23280         (output_asm_line_debug_info): Use option variables.
23281         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
23282         (output_loc_list): Likewise.
23283         (add_high_low_attributes): Check option variables.
23284         Don't output entry view attribute in strict mode.
23285         (gen_inlined_subroutine_die): Check option variables.
23286         (dwarf2out_inline_entry): Likewise.
23287         (init_sections_and_labels): Likewise.
23288         (dwarf2out_early_finish): Likewise.
23289         (maybe_reset_location_view): New, from...
23290         (dwarf2out_var_location): ... here.  Call it.
23291         * debug.h (dwarf2out_default_as_loc_support): Declare.
23292         (dwarf2out_default_as_locview_support): Declare.
23293         * hooks.c (hook_int_rtx_insn_0): New.
23294         * hooks.h (hook_int_rtx_insn_0): Declare.
23295         * toplev.c (process_options): Take -gas-loc-support and
23296         -gas-locview-support from dwarf2out.  Enable
23297         -gvariable-location-views by default only with locview
23298         assembler support.  Enable -ginternal-reset-location-views by
23299         default only if the target defines the corresponding hook.
23300         Enable -ginline-points by default if location views are
23301         enabled; force it disabled if statement frontiers are
23302         disabled.
23303         * tree-inline.c (expand_call_inline): Check option variables.
23304         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
23306 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
23308         PR tree-optimization/84321
23309         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
23310         handling.  Also check whether the anti-range contains any values
23311         that satisfy the mask; switch to a VR_RANGE if not.
23313 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
23315         PR sanitizer/84340
23316         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
23318 2018-02-13  Martin Jambor  <mjambor@suse.cz>
23320         PR c++/83990
23321         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
23322         of call statements, also set location of a load to a temporary.
23324 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
23326         * config/rl78/rl78.c (add_vector_labels): New function.
23327         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
23328         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
23329         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
23330         which checks that no arguments are passed.
23331         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
23332         * doc/extend.texi: Documentation for the new attribute.
23334 2018-02-13  Andreas Schwab  <schwab@suse.de>
23336         * config/riscv/linux.h (CPP_SPEC): Define.
23338 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23340         PR target/84335
23341         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
23342         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
23343         OPTION_MASK_ISA_AES as first argument to def_builtin_const
23344         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
23345         instead of OPTION_MASK_ISA_PCLMUL as first argument to
23346         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
23347         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
23348         temporarily for AES and PCLMUL builtins.
23350         PR tree-optimization/84339
23351         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
23352         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
23353         Formatting fixes.
23355         PR middle-end/84309
23356         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
23357         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
23358         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
23359         inline function.
23360         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
23361         inline function.
23362         * omp-simd-clone.h: New file.
23363         * omp-simd-clone.c: Include omp-simd-clone.h.
23364         (expand_simd_clones): No longer static.
23365         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
23366         cgraph.h and omp-simd-clone.h.
23367         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
23368         (vect_recog_widen_shift_pattern): Formatting fix.
23369         (vect_pattern_recog_1): Don't check optab for calls.
23371         PR target/84336
23372         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
23373         operands[2] into a REG before using gen_lowpart on it.
23375 2018-02-12  Jeff Law  <law@redhat.com>
23377         PR target/83760
23378         * config/sh/sh.c (find_barrier): Consider a sibling call
23379         a barrier as well.
23381         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
23382         successfully back substituting a reg.
23384 2018-02-12  Richard Biener  <rguenther@suse.de>
23386         PR tree-optimization/84037
23387         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
23388         parameter, move visited init to caller.
23389         (vect_slp_analyze_operations): Separate cost from validity
23390         check, initialize visited once for all instances.
23391         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
23392         for all instances.
23393         * tree-vect-stmts.c (vect_model_simple_cost): Make early
23394         out an assert.
23395         (vect_model_promotion_demotion_cost): Likewise.
23396         (vectorizable_bswap): Guard cost modeling with !slp_node
23397         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23398         SLP stmts.
23399         (vectorizable_call): Likewise.
23400         (vectorizable_conversion): Likewise.
23401         (vectorizable_assignment): Likewise.
23402         (vectorizable_shift): Likewise.
23403         (vectorizable_operation): Likewise.
23404         (vectorizable_store): Likewise.
23405         (vectorizable_load): Likewise.
23406         (vectorizable_condition): Likewise.
23407         (vectorizable_comparison): Likewise.
23409 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
23411         PR sanitizer/84307
23412         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23413         (ASAN_MARK): Fix fnspec to account for return value, change pointer
23414         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23416 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23418         PR middle-end/83665
23419         * params.def (inline-min-speedup): Increase from 8 to 15.
23420         (max-inline-insns-auto): Decrease from 40 to 30.
23421         * ipa-split.c (consider_split): Add some buffer for function to
23422         be considered inlining candidate.
23423         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23424         default values.
23426 2018-02-12  Richard Biener  <rguenther@suse.de>
23428         PR tree-optimization/84037
23429         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23430         matched stmts if we cannot swap the non-matched ones.
23432 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23434         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23435         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23436         _mm_maskz_scalef_round_ss): New intrinsics.
23437         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23438         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23439         __builtin_ia32_scalefss_round): Remove.
23440         (__builtin_ia32_scalefsd_mask_round,
23441         __builtin_ia32_scalefss_mask_round): New intrinsics.
23442         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23443         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23444         ((match_operand:VF_128 2 "<round_nimm_predicate>"
23445         "<round_constraint>")): Changed to ...
23446         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23447         "<round_scalar_constraint>")): ... this.
23448         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23449         %0, %1, %2<round_op3>}"): Changed to ...
23450         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23451         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23452         %2<round_scalar_mask_op3>}"): ... this.
23453         * config/i386/subst.md (round_scalar_nimm_predicate): New.
23455 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23457         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23458         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23459         (_mm_maskz_sqrt_round_ss): New intrinsics.
23460         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23461         (__builtin_ia32_sqrtsd_mask_round)
23462         (__builtin_ia32_sqrtss_mask_round): New builtins.
23463         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23464         (__builtin_ia32_sqrtss_round): Remove.
23465         (__builtin_ia32_sqrtsd_mask_round)
23466         (__builtin_ia32_sqrtss_mask_round): New builtins.
23467         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23468         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23469         ((match_operand:VF_128 1 "vector_operand"
23470         "xBm,<round_constraint>")): Changed to ...
23471         ((match_operand:VF_128 1 "vector_operand"
23472         "xBm,<round_scalar_constraint>")): ... this.
23473         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23474         %0, %2, %<iptr>1<round_op3>}): Changed to ...
23475         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23476         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23477         %<iptr>1<round_scalar_mask_op3>}): ... this.
23478         ((set_attr "prefix" "<round_prefix>")): Changed to ...
23479         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23481 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
23483         PR target/84266
23484         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23485         Cast vec_cmpeq result to correct type.
23486         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23487         Cast vec_cmpgt result to correct type.
23489 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
23491         * final.c (final_scan_insn_1): Renamed from...
23492         (final_scan_insn): ... this.  New wrapper, to recover
23493         seen from the outermost call in recursive ones.
23494         * config/sparc/sparc.c (output_return): Drop seen from call.
23495         (output_sibcall): Likewise.
23496         * config/visium/visium.c (output_branch): Likewise.
23498 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
23500         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23501         function label.
23503 2018-02-10  Alan Modra  <amodra@gmail.com>
23505         PR target/84300
23506         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23507         Specify LR as an input.
23509 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
23511         PR sanitizer/83987
23512         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23513         remove_member_access_dummy_vars): New functions.
23514         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23515         lower_omp_1, execute_lower_omp): Use them.
23517         PR rtl-optimization/84308
23518         * shrink-wrap.c (spread_components): Release todo vector.
23520 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
23522         PR rtl-optimization/57193
23523         * ira-color.c (struct allocno_color_data): Add member
23524         conflict_allocno_hard_prefs.
23525         (update_conflict_allocno_hard_prefs): New.
23526         (bucket_allocno_compare_func): Add a preference based on
23527         conflict_allocno_hard_prefs.
23528         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23529         (color_allocnos): Remove a dead code.  Initiate
23530         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
23532 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23534         PR target/84226
23535         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23536         constraint from =wa to wa.  Avoid a subreg on the output operand,
23537         instead use a pseudo and subreg it in a move.
23538         (p9_xxbrd_<mode>): Changed to ...
23539         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23540         (p9_xxbrd_v2df): New expander.
23541         (p9_xxbrw_<mode>): Changed to ...
23542         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23543         (p9_xxbrw_v4sf): New expander.
23545 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23547         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23549 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
23551         PR target/83926
23552         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23553         multiply in 32-bit mode.
23554         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23555         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23556         mode.
23558 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23560         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23561         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23562         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23563         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23565 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
23567         PR lto/84213
23568         * dwarf2out.c (is_trivial_indirect_ref): New function.
23569         (dwarf2out_late_global_decl): Do not generate a location
23570         attribute for variables that have a non-trivial DECL_VALUE_EXPR
23571         and that are not defined in the current unit.
23573 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23575         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23576         instead of a libcall for UNORDERED.
23578 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
23580         PR target/82641
23581         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23582         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23584 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23586         PR target/PR84295
23587         * config/s390/s390.c (s390_set_current_function): Invoke
23588         s390_indirect_branch_settings also if fndecl didn't change.
23590 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23592         * config/rs6000/rs6000.md (blockage): Set length to zero.
23594 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23596         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23598 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23600         PR sanitizer/84285
23601         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23602         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23603         -static-lib*san.
23605         PR debug/84252
23606         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23607         PARALLEL incoming that failed vt_get_decl_and_offset check.
23609         PR middle-end/84237
23610         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23611         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23612         TREE_READONLY bit.
23613         (get_variable_section): For decls in named .bss* sections pass true as
23614         second argument to bss_initializer_p.
23616 2018-02-09  Marek Polacek  <polacek@redhat.com>
23617             Jakub Jelinek  <jakub@redhat.com>
23619         PR c++/83659
23620         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23621         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
23622         Sync some changes from cxx_fold_indirect_ref.
23624 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23626         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23627         markers.
23628         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23629         (BLOCK_INLINE_ENTRY_LABEL): New.
23630         (dwarf2out_var_location): Disregard inline entry markers.
23631         (inline_entry_data): New struct.
23632         (inline_entry_data_hasher): New hashtable type.
23633         (inline_entry_data_hasher::hash): New.
23634         (inline_entry_data_hasher::equal): New.
23635         (inline_entry_data_table): New variable.
23636         (add_high_low_attributes): Add DW_AT_entry_pc and
23637         DW_AT_GNU_entry_view attributes if a pending entry is found
23638         in inline_entry_data_table.  Add old entry_pc attribute only
23639         if debug nonbinding markers are disabled.
23640         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23641         markers are enabled.
23642         (block_within_block_p, dwarf2out_inline_entry): New.
23643         (dwarf2out_finish): Check that no entries remained in
23644         inline_entry_data_table.
23645         * final.c (reemit_insn_block_notes): Handle inline entry notes.
23646         (final_scan_insn, notice_source_line): Likewise.
23647         (rest_of_clean_state): Skip inline entry markers.
23648         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23649         markers.
23650         * gimple.c (gimple_build_debug_inline_entry): New.
23651         * gimple.h (enum gimple_debug_subcode): Add
23652         GIMPLE_DEBUG_INLINE_ENTRY.
23653         (gimple_build_debug_inline_entry): Declare.
23654         (gimple_debug_inline_entry_p): New.
23655         (gimple_debug_nonbind_marker_p): Adjust.
23656         * insn-notes.def (INLINE_ENTRY): New.
23657         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23658         inline entry marker notes.
23659         (print_insn): Likewise.
23660         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23661         (INSN_DEBUG_MARKER_KIND): Likewise.
23662         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23663         * tree-inline.c (expand_call_inline): Build and insert
23664         debug_inline_entry stmt.
23665         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23666         inline entry blocks early, if nonbind markers are enabled.
23667         (dump_scope_block): Dump fragment info.
23668         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23669         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23670         (gimple_build_debug_inline_entry): New.
23671         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23672         Enable/disable inline entry points too.
23673         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23674         (DEBUG_INSN): Describe inline entry markers.
23676         * common.opt (gvariable-location-views): New.
23677         (gvariable-location-views=incompat5): New.
23678         * config.in: Rebuilt.
23679         * configure: Rebuilt.
23680         * configure.ac: Test assembler for view support.
23681         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23682         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23683         * dwarf2out.c (var_loc_view): New typedef.
23684         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23685         (dwarf2out_locviews_in_attribute): New.
23686         (dwarf2out_locviews_in_loclist): New.
23687         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23688         (enum dw_line_info_opcode): Add LI_adv_address.
23689         (struct dw_line_info_table): Add view.
23690         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23691         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23692         (zero_view_p): New variable.
23693         (ZERO_VIEW_P): New macro.
23694         (output_asm_line_debug_info): New.
23695         (struct var_loc_node): Add view.
23696         (add_AT_view_list, AT_loc_list): New.
23697         (add_var_loc_to_decl): Add view param.  Test it against last.
23698         (new_loc_list): Add view params.  Record them.
23699         (AT_loc_list_ptr): Handle loc and view lists.
23700         (view_list_to_loc_list_val_node): New.
23701         (print_dw_val): Handle dw_val_class_view_list.
23702         (size_of_die): Likewise.
23703         (value_format): Likewise.
23704         (loc_list_has_views): New.
23705         (gen_llsym): Set vl_symbol too.
23706         (maybe_gen_llsym, skip_loc_list_entry): New.
23707         (dwarf2out_maybe_output_loclist_view_pair): New.
23708         (output_loc_list): Output view list or entries too.
23709         (output_view_list_offset): New.
23710         (output_die): Handle dw_val_class_view_list.
23711         (output_dwarf_version): New.
23712         (output_compilation_unit_header): Use it.
23713         (output_skeleton_debug_sections): Likewise.
23714         (output_rnglists, output_line_info): Likewise.
23715         (output_pubnames, output_aranges): Update version comments.
23716         (output_one_line_info_table): Output view numbers in asm comments.
23717         (dw_loc_list): Determine current endview, pass it to new_loc_list.
23718         Call maybe_gen_llsym.
23719         (loc_list_from_tree_1): Adjust.
23720         (add_AT_location_description): Create view list attribute if
23721         needed, check it's absent otherwise.
23722         (convert_cfa_to_fb_loc_list): Adjust.
23723         (maybe_emit_file): Call output_asm_line_debug_info for test.
23724         (dwarf2out_var_location): Reset views as needed.  Precompute
23725         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
23726         attribute.  Set view.
23727         (new_line_info_table): Reset next view.
23728         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23729         (dwarf2out_source_line): Likewise.  Output view resets and labels to
23730         the assembler, or select appropriate line info opcodes.
23731         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23732         (optimize_string_length): Catch it.  Adjust.
23733         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
23734         dw_val_class_view_list, and remove it if no longer needed.
23735         (hash_loc_list): Hash view numbers.
23736         (loc_list_hasher::equal): Compare them.
23737         (optimize_location_lists): Check whether a view list symbol is
23738         needed, and whether the locview attribute is present, and
23739         whether they match.  Remove the locview attribute if no longer
23740         needed.
23741         (index_location_lists): Call skip_loc_list_entry for test.
23742         (dwarf2out_finish): Call output_asm_line_debug_info for test.
23743         Use output_dwarf_version.
23744         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23745         (struct dw_val_node): Add val_view_list.
23746         * final.c (SEEN_NEXT_VIEW): New.
23747         (set_next_view_needed): New.
23748         (clear_next_view_needed): New.
23749         (maybe_output_next_view): New.
23750         (final_start_function): Rename to...
23751         (final_start_function_1): ... this.  Take pointer to FIRST,
23752         add SEEN parameter.  Emit param bindings in the initial view.
23753         (final_start_function): Reintroduce SEEN-less interface.
23754         (final): Rename to...
23755         (final_1): ... this.  Take SEEN parameter.  Output final pending
23756         next view at the end.
23757         (final): Reintroduce seen-less interface.
23758         (final_scan_insn): Output pending next view before switching
23759         sections or ending a block.  Mark the next view as needed when
23760         outputting variable locations.  Notify debug backend of section
23761         changes, and of location view changes.
23762         (rest_of_handle_final): Adjust.
23763         * toplev.c (process_options): Autodetect value for debug variable
23764         location views option.  Warn on incompat5 without -gdwarf-5.
23765         * doc/invoke.texi (gvariable-location-views): New.
23766         (gvariable-location-views=incompat5): New.
23767         (gno-variable-location-views): New.
23769 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
23771         PR tree-optimization/84136
23772         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23773         that the result of find_edge is non-NULL.
23775 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23777         PR target/83008
23778         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
23779         storing integer register in SImode.  Fix cost of 256 and 512
23780         byte aligned SSE register store.
23782 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23784         * config/i386/i386.c (ix86_multiplication_cost): Fix
23785         multiplication cost for TARGET_AVX512DQ.
23787 2018-02-08  Marek Polacek  <polacek@redhat.com>
23789         PR tree-optimization/84238
23790         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
23791         get_range_strlen.
23793 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23795         PR tree-optimization/84265
23796         * tree-vect-stmts.c (vectorizable_store): Don't treat
23797         VMAT_CONTIGUOUS accesses as grouped.
23798         (vectorizable_load): Likewise.
23800 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23802         PR tree-optimization/81635
23803         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
23804         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
23805         (test_round_for_mask): New functions.
23806         (wide_int_cc_tests): Call test_round_for_mask.
23807         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
23808         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
23809         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
23810         range returned by get_range_info.
23812 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23814         PR ipa/81360
23815         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
23816         * symtab.c: Include builtins.h
23817         (symtab_node::output_to_lto_symbol_table_p): Move here
23818         from lto-streamer-out.c:output_symbol_p.
23819         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
23820         (output_symbol_p): Move all logic to symtab.c
23821         (produce_symtab): Update.
23823 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23825         * config/s390/s390-opts.h (enum indirect_branch): Define.
23826         * config/s390/s390-protos.h (s390_return_addr_from_memory)
23827         (s390_indirect_branch_via_thunk)
23828         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
23829         (enum s390_indirect_branch_type): Define.
23830         * config/s390/s390.c (struct s390_frame_layout, struct
23831         machine_function): Remove.
23832         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
23833         (indirect_branch_table_label_no, indirect_branch_table_name):
23834         Define variables.
23835         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
23836         (enum s390_indirect_branch_option): Define.
23837         (s390_return_addr_from_memory): New function.
23838         (s390_handle_string_attribute): New function.
23839         (s390_attribute_table): Add new attribute handler.
23840         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
23841         (s390_indirect_branch_via_thunk): New function.
23842         (s390_indirect_branch_via_inline_thunk): New function.
23843         (s390_function_ok_for_sibcall): When jumping via thunk disallow
23844         sibling call optimization for non z10 compiles.
23845         (s390_emit_call): Force indirect branch target to be a single
23846         register.  Add r1 clobber for non-z10 compiles.
23847         (s390_emit_epilogue): Emit return jump via return_use expander.
23848         (s390_reorg): Handle JUMP_INSNs as execute targets.
23849         (s390_option_override_internal): Perform validity checks for the
23850         new command line options.
23851         (s390_indirect_branch_attrvalue): New function.
23852         (s390_indirect_branch_settings): New function.
23853         (s390_set_current_function): Invoke s390_indirect_branch_settings.
23854         (s390_output_indirect_thunk_function):  New function.
23855         (s390_code_end): Implement target hook.
23856         (s390_case_values_threshold): Implement target hook.
23857         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
23858         macros.
23859         * config/s390/s390.h (struct s390_frame_layout)
23860         (struct machine_function): Move here from s390.c.
23861         (TARGET_INDIRECT_BRANCH_NOBP_RET)
23862         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
23863         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
23864         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
23865         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
23866         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
23867         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
23868         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
23869         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
23870         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
23871         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
23872         (mnemonic attribute): Add values which aren't recognized
23873         automatically.
23874         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
23875         pattern for branch conversion.  Fix mnemonic attribute.
23876         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
23877         indirect branch via thunk if requested.
23878         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
23879         ("*indirect_jump"): Disable for branch conversion using out of
23880         line thunks.
23881         ("indirect_jump_via_thunk<mode>_z10")
23882         ("indirect_jump_via_thunk<mode>")
23883         ("indirect_jump_via_inlinethunk<mode>_z10")
23884         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
23885         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
23886         ("casesi_jump_via_inlinethunk<mode>_z10")
23887         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
23888         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
23889         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
23890         ("*indirect2_jump"): Disable for branch conversion.
23891         ("casesi_jump"): Turn into expander and expand patterns for branch
23892         conversion.
23893         ("return_use"): New expander.
23894         ("*return"): Emit return via thunk and rename it to ...
23895         ("*return<mode>"): ... this one.
23896         * config/s390/s390.opt: Add new options and and enum for the
23897         option values.
23899 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23901         * lra-constraints.c (match_reload): Unconditionally use
23902         gen_lowpart_SUBREG, rather than selecting between that
23903         and equivalent gen_rtx_SUBREG code.
23905 2018-02-08  Richard Biener  <rguenther@suse.de>
23907         PR tree-optimization/84233
23908         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
23909         changed flag instead of boguously re-using phi_inserted.
23911 2018-02-08  Martin Jambor  <mjambor@suse.cz>
23913         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
23914         static local variables.
23916 2018-02-08  Richard Biener  <rguenther@suse.de>
23918         PR tree-optimization/84278
23919         * tree-vect-stmts.c (vectorizable_store): When looking for
23920         smaller vector types to perform grouped strided loads/stores
23921         make sure the mode is supported by the target.
23922         (vectorizable_load): Likewise.
23924 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23926         * config/aarch64/aarch64.c (aarch64_components_for_bb):
23927         Increase LDP/STP opportunities by adding adjacent callee-saves.
23929 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23931         PR rtl-optimization/84068
23932         PR rtl-optimization/83459
23933         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
23935 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
23937         PR tree-optimization/84224
23938         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
23939         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
23940         non-zero arguments.
23942 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
23944         PR target/84113
23945         * config/rs6000/altivec.md (*restore_world): Remove LR use.
23946         * config/rs6000/predicates.md (restore_world_operation): Adjust op
23947         count, remove one USE.
23949 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
23951         * doc/install.texi (Configuration): Document the
23952         --with-long-double-format={ibm,ieee} PowerPC configuration
23953         options.
23955         PR target/84154
23956         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
23957         Convert from define_expand to be define_insn_and_split.  Rework
23958         float/double/_Float128 conversions to QI/HI/SImode to work with
23959         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
23960         conversions to QI/HImode types did a store and then a load to
23961         truncate the value.  For conversions to VSX registers, don't split
23962         the insn, instead emit the code directly.  Use the code iterator
23963         any_fix to combine signed and unsigned conversions.
23964         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
23965         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
23966         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23967         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
23968         (fix_<mode>di2_hw): Likewise.
23969         (fixuns_<mode>di2_hw): Likewise.
23970         (fix_<mode>si2_hw): Likewise.
23971         (fixuns_<mode>si2_hw): Likewise.
23972         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
23973         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23974         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
23975         fix<uns>_trunc<SFDF:mode>si2_p8.
23976         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
23977         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23978         (fix<uns>_<mode>_mem): Likewise.
23979         (fctiw<u>z_<mode>_mem): Likewise.
23980         (fix<uns>_<mode>_mem): Likewise.
23981         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
23982         the register allocator from doing a direct move to the GPRs to do
23983         a store, and instead use the ISA 3.0 store byte/half-word from
23984         vector register instruction.  For IEEE 128-bit floating point,
23985         also optimize stores of 32-bit ints.
23986         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
23988 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
23990         * genextract.c (push_pathstr_operand): New function to support
23991         [a-zA-Z].
23992         (walk_rtx): Call push_pathstr_operand.
23993         (print_path): Support [a-zA-Z].
23995 2018-02-07  Richard Biener  <rguenther@suse.de>
23997         PR tree-optimization/84037
23998         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
23999         (cse_and_gimplify_to_preheader): Declare.
24000         (vect_get_place_in_interleaving_chain): Likewise.
24001         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24002         ivexpr_map.
24003         (_loop_vec_info::~_loop_vec_info): Delete it.
24004         (cse_and_gimplify_to_preheader): New function.
24005         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
24006         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
24007         (vectorizable_load): Likewise.  For grouped stores always base
24008         the IV on the first element.
24009         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
24010         condition before gimplifying.
24012 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24014         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
24015         *DIV_EXPR and *MOD_EXPR.
24017 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
24019         PR target/84248
24020         * config/i386/i386.c (ix86_option_override_internal): Mask out
24021         the CF_SET bit when checking -fcf-protection.
24023 2018-02-07  Tom de Vries  <tom@codesourcery.com>
24025         PR libgomp/84217
24026         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
24027         enough.
24029 2018-02-07  Richard Biener  <rguenther@suse.de>
24031         PR tree-optimization/84204
24032         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
24033         this place.
24035         PR tree-optimization/84205
24036         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
24037         special-case isl_ast_op_zdiv_r.
24039         PR tree-optimization/84223
24040         * graphite-scop-detection.c (gather_bbs::before_dom_children):
24041         Only add conditions from within the region.
24042         (gather_bbs::after_dom_children): Adjust.
24044 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
24046         PR target/84209
24047         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
24048         * config/avr/avr.md: Only post-reload split REG-REG moves if
24049         either register is GENERAL_REG_P.
24051 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24053         PR tree-optimization/84235
24054         * tree-ssa-scopedtables.c
24055         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
24056         if the subtraction is performed in floating point type where NaNs are
24057         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
24058         build 1.  Formatting fix.
24060 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
24062         PR target/84146
24063         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
24064         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
24065         and skip it regardless of bb boundaries.  Use CALL_P macro,
24066         don't test INSN_P (insn) together with CALL_P or JUMP_P check
24067         unnecessarily, formatting fix.
24069 2018-02-06  Michael Collison  <michael.collison@arm.com>
24071         * config/arm/thumb2.md:
24072         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
24073         (*thumb_mov_notscc): Ditto.
24075 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
24077         PR target/84154
24078         * config/rs6000/rs6000.md (su code attribute): Use "u" for
24079         unsigned_fix, not "s".
24081 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24083         * configure.ac (gcc_fn_eh_frame_ro): New function.
24084         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
24085         correct .eh_frame permissions.
24086         * configure: Regenerate.
24088 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
24090         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
24091         irrelevant options.
24093 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24095         * config/rs6000/rs6000.c (rs6000_option_override_internal):
24096         Display warning message for -mno-speculate-indirect-jumps.
24098 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
24100         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
24101         Undocumented.
24102         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
24104 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
24106         PR tree-optimization/84225
24107         * tree-eh.c (find_trapping_overflow): Only call
24108         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
24110 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24112         PR target/84145
24113         * config/i386/i386.c: Reimplement the check of possible options
24114         -mibt/-mshstk conbination. Change error messages.
24115         * doc/invoke.texi: Fix a typo: remove extra '='.
24117 2018-02-06  Marek Polacek  <polacek@redhat.com>
24119         PR tree-optimization/84228
24120         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
24122 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
24124         PR target/82641
24125         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
24126         emitted arch directives.
24127         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
24128         __ARM_FEATURE_COPROC before changing architectures.
24130 2018-02-06  Richard Biener  <rguenther@suse.de>
24132         * config/i386/i386.c (print_reg): Fix typo.
24133         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
24135 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
24137         * configure: Regenerate.
24139 2018-02-05  Martin Sebor  <msebor@redhat.com>
24141         PR tree-optimization/83369
24142         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
24143         inlining context.
24145 2018-02-05  Martin Liska  <mliska@suse.cz>
24147         * doc/invoke.texi: Cherry-pick upstream r323995.
24149 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
24151         * ira.c (ira_init_register_move_cost): Adjust comment.
24153 2018-02-05  Martin Liska  <mliska@suse.cz>
24155         PR gcov-profile/84137
24156         * doc/gcov.texi: Fix typo in documentation.
24158 2018-02-05  Martin Liska  <mliska@suse.cz>
24160         PR gcov-profile/83879
24161         * doc/gcov.texi: Document necessity of --dynamic-list-data when
24162         using dlopen functionality.
24164 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
24166         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
24167         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
24168         _mm_maskz_range_ss, _mm_mask_range_round_ss,
24169         _mm_maskz_range_round_ss): New intrinsics.
24170         (__builtin_ia32_rangesd128_round)
24171         (__builtin_ia32_rangess128_round): Remove.
24172         (__builtin_ia32_rangesd128_mask_round,
24173         __builtin_ia32_rangess128_mask_round): New builtins.
24174         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
24175         __builtin_ia32_rangess128_round): Remove.
24176         (__builtin_ia32_rangesd128_mask_round,
24177         __builtin_ia32_rangess128_mask_round): New builtins.
24178         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
24179         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
24180         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
24181         "<round_saeonly_constraint>")): Changed to ...
24182         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
24183         "<round_saeonly_scalar_constraint>")): ... this.
24184         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
24185         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
24186         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
24187         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
24188         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
24190 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
24192         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
24193         options.
24194         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
24195         Remove all values except native, 8540 and 8548.
24197 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
24199         * config/i386/i386.c (ix86_output_function_return): Pass
24200         INVALID_REGNUM, instead of -1, as invalid register number to
24201         indirect_thunk_name and output_indirect_thunk.
24203 2018-02-02  Julia Koval  <julia.koval@intel.com>
24205         * config.gcc: Add -march=icelake.
24206         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
24207         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
24208         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
24209         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
24210         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
24211         (processor_target_table): Add icelake.
24212         (ix86_option_override_internal): Handle new PTAs.
24213         (get_builtin_code_for_version): Handle icelake.
24214         (M_INTEL_COREI7_ICELAKE): New.
24215         (fold_builtin_cpu): Handle icelake.
24216         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
24217         * doc/invoke.texi: Add -march=icelake.
24219 2018-02-02  Julia Koval  <julia.koval@intel.com>
24221         * config/i386/i386.c (ix86_option_override_internal): Change flags type
24222         to wide_int_bitmask.
24223         * wide-int-bitmask.h: New.
24225 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24227         PR target/84066
24228         * config/i386/i386.md: Replace Pmode with word_mode in
24229         builtin_setjmp_setup and builtin_longjmp to support x32.
24231 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
24233         PR target/56010
24234         PR target/83743
24235         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
24236         #include "opts.h".
24237         (rs6000_supported_cpu_names): New static variable.
24238         (linux_cpu_translation_table): Likewise.
24239         (elf_platform) <cpu>: Define new static variable and use it.
24240         Translate kernel AT_PLATFORM name to canonical name if needed.
24241         Error if platform name is unknown.
24243 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
24245         PR target/84089
24246         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
24248 2018-02-01  Jeff Law  <law@redhat.com>
24250         PR target/84128
24251         * config/i386/i386.c (release_scratch_register_on_entry): Add new
24252         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
24253         the scratch if RELEASE_VIA_POP is false.
24254         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
24255         If we have to save a temporary register, decrement SIZE appropriately.
24256         Pass new arguments to release_scratch_register_on_entry.
24257         (ix86_adjust_stack_and_probe): Likewise.
24258         (ix86_emit_probe_stack_range): Pass new arguments to
24259         release_scratch_register_on_entry.
24261 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
24263         PR rtl-optimization/84157
24264         * combine.c (change_zero_ext): Use REG_P predicate in
24265         front of HARD_REGISTER_P predicate.
24267 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
24269         * config/avr/avr.c (avr_option_override): Move disabling of
24270         -fdelete-null-pointer-checks to...
24271         * common/config/avr/avr-common.c (avr_option_optimization_table):
24272         ...here.
24274 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24276         PR tree-optimization/81635
24277         * tree-data-ref.c (split_constant_offset_1): For types that
24278         wrap on overflow, try to use range info to prove that wrapping
24279         cannot occur.
24281 2018-02-01  Renlin Li  <renlin.li@arm.com>
24283         PR target/83370
24284         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
24285         TAILCALL_ADDR_REGS.
24286         (aarch64_register_move_cost): Likewise.
24287         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
24288         TAILCALL_ADDR_REGS.
24289         (REG_CLASS_NAMES): Likewise.
24290         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
24291         TAILCALL_ADDR_REGS. Remove IP registers.
24292         * config/aarch64/aarch64.md (Ucs): Update register constraint.
24294 2018-02-01  Richard Biener  <rguenther@suse.de>
24296         * domwalk.h (dom_walker::dom_walker): Add additional constructor
24297         for specifying RPO order and allow NULL for that.
24298         * domwalk.c (dom_walker::dom_walker): Likewise.
24299         (dom_walker::walk): Handle NULL RPO order.
24300         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
24301         in RPO order.
24302         (rewrite_update_dom_walker): Likewise.
24303         (mark_def_dom_walker): Likewise.
24305 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24307         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
24308         (aarch64_maybe_expand_sve_subreg_move): Declare.
24309         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
24310         * config/aarch64/predicates.md (aarch64_any_register_operand): New
24311         predicate.
24312         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
24313         that are semantically a reverse operation.
24314         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
24315         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
24316         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
24317         functions.
24318         (aarch64_can_change_mode_class): For big-endian, forbid changes
24319         between two SVE modes if they have different element sizes.
24321 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24323         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
24324         the TImode handling for big-endian targets.
24326 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24328         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
24329         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
24330         not just bytes.
24331         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
24332         Remove BSWAP handing for big-endian targets and use the form of
24333         LD1RQ appropariate for the mode.
24335 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24337         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
24338         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
24339         duplicated element.
24341 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24343         PR tearget/83845
24344         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
24345         check for operands that need to go through aarch64_sve_reload_be.
24347 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
24349         PR tree-optimization/81661
24350         PR tree-optimization/84117
24351         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
24352         * tree-eh.c: Include gimplify.h.
24353         (find_trapping_overflow, replace_trapping_overflow,
24354         rewrite_to_non_trapping_overflow): New functions.
24355         * tree-vect-loop.c: Include tree-eh.h.
24356         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
24357         * tree-data-ref.c: Include tree-eh.h.
24358         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
24360 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
24362         PR rtl-optimization/84123
24363         * combine.c (change_zero_ext): Check if hard register satisfies
24364         can_change_dest_mode before calling gen_lowpart_SUBREG.
24366 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
24368         PR target/82444
24369         * ira.c (ira_init_register_move_cost): Remove assert.
24371 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24373         PR rtl-optimization/84071
24374         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
24375         * doc/tm.texi: Regenerate.
24377 2018-01-31  Richard Biener  <rguenther@suse.de>
24379         PR tree-optimization/84132
24380         * tree-data-ref.c (analyze_miv_subscript): Properly
24381         check whether evolution_function_is_affine_multivariate_p
24382         before calling gcd_of_steps_may_divide_p.
24384 2018-01-31  Julia Koval  <julia.koval@intel.com>
24386         PR target/83618
24387         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
24388         * config/i386/i386.md (rdpid_rex64) New.
24389         (rdpid): Make 32bit only.
24391 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
24393         PR lto/84105
24394         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
24395         an IDENTIFIER_NODE for FUNCTION_TYPE's.
24397 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24399         Revert
24400         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24402         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24404 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24406         PR rtl-optimization/84071
24407         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24408         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24410 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24412         * config/arc/arc.c (arc_handle_aux_attribute): New function.
24413         (arc_attribute_table): Add 'aux' attribute.
24414         (arc_in_small_data_p): Consider aux like variables.
24415         (arc_is_aux_reg_p): New function.
24416         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24417         (arc_get_aux_arg): New function.
24418         (prepare_move_operands): Handle aux-register access.
24419         (arc_handle_aux_attribute): New function.
24420         * doc/extend.texi (ARC Variable attributes): Add subsection.
24422 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24424         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24425         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24426         (arc_attribute_table): Add 'uncached' attribute.
24427         (arc_print_operand): Print '.di' flag for uncached memory
24428         accesses.
24429         (arc_in_small_data_p): Do not consider for small data the uncached
24430         types.
24431         (arc_is_uncached_mem_p): New function.
24432         * config/arc/predicates.md (compact_store_memory_operand): Check
24433         for uncached memory accesses.
24434         (nonvol_nonimm_operand): Likewise.
24435         * doc/extend.texi (ARC Type Attribute): New subsection.
24437 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
24439         PR c/84100
24440         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24441         falign-loops=): Add Optimization flag.
24443 2018-01-30  Jeff Law  <law@redhat.com>
24445         PR target/84064
24446         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24447         INT_REGISTERS_SAVED.  Check it prior to calling
24448         get_scratch_register_on_entry.
24449         (ix86_adjust_stack_and_probe): Similarly.
24450         (ix86_emit_probe_stack_range): Similarly.
24451         (ix86_expand_prologue): Corresponding changes.
24453 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24455         PR target/40411
24456         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24457         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24459 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
24461         PR target/84112
24462         * lra-constraints.c (curr_insn_transform): Process AND in the
24463         address.
24465 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
24467         PR rtl-optimization/83986
24468         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24469         dependence against last_pending_memory_flush in addition to
24470         pending_jump_insns.
24472 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
24474         PR tree-optimization/81611
24475         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24476         copies.
24478 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24480         PR target/83758
24481         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24482         a reg rtx.
24484 2018-01-30  Richard Biener  <rguenther@suse.de>
24485             Jakub Jelinek  <jakub@redhat.com>
24487         PR tree-optimization/84111
24488         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24489         inner loops added during recursion, as they don't have up-to-date
24490         SSA form.
24492 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24494         PR ipa/81360
24495         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24496         (can_inline_edge_by_limits_p): ... here.
24497         (can_early_inline_edge_p, check_callers,
24498         update_caller_keys, update_callee_keys, recursive_inlining,
24499         add_new_edges_to_heap, speculation_useful_p,
24500         inline_small_functions,
24501         inline_small_functions, flatten_function,
24502         inline_to_all_callers_1): Update.
24504 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24506         * profile-count.c (profile_count::combine_with_ipa_count): Handle
24507         zeros correctly.
24509 2018-01-30  Richard Biener  <rguenther@suse.de>
24511         PR tree-optimization/83008
24512         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24513         invariant and constant vector uses in stmts when they need
24514         more than one stmt.
24516 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24518         PR bootstrap/84017
24519         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24520         * configure: Regenerate.
24522 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24524         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24525         pattern.
24526         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24527         Use gen_rtx_REG rather than gen_lowpart.
24529 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24531         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24532         rather than 0 when creating partial subregs.
24534 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24536         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24537         of usage.
24539 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
24541         PR target/81550
24542         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24543         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24544         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24545         flags.  This restores the settings used before the 2017-07-24.
24546         Turning off pre increment/decrement/modify allows IVOPTS to
24547         optimize DF/SF loops where the index is an int.
24549 2018-01-29  Richard Biener  <rguenther@suse.de>
24550             Kelvin Nilsen  <kelvin@gcc.gnu.org>
24552         PR bootstrap/80867
24553         * tree-vect-stmts.c (vectorizable_call): Don't call
24554         targetm.vectorize_builtin_md_vectorized_function if callee is
24555         NULL.
24557 2018-01-22  Carl Love  <cel@us.ibm.com>
24559         * doc/extend.tex: Fix typo in second arg in
24560         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24562 2018-01-29  Richard Biener  <rguenther@suse.de>
24564         PR tree-optimization/84086
24565         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24566         (flush_ssaname_freelist): When SSA names were released reset
24567         the SCEV hash table.
24569 2018-01-29  Richard Biener  <rguenther@suse.de>
24571         PR tree-optimization/84057
24572         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24573         removed paths when removing edges.
24575 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
24577         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24578         -mfunction-return=@var{choice}.
24580 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
24582         PR diagnostic/84034
24583         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24584         Handle CR like TAB.
24585         (layout::print_source_line): Likewise.
24586         (test_get_line_width_without_trailing_whitespace): Add test cases.
24588 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
24590         PR middle-end/84040
24591         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24592         debug insns.
24594 2018-01-26  Jim Wilson  <jimw@sifive.com>
24596         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24598         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24599         specified.
24601 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24603         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24604         and CMP + SUB-immediate -> SUBS.
24606 2018-01-26  Martin Sebor  <msebor@redhat.com>
24608         PR tree-optimization/83896
24609         * tree-ssa-strlen.c (get_string_len): Rename...
24610         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
24611         Avoid assuming length is constant.
24612         (handle_char_store): Use HOST_WIDE_INT for string length.
24614 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
24616         PR target/81763
24617         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24618         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24620 2018-01-26  Richard Biener  <rguenther@suse.de>
24622         PR rtl-optimization/84003
24623         * dse.c (record_store): Only record redundant stores when
24624         the earlier store aliases at least all accesses the later one does.
24626 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
24628         PR rtl-optimization/83985
24629         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24630         REG_CFA_RESTORE insns.
24631         (delete_unmarked_insns): Don't ignore separate shrink wrapping
24632         REG_CFA_RESTORE insns here.
24634         PR c/83989
24635         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24636         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24638 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24640         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24641         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24642         (arc_init): Likewise.
24643         (arc_override_options): Likewise.
24644         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24645         value.
24646         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24647         support.
24648         * config/arc/arc.h (TARGET_DBNZ): Define.
24649         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24650         properly set the tune attribute.
24651         (dbnz): Use TARGET_DBNZ guard.
24652         * config/arc/arc.opt (mtune): Add core3 option.
24654 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24656         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24657         recognize new pic like addresses.
24658         (arc_delegitimize_address): Clean up.
24660 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24662         * config/arc/arc-arches.def: Option mrf16 valid for all
24663         architectures.
24664         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24665         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24666         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24667         * config/arc/arc-tables.opt: Regenerate.
24668         * config/arc/arc.c (arc_conditional_register_usage): Handle
24669         reduced register file case.
24670         (arc_file_start): Set must have build attributes.
24671         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24672         mrf16 option value.
24673         * config/arc/arc.opt (mrf16): Add new option.
24674         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24675         * config/arc/genmultilib.awk: Handle new mrf16 option.
24676         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24677         * config/arc/t-multilib: Regenerate.
24678         * doc/invoke.texi (ARC Options): Document mrf16 option.
24680 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24682         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24683         * config/arc/arc.c (arc_handle_secure_attribute): New function.
24684         (arc_attribute_table): Add 'secure_call' attribute.
24685         (arc_print_operand): Print secure call operand.
24686         (arc_function_ok_for_sibcall): Don't optimize tail calls when
24687         secure.
24688         (arc_is_secure_call_p): New function.  * config/arc/arc.md
24689         (call_i): Add support for sjli instruction.
24690         (call_value_i): Likewise.
24691         * config/arc/constraints.md (Csc): New constraint.
24693 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24694             John Eric Martin  <John.Martin@emmicro-us.com>
24696         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24697         * config/arc/arc.c (_arc_jli_section): New struct.
24698         (arc_jli_section): New type.
24699         (rc_jli_sections): New static variable.
24700         (arc_handle_jli_attribute): New function.
24701         (arc_attribute_table): Add jli_always and jli_fixed attribute.
24702         (arc_file_end): New function.
24703         (TARGET_ASM_FILE_END): Define.
24704         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24705         (arc_add_jli_section): New function.
24706         (jli_call_scan): Likewise.
24707         (arc_reorg): Call jli_call_scan.
24708         (arc_output_addsi): Remove 'S' from printing asm operand.
24709         (arc_is_jli_call_p): New function.
24710         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24711         operand.
24712         (movhi_insn): Likewise.
24713         (movsi_insn): Likewise.
24714         (movsi_set_cc_insn): Likewise.
24715         (loadqi_update): Likewise.
24716         (load_zeroextendqisi_update): Likewise.
24717         (load_signextendqisi_update): Likewise.
24718         (loadhi_update): Likewise.
24719         (load_zeroextendhisi_update): Likewise.
24720         (load_signextendhisi_update): Likewise.
24721         (loadsi_update): Likewise.
24722         (loadsf_update): Likewise.
24723         (movsicc_insn): Likewise.
24724         (bset_insn): Likewise.
24725         (bxor_insn): Likewise.
24726         (bclr_insn): Likewise.
24727         (bmsk_insn): Likewise.
24728         (bicsi3_insn): Likewise.
24729         (cmpsi_cc_c_insn): Likewise.
24730         (movsi_ne): Likewise.
24731         (movsi_cond_exec): Likewise.
24732         (clrsbsi2): Likewise.
24733         (norm_f): Likewise.
24734         (normw): Likewise.
24735         (swap): Likewise.
24736         (divaw): Likewise.
24737         (flag): Likewise.
24738         (sr): Likewise.
24739         (kflag): Likewise.
24740         (ffs): Likewise.
24741         (ffs_f): Likewise.
24742         (fls): Likewise.
24743         (call_i): Remove 'S' asm letter, add jli instruction.
24744         (call_value_i): Likewise.
24745         * config/arc/arc.op (mjli-always): New option.
24746         * config/arc/constraints.md (Cji): New constraint.
24747         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24748         operand.
24749         (subsf3_fpx): Likewise.
24750         (mulsf3_fpx): Likewise.
24751         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24752         asm operand.
24753         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24754         function attrbutes.
24755         * doc/invoke.texi (ARC): Document mjli-always option.
24757 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
24759         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24760         avoid addition with 0 and use incw and decw where possible.
24762 2018-01-26  Richard Biener  <rguenther@suse.de>
24764         PR tree-optimization/81082
24765         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24766         association if it requires casting to unsigned.
24767         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24768         from fold_plusminus_mult_expr to catch important cases late when
24769         range info is available.
24771 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24773         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24774         * configure.ac (hidden_linkonce): New test.
24775         * configure: Regenerate.
24776         * config.in: Regenerate.
24778 2018-01-26  Julia Koval  <julia.koval@intel.com>
24780         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
24781         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
24782         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
24783         _mm_mask_bitshuffle_epi64_mask): Fix type.
24784         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
24785         USI_FTYPE_V4DI_V4DI_USI): Remove.
24786         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
24787         __builtin_ia32_vpshufbitqmb256_mask,
24788         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
24789         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
24790         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
24792 2018-01-26  Alan Modra  <amodra@gmail.com>
24794         PR target/84033
24795         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
24796         UNSPEC_VBPERMQ.  Sort other unspecs.
24798 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
24800         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
24802 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
24804         PR middle-end/83055
24805         * predict.c (drop_profile): Do not push/pop cfun; update also
24806         node->count.
24807         (handle_missing_profiles): Fix logic looking for zero profiles.
24809 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
24811         PR middle-end/83977
24812         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
24813         on functions with #pragma omp declare simd or functions with simd
24814         attribute.
24815         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
24816         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
24817         Remove trailing \n from warning_at calls.
24819 2018-01-25  Tom de Vries  <tom@codesourcery.com>
24821         PR target/84028
24822         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24823         for neutered workers.
24825 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
24827         PR target/68467
24828         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
24829         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
24831 2018-01-24  Jeff Law  <law@redhat.com>
24833         PR target/83994
24834         * i386.c (get_probe_interval): Move to earlier point.
24835         (ix86_compute_frame_layout): If -fstack-clash-protection and
24836         the frame is larger than the probe interval, then use pushes
24837         to save registers rather than reg->mem moves.
24838         (ix86_expand_prologue): Remove conditional for int_registers_saved
24839         assertion.
24841 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
24843         PR target/84014
24844         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
24845         min/max for never referenced object.
24847 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
24849         PR middle-end/83977
24850         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
24851         here.
24852         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
24853         attributes from DECL_ATTRIBUTES (decl) without affecting
24854         DECL_ATTRIBUTES (current_function_decl).
24855         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
24856         functions with non-NULL DECL_ABSTRACT_ORIGIN.
24858 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
24860         PR tree-optimization/83979
24861         * fold-const.c (fold_comparison): Use constant_boolean_node
24862         instead of boolean_{true,false}_node.
24864 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
24866         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
24867         with zero counts.
24869 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24871         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24872         Simplify the clause that sets the length attribute.
24873         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24874         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
24875         clause that sets the length attribute.
24876         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24878 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24880         PR target/83589
24881         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
24882         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
24883         Add strict parameter.
24884         (prevent_branch_around_nothing): Insert dummy insn between branch to
24885         label and label with no ptx insn inbetween.
24886         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
24888 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24890         PR target/81352
24891         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24892         for neutered threads in warp.
24893         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
24895 2018-01-24  Richard Biener  <rguenther@suse.de>
24897         PR tree-optimization/83176
24898         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
24899         operands.
24901 2018-01-24  Richard Biener  <rguenther@suse.de>
24903         PR tree-optimization/82819
24904         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
24905         code generating pluses that are no-ops in the target precision.
24907 2018-01-24  Richard Biener  <rguenther@suse.de>
24909         PR middle-end/84000
24910         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
24912 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24914         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
24915         to merge probabilities.
24916         * predict.c (probably_never_executed): Also mark as cold functions
24917         with global 0 profile and guessed local profile.
24918         * profile-count.c (profile_probability::combine_with_count): New
24919         member function.
24920         * profile-count.h (profile_probability::operator*,
24921         profile_probability::operator*=, profile_probability::operator/,
24922         profile_probability::operator/=): Reduce precision to adjusted
24923         and set value to guessed on contradictory divisions.
24924         (profile_probability::combine_with_freq): Remove.
24925         (profile_probability::combine_wiht_count): Declare.
24926         (profile_count::force_nonzero):: Set to adjusted.
24927         (profile_count::probability_in):: Set quality to adjusted.
24928         * tree-ssa-tail-merge.c (replace_block_by): Use
24929         combine_with_count.
24931 2018-01-23  Andrew Waterman  <andrew@sifive.com>
24932             Jim Wilson  <jimw@sifive.com>
24934         * config/riscv/riscv.c (riscv_stack_boundary): New.
24935         (riscv_option_override): Set riscv_stack_boundary.  Handle
24936         riscv_preferred_stack_boundary_arg.
24937         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
24938         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
24939         (STACK_BOUNDARY): Set to riscv_stack_boundary.
24940         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
24941         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
24942         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
24944 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
24946         PR target/83905
24947         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
24948         of struct ix86_frame.
24949         (ix86_expand_epilogue): Likewise.  Add a local variable for
24950         the reg_save_offset field in struct ix86_frame.
24952 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
24954         PR tree-optimization/82604
24955         * tree-loop-distribution.c (enum partition_kind): New enum item
24956         PKIND_PARTIAL_MEMSET.
24957         (partition_builtin_p): Support above new enum item.
24958         (generate_code_for_partition): Ditto.
24959         (compute_access_range): Differentiate cases that equality can be
24960         proven at all loops, the innermost loops or no loops.
24961         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
24962         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
24963         (finalize_partitions, distribute_loop): Don't fuse partition of
24964         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
24965         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
24966         parloop is enabled.
24968 2018-01-23  Martin Liska  <mliska@suse.cz>
24970         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
24971         order to ignore the predictor.
24972         (PRED_POLYMORPHIC_CALL): Likewise.
24973         (PRED_RECURSIVE_CALL): Likewise.
24975 2018-01-23  Martin Liska  <mliska@suse.cz>
24977         * tree-profile.c (tree_profiling): Print function header to
24978         aware reader which function we are working on.
24979         * value-prof.c (gimple_find_values_to_profile): Do not print
24980         not interesting value histograms.
24982 2018-01-23  Martin Liska  <mliska@suse.cz>
24984         * profile-count.h (enum profile_quality): Add
24985         profile_uninitialized as the first value. Do not number values
24986         as they are zero based.
24987         (profile_count::verify): Update sanity check.
24988         (profile_probability::verify): Likewise.
24990 2018-01-23  Nathan Sidwell  <nathan@acm.org>
24992         * doc/invoke.texi (ffor-scope): Deprecate.
24994 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24996         PR tree-optimization/83510
24997         * domwalk.c (set_all_edges_as_executable): New function.
24998         (dom_walker::dom_walker): Convert bool param
24999         "skip_unreachable_blocks" to enum reachability.  Move setup of
25000         edge flags to set_all_edges_as_executable and only do it when
25001         reachability is REACHABLE_BLOCKS.
25002         * domwalk.h (enum dom_walker::reachability): New enum.
25003         (dom_walker::dom_walker): Convert bool param
25004         "skip_unreachable_blocks" to enum reachability.
25005         (set_all_edges_as_executable): New decl.
25006         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
25007         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
25008         "reachability".
25009         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
25010         but converting true to REACHABLE_BLOCKS.
25011         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
25012         * tree-vrp.c
25013         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
25014         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
25015         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
25016         REACHABLE_BLOCKS.
25017         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
25018         if check_all_array_refs will be called.
25020 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25022         * tree.c (selftest::test_location_wrappers): Add more test
25023         coverage.
25025 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25027         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
25028         (selftest::test_bit_in_range): Likewise.
25030 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25032         PR testsuite/83888
25033         * doc/sourcebuild.texi (vect_float): Say that the selector
25034         only describes the situation when -funsafe-math-optimizations is on.
25035         (vect_float_strict): Document.
25037 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25039         PR tree-optimization/83965
25040         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
25041         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
25042         instead of checking only for a reduction.
25043         (vect_recog_widen_sum_pattern): Likewise.
25045 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25047         * predict.c (probably_never_executed): Only use precise profile info.
25048         (compute_function_frequency): Skip after inlining hack since we now
25049         have quality checking.
25051 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25053         * profile-count.h (profile_probability::very_unlikely,
25054         profile_probability::unlikely, profile_probability::even): Set
25055         precision to guessed.
25057 2018-01-23  Richard Biener  <rguenther@suse.de>
25059         PR tree-optimization/83963
25060         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
25061         Properly terminate dominator walk when crossing the exit edge not
25062         when visiting its source block.
25064 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
25066         PR c++/83918
25067         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
25068         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
25070 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
25072         PR tree-optimization/83957
25073         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
25074         semicolon after for body surrounded by braces.
25076         PR tree-optimization/83081
25077         * profile-count.h (profile_probability::split): New method.
25078         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
25079         Use profile_probability::split.
25080         (do_compare_rtx_and_jump): Fix adjustment of probabilities
25081         when splitting a single conditional jump into 2.
25083 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
25085         PR tree-optimization/69452
25086         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
25087         decl.
25089 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25091         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
25092         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
25093         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
25095 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25097         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
25098         declaration.
25099         * config/rl78/rl78.md (movdi): New define_expand.
25100         * config/rl78/rl78.c (rl78_split_movdi): New function.
25102 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
25104         PR target/83862
25105         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
25106         no longer used.
25107         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
25108         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
25109         128-bit to produce an UNSPEC move to get the double word with the
25110         signbit and then a shift directly to do signbit.
25111         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
25112         implementation with a new version that just does either a direct
25113         move or a regular move.  Move memory interface to separate insns.
25114         Move insns so they are next to the expander.
25115         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
25116         with signbit move.  Split big and little endian case.
25117         (signbit<mode>2_dm_mem_le): Likewise.
25118         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
25119         (signbit<mode>2_dm2): Likewise.
25121 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25123         * config/rl78/rl78.md (anddi3): New define_expand.
25125 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25127         * config/rl78/rl78.md (umindi3): New define_expand.
25129 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25131         * config/rl78/rl78.md (smindi3): New define_expand.
25133 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25135         * config/rl78/rl78.md (smaxdi3): New define_expand.
25137 2018-01-22  Carl Love  <cel@us.ibm.com>
25139         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
25140         LVX_V1TI): Add macro expansion.
25141         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
25142         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
25143         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
25144         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
25145         Change check to determine if the instruction is a byte reversing
25146         entry.  Fix typo in comment.
25147         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
25148         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
25149         Add def_builtin calls for new builtins.
25150         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
25151         Add define_insn expansion.
25153 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25155         * config/rl78/rl78.md (umaxdi3): New define_expand.
25157 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25159         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
25160         for non-QImode registers.
25162 2018-01-22  Richard Biener  <rguenther@suse.de>
25164         PR tree-optimization/83963
25165         * graphite-scop-detection.c (scop_detection::get_sese): Delay
25166         including the loop exit block.
25167         (scop_detection::merge_sese): Likewise.
25168         (scop_detection::add_scop): Do it here instead.
25170 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25172         * doc/sourcebuild.texi (arm_softfloat): Document.
25174 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
25176         PR gcc/77734
25177         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
25178         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
25179         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
25181 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25182             David Edelsohn  <dje.gcc@gmail.com>
25184         PR target/83946
25185         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25186         Change "crset eq" to "crset 2".
25187         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25188         (*call_indirect_aix<mode>_nospec): Likewise.
25189         (*call_value_indirect_aix<mode>_nospec): Likewise.
25190         (*call_indirect_elfv2<mode>_nospec): Likewise.
25191         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
25192         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
25193         change assembly output from . to $.
25194         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25195         (indirect_jump<mode>_nospec): Change assembly output from . to $.
25196         (*tablejump<mode>_internal1_nospec): Likewise.
25198 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
25200         PR target/80870
25201         * config/sh/sh_optimize_sett_clrt.cc:
25202         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
25204 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
25206         PR tree-optimization/83940
25207         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
25208         offset_dt to vect_constant_def rather than vect_unknown_def_type.
25209         (vect_check_load_store_mask): Add a mask_dt_out parameter and
25210         use it to pass back the definition type.
25211         (vect_check_store_rhs): Likewise rhs_dt_out.
25212         (vect_build_gather_load_calls): Add a mask_dt argument and use
25213         it instead of a call to vect_is_simple_use.
25214         (vectorizable_store): Update calls to vect_check_load_store_mask
25215         and vect_check_store_rhs.  Use the dt returned by the latter instead
25216         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
25217         instead of calls to vect_is_simple_use.  Pass the scalar rather
25218         than the vector operand to vect_is_simple_use when handling
25219         second and subsequent copies of an rhs value.
25220         (vectorizable_load): Update calls to vect_check_load_store_mask
25221         and vect_build_gather_load_calls.  Use the cached mask_dt and
25222         gs_info.offset_dt instead of calls to vect_is_simple_use.
25224 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
25226         PR middle-end/83945
25227         * tree-emutls.c: Include gimplify.h.
25228         (lower_emutls_2): New function.
25229         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
25230         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
25231         it before further processing.
25233         PR target/83930
25234         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
25235         UINTVAL (trueop1) instead of INTVAL (op1).
25237 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
25239         PR debug/81570
25240         PR debug/83728
25241         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
25242         INCOMING_FRAME_SP_OFFSET if not defined.
25243         (scan_trace): Add ENTRY argument.  If true and
25244         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
25245         emit a note to adjust the CFA offset.
25246         (create_cfi_notes): Adjust scan_trace callers.
25247         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
25248         INCOMING_FRAME_SP_OFFSET in the CIE.
25249         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
25250         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
25251         Likewise.
25252         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
25253         * doc/tm.texi: Regenerated.
25255 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25257         PR rtl-optimization/83147
25258         * lra-constraints.c (remove_inheritance_pseudos): Use
25259         lra_substitute_pseudo_within_insn.
25261 2018-01-19  Tom de Vries  <tom@codesourcery.com>
25262             Cesar Philippidis  <cesar@codesourcery.com>
25264         PR target/83920
25265         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
25267 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
25269         PR target/83790
25270         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
25271         spaces for function labels.
25273 2018-01-19  Martin Liska  <mliska@suse.cz>
25275         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
25276         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
25277         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
25278         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
25279         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
25280         (PRED_CONST_RETURN): Change from 69 to 65.
25281         (PRED_NULL_RETURN): Change from 91 to 71.
25282         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
25283         (PRED_LOOP_GUARD): Change from 66 to 73.
25285 2018-01-19  Martin Liska  <mliska@suse.cz>
25287         * predict.c (predict_insn_def): Add new assert.
25288         (struct branch_predictor): Change type to signed integer.
25289         (test_prediction_value_range): Amend test to cover
25290         PROB_UNINITIALIZED.
25291         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
25292         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
25293         (PRED_LOOP_ITERATIONS_MAX): Likewise.
25294         (PRED_LOOP_IV_COMPARE): Likewise.
25295         * predict.h (PROB_UNINITIALIZED): Define new constant.
25297 2018-01-19  Martin Liska  <mliska@suse.cz>
25299         * predict.c (dump_prediction): Add new format for
25300         analyze_brprob.py script which is enabled with -details
25301         suboption.
25302         * profile-count.h (precise_p): New function.
25304 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25306         PR tree-optimization/83922
25307         * tree-vect-loop.c (vect_verify_full_masking): Return false if
25308         there are no statements that need masking.
25309         (vect_active_double_reduction_p): New function.
25310         (vect_analyze_loop_operations): Use it when handling phis that
25311         are not in the loop header.
25313 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25315         PR tree-optimization/83914
25316         * tree-vect-loop.c (vectorizable_induction): Don't convert
25317         init_expr or apply the peeling adjustment for inductions
25318         that are nested within the vectorized loop.
25320 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25322         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
25323         instead of NEG.
25325 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25327         PR sanitizer/81715
25328         PR testsuite/83882
25329         * function.h (gimplify_parameters): Add gimple_seq * argument.
25330         * function.c: Include gimple.h and options.h.
25331         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
25332         for the added local temporaries if needed.
25333         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
25334         if there are any parameter cleanups, wrap whole body into a
25335         try/finally with the cleanups.
25337 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
25339         PR target/82964
25340         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
25341         Use GET_MODE_CLASS for scalar floating point.
25343 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25345         PR ipa/82256
25346         patch by PaX Team
25347         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
25348         Fix call of call_cgraph_insertion_hooks.
25350 2018-01-18  Martin Sebor  <msebor@redhat.com>
25352         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
25354 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25356         PR ipa/83619
25357         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
25358         frequencies.
25360 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
25362         PR other/70268
25363         * common.opt: (-ffile-prefix-map): New option.
25364         * opts.c (common_handle_option): Defer it.
25365         * opts-global.c (handle_common_deferred_options): Handle it.
25366         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
25367         * file-prefix-map.h: New file.
25368         (remap_debug_filename, add_debug_prefix_map): ...here.
25369         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
25370         * final.c (debug_prefix_map, add_debug_prefix_map
25371         remap_debug_filename): Move to...
25372         * file-prefix-map.c: New file.
25373         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
25374         generalize, get rid of alloca(), use strrchr() instead of strchr().
25375         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
25376         Implement in terms of add_prefix_map().
25377         (remap_macro_filename, remap_debug_filename): Implement in term of
25378         remap_filename().
25379         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
25380         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
25381         * dbxout.c: Include file-prefix-map.h.
25382         * varasm.c: Likewise.
25383         * vmsdbgout.c: Likewise.
25384         * xcoffout.c: Likewise.
25385         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
25386         * doc/cppopts.texi (-fmacro-prefix-map): Document.
25387         * doc/invoke.texi (-ffile-prefix-map): Document.
25388         (-fdebug-prefix-map): Update description.
25390 2018-01-18  Martin Liska  <mliska@suse.cz>
25392         * config/i386/i386.c (indirect_thunk_name): Document that also
25393         lfence is emitted.
25394         (output_indirect_thunk): Document why both instructions
25395         (pause and lfence) are generated.
25397 2018-01-18  Richard Biener  <rguenther@suse.de>
25399         PR tree-optimization/83887
25400         * graphite-scop-detection.c
25401         (scop_detection::get_nearest_dom_with_single_entry): Remove.
25402         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25403         (scop_detection::merge_sese): Re-implement with a flood-fill
25404         algorithm that properly finds a SESE region if it exists.
25406 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25408         PR c/61240
25409         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25410         pointer_diff optimizations use view_convert instead of convert.
25412 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25414         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25415         Generate different code for -mno-speculate-indirect-jumps.
25416         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25417         (*call_indirect_aix<mode>): Disable for
25418         -mno-speculate-indirect-jumps.
25419         (*call_indirect_aix<mode>_nospec): New define_insn.
25420         (*call_value_indirect_aix<mode>): Disable for
25421         -mno-speculate-indirect-jumps.
25422         (*call_value_indirect_aix<mode>_nospec): New define_insn.
25423         (*sibcall_nonlocal_sysv<mode>): Generate different code for
25424         -mno-speculate-indirect-jumps.
25425         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25427 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
25429         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25430         long double type, set the flags for noting the default long double
25431         type, even if we don't pass or return a long double type.
25433 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
25435         PR ipa/83051
25436         * ipa-inline.c (flatten_function): Do not overwrite final inlining
25437         failure.
25439 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
25441         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25442         support for merge[hl].
25443         (fold_mergehl_helper): New helper function.
25444         (tree-vector-builder.h): New #include for tree_vector_builder usage.
25445         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25446         (altivec_vmrglw_direct): Add xxmrglw insn.
25448 2018-01-17  Andrew Waterman  <andrew@sifive.com>
25450         * config/riscv/riscv.c (riscv_conditional_register_usage): If
25451         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25453 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
25455         PR lto/83121
25456         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25457         call the lto_location_cache before reading the
25458         DECL_SOURCE_LOCATION of the types.
25460 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
25461             Richard Sandiford  <richard.sandiford@linaro.org>
25463         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25464         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25465         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25466         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25467         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25468         Add declaration.
25469         * config/aarch64/constraints.md (aarch64_movti_operand):
25470         Limit immediates.
25471         * config/aarch64/predicates.md (Uti): Add new constraint.
25473 2018-01-17  Carl Love  <cel@us.ibm.com>
25475         * config/rs6000/vsx.md (define_expand xl_len_r,
25476         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25477         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25478         lxvll.
25479         (define_expand, define_insn): Move the shift left from  the
25480         define_insn to the define_expand for lxvl and stxvl instructions.
25481         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25482         and XL_LEN_R definitions to PURE.
25484 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
25486         * config/i386/i386.c (indirect_thunk_name): Declare regno
25487         as unsigned int.  Compare regno with INVALID_REGNUM.
25488         (output_indirect_thunk): Ditto.
25489         (output_indirect_thunk_function): Ditto.
25490         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
25491         in the call to output_indirect_thunk_function.
25493 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
25495         PR middle-end/83884
25496         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25497         rather than the size of inner_type to determine the stack slot size
25498         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25500 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
25502         PR target/83546
25503         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25504         to PTA_SILVERMONT.
25506 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
25508         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25509         endian Linux systems to optionally enable multilibs for selecting
25510         the long double type if the user configured an explicit type.
25511         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25512         have no long double multilibs if not defined.
25513         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25514         warn if the user used -mabi={ieee,ibm}longdouble and we built
25515         multilibs for long double.
25516         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25517         appropriate multilib option.
25518         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25519         multilib options.
25520         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25521         for building long double multilibs.
25522         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25524 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
25526         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25527         copies.
25529         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25530         64 bits.
25531         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25532         128 bits.
25534         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25535         variables.
25537         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25538         return value.
25540 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25542         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25543         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25545 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25547         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25548         different rtl trees depending on TARGET_64BIT.
25549         (rs6000_gen_lvx): Likewise.
25551 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25553         * config/visium/visium.md (nop): Tweak comment.
25554         (hazard_nop): Likewise.
25556 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25558         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25559         -mspeculate-indirect-jumps.
25560         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25561         for -mno-speculate-indirect-jumps.
25562         (*call_indirect_elfv2<mode>_nospec): New define_insn.
25563         (*call_value_indirect_elfv2<mode>): Disable for
25564         -mno-speculate-indirect-jumps.
25565         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25566         (indirect_jump): Emit different RTL for
25567         -mno-speculate-indirect-jumps.
25568         (*indirect_jump<mode>): Disable for
25569         -mno-speculate-indirect-jumps.
25570         (*indirect_jump<mode>_nospec): New define_insn.
25571         (tablejump): Emit different RTL for
25572         -mno-speculate-indirect-jumps.
25573         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25574         (tablejumpsi_nospec): New define_expand.
25575         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25576         (tablejumpdi_nospec): New define_expand.
25577         (*tablejump<mode>_internal1): Disable for
25578         -mno-speculate-indirect-jumps.
25579         (*tablejump<mode>_internal1_nospec): New define_insn.
25580         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25581         option.
25583 2018-01-16  Artyom Skrobov tyomitch@gmail.com
25585         * caller-save.c (insert_save): Drop unnecessary parameter.  All
25586         callers updated.
25588 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25589             Richard Biener  <rguenth@suse.de>
25591         PR libgomp/83590
25592         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25593         return early, inline manually is_gimple_sizepos.  Make sure if we
25594         call gimplify_expr we don't end up with a gimple constant.
25595         * tree.c (variably_modified_type_p): Don't return true for
25596         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
25597         * gimplify.h (is_gimple_sizepos): Remove.
25599 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25601         PR tree-optimization/83857
25602         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25603         vectorizable_live_operation for pure SLP statements.
25604         (vectorizable_live_operation): Handle PHIs.
25606 2018-01-16  Richard Biener  <rguenther@suse.de>
25608         PR tree-optimization/83867
25609         * tree-vect-stmts.c (vect_transform_stmt): Precompute
25610         nested_in_vect_loop_p since the scalar stmt may get invalidated.
25612 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25614         PR c/83844
25615         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25616         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25617         If off is not INTEGER_CST, issue a may not be aligned warning
25618         rather than isn't aligned.  Use isn%'t rather than isn't.
25619         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25620         into MULT_EXPR.
25621         <case MULT_EXPR>: Improve the case when bottom and one of the
25622         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25623         operand, in that case check if the other operand is multiple of
25624         bottom divided by the INTEGER_CST operand.
25626 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25628         PR target/83858
25629         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25630         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25631         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25632         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25633         * config/pa/pa.c (pa_function_arg_advance): Likewise.
25634         (pa_function_arg, pa_arg_partial_bytes): Likewise.
25635         (pa_function_arg_size): New function.
25637 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25639         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25640         in a separate statement.
25642 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25644         PR tree-optimization/83847
25645         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25646         group gathers and scatters.
25648 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25650         PR rtl-optimization/86620
25651         * params.def (max-sched-ready-insns): Bump minimum value to 1.
25653         PR rtl-optimization/83213
25654         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25655         to last if both are JUMP_INSNs.
25657         PR tree-optimization/83843
25658         * gimple-ssa-store-merging.c
25659         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25660         store_immediate_info for bswap/nop orig_stores.
25662 2018-01-15  Andrew Waterman  <andrew@sifive.com>
25664         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25665         !TARGET_MUL.
25666         <UDIV>: Increase cost if !TARGET_DIV.
25668 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
25670         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25671         (define_attr "cr_logical_3op"): New.
25672         (cceq_ior_compare): Adjust.
25673         (cceq_ior_compare_complement): Adjust.
25674         (*cceq_rev_compare): Adjust.
25675         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25676         (is_cracked_insn): Adjust.
25677         (insn_must_be_first_in_group): Adjust.
25678         * config/rs6000/40x.md: Adjust.
25679         * config/rs6000/440.md: Adjust.
25680         * config/rs6000/476.md: Adjust.
25681         * config/rs6000/601.md: Adjust.
25682         * config/rs6000/603.md: Adjust.
25683         * config/rs6000/6xx.md: Adjust.
25684         * config/rs6000/7450.md: Adjust.
25685         * config/rs6000/7xx.md: Adjust.
25686         * config/rs6000/8540.md: Adjust.
25687         * config/rs6000/cell.md: Adjust.
25688         * config/rs6000/e300c2c3.md: Adjust.
25689         * config/rs6000/e500mc.md: Adjust.
25690         * config/rs6000/e500mc64.md: Adjust.
25691         * config/rs6000/e5500.md: Adjust.
25692         * config/rs6000/e6500.md: Adjust.
25693         * config/rs6000/mpc.md: Adjust.
25694         * config/rs6000/power4.md: Adjust.
25695         * config/rs6000/power5.md: Adjust.
25696         * config/rs6000/power6.md: Adjust.
25697         * config/rs6000/power7.md: Adjust.
25698         * config/rs6000/power8.md: Adjust.
25699         * config/rs6000/power9.md: Adjust.
25700         * config/rs6000/rs64.md: Adjust.
25701         * config/rs6000/titan.md: Adjust.
25703 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25705         * config/i386/predicates.md (indirect_branch_operand): Rewrite
25706         ix86_indirect_branch_register logic.
25708 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25710         * config/i386/constraints.md (Bs): Update
25711         ix86_indirect_branch_register check.  Don't check
25712         ix86_indirect_branch_register with GOT_memory_operand.
25713         (Bw): Likewise.
25714         * config/i386/predicates.md (GOT_memory_operand): Don't check
25715         ix86_indirect_branch_register here.
25716         (GOT32_symbol_operand): Likewise.
25718 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25720         * config/i386/predicates.md (constant_call_address_operand):
25721         Rewrite ix86_indirect_branch_register logic.
25722         (sibcall_insn_operand): Likewise.
25724 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25726         * config/i386/constraints.md (Bs): Replace
25727         ix86_indirect_branch_thunk_register with
25728         ix86_indirect_branch_register.
25729         (Bw): Likewise.
25730         * config/i386/i386.md (indirect_jump): Likewise.
25731         (tablejump): Likewise.
25732         (*sibcall_memory): Likewise.
25733         (*sibcall_value_memory): Likewise.
25734         Peepholes of indirect call and jump via memory: Likewise.
25735         * config/i386/i386.opt: Likewise.
25736         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25737         (GOT_memory_operand): Likewise.
25738         (call_insn_operand): Likewise.
25739         (sibcall_insn_operand): Likewise.
25740         (GOT32_symbol_operand): Likewise.
25742 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25744         PR middle-end/83837
25745         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25746         type rather than type addr's type points to.
25747         (expand_omp_atomic_mutex): Likewise.
25748         (expand_omp_atomic): Likewise.
25750 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25752         PR target/83839
25753         * config/i386/i386.c (output_indirect_thunk_function): Use
25754         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25755         for  __x86_return_thunk.
25757 2018-01-15  Richard Biener  <rguenther@suse.de>
25759         PR middle-end/83850
25760         * expmed.c (extract_bit_field_1): Fix typo.
25762 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25764         PR target/83687
25765         * config/arm/iterators.md (VF): New mode iterator.
25766         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25767         Remove integer-related logic from pattern.
25768         (neon_vabd<mode>_3): Likewise.
25770 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25772         PR middle-end/82694
25773         * common.opt (fstrict-overflow): No longer an alias.
25774         (fwrapv-pointer): New option.
25775         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25776         also for pointer types based on flag_wrapv_pointer.
25777         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25778         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
25779         opts->x_flag_wrapv got set.
25780         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
25781         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
25782         POINTER_TYPE_OVERFLOW_UNDEFINED.
25783         * match.pd: Likewise in address comparison pattern.
25784         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
25786 2018-01-15  Richard Biener  <rguenther@suse.de>
25788         PR lto/83804
25789         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
25790         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
25791         Reset type names to their identifier if their TYPE_DECL doesn't
25792         have linkage (and thus is used for ODR and devirt).
25793         (save_debug_info_for_decl): Remove.
25794         (save_debug_info_for_type): Likewise.
25795         (add_tree_to_fld_list): Adjust.
25796         * tree-pretty-print.c (dump_generic_node): Make dumping of
25797         type names more robust.
25799 2018-01-15  Richard Biener  <rguenther@suse.de>
25801         * BASE-VER: Bump to 8.0.1.
25803 2018-01-14  Martin Sebor  <msebor@redhat.com>
25805         PR other/83508
25806         * builtins.c (check_access): Avoid warning when the no-warning bit
25807         is set.
25809 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
25811         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
25812         * ira-color (allocno_hard_regs_compare): Likewise.
25814 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
25816         PR target/83013
25817         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
25818         Use .pushsection/.popsection.
25820 2018-01-14  Martin Sebor  <msebor@redhat.com>
25822         PR c++/81327
25823         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
25825 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
25827         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
25828         entry from extra_headers.
25829         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
25830         extra_headers, make the list bitwise identical to the i?86-*-* one.
25832 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25834         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
25835         -mcmodel=large with -mindirect-branch=thunk,
25836         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
25837         -mfunction-return=thunk-extern.
25838         * doc/invoke.texi: Document -mcmodel=large is incompatible with
25839         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
25840         -mfunction-return=thunk and -mfunction-return=thunk-extern.
25842 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25844         * config/i386/i386.c (print_reg): Print the name of the full
25845         integer register without '%'.
25846         (ix86_print_operand): Handle 'V'.
25847         * doc/extend.texi: Document 'V' modifier.
25849 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25851         * config/i386/constraints.md (Bs): Disallow memory operand for
25852         -mindirect-branch-register.
25853         (Bw): Likewise.
25854         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25855         (GOT_memory_operand): Likewise.
25856         (call_insn_operand): Likewise.
25857         (sibcall_insn_operand): Likewise.
25858         (GOT32_symbol_operand): Likewise.
25859         * config/i386/i386.md (indirect_jump): Call convert_memory_address
25860         for -mindirect-branch-register.
25861         (tablejump): Likewise.
25862         (*sibcall_memory): Likewise.
25863         (*sibcall_value_memory): Likewise.
25864         Disallow peepholes of indirect call and jump via memory for
25865         -mindirect-branch-register.
25866         (*call_pop): Replace m with Bw.
25867         (*call_value_pop): Likewise.
25868         (*sibcall_pop_memory): Replace m with Bs.
25869         * config/i386/i386.opt (mindirect-branch-register): New option.
25870         * doc/invoke.texi: Document -mindirect-branch-register option.
25872 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25874         * config/i386/i386-protos.h (ix86_output_function_return): New.
25875         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
25876         set function_return_type.
25877         (indirect_thunk_name): Add ret_p to indicate thunk for function
25878         return.
25879         (output_indirect_thunk_function): Pass false to
25880         indirect_thunk_name.
25881         (ix86_output_indirect_branch_via_reg): Likewise.
25882         (ix86_output_indirect_branch_via_push): Likewise.
25883         (output_indirect_thunk_function): Create alias for function
25884         return thunk if regno < 0.
25885         (ix86_output_function_return): New function.
25886         (ix86_handle_fndecl_attribute): Handle function_return.
25887         (ix86_attribute_table): Add function_return.
25888         * config/i386/i386.h (machine_function): Add
25889         function_return_type.
25890         * config/i386/i386.md (simple_return_internal): Use
25891         ix86_output_function_return.
25892         (simple_return_internal_long): Likewise.
25893         * config/i386/i386.opt (mfunction-return=): New option.
25894         (indirect_branch): Mention -mfunction-return=.
25895         * doc/extend.texi: Document function_return function attribute.
25896         * doc/invoke.texi: Document -mfunction-return= option.
25898 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25900         * config/i386/i386-opts.h (indirect_branch): New.
25901         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
25902         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
25903         with local indirect jump when converting indirect call and jump.
25904         (ix86_set_indirect_branch_type): New.
25905         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
25906         (indirectlabelno): New.
25907         (indirect_thunk_needed): Likewise.
25908         (indirect_thunk_bnd_needed): Likewise.
25909         (indirect_thunks_used): Likewise.
25910         (indirect_thunks_bnd_used): Likewise.
25911         (INDIRECT_LABEL): Likewise.
25912         (indirect_thunk_name): Likewise.
25913         (output_indirect_thunk): Likewise.
25914         (output_indirect_thunk_function): Likewise.
25915         (ix86_output_indirect_branch_via_reg): Likewise.
25916         (ix86_output_indirect_branch_via_push): Likewise.
25917         (ix86_output_indirect_branch): Likewise.
25918         (ix86_output_indirect_jmp): Likewise.
25919         (ix86_code_end): Call output_indirect_thunk_function if needed.
25920         (ix86_output_call_insn): Call ix86_output_indirect_branch if
25921         needed.
25922         (ix86_handle_fndecl_attribute): Handle indirect_branch.
25923         (ix86_attribute_table): Add indirect_branch.
25924         * config/i386/i386.h (machine_function): Add indirect_branch_type
25925         and has_local_indirect_jump.
25926         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
25927         to true.
25928         (tablejump): Likewise.
25929         (*indirect_jump): Use ix86_output_indirect_jmp.
25930         (*tablejump_1): Likewise.
25931         (simple_return_indirect_internal): Likewise.
25932         * config/i386/i386.opt (mindirect-branch=): New option.
25933         (indirect_branch): New.
25934         (keep): Likewise.
25935         (thunk): Likewise.
25936         (thunk-inline): Likewise.
25937         (thunk-extern): Likewise.
25938         * doc/extend.texi: Document indirect_branch function attribute.
25939         * doc/invoke.texi: Document -mindirect-branch= option.
25941 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
25943         PR ipa/83051
25944         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
25946 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
25948         * ipa-inline.c (want_inline_small_function_p): Return false if
25949         inlining has already failed with CIF_FINAL_ERROR.
25950         (update_caller_keys): Call want_inline_small_function_p before
25951         can_inline_edge_p.
25952         (update_callee_keys): Likewise.
25954 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25956         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
25957         New function.
25958         (rs6000_quadword_masked_address_p): Likewise.
25959         (quad_aligned_load_p): Likewise.
25960         (quad_aligned_store_p): Likewise.
25961         (const_load_sequence_p): Add comment to describe the outer-most loop.
25962         (mimic_memory_attributes_and_flags): New function.
25963         (rs6000_gen_stvx): Likewise.
25964         (replace_swapped_aligned_store): Likewise.
25965         (rs6000_gen_lvx): Likewise.
25966         (replace_swapped_aligned_load): Likewise.
25967         (replace_swapped_load_constant): Capitalize argument name in
25968         comment describing this function.
25969         (rs6000_analyze_swaps): Add a third pass to search for vector loads
25970         and stores that access quad-word aligned addresses and replace
25971         with stvx or lvx instructions when appropriate.
25972         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
25973         New function prototype.
25974         (rs6000_quadword_masked_address_p): Likewise.
25975         (rs6000_gen_lvx): Likewise.
25976         (rs6000_gen_stvx): Likewise.
25977         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
25978         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
25979         when memory address is aligned.
25980         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
25981         this split to select lvx instruction when memory address is aligned.
25982         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
25983         instruction when memory address is aligned.
25984         (*vsx_le_perm_load_v16qi): Likewise.
25985         (four unnamed splitters): Modify to select the stvx instruction
25986         when memory is aligned.
25988 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
25990         * predict.c (determine_unlikely_bbs): Handle correctly BBs
25991         which appears in the queue multiple times.
25993 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25994             Alan Hayward  <alan.hayward@arm.com>
25995             David Sherwood  <david.sherwood@arm.com>
25997         * tree-vectorizer.h (vec_lower_bound): New structure.
25998         (_loop_vec_info): Add check_nonzero and lower_bounds.
25999         (LOOP_VINFO_CHECK_NONZERO): New macro.
26000         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
26001         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
26002         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
26003         fields.  Make seg_len the distance travelled, not including the
26004         access size.
26005         (dr_direction_indicator): Declare.
26006         (dr_zero_step_indicator): Likewise.
26007         (dr_known_forward_stride_p): Likewise.
26008         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
26009         tree-ssanames.h.
26010         (runtime_alias_check_p): Allow runtime alias checks with
26011         variable strides.
26012         (operator ==): Compare access_size and align.
26013         (prune_runtime_alias_test_list): Rework for new distinction between
26014         the access_size and seg_len.
26015         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
26016         segment lengths.
26017         (get_segment_min_max): New function.
26018         (create_intersect_range_checks): Use it.
26019         (dr_step_indicator): New function.
26020         (dr_direction_indicator): Likewise.
26021         (dr_zero_step_indicator): Likewise.
26022         (dr_known_forward_stride_p): Likewise.
26023         * tree-loop-distribution.c (data_ref_segment_size): Return
26024         DR_STEP * (niters - 1).
26025         (compute_alias_check_pairs): Update call to the dr_with_seg_len
26026         constructor.
26027         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
26028         (vect_preserves_scalar_order_p): New function, split out from...
26029         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
26030         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
26031         (vect_vfa_access_size): New function.
26032         (vect_vfa_align): Likewise.
26033         (vect_compile_time_alias): Take access_size_a and access_b arguments.
26034         (dump_lower_bound): New function.
26035         (vect_check_lower_bound): Likewise.
26036         (vect_small_gap_p): Likewise.
26037         (vectorizable_with_step_bound_p): Likewise.
26038         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
26039         depencies if the vectorization factor is 1.  Convert the checks
26040         for nonzero steps into checks on the bounds of DR_STEP.  Try using
26041         a bunds check for variable steps if the minimum required step is
26042         relatively small. Update calls to the dr_with_seg_len
26043         constructor and to vect_compile_time_alias.
26044         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
26045         function.
26046         (vect_loop_versioning): Call it.
26047         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
26048         when retrying.
26049         (vect_estimate_min_profitable_iters): Account for any bounds checks.
26051 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26052             Alan Hayward  <alan.hayward@arm.com>
26053             David Sherwood  <david.sherwood@arm.com>
26055         * doc/sourcebuild.texi (vect_scatter_store): Document.
26056         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
26057         optabs.
26058         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
26059         Document.
26060         * genopinit.c (main): Add supports_vec_scatter_store and
26061         supports_vec_scatter_store_cached to target_optabs.
26062         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
26063         IFN_MASK_SCATTER_STORE.
26064         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
26065         functions.
26066         * internal-fn.h (internal_store_fn_p): Declare.
26067         (internal_fn_stored_value_index): Likewise.
26068         * internal-fn.c (scatter_store_direct): New macro.
26069         (expand_scatter_store_optab_fn): New function.
26070         (direct_scatter_store_optab_supported_p): New macro.
26071         (internal_store_fn_p): New function.
26072         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
26073         IFN_MASK_SCATTER_STORE.
26074         (internal_fn_mask_index): Likewise.
26075         (internal_fn_stored_value_index): New function.
26076         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
26077         for scatter stores.
26078         * optabs-query.h (supports_vec_scatter_store_p): Declare.
26079         * optabs-query.c (supports_vec_scatter_store_p): New function.
26080         * tree-vectorizer.h (vect_get_store_rhs): Declare.
26081         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
26082         true for scatter stores.
26083         (vect_gather_scatter_fn_p): Handle scatter stores too.
26084         (vect_check_gather_scatter): Consider using scatter stores if
26085         supports_vec_scatter_store_p.
26086         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
26087         scatter stores too.
26088         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26089         internal_fn_stored_value_index.
26090         (check_load_store_masking): Handle scatter stores too.
26091         (vect_get_store_rhs): Make public.
26092         (vectorizable_call): Use internal_store_fn_p.
26093         (vectorizable_store): Handle scatter store internal functions.
26094         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
26095         when deciding whether the end of the group has been reached.
26096         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
26097         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
26098         (mask_scatter_store<mode>): New insns.
26100 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26101             Alan Hayward  <alan.hayward@arm.com>
26102             David Sherwood  <david.sherwood@arm.com>
26104         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
26105         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
26106         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
26107         function.
26108         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
26109         Use vect_truncate_gather_scatter_offset if we can't treat the
26110         operation as a normal gather load or scatter store.
26111         (get_group_load_store_type): Take the gather_scatter_info
26112         as argument.  Try using a gather load or scatter store for
26113         single-element groups.
26114         (get_load_store_type): Update calls to get_group_load_store_type
26115         and vect_use_strided_gather_scatters_p.
26117 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26118             Alan Hayward  <alan.hayward@arm.com>
26119             David Sherwood  <david.sherwood@arm.com>
26121         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
26122         optional tree argument.
26123         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
26124         null target hooks.
26125         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
26126         but continue to use the current value as a fallback.
26127         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
26128         to compare the updates.
26129         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
26130         (get_load_store_type): Use it when handling a strided access.
26131         (vect_get_strided_load_store_ops): New function.
26132         (vect_get_data_ptr_increment): Likewise.
26133         (vectorizable_load): Handle strided gather loads.  Always pass
26134         a step to vect_create_data_ref_ptr and bump_vector_ptr.
26136 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26137             Alan Hayward  <alan.hayward@arm.com>
26138             David Sherwood  <david.sherwood@arm.com>
26140         * doc/md.texi (gather_load@var{m}): Document.
26141         (mask_gather_load@var{m}): Likewise.
26142         * genopinit.c (main): Add supports_vec_gather_load and
26143         supports_vec_gather_load_cached to target_optabs.
26144         * optabs-tree.c (init_tree_optimization_optabs): Use
26145         ggc_cleared_alloc to allocate target_optabs.
26146         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
26147         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
26148         functions.
26149         * internal-fn.h (internal_load_fn_p): Declare.
26150         (internal_gather_scatter_fn_p): Likewise.
26151         (internal_fn_mask_index): Likewise.
26152         (internal_gather_scatter_fn_supported_p): Likewise.
26153         * internal-fn.c (gather_load_direct): New macro.
26154         (expand_gather_load_optab_fn): New function.
26155         (direct_gather_load_optab_supported_p): New macro.
26156         (direct_internal_fn_optab): New function.
26157         (internal_load_fn_p): Likewise.
26158         (internal_gather_scatter_fn_p): Likewise.
26159         (internal_fn_mask_index): Likewise.
26160         (internal_gather_scatter_fn_supported_p): Likewise.
26161         * optabs-query.c (supports_at_least_one_mode_p): New function.
26162         (supports_vec_gather_load_p): Likewise.
26163         * optabs-query.h (supports_vec_gather_load_p): Declare.
26164         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
26165         and memory_type field.
26166         (NUM_PATTERNS): Bump to 15.
26167         * tree-vect-data-refs.c: Include internal-fn.h.
26168         (vect_gather_scatter_fn_p): New function.
26169         (vect_describe_gather_scatter_call): Likewise.
26170         (vect_check_gather_scatter): Try using internal functions for
26171         gather loads.  Recognize existing calls to a gather load function.
26172         (vect_analyze_data_refs): Consider using gather loads if
26173         supports_vec_gather_load_p.
26174         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
26175         (vect_get_gather_scatter_offset_type): Likewise.
26176         (vect_convert_mask_for_vectype): Likewise.
26177         (vect_add_conversion_to_patterm): Likewise.
26178         (vect_try_gather_scatter_pattern): Likewise.
26179         (vect_recog_gather_scatter_pattern): New pattern recognizer.
26180         (vect_vect_recog_func_ptrs): Add it.
26181         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26182         internal_fn_mask_index and internal_gather_scatter_fn_p.
26183         (check_load_store_masking): Take the gather_scatter_info as an
26184         argument and handle gather loads.
26185         (vect_get_gather_scatter_ops): New function.
26186         (vectorizable_call): Check internal_load_fn_p.
26187         (vectorizable_load): Likewise.  Handle gather load internal
26188         functions.
26189         (vectorizable_store): Update call to check_load_store_masking.
26190         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
26191         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
26192         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
26193         (aarch64_gather_scale_operand_d): New predicates.
26194         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
26195         (mask_gather_load<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         * optabs.def (fold_left_plus_optab): New optab.
26202         * doc/md.texi (fold_left_plus_@var{m}): Document.
26203         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
26204         * internal-fn.c (fold_left_direct): Define.
26205         (expand_fold_left_optab_fn): Likewise.
26206         (direct_fold_left_optab_supported_p): Likewise.
26207         * fold-const-call.c (fold_const_fold_left): New function.
26208         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
26209         * tree-parloops.c (valid_reduction_p): New function.
26210         (gather_scalar_reductions): Use it.
26211         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
26212         (vect_finish_replace_stmt): Declare.
26213         * tree-vect-loop.c (fold_left_reduction_fn): New function.
26214         (needs_fold_left_reduction_p): New function, split out from...
26215         (vect_is_simple_reduction): ...here.  Accept reductions that
26216         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
26217         (vect_force_simple_reduction): Also store the reduction type in
26218         the assignment's STMT_VINFO_REDUC_TYPE.
26219         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
26220         (merge_with_identity): New function.
26221         (vect_expand_fold_left): Likewise.
26222         (vectorize_fold_left_reduction): Likewise.
26223         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
26224         scalar phi in place for it.  Check for target support and reject
26225         cases that would reassociate the operation.  Defer the transform
26226         phase to vectorize_fold_left_reduction.
26227         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
26228         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
26229         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
26231 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26233         * tree-if-conv.c (predicate_mem_writes): Remove redundant
26234         call to ifc_temp_var.
26236 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26237             Alan Hayward  <alan.hayward@arm.com>
26238             David Sherwood  <david.sherwood@arm.com>
26240         * target.def (legitimize_address_displacement): Take the original
26241         offset as a poly_int.
26242         * targhooks.h (default_legitimize_address_displacement): Update
26243         accordingly.
26244         * targhooks.c (default_legitimize_address_displacement): Likewise.
26245         * doc/tm.texi: Regenerate.
26246         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
26247         as an argument, moving assert of ad->disp == ad->disp_term to...
26248         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
26249         Try calling targetm.legitimize_address_displacement before expanding
26250         the address rather than afterwards, and adjust for the new interface.
26251         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
26252         Match the new hook interface.  Handle SVE addresses.
26253         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
26254         new hook interface.
26256 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26258         * Makefile.in (OBJS): Add early-remat.o.
26259         * target.def (select_early_remat_modes): New hook.
26260         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
26261         * doc/tm.texi: Regenerate.
26262         * targhooks.h (default_select_early_remat_modes): Declare.
26263         * targhooks.c (default_select_early_remat_modes): New function.
26264         * timevar.def (TV_EARLY_REMAT): New timevar.
26265         * passes.def (pass_early_remat): New pass.
26266         * tree-pass.h (make_pass_early_remat): Declare.
26267         * early-remat.c: New file.
26268         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
26269         function.
26270         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
26272 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26273             Alan Hayward  <alan.hayward@arm.com>
26274             David Sherwood  <david.sherwood@arm.com>
26276         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
26277         vfm1 with a bound_epilog parameter.
26278         (vect_do_peeling): Update calls accordingly, and move the prologue
26279         call earlier in the function.  Treat the base bound_epilog as 0 for
26280         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
26281         this base when peeling for gaps.
26282         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
26283         with fully-masked loops.
26284         (vect_estimate_min_profitable_iters): Handle the single peeled
26285         iteration in that case.
26287 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26288             Alan Hayward  <alan.hayward@arm.com>
26289             David Sherwood  <david.sherwood@arm.com>
26291         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
26292         single-element interleaving even if the size is not a power of 2.
26293         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
26294         accesses for single-element interleaving if the group size is
26295         not a power of 2.
26297 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26298             Alan Hayward  <alan.hayward@arm.com>
26299             David Sherwood  <david.sherwood@arm.com>
26301         * doc/md.texi (fold_extract_last_@var{m}): Document.
26302         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
26303         * optabs.def (fold_extract_last_optab): New optab.
26304         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
26305         * internal-fn.c (fold_extract_direct): New macro.
26306         (expand_fold_extract_optab_fn): Likewise.
26307         (direct_fold_extract_optab_supported_p): Likewise.
26308         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
26309         * tree-vect-loop.c (vect_model_reduction_cost): Handle
26310         EXTRACT_LAST_REDUCTION.
26311         (get_initial_def_for_reduction): Do not create an initial vector
26312         for EXTRACT_LAST_REDUCTION reductions.
26313         (vectorizable_reduction): Leave the scalar phi in place for
26314         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
26315         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
26316         epilogue code for EXTRACT_LAST_REDUCTION and defer the
26317         transform phase to vectorizable_condition.
26318         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
26319         split out from...
26320         (vect_finish_stmt_generation): ...here.
26321         (vect_finish_replace_stmt): New function.
26322         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
26323         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
26324         pattern.
26325         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
26327 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26328             Alan Hayward  <alan.hayward@arm.com>
26329             David Sherwood  <david.sherwood@arm.com>
26331         * doc/md.texi (extract_last_@var{m}): Document.
26332         * optabs.def (extract_last_optab): New optab.
26333         * internal-fn.def (EXTRACT_LAST): New internal function.
26334         * internal-fn.c (cond_unary_direct): New macro.
26335         (expand_cond_unary_optab_fn): Likewise.
26336         (direct_cond_unary_optab_supported_p): Likewise.
26337         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
26338         loops using EXTRACT_LAST.
26339         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
26340         (extract_last_<mode>): ...this optab.
26341         (vec_extract<mode><Vel>): Update accordingly.
26343 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26344             Alan Hayward  <alan.hayward@arm.com>
26345             David Sherwood  <david.sherwood@arm.com>
26347         * target.def (empty_mask_is_expensive): New hook.
26348         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
26349         * doc/tm.texi: Regenerate.
26350         * targhooks.h (default_empty_mask_is_expensive): Declare.
26351         * targhooks.c (default_empty_mask_is_expensive): New function.
26352         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
26353         if the target says that empty masks are expensive.
26354         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
26355         New function.
26356         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
26358 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26359             Alan Hayward  <alan.hayward@arm.com>
26360             David Sherwood  <david.sherwood@arm.com>
26362         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
26363         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
26364         (vect_use_loop_mask_for_alignment_p): New function.
26365         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
26366         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
26367         niters_skip argument.  Make sure that the first niters_skip elements
26368         of the first iteration are inactive.
26369         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
26370         Update call to vect_set_loop_masks_directly.
26371         (get_misalign_in_elems): New function, split out from...
26372         (vect_gen_prolog_loop_niters): ...here.
26373         (vect_update_init_of_dr): Take a code argument that specifies whether
26374         the adjustment should be added or subtracted.
26375         (vect_update_init_of_drs): Likewise.
26376         (vect_prepare_for_masked_peels): New function.
26377         (vect_do_peeling): Skip prologue peeling if we're using a mask
26378         instead.  Update call to vect_update_inits_of_drs.
26379         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26380         mask_skip_niters.
26381         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
26382         alignment.  Do not include the number of peeled iterations in
26383         the minimum threshold in that case.
26384         (vectorizable_induction): Adjust the start value down by
26385         LOOP_VINFO_MASK_SKIP_NITERS iterations.
26386         (vect_transform_loop): Call vect_prepare_for_masked_peels.
26387         Take the number of skipped iterations into account when calculating
26388         the loop bounds.
26389         * tree-vect-stmts.c (vect_gen_while_not): New function.
26391 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26392             Alan Hayward  <alan.hayward@arm.com>
26393             David Sherwood  <david.sherwood@arm.com>
26395         * doc/sourcebuild.texi (vect_fully_masked): Document.
26396         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26397         default value to 0.
26398         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26399         split out from...
26400         (vect_analyze_loop_2): ...here. Don't check the vectorization
26401         factor against the number of loop iterations if the loop is
26402         fully-masked.
26404 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26405             Alan Hayward  <alan.hayward@arm.com>
26406             David Sherwood  <david.sherwood@arm.com>
26408         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26409         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26410         (dump_groups): Update accordingly.
26411         (iv_use::mem_type): New member variable.
26412         (address_p): New function.
26413         (record_use): Add a mem_type argument and initialize the new
26414         mem_type field.
26415         (record_group_use): Add a mem_type argument.  Use address_p.
26416         Remove obsolete null checks of base_object.  Update call to record_use.
26417         (find_interesting_uses_op): Update call to record_group_use.
26418         (find_interesting_uses_cond): Likewise.
26419         (find_interesting_uses_address): Likewise.
26420         (get_mem_type_for_internal_fn): New function.
26421         (find_address_like_use): Likewise.
26422         (find_interesting_uses_stmt): Try find_address_like_use before
26423         calling find_interesting_uses_op.
26424         (addr_offset_valid_p): Use the iv mem_type field as the type
26425         of the addressed memory.
26426         (add_autoinc_candidates): Likewise.
26427         (get_address_cost): Likewise.
26428         (split_small_address_groups_p): Use address_p.
26429         (split_address_groups): Likewise.
26430         (add_iv_candidate_for_use): Likewise.
26431         (autoinc_possible_for_pair): Likewise.
26432         (rewrite_groups): Likewise.
26433         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26434         (determine_group_iv_cost): Update after split of USE_ADDRESS.
26435         (get_alias_ptr_type_for_ptr_address): New function.
26436         (rewrite_use_address): Rewrite address uses in calls that were
26437         identified by find_address_like_use.
26439 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26440             Alan Hayward  <alan.hayward@arm.com>
26441             David Sherwood  <david.sherwood@arm.com>
26443         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26444         TARGET_MEM_REFs.
26445         * gimple-expr.h (is_gimple_addressable: Likewise.
26446         * gimple-expr.c (is_gimple_address): Likewise.
26447         * internal-fn.c (expand_call_mem_ref): New function.
26448         (expand_mask_load_optab_fn): Use it.
26449         (expand_mask_store_optab_fn): Likewise.
26451 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26452             Alan Hayward  <alan.hayward@arm.com>
26453             David Sherwood  <david.sherwood@arm.com>
26455         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26456         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26457         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26458         (cond_umax@var{mode}): Document.
26459         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26460         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26461         (cond_umin_optab, cond_umax_optab): New optabs.
26462         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26463         (COND_IOR, COND_XOR): New internal functions.
26464         * internal-fn.h (get_conditional_internal_fn): Declare.
26465         * internal-fn.c (cond_binary_direct): New macro.
26466         (expand_cond_binary_optab_fn): Likewise.
26467         (direct_cond_binary_optab_supported_p): Likewise.
26468         (get_conditional_internal_fn): New function.
26469         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26470         Cope with reduction statements that are vectorized as calls rather
26471         than assignments.
26472         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26473         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26474         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26475         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26476         (UNSPEC_COND_EOR): New unspecs.
26477         (optab): Add mappings for them.
26478         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26479         (sve_int_op, sve_fp_op): New int attributes.
26481 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26482             Alan Hayward  <alan.hayward@arm.com>
26483             David Sherwood  <david.sherwood@arm.com>
26485         * optabs.def (while_ult_optab): New optab.
26486         * doc/md.texi (while_ult@var{m}@var{n}): Document.
26487         * internal-fn.def (WHILE_ULT): New internal function.
26488         * internal-fn.h (direct_internal_fn_supported_p): New override
26489         that takes two types as argument.
26490         * internal-fn.c (while_direct): New macro.
26491         (expand_while_optab_fn): New function.
26492         (convert_optab_supported_p): Likewise.
26493         (direct_while_optab_supported_p): New macro.
26494         * wide-int.h (wi::udiv_ceil): New function.
26495         * tree-vectorizer.h (rgroup_masks): New structure.
26496         (vec_loop_masks): New typedef.
26497         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26498         and fully_masked_p.
26499         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26500         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26501         (vect_max_vf): New function.
26502         (slpeel_make_loop_iterate_ntimes): Delete.
26503         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26504         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26505         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26506         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26507         internal-fn.h, stor-layout.h and optabs-query.h.
26508         (vect_set_loop_mask): New function.
26509         (add_preheader_seq): Likewise.
26510         (add_header_seq): Likewise.
26511         (interleave_supported_p): Likewise.
26512         (vect_maybe_permute_loop_masks): Likewise.
26513         (vect_set_loop_masks_directly): Likewise.
26514         (vect_set_loop_condition_masked): Likewise.
26515         (vect_set_loop_condition_unmasked): New function, split out from
26516         slpeel_make_loop_iterate_ntimes.
26517         (slpeel_make_loop_iterate_ntimes): Rename to..
26518         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
26519         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26520         (vect_do_peeling): Update call accordingly.
26521         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26522         loops.
26523         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26524         mask_compare_type, can_fully_mask_p and fully_masked_p.
26525         (release_vec_loop_masks): New function.
26526         (_loop_vec_info): Use it to free the loop masks.
26527         (can_produce_all_loop_masks_p): New function.
26528         (vect_get_max_nscalars_per_iter): Likewise.
26529         (vect_verify_full_masking): Likewise.
26530         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26531         retries, and free the mask rgroups before retrying.  Check loop-wide
26532         reasons for disallowing fully-masked loops.  Make the final decision
26533         about whether use a fully-masked loop or not.
26534         (vect_estimate_min_profitable_iters): Do not assume that peeling
26535         for the number of iterations will be needed for fully-masked loops.
26536         (vectorizable_reduction): Disable fully-masked loops.
26537         (vectorizable_live_operation): Likewise.
26538         (vect_halve_mask_nunits): New function.
26539         (vect_double_mask_nunits): Likewise.
26540         (vect_record_loop_mask): Likewise.
26541         (vect_get_loop_mask): Likewise.
26542         (vect_transform_loop): Handle the case in which the final loop
26543         iteration might handle a partial vector.  Call vect_set_loop_condition
26544         instead of slpeel_make_loop_iterate_ntimes.
26545         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26546         (check_load_store_masking): New function.
26547         (prepare_load_store_mask): Likewise.
26548         (vectorizable_store): Handle fully-masked loops.
26549         (vectorizable_load): Likewise.
26550         (supportable_widening_operation): Use vect_halve_mask_nunits for
26551         booleans.
26552         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26553         (vect_gen_while): New function.
26554         * config/aarch64/aarch64.md (umax<mode>3): New expander.
26555         (aarch64_uqdec<mode>): New insn.
26557 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26558             Alan Hayward  <alan.hayward@arm.com>
26559             David Sherwood  <david.sherwood@arm.com>
26561         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26562         (reduc_xor_scal_optab): New optabs.
26563         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26564         (reduc_xor_scal_@var{m}): Document.
26565         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26566         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26567         internal functions.
26568         * fold-const-call.c (fold_const_call): Handle them.
26569         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26570         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26571         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26572         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26573         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26574         (UNSPEC_XORV): New unspecs.
26575         (optab): Add entries for them.
26576         (BITWISEV): New int iterator.
26577         (bit_reduc_op): New int attributes.
26579 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26580             Alan Hayward  <alan.hayward@arm.com>
26581             David Sherwood  <david.sherwood@arm.com>
26583         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26584         * internal-fn.def (VEC_SHL_INSERT): New internal function.
26585         * optabs.def (vec_shl_insert_optab): New optab.
26586         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26587         (duplicate_and_interleave): Likewise.
26588         * tree-vect-loop.c: Include internal-fn.h.
26589         (neutral_op_for_slp_reduction): New function, split out from
26590         get_initial_defs_for_reduction.
26591         (get_initial_def_for_reduction): Handle option 2 for variable-length
26592         vectors by loading the neutral value into a vector and then shifting
26593         the initial value into element 0.
26594         (get_initial_defs_for_reduction): Replace the code argument with
26595         the neutral value calculated by neutral_op_for_slp_reduction.
26596         Use gimple_build_vector for constant-length vectors.
26597         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26598         but the first group_size elements have a neutral value.
26599         Use duplicate_and_interleave otherwise.
26600         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26601         Update call to get_initial_defs_for_reduction.  Handle SLP
26602         reductions for variable-length vectors by creating one vector
26603         result for each scalar result, with the elements associated
26604         with other scalar results stubbed out with the neutral value.
26605         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26606         Require IFN_VEC_SHL_INSERT for double reductions on
26607         variable-length vectors, or SLP reductions that have
26608         a neutral value.  Require can_duplicate_and_interleave_p
26609         support for variable-length unchained SLP reductions if there
26610         is no neutral value, such as for MIN/MAX reductions.  Also require
26611         the number of vector elements to be a multiple of the number of
26612         SLP statements when doing variable-length unchained SLP reductions.
26613         Update call to vect_create_epilog_for_reduction.
26614         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26615         and remove initial values.
26616         (duplicate_and_interleave): Make public.
26617         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26618         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26620 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26621             Alan Hayward  <alan.hayward@arm.com>
26622             David Sherwood  <david.sherwood@arm.com>
26624         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26625         (can_duplicate_and_interleave_p): New function.
26626         (vect_get_and_check_slp_defs): Take the vector of statements
26627         rather than just the current one.  Remove excess parentheses.
26628         Restriction rejectinon of vect_constant_def and vect_external_def
26629         for variable-length vectors to boolean types, or types for which
26630         can_duplicate_and_interleave_p is false.
26631         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26632         (duplicate_and_interleave): New function.
26633         (vect_get_constant_vectors): Use gimple_build_vector for
26634         constant-length vectors and suitable variable-length constant
26635         vectors.  Use duplicate_and_interleave for other variable-length
26636         vectors.  Don't defer the update when inserting new statements.
26638 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26639             Alan Hayward  <alan.hayward@arm.com>
26640             David Sherwood  <david.sherwood@arm.com>
26642         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26643         min_profitable_iters doesn't go negative.
26645 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26646             Alan Hayward  <alan.hayward@arm.com>
26647             David Sherwood  <david.sherwood@arm.com>
26649         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26650         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26651         * optabs.def (vec_mask_load_lanes_optab): New optab.
26652         (vec_mask_store_lanes_optab): Likewise.
26653         * internal-fn.def (MASK_LOAD_LANES): New internal function.
26654         (MASK_STORE_LANES): Likewise.
26655         * internal-fn.c (mask_load_lanes_direct): New macro.
26656         (mask_store_lanes_direct): Likewise.
26657         (expand_mask_load_optab_fn): Handle masked operations.
26658         (expand_mask_load_lanes_optab_fn): New macro.
26659         (expand_mask_store_optab_fn): Handle masked operations.
26660         (expand_mask_store_lanes_optab_fn): New macro.
26661         (direct_mask_load_lanes_optab_supported_p): Likewise.
26662         (direct_mask_store_lanes_optab_supported_p): Likewise.
26663         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26664         parameter.
26665         (vect_load_lanes_supported): Likewise.
26666         * tree-vect-data-refs.c (strip_conversion): New function.
26667         (can_group_stmts_p): Likewise.
26668         (vect_analyze_data_ref_accesses): Use it instead of checking
26669         for a pair of assignments.
26670         (vect_store_lanes_supported): Take a masked_p parameter.
26671         (vect_load_lanes_supported): Likewise.
26672         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26673         vect_store_lanes_supported and vect_load_lanes_supported.
26674         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26675         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26676         parameter.  Don't allow gaps for masked accesses.
26677         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
26678         and vect_load_lanes_supported.
26679         (get_load_store_type): Take a masked_p parameter and update
26680         call to get_group_load_store_type.
26681         (vectorizable_store): Update call to get_load_store_type.
26682         Handle IFN_MASK_STORE_LANES.
26683         (vectorizable_load): Update call to get_load_store_type.
26684         Handle IFN_MASK_LOAD_LANES.
26686 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26687             Alan Hayward  <alan.hayward@arm.com>
26688             David Sherwood  <david.sherwood@arm.com>
26690         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26691         modes for SVE.
26692         * config/aarch64/aarch64-protos.h
26693         (aarch64_sve_struct_memory_operand_p): Declare.
26694         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26695         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26696         (VPRED, vpred): Handle SVE structure modes.
26697         * config/aarch64/constraints.md (Utx): New constraint.
26698         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26699         (aarch64_sve_struct_nonimmediate_operand): New predicates.
26700         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26701         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26702         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26703         structure modes.  Split into pieces after RA.
26704         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26705         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26706         New patterns.
26707         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26708         SVE structure modes.
26709         (aarch64_classify_address): Likewise.
26710         (sizetochar): Move earlier in file.
26711         (aarch64_print_operand): Handle SVE register lists.
26712         (aarch64_array_mode): New function.
26713         (aarch64_sve_struct_memory_operand_p): Likewise.
26714         (TARGET_ARRAY_MODE): Redefine.
26716 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26717             Alan Hayward  <alan.hayward@arm.com>
26718             David Sherwood  <david.sherwood@arm.com>
26720         * target.def (array_mode): New target hook.
26721         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26722         * doc/tm.texi: Regenerate.
26723         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26724         * hooks.c (hook_optmode_mode_uhwi_none): New function.
26725         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26726         targetm.array_mode.
26727         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
26728         type sizes.
26730 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26731             Alan Hayward  <alan.hayward@arm.com>
26732             David Sherwood  <david.sherwood@arm.com>
26734         * fold-const.c (fold_binary_loc): Check the argument types
26735         rather than the result type when testing for a vector operation.
26737 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26739         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26740         * doc/tm.texi: Regenerate.
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/invoke.texi (-msve-vector-bits=): Document new option.
26747         (sve): Document new AArch64 extension.
26748         * doc/md.texi (w): Extend the description of the AArch64
26749         constraint to include SVE vectors.
26750         (Upl, Upa): Document new AArch64 predicate constraints.
26751         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26752         enum.
26753         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26754         (msve-vector-bits=): New option.
26755         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26756         SVE when these are disabled.
26757         (sve): New extension.
26758         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26759         modes.  Adjust their number of units based on aarch64_sve_vg.
26760         (MAX_BITSIZE_MODE_ANY_MODE): Define.
26761         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26762         aarch64_addr_query_type.
26763         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26764         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26765         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26766         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26767         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26768         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26769         (aarch64_simd_imm_zero_p): Delete.
26770         (aarch64_check_zero_based_sve_index_immediate): Declare.
26771         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26772         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26773         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26774         (aarch64_sve_float_mul_immediate_p): Likewise.
26775         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26776         rather than an rtx.
26777         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26778         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
26779         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
26780         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
26781         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
26782         (aarch64_regmode_natural_size): Likewise.
26783         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
26784         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
26785         left one place.
26786         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
26787         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
26788         for VG and the SVE predicate registers.
26789         (V_ALIASES): Add a "z"-prefixed alias.
26790         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
26791         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
26792         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
26793         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
26794         (REG_CLASS_NAMES): Add entries for them.
26795         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
26796         and the predicate registers.
26797         (aarch64_sve_vg): Declare.
26798         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
26799         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
26800         (REGMODE_NATURAL_SIZE): Define.
26801         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
26802         SVE macros.
26803         * config/aarch64/aarch64.c: Include cfgrtl.h.
26804         (simd_immediate_info): Add a constructor for series vectors,
26805         and an associated step field.
26806         (aarch64_sve_vg): New variable.
26807         (aarch64_dbx_register_number): Handle VG and the predicate registers.
26808         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
26809         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
26810         (VEC_ANY_DATA, VEC_STRUCT): New constants.
26811         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
26812         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
26813         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
26814         (aarch64_get_mask_mode): New functions.
26815         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
26816         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
26817         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
26818         predicate modes and predicate registers.  Explicitly restrict
26819         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
26820         to store a vector mode if it is recognized by
26821         aarch64_classify_vector_mode.
26822         (aarch64_regmode_natural_size): New function.
26823         (aarch64_hard_regno_caller_save_mode): Return the original mode
26824         for predicates.
26825         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
26826         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
26827         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
26828         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
26829         functions.
26830         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
26831         does not overlap dest if the function is frame-related.  Handle
26832         SVE constants.
26833         (aarch64_split_add_offset): New function.
26834         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
26835         them aarch64_add_offset.
26836         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
26837         and update call to aarch64_sub_sp.
26838         (aarch64_add_cfa_expression): New function.
26839         (aarch64_expand_prologue): Pass extra temporary registers to the
26840         functions above.  Handle the case in which we need to emit new
26841         DW_CFA_expressions for registers that were originally saved
26842         relative to the stack pointer, but now have to be expressed
26843         relative to the frame pointer.
26844         (aarch64_output_mi_thunk): Pass extra temporary registers to the
26845         functions above.
26846         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
26847         IP0 and IP1 values for SVE frames.
26848         (aarch64_expand_vec_series): New function.
26849         (aarch64_expand_sve_widened_duplicate): Likewise.
26850         (aarch64_expand_sve_const_vector): Likewise.
26851         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
26852         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
26853         into the register, rather than emitting a SET directly.
26854         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
26855         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
26856         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
26857         (offset_9bit_signed_scaled_p): New functions.
26858         (aarch64_replicate_bitmask_imm): New function.
26859         (aarch64_bitmask_imm): Use it.
26860         (aarch64_cannot_force_const_mem): Reject expressions involving
26861         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
26862         (aarch64_classify_index): Handle SVE indices, by requiring
26863         a plain register index with a scale that matches the element size.
26864         (aarch64_classify_address): Handle SVE addresses.  Assert that
26865         the mode of the address is VOIDmode or an integer mode.
26866         Update call to aarch64_classify_symbol.
26867         (aarch64_classify_symbolic_expression): Update call to
26868         aarch64_classify_symbol.
26869         (aarch64_const_vec_all_in_range_p): New function.
26870         (aarch64_print_vector_float_operand): Likewise.
26871         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
26872         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
26873         and the FP immediates 1.0 and 0.5.
26874         (aarch64_print_address_internal): Handle SVE addresses.
26875         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
26876         (aarch64_regno_regclass): Handle predicate registers.
26877         (aarch64_secondary_reload): Handle big-endian reloads of SVE
26878         data modes.
26879         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
26880         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
26881         (aarch64_convert_sve_vector_bits): New function.
26882         (aarch64_override_options): Use it to handle -msve-vector-bits=.
26883         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26884         rather than an rtx.
26885         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
26886         Handle SVE vector and predicate modes.  Accept VL-based constants
26887         that need only one temporary register, and VL offsets that require
26888         no temporary registers.
26889         (aarch64_conditional_register_usage): Mark the predicate registers
26890         as fixed if SVE isn't available.
26891         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
26892         Return true for SVE vector and predicate modes.
26893         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
26894         rather than an unsigned int.  Handle SVE modes.
26895         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
26896         SVE modes.
26897         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
26898         if SVE is enabled.
26899         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26900         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26901         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26902         (aarch64_sve_float_mul_immediate_p): New functions.
26903         (aarch64_sve_valid_immediate): New function.
26904         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
26905         Explicitly reject structure modes.  Check for INDEX constants.
26906         Handle PTRUE and PFALSE constants.
26907         (aarch64_check_zero_based_sve_index_immediate): New function.
26908         (aarch64_simd_imm_zero_p): Delete.
26909         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
26910         vector modes.  Accept constants in the range of CNT[BHWD].
26911         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
26912         ask for an Advanced SIMD mode.
26913         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
26914         (aarch64_simd_vector_alignment): Handle SVE predicates.
26915         (aarch64_vectorize_preferred_vector_alignment): New function.
26916         (aarch64_simd_vector_alignment_reachable): Use it instead of
26917         the vector size.
26918         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
26919         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
26920         functions.
26921         (MAX_VECT_LEN): Delete.
26922         (expand_vec_perm_d): Add a vec_flags field.
26923         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
26924         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
26925         (aarch64_evpc_ext): Don't apply a big-endian lane correction
26926         for SVE modes.
26927         (aarch64_evpc_rev): Rename to...
26928         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
26929         (aarch64_evpc_rev_global): New function.
26930         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
26931         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
26932         MAX_VECT_LEN.
26933         (aarch64_evpc_sve_tbl): New function.
26934         (aarch64_expand_vec_perm_const_1): Update after rename of
26935         aarch64_evpc_rev.  Handle SVE permutes too, trying
26936         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
26937         than aarch64_evpc_tbl.
26938         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
26939         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
26940         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
26941         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
26942         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
26943         (aarch64_expand_sve_vcond): New functions.
26944         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
26945         of aarch64_vector_mode_p.
26946         (aarch64_dwarf_poly_indeterminate_value): New function.
26947         (aarch64_compute_pressure_classes): Likewise.
26948         (aarch64_can_change_mode_class): Likewise.
26949         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
26950         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
26951         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
26952         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
26953         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
26954         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
26955         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
26956         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
26957         constraints.
26958         (Dn, Dl, Dr): Accept const as well as const_vector.
26959         (Dz): Likewise.  Compare against CONST0_RTX.
26960         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
26961         of "vector" where appropriate.
26962         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
26963         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
26964         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
26965         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
26966         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
26967         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
26968         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
26969         (v_int_equiv): Extend to SVE modes.
26970         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
26971         mode attributes.
26972         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
26973         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
26974         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
26975         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
26976         (SVE_COND_FP_CMP): New int iterators.
26977         (perm_hilo): Handle the new unpack unspecs.
26978         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
26979         attributes.
26980         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
26981         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
26982         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
26983         (aarch64_equality_operator, aarch64_constant_vector_operand)
26984         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
26985         (aarch64_sve_nonimmediate_operand): Likewise.
26986         (aarch64_sve_general_operand): Likewise.
26987         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
26988         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
26989         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
26990         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
26991         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
26992         (aarch64_sve_float_arith_immediate): Likewise.
26993         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
26994         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
26995         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
26996         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
26997         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
26998         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
26999         (aarch64_sve_float_arith_operand): Likewise.
27000         (aarch64_sve_float_arith_with_sub_operand): Likewise.
27001         (aarch64_sve_float_mul_operand): Likewise.
27002         (aarch64_sve_vec_perm_operand): Likewise.
27003         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
27004         (aarch64_mov_operand): Accept const_poly_int and const_vector.
27005         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
27006         as well as const_vector.
27007         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
27008         in file.  Use CONST0_RTX and CONSTM1_RTX.
27009         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
27010         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
27011         Use aarch64_simd_imm_zero.
27012         * config/aarch64/aarch64-sve.md: New file.
27013         * config/aarch64/aarch64.md: Include it.
27014         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
27015         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
27016         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
27017         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
27018         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
27019         (sve): New attribute.
27020         (enabled): Disable instructions with the sve attribute unless
27021         TARGET_SVE.
27022         (movqi, movhi): Pass CONST_POLY_INT operaneds through
27023         aarch64_expand_mov_immediate.
27024         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
27025         CNT[BHSD] immediates.
27026         (movti): Split CONST_POLY_INT moves into two halves.
27027         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
27028         Split additions that need a temporary here if the destination
27029         is the stack pointer.
27030         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
27031         (*add<mode>3_poly_1): New instruction.
27032         (set_clobber_cc): New expander.
27034 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27036         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
27037         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
27038         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
27039         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
27040         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
27041         Change innermode from fixed_mode_size to machine_mode.
27042         (simplify_subreg): Update call accordingly.  Handle a constant-sized
27043         subreg of a variable-length CONST_VECTOR.
27045 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27046             Alan Hayward  <alan.hayward@arm.com>
27047             David Sherwood  <david.sherwood@arm.com>
27049         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
27050         (add_offset_to_base): New function, split out from...
27051         (create_mem_ref): ...here.  When handling a scale other than 1,
27052         check first whether the address is valid without the offset.
27053         Add it into the base if so, leaving the index and scale as-is.
27055 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27057         PR c++/83778
27058         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
27059         fold_for_warn before checking if arg2 is INTEGER_CST.
27061 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
27063         * config/rs6000/predicates.md (load_multiple_operation): Delete.
27064         (store_multiple_operation): Delete.
27065         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
27066         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
27067         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
27068         guarded by TARGET_STRING.
27069         (rs6000_output_load_multiple): Delete.
27070         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
27071         OPTION_MASK_STRING / TARGET_STRING handling.
27072         (print_operand) <'N', 'O'>: Add comment that these are unused now.
27073         (const rs6000_opt_masks) <"string">: Change mask to 0.
27074         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
27075         (MASK_STRING): Delete.
27076         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
27077         parts.  Simplify.
27078         (load_multiple): Delete.
27079         (*ldmsi8): Delete.
27080         (*ldmsi7): Delete.
27081         (*ldmsi6): Delete.
27082         (*ldmsi5): Delete.
27083         (*ldmsi4): Delete.
27084         (*ldmsi3): Delete.
27085         (store_multiple): Delete.
27086         (*stmsi8): Delete.
27087         (*stmsi7): Delete.
27088         (*stmsi6): Delete.
27089         (*stmsi5): Delete.
27090         (*stmsi4): Delete.
27091         (*stmsi3): Delete.
27092         (movmemsi_8reg): Delete.
27093         (corresponding unnamed define_insn): Delete.
27094         (movmemsi_6reg): Delete.
27095         (corresponding unnamed define_insn): Delete.
27096         (movmemsi_4reg): Delete.
27097         (corresponding unnamed define_insn): Delete.
27098         (movmemsi_2reg): Delete.
27099         (corresponding unnamed define_insn): Delete.
27100         (movmemsi_1reg): Delete.
27101         (corresponding unnamed define_insn): Delete.
27102         * config/rs6000/rs6000.opt (mno-string): New.
27103         (mstring): Replace by deprecation warning stub.
27104         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
27106 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27108         * regrename.c (regrename_do_replace): If replacing the same
27109         reg multiple times, try to reuse last created gen_raw_REG.
27111         PR debug/81155
27112         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
27113         main to workaround a bug in GDB.
27115 2018-01-12  Tom de Vries  <tom@codesourcery.com>
27117         PR target/83737
27118         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
27120 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
27122         PR rtl-optimization/80481
27123         * ira-color.c (get_cap_member): New function.
27124         (allocnos_conflict_by_live_ranges_p): Use it.
27125         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
27126         (setup_slot_coalesced_allocno_live_ranges): Ditto.
27128 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
27130         PR target/83628
27131         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
27132         (*saddl_se_1): Ditto.
27133         (*ssubsi_1): Ditto.
27134         (*ssubl_se_1): Ditto.
27136 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27138         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
27139         rather than wi::to_widest for DR_INITs.
27140         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
27141         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
27142         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
27143         INTEGER_CSTs.
27144         (vect_analyze_group_access_1): Note that here.
27146 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27148         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
27149         polynomial type sizes.
27151 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27153         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
27154         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
27155         (gimple_add_tmp_var): Likewise.
27157 2018-01-12  Martin Liska  <mliska@suse.cz>
27159         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
27160         (gimple_alloc_sizes): Likewise.
27161         (dump_gimple_statistics): Use PRIu64 in printf format.
27162         * gimple.h: Change uint64_t to int.
27164 2018-01-12  Martin Liska  <mliska@suse.cz>
27166         * tree-core.h: Use uint64_t instead of int.
27167         * tree.c (tree_node_counts): Likewise.
27168         (tree_node_sizes): Likewise.
27169         (dump_tree_statistics): Use PRIu64 in printf format.
27171 2018-01-12  Martin Liska  <mliska@suse.cz>
27173         * Makefile.in: As qsort_chk is implemented in vec.c, add
27174         vec.o to linkage of gencfn-macros.
27175         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
27176         passing the info to record_node_allocation_statistics.
27177         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
27178         and pass the info.
27179         * ggc-common.c (struct ggc_usage): Add operator== and use
27180         it in operator< and compare function.
27181         * mem-stats.h (struct mem_usage): Likewise.
27182         * vec.c (struct vec_usage): Remove operator< and compare
27183         function. Can be simply inherited.
27185 2018-01-12  Martin Jambor  <mjambor@suse.cz>
27187         PR target/81616
27188         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
27189         * tree-ssa-math-opts.c: Include domwalk.h.
27190         (convert_mult_to_fma_1): New function.
27191         (fma_transformation_info): New type.
27192         (fma_deferring_state): Likewise.
27193         (cancel_fma_deferring): New function.
27194         (result_of_phi): Likewise.
27195         (last_fma_candidate_feeds_initial_phi): Likewise.
27196         (convert_mult_to_fma): Added deferring logic, split actual
27197         transformation to convert_mult_to_fma_1.
27198         (math_opts_dom_walker): New type.
27199         (math_opts_dom_walker::after_dom_children): New method, body moved
27200         here from pass_optimize_widening_mul::execute, added deferring logic
27201         bits.
27202         (pass_optimize_widening_mul::execute): Moved most of code to
27203         math_opts_dom_walker::after_dom_children.
27204         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
27205         * config/i386/i386.c (ix86_option_override_internal): Added
27206         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
27208 2018-01-12  Richard Biener  <rguenther@suse.de>
27210         PR debug/83157
27211         * dwarf2out.c (gen_variable_die): Do not reset old_die for
27212         inline instance vars.
27214 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
27216         PR target/81819
27217         * config/rx/rx.c (rx_is_restricted_memory_address):
27218         Handle SUBREG case.
27220 2018-01-12  Richard Biener  <rguenther@suse.de>
27222         PR tree-optimization/80846
27223         * target.def (split_reduction): New target hook.
27224         * targhooks.c (default_split_reduction): New function.
27225         * targhooks.h (default_split_reduction): Declare.
27226         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
27227         target requests first reduce vectors by combining low and high
27228         parts.
27229         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
27230         (get_vectype_for_scalar_type_and_size): Export.
27231         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
27232         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
27233         * doc/tm.texi: Regenerate.
27234         * config/i386/i386.c (ix86_split_reduction): Implement
27235         TARGET_VECTORIZE_SPLIT_REDUCTION.
27237 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27239         PR target/83368
27240         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
27241         in PIC mode except for TARGET_VXWORKS_RTP.
27242         * config/sparc/sparc.c: Include cfgrtl.h.
27243         (TARGET_INIT_PIC_REG): Define.
27244         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
27245         (sparc_pic_register_p): New predicate.
27246         (sparc_legitimate_address_p): Use it.
27247         (sparc_legitimize_pic_address): Likewise.
27248         (sparc_delegitimize_address): Likewise.
27249         (sparc_mode_dependent_address_p): Likewise.
27250         (gen_load_pcrel_sym): Remove 4th parameter.
27251         (load_got_register): Adjust call to above.  Remove obsolete stuff.
27252         (sparc_expand_prologue): Do not call load_got_register here.
27253         (sparc_flat_expand_prologue): Likewise.
27254         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
27255         (sparc_use_pseudo_pic_reg): New function.
27256         (sparc_init_pic_reg): Likewise.
27257         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27258         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
27260 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
27262         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
27263         Add item for branch_cost.
27265 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27267         PR rtl-optimization/83565
27268         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
27269         not extend the result to a larger mode for rotate operations.
27270         (num_sign_bit_copies1): Likewise.
27272 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27274         PR target/40411
27275         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
27276         -symbolic.
27277         Use values-Xc.o for -pedantic.
27278         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
27280 2018-01-12  Martin Liska  <mliska@suse.cz>
27282         PR ipa/83054
27283         * ipa-devirt.c (final_warning_record::grow_type_warnings):
27284         New function.
27285         (possible_polymorphic_call_targets): Use it.
27286         (ipa_devirt): Likewise.
27288 2018-01-12  Martin Liska  <mliska@suse.cz>
27290         * profile-count.h (enum profile_quality): Use 0 as invalid
27291         enum value of profile_quality.
27293 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
27295         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
27296         -mext-string options.
27298 2018-01-12  Richard Biener  <rguenther@suse.de>
27300         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
27301         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
27302         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
27303         Likewise.
27304         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
27306 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
27308         * configure.ac (--with-long-double-format): Add support for the
27309         configuration option to change the default long double format on
27310         PowerPC systems.
27311         * config.gcc (powerpc*-linux*-*): Likewise.
27312         * configure: Regenerate.
27313         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
27314         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
27315         used without modification.
27317 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27319         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
27320         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
27321         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
27322         MISC_BUILTIN_SPEC_BARRIER.
27323         (rs6000_init_builtins): Likewise.
27324         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
27325         enum value.
27326         (speculation_barrier): New define_insn.
27327         * doc/extend.texi: Document __builtin_speculation_barrier.
27329 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27331         PR target/83203
27332         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
27333         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
27334         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
27335         iterators.
27336         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
27337         integral modes instead of "ss" and "sd".
27338         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
27339         vectors with 32-bit and 64-bit elements.
27340         (vecdupssescalarmodesuffix): New mode attribute.
27341         (vec_dup<mode>): Use it.
27343 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
27345         PR target/83330
27346         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
27347         frame if argument is passed on stack.
27349 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27351         PR target/82682
27352         * ree.c (combine_reaching_defs): Optimize also
27353         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
27354         reg2=any_extend(exp); reg1=reg2;, formatting fix.
27356 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27358         PR middle-end/83189
27359         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
27361 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27363         PR middle-end/83718
27364         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
27365         after they are computed.
27367 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
27369         PR tree-optimization/83695
27370         * gimple-loop-linterchange.cc
27371         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
27372         reset cached scev information after interchange.
27373         (pass_linterchange::execute): Remove call to scev_reset_htab.
27375 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27377         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
27378         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
27379         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
27380         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
27381         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
27382         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
27383         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
27384         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
27385         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
27386         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
27387         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
27388         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
27389         (V_lane_reg): Likewise.
27390         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
27391         New define_expand.
27392         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
27393         (vfmal_lane_low<mode>_intrinsic,
27394         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
27395         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27396         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27397         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27398         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27399         vfmsl_lane_high<mode>_intrinsic): New define_insns.
27401 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27403         * config/arm/arm-cpus.in (fp16fml): New feature.
27404         (ALL_SIMD): Add fp16fml.
27405         (armv8.2-a): Add fp16fml as an option.
27406         (armv8.3-a): Likewise.
27407         (armv8.4-a): Add fp16fml as part of fp16.
27408         * config/arm/arm.h (TARGET_FP16FML): Define.
27409         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27410         when appropriate.
27411         * config/arm/arm-modes.def (V2HF): Define.
27412         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27413         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27414         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27415         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27416         vfmsl_low, vfmsl_high): New set of builtins.
27417         * config/arm/iterators.md (PLUSMINUS): New code iterator.
27418         (vfml_op): New code attribute.
27419         (VFMLHALVES): New int iterator.
27420         (VFML, VFMLSEL): New mode attributes.
27421         (V_reg): Define mapping for V2HF.
27422         (V_hi, V_lo): New mode attributes.
27423         (VF_constraint): Likewise.
27424         (vfml_half, vfml_half_selector): New int attributes.
27425         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27426         define_expand.
27427         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27428         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27429         New define_insn.
27430         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27431         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27432         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27433         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
27434         documentation.
27435         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27436         Document new effective target and option set.
27438 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27440         * config/arm/arm-cpus.in (armv8_4): New feature.
27441         (ARMv8_4a): New fgroup.
27442         (armv8.4-a): New arch.
27443         * config/arm/arm-tables.opt: Regenerate.
27444         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27445         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27446         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27447         Add matching rules for -march=armv8.4-a and extensions.
27448         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27450 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
27452         PR target/81821
27453         * config/rx/rx.md (BW): New mode attribute.
27454         (sync_lock_test_and_setsi): Add mode suffix to insn output.
27456 2018-01-11  Richard Biener  <rguenther@suse.de>
27458         PR tree-optimization/83435
27459         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27460         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27461         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27463 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27464             Alan Hayward  <alan.hayward@arm.com>
27465             David Sherwood  <david.sherwood@arm.com>
27467         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27468         field.
27469         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
27470         (aarch64_print_address_internal): Use it to check for a zero offset.
27472 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27473             Alan Hayward  <alan.hayward@arm.com>
27474             David Sherwood  <david.sherwood@arm.com>
27476         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27477         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27478         Return a poly_int64 rather than a HOST_WIDE_INT.
27479         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27480         rather than a HOST_WIDE_INT.
27481         * config/aarch64/aarch64.h (aarch64_frame): Protect with
27482         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
27483         hard_fp_offset, frame_size, initial_adjust, callee_offset and
27484         final_offset from HOST_WIDE_INT to poly_int64.
27485         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27486         to_constant when getting the number of units in an Advanced SIMD
27487         mode.
27488         (aarch64_builtin_vectorized_function): Check for a constant number
27489         of units.
27490         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27491         GET_MODE_SIZE.
27492         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27493         attribute instead of GET_MODE_NUNITS.
27494         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27495         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27496         GET_MODE_SIZE for fixed-size registers.
27497         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27498         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27499         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27500         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27501         (aarch64_print_operand, aarch64_print_address_internal)
27502         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27503         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27504         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27505         Handle polynomial GET_MODE_SIZE.
27506         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
27507         wider than SImode without modification.
27508         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27509         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27510         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27511         passing and returning SVE modes.
27512         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27513         rather than GEN_INT.
27514         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27515         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27516         (aarch64_allocate_and_probe_stack_space): Likewise.
27517         (aarch64_layout_frame): Cope with polynomial offsets.
27518         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27519         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
27520         polynomial offsets.
27521         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27522         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27523         poly_int64 rather than a HOST_WIDE_INT.
27524         (aarch64_get_separate_components, aarch64_process_components)
27525         (aarch64_expand_prologue, aarch64_expand_epilogue)
27526         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27527         (aarch64_anchor_offset): New function, split out from...
27528         (aarch64_legitimize_address): ...here.
27529         (aarch64_builtin_vectorization_cost): Handle polynomial
27530         TYPE_VECTOR_SUBPARTS.
27531         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27532         GET_MODE_NUNITS.
27533         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27534         number of elements from the PARALLEL rather than the mode.
27535         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27536         rather than GET_MODE_BITSIZE.
27537         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27538         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27539         (aarch64_expand_vec_perm_const_1): Handle polynomial
27540         d->perm.length () and d->perm elements.
27541         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
27542         Apply to_constant to d->perm elements.
27543         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27544         polynomial CONST_VECTOR_NUNITS.
27545         (aarch64_move_pointer): Take amount as a poly_int64 rather
27546         than an int.
27547         (aarch64_progress_pointer): Avoid temporary variable.
27548         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27549         the mode attribute instead of GET_MODE.
27551 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27552             Alan Hayward  <alan.hayward@arm.com>
27553             David Sherwood  <david.sherwood@arm.com>
27555         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27556         x exists before using it.
27557         (aarch64_add_constant_internal): Rename to...
27558         (aarch64_add_offset_1): ...this.  Replace regnum with separate
27559         src and dest rtxes.  Handle the case in which they're different,
27560         including when the offset is zero.  Replace scratchreg with an rtx.
27561         Use 2 additions if there is no spare register into which we can
27562         move a 16-bit constant.
27563         (aarch64_add_constant): Delete.
27564         (aarch64_add_offset): Replace reg with separate src and dest
27565         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
27566         Use aarch64_add_offset_1.
27567         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27568         an rtx rather than an int.  Take the delta as a poly_int64
27569         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
27570         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27571         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27572         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27573         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27574         and aarch64_add_sp.
27575         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27576         aarch64_add_constant.
27578 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27580         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27581         Use scalar_float_mode.
27583 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27585         * config/aarch64/aarch64-simd.md
27586         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27587         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27588         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27589         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27590         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27591         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27592         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27593         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27594         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27595         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27597 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27599         PR target/83514
27600         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27601         targ_options->x_arm_arch_string is non NULL.
27603 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
27605         * config/aarch64/aarch64.h
27606         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
27608 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
27610         PR target/82096
27611         * expmed.c (emit_store_flag_force): Swap if const op0
27612         and change VOIDmode to mode of op0.
27614 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27616         PR rtl-optimization/83761
27617         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27618         than bytes to mode_for_size.
27620 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27622         PR middle-end/83189
27623         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27624         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27625         profile.
27627 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27629         PR middle-end/83575
27630         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27631         when in layout mode.
27632         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27633         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27634         partition fixup.
27636 2018-01-10  Michael Collison  <michael.collison@arm.com>
27638         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27639         * config/aarch64/aarch64-option-extension.def: Add
27640         AARCH64_OPT_EXTENSION of 'fp16fml'.
27641         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27642         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27643         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27644         * config/aarch64/constraints.md (Ui7): New constraint.
27645         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27646         (VFMLA_SEL_W): Ditto.
27647         (f16quad): Ditto.
27648         (f16mac1): Ditto.
27649         (VFMLA16_LOW): New int iterator.
27650         (VFMLA16_HIGH): Ditto.
27651         (UNSPEC_FMLAL): New unspec.
27652         (UNSPEC_FMLSL): Ditto.
27653         (UNSPEC_FMLAL2): Ditto.
27654         (UNSPEC_FMLSL2): Ditto.
27655         (f16mac): New code attribute.
27656         * config/aarch64/aarch64-simd-builtins.def
27657         (aarch64_fmlal_lowv2sf): Ditto.
27658         (aarch64_fmlsl_lowv2sf): Ditto.
27659         (aarch64_fmlalq_lowv4sf): Ditto.
27660         (aarch64_fmlslq_lowv4sf): Ditto.
27661         (aarch64_fmlal_highv2sf): Ditto.
27662         (aarch64_fmlsl_highv2sf): Ditto.
27663         (aarch64_fmlalq_highv4sf): Ditto.
27664         (aarch64_fmlslq_highv4sf): Ditto.
27665         (aarch64_fmlal_lane_lowv2sf): Ditto.
27666         (aarch64_fmlsl_lane_lowv2sf): Ditto.
27667         (aarch64_fmlal_laneq_lowv2sf): Ditto.
27668         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27669         (aarch64_fmlalq_lane_lowv4sf): Ditto.
27670         (aarch64_fmlsl_lane_lowv4sf): Ditto.
27671         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27672         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27673         (aarch64_fmlal_lane_highv2sf): Ditto.
27674         (aarch64_fmlsl_lane_highv2sf): Ditto.
27675         (aarch64_fmlal_laneq_highv2sf): Ditto.
27676         (aarch64_fmlsl_laneq_highv2sf): Ditto.
27677         (aarch64_fmlalq_lane_highv4sf): Ditto.
27678         (aarch64_fmlsl_lane_highv4sf): Ditto.
27679         (aarch64_fmlalq_laneq_highv4sf): Ditto.
27680         (aarch64_fmlsl_laneq_highv4sf): Ditto.
27681         * config/aarch64/aarch64-simd.md:
27682         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27683         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27684         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27685         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27686         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27687         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27688         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27689         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27690         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27691         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27692         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27693         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27694         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27695         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27696         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27697         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27698         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27699         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27700         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27701         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27702         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27703         (vfmlsl_low_u32): Ditto.
27704         (vfmlalq_low_u32): Ditto.
27705         (vfmlslq_low_u32): Ditto.
27706         (vfmlal_high_u32): Ditto.
27707         (vfmlsl_high_u32): Ditto.
27708         (vfmlalq_high_u32): Ditto.
27709         (vfmlslq_high_u32): Ditto.
27710         (vfmlal_lane_low_u32): Ditto.
27711         (vfmlsl_lane_low_u32): Ditto.
27712         (vfmlal_laneq_low_u32): Ditto.
27713         (vfmlsl_laneq_low_u32): Ditto.
27714         (vfmlalq_lane_low_u32): Ditto.
27715         (vfmlslq_lane_low_u32): Ditto.
27716         (vfmlalq_laneq_low_u32): Ditto.
27717         (vfmlslq_laneq_low_u32): Ditto.
27718         (vfmlal_lane_high_u32): Ditto.
27719         (vfmlsl_lane_high_u32): Ditto.
27720         (vfmlal_laneq_high_u32): Ditto.
27721         (vfmlsl_laneq_high_u32): Ditto.
27722         (vfmlalq_lane_high_u32): Ditto.
27723         (vfmlslq_lane_high_u32): Ditto.
27724         (vfmlalq_laneq_high_u32): Ditto.
27725         (vfmlslq_laneq_high_u32): Ditto.
27726         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27727         (AARCH64_FL_FOR_ARCH8_4): New.
27728         (AARCH64_ISA_F16FML): New ISA flag.
27729         (TARGET_F16FML): New feature flag for fp16fml.
27730         (doc/invoke.texi): Document new fp16fml option.
27732 2018-01-10  Michael Collison  <michael.collison@arm.com>
27734         * config/aarch64/aarch64-builtins.c:
27735         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27736         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27737         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27738         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27739         (AARCH64_ISA_SHA3): New ISA flag.
27740         (TARGET_SHA3): New feature flag for sha3.
27741         * config/aarch64/iterators.md (sha512_op): New int attribute.
27742         (CRYPTO_SHA512): New int iterator.
27743         (UNSPEC_SHA512H): New unspec.
27744         (UNSPEC_SHA512H2): Ditto.
27745         (UNSPEC_SHA512SU0): Ditto.
27746         (UNSPEC_SHA512SU1): Ditto.
27747         * config/aarch64/aarch64-simd-builtins.def
27748         (aarch64_crypto_sha512hqv2di): New builtin.
27749         (aarch64_crypto_sha512h2qv2di): Ditto.
27750         (aarch64_crypto_sha512su0qv2di): Ditto.
27751         (aarch64_crypto_sha512su1qv2di): Ditto.
27752         (aarch64_eor3qv8hi): Ditto.
27753         (aarch64_rax1qv2di): Ditto.
27754         (aarch64_xarqv2di): Ditto.
27755         (aarch64_bcaxqv8hi): Ditto.
27756         * config/aarch64/aarch64-simd.md:
27757         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27758         (aarch64_crypto_sha512su0qv2di): Ditto.
27759         (aarch64_crypto_sha512su1qv2di): Ditto.
27760         (aarch64_eor3qv8hi): Ditto.
27761         (aarch64_rax1qv2di): Ditto.
27762         (aarch64_xarqv2di): Ditto.
27763         (aarch64_bcaxqv8hi): Ditto.
27764         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27765         (vsha512h2q_u64): Ditto.
27766         (vsha512su0q_u64): Ditto.
27767         (vsha512su1q_u64): Ditto.
27768         (veor3q_u16): Ditto.
27769         (vrax1q_u64): Ditto.
27770         (vxarq_u64): Ditto.
27771         (vbcaxq_u16): Ditto.
27772         * config/arm/types.md (crypto_sha512): New type attribute.
27773         (crypto_sha3): Ditto.
27774         (doc/invoke.texi): Document new sha3 option.
27776 2018-01-10  Michael Collison  <michael.collison@arm.com>
27778         * config/aarch64/aarch64-builtins.c:
27779         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
27780         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27781         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
27782         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
27783         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
27784         (AARCH64_ISA_SM4): New ISA flag.
27785         (TARGET_SM4): New feature flag for sm4.
27786         * config/aarch64/aarch64-simd-builtins.def
27787         (aarch64_sm3ss1qv4si): Ditto.
27788         (aarch64_sm3tt1aq4si): Ditto.
27789         (aarch64_sm3tt1bq4si): Ditto.
27790         (aarch64_sm3tt2aq4si): Ditto.
27791         (aarch64_sm3tt2bq4si): Ditto.
27792         (aarch64_sm3partw1qv4si): Ditto.
27793         (aarch64_sm3partw2qv4si): Ditto.
27794         (aarch64_sm4eqv4si): Ditto.
27795         (aarch64_sm4ekeyqv4si): Ditto.
27796         * config/aarch64/aarch64-simd.md:
27797         (aarch64_sm3ss1qv4si): Ditto.
27798         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
27799         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
27800         (aarch64_sm4eqv4si): Ditto.
27801         (aarch64_sm4ekeyqv4si): Ditto.
27802         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
27803         (sm3part_op): Ditto.
27804         (CRYPTO_SM3TT): Ditto.
27805         (CRYPTO_SM3PART): Ditto.
27806         (UNSPEC_SM3SS1): New unspec.
27807         (UNSPEC_SM3TT1A): Ditto.
27808         (UNSPEC_SM3TT1B): Ditto.
27809         (UNSPEC_SM3TT2A): Ditto.
27810         (UNSPEC_SM3TT2B): Ditto.
27811         (UNSPEC_SM3PARTW1): Ditto.
27812         (UNSPEC_SM3PARTW2): Ditto.
27813         (UNSPEC_SM4E): Ditto.
27814         (UNSPEC_SM4EKEY): Ditto.
27815         * config/aarch64/constraints.md (Ui2): New constraint.
27816         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
27817         * config/arm/types.md (crypto_sm3): New type attribute.
27818         (crypto_sm4): Ditto.
27819         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
27820         (vsm3tt1aq_u32): Ditto.
27821         (vsm3tt1bq_u32): Ditto.
27822         (vsm3tt2aq_u32): Ditto.
27823         (vsm3tt2bq_u32): Ditto.
27824         (vsm3partw1q_u32): Ditto.
27825         (vsm3partw2q_u32): Ditto.
27826         (vsm4eq_u32): Ditto.
27827         (vsm4ekeyq_u32): Ditto.
27828         (doc/invoke.texi): Document new sm4 option.
27830 2018-01-10  Michael Collison  <michael.collison@arm.com>
27832         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
27833         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
27834         (AARCH64_FL_FOR_ARCH8_4): New.
27835         (AARCH64_FL_V8_4): New flag.
27836         (doc/invoke.texi): Document new armv8.4-a option.
27838 2018-01-10  Michael Collison  <michael.collison@arm.com>
27840         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27841         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
27842         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
27843         * config/aarch64/aarch64-option-extension.def: Add
27844         AARCH64_OPT_EXTENSION of 'sha2'.
27845         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
27846         (crypto): Disable sha2 and aes if crypto disabled.
27847         (crypto): Enable aes and sha2 if enabled.
27848         (simd): Disable sha2 and aes if simd disabled.
27849         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
27850         New flags.
27851         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
27852         (TARGET_SHA2): New feature flag for sha2.
27853         (TARGET_AES): New feature flag for aes.
27854         * config/aarch64/aarch64-simd.md:
27855         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
27856         conditional on TARGET_AES.
27857         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
27858         (aarch64_crypto_sha1hsi): Make pattern conditional
27859         on TARGET_SHA2.
27860         (aarch64_crypto_sha1hv4si): Ditto.
27861         (aarch64_be_crypto_sha1hv4si): Ditto.
27862         (aarch64_crypto_sha1su1v4si): Ditto.
27863         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
27864         (aarch64_crypto_sha1su0v4si): Ditto.
27865         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
27866         (aarch64_crypto_sha256su0v4si): Ditto.
27867         (aarch64_crypto_sha256su1v4si): Ditto.
27868         (doc/invoke.texi): Document new aes and sha2 options.
27870 2018-01-10  Martin Sebor  <msebor@redhat.com>
27872         PR tree-optimization/83781
27873         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
27874         as string arrays.
27876 2018-01-11  Martin Sebor  <msebor@gmail.com>
27877             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27879         PR tree-optimization/83501
27880         PR tree-optimization/81703
27882         * tree-ssa-strlen.c (get_string_cst): Rename...
27883         (get_string_len): ...to this.  Handle global constants.
27884         (handle_char_store): Adjust.
27886 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
27887             Jim Wilson  <jimw@sifive.com>
27889         * config/riscv/riscv-protos.h (riscv_output_return): New.
27890         * config/riscv/riscv.c (struct machine_function): New naked_p field.
27891         (riscv_attribute_table, riscv_output_return),
27892         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
27893         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
27894         (riscv_compute_frame_info): Only compute frame->mask if not a naked
27895         function.
27896         (riscv_expand_prologue): Add early return for naked function.
27897         (riscv_expand_epilogue): Likewise.
27898         (riscv_function_ok_for_sibcall): Return false for naked function.
27899         (riscv_set_current_function): New.
27900         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
27901         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
27902         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
27903         * doc/extend.texi (RISC-V Function Attributes): New.
27905 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
27907         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
27908         check for 128-bit long double before checking TCmode.
27909         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
27910         128-bit long doubles before checking TFmode or TCmode.
27911         (FLOAT128_IBM_P): Likewise.
27913 2018-01-10  Martin Sebor  <msebor@redhat.com>
27915         PR tree-optimization/83671
27916         * builtins.c (c_strlen): Unconditionally return zero for the empty
27917         string.
27918         Use -Warray-bounds for warnings.
27919         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
27920         for non-constant array indices with COMPONENT_REF, arrays of
27921         arrays, and pointers to arrays.
27922         (gimple_fold_builtin_strlen): Determine and set length range for
27923         non-constant character arrays.
27925 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
27927         PR middle-end/81897
27928         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
27929         empty blocks.
27931 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
27933         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
27935 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27937         PR target/83399
27938         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
27939         VECTOR_MEM_ALTIVEC_OR_VSX_P.
27940         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
27941         indexed_or_indirect_operand predicate.
27942         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
27943         (*vsx_le_perm_load_v8hi): Likewise.
27944         (*vsx_le_perm_load_v16qi): Likewise.
27945         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
27946         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
27947         (*vsx_le_perm_store_v8hi): Likewise.
27948         (*vsx_le_perm_store_v16qi): Likewise.
27949         (eight unnamed splitters): Likewise.
27951 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27953         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
27954         * config/rs6000/emmintrin.h: Likewise.
27955         * config/rs6000/mmintrin.h: Likewise.
27956         * config/rs6000/xmmintrin.h: Likewise.
27958 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
27960         PR c++/43486
27961         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
27962         "public_flag".
27963         * tree.c (tree_nop_conversion): Return true for location wrapper
27964         nodes.
27965         (maybe_wrap_with_location): New function.
27966         (selftest::check_strip_nops): New function.
27967         (selftest::test_location_wrappers): New function.
27968         (selftest::tree_c_tests): Call it.
27969         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
27970         (maybe_wrap_with_location): New decl.
27971         (EXPR_LOCATION_WRAPPER_P): New macro.
27972         (location_wrapper_p): New inline function.
27973         (tree_strip_any_location_wrapper): New inline function.
27975 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
27977         PR target/83735
27978         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
27979         stack_realign_offset for the largest alignment of stack slot
27980         actually used.
27981         (ix86_find_max_used_stack_alignment): New function.
27982         (ix86_finalize_stack_frame_flags): Use it.  Set
27983         max_used_stack_alignment if we don't realign stack.
27984         * config/i386/i386.h (machine_function): Add
27985         max_used_stack_alignment.
27987 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
27989         * config/arm/arm.opt (-mbranch-cost): New option.
27990         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
27991         account.
27993 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
27995         PR target/83629
27996         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
27997         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
27999 2018-01-10  Richard Biener  <rguenther@suse.de>
28001         PR debug/83765
28002         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
28003         early out so it also covers the case where we have a non-NULL
28004         origin.
28006 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28008         PR tree-optimization/83753
28009         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
28010         for non-strided grouped accesses if the number of elements is 1.
28012 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28014         PR target/81616
28015         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
28016         * i386.h (TARGET_USE_GATHER): Define.
28017         * x86-tune.def (X86_TUNE_USE_GATHER): New.
28019 2018-01-10  Martin Liska  <mliska@suse.cz>
28021         PR bootstrap/82831
28022         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
28023         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
28024         partitioning.
28025         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
28026         CLEANUP_NO_PARTITIONING is not set.
28028 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28030         * doc/rtl.texi: Remove documentation of (const ...) wrappers
28031         for vectors, as a partial revert of r254296.
28032         * rtl.h (const_vec_p): Delete.
28033         (const_vec_duplicate_p): Don't test for vector CONSTs.
28034         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
28035         * expmed.c (make_tree): Likewise.
28037         Revert:
28038         * common.md (E, F): Use CONSTANT_P instead of checking for
28039         CONST_VECTOR.
28040         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
28041         checking for CONST_VECTOR.
28043 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28045         PR middle-end/83575
28046         * predict.c (force_edge_cold): Handle in more sane way edges
28047         with no prediction.
28049 2018-01-09  Carl Love  <cel@us.ibm.com>
28051         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
28052         V4SI, V4SF types.
28053         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
28054         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
28055         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
28056         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
28057         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
28058         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
28059         * config/rs6000/rs6000-protos.h: Add extern defition for
28060         rs6000_generate_float2_double_code.
28061         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
28062         function.
28063         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
28064         (float2_v2df): Add define_expand.
28066 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
28068         PR target/83628
28069         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
28070         op_mode in the force_to_mode call.
28072 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28074         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
28075         instead of checking each element individually.
28076         (aarch64_evpc_uzp): Likewise.
28077         (aarch64_evpc_zip): Likewise.
28078         (aarch64_evpc_ext): Likewise.
28079         (aarch64_evpc_rev): Likewise.
28080         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
28081         instead of checking each element individually.  Return true without
28082         generating rtl if
28083         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
28084         whether all selected elements come from the same input, instead of
28085         checking each element individually.  Remove calls to gen_rtx_REG,
28086         start_sequence and end_sequence and instead assert that no rtl is
28087         generated.
28089 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28091         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
28092         order of HIGH and CONST checks.
28094 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28096         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
28097         if the destination isn't an SSA_NAME.
28099 2018-01-09  Richard Biener  <rguenther@suse.de>
28101         PR tree-optimization/83668
28102         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
28103         move prologue...
28104         (canonicalize_loop_form): ... here, renamed from ...
28105         (canonicalize_loop_closed_ssa_form): ... this and amended to
28106         swap successor edges for loop exit blocks to make us use
28107         the RPO order we need for initial schedule generation.
28109 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
28111         PR tree-optimization/64811
28112         * match.pd: When optimizing comparisons with Inf, avoid
28113         introducing or losing exceptions from comparisons with NaN.
28115 2018-01-09  Martin Liska  <mliska@suse.cz>
28117         PR sanitizer/82517
28118         * asan.c (shadow_mem_size): Add gcc_assert.
28120 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
28122         Don't save registers in main().
28124         PR target/83738
28125         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
28126         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
28127         * config/avr/avr.c (avr_set_current_function): Don't error if
28128         naked, OS_task or OS_main are specified at the same time.
28129         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
28130         OS_main.
28131         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
28132         attribute.
28133         * common/config/avr/avr-common.c (avr_option_optimization_table):
28134         Switch on -mmain-is-OS_task for optimizing compilations.
28136 2018-01-09  Richard Biener  <rguenther@suse.de>
28138         PR tree-optimization/83572
28139         * graphite.c: Include cfganal.h.
28140         (graphite_transform_loops): Connect infinite loops to exit
28141         and remove fake edges at the end.
28143 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28145         * ipa-inline.c (edge_badness): Revert accidental checkin.
28147 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28149         PR ipa/80763
28150         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
28151         symbols; not inline clones.
28153 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
28155         PR target/83507
28156         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
28157         hard registers.  Formatting fixes.
28159         PR preprocessor/83722
28160         * gcc.c (try_generate_repro): Pass
28161         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
28162         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
28163         do_report_bug.
28165 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
28166             Kito Cheng  <kito.cheng@gmail.com>
28168         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
28169         (riscv_leaf_function_p): Delete.
28170         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
28172 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28174         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
28175         function.
28176         (do_ifelse): New function.
28177         (do_isel): New function.
28178         (do_sub3): New function.
28179         (do_add3): New function.
28180         (do_load_mask_compare): New function.
28181         (do_overlap_load_compare): New function.
28182         (expand_compare_loop): New function.
28183         (expand_block_compare): Call expand_compare_loop() when appropriate.
28184         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
28185         option description.
28186         (-mblock-compare-inline-loop-limit): New option.
28188 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28190         PR target/83677
28191         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
28192         Reverse order of second and third operands in first alternative.
28193         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
28194         of first and second elements in UNSPEC_VPERMR vector.
28195         (altivec_expand_vec_perm_le): Likewise.
28197 2018-01-08  Jeff Law  <law@redhat.com>
28199         PR rtl-optimizatin/81308
28200         * tree-switch-conversion.c (cfg_altered): New file scoped static.
28201         (process_switch): If group_case_labels makes a change, then set
28202         cfg_altered.
28203         (pass_convert_switch::execute): If a switch is converted, then
28204         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
28206         PR rtl-optimization/81308
28207         * recog.c (split_all_insns): Conditionally cleanup the CFG after
28208         splitting insns.
28210 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
28212         PR target/83663 - Revert r255946
28213         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
28214         generation for cases where splatting a value is not useful.
28215         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
28216         across a vec_duplicate and a paradoxical subreg forming a vector
28217         mode to a vec_concat.
28219 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28221         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
28222         -march=armv8.3-a variants.
28223         * config/arm/t-multilib: Likewise.
28224         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
28226 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28228         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
28229         to generate rtl.
28230         (cceq_ior_compare_complement): Give it a name so I can use it, and
28231         change boolean_or_operator predicate to boolean_operator so it can
28232         be used to generate a crand.
28233         (eqne): New code iterator.
28234         (bd/bd_neg): New code_attrs.
28235         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
28236         a single define_insn.
28237         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
28238         decrement (bdnzt/bdnzf/bdzt/bdzf).
28239         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
28240         with the new names of the branch decrement patterns, and added the
28241         names of the branch decrement conditional patterns.
28243 2018-01-08  Richard Biener  <rguenther@suse.de>
28245         PR tree-optimization/83563
28246         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
28247         cache.
28249 2018-01-08  Richard Biener  <rguenther@suse.de>
28251         PR middle-end/83713
28252         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
28254 2018-01-08  Richard Biener  <rguenther@suse.de>
28256         PR tree-optimization/83685
28257         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
28258         references to abnormals.
28260 2018-01-08  Richard Biener  <rguenther@suse.de>
28262         PR lto/83719
28263         * dwarf2out.c (output_indirect_strings): Handle empty
28264         skeleton_debug_str_hash.
28265         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
28267 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28269         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
28270         (emit_store_direct): Likewise.
28271         (arc_trampoline_adjust_address): Likewise.
28272         (arc_asm_trampoline_template): New function.
28273         (arc_initialize_trampoline): Use asm_trampoline_template.
28274         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
28275         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
28276         * config/arc/arc.md (flush_icache): Delete pattern.
28278 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28280         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
28281         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
28282         munaligned-access.
28284 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28286         PR target/83681
28287         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
28288         by not USED_FOR_TARGET.
28289         (make_pass_resolve_sw_modes): Likewise.
28291 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28293         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
28294         USED_FOR_TARGET.
28296 2018-01-08  Richard Biener  <rguenther@suse.de>
28298         PR middle-end/83580
28299         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
28301 2018-01-08  Richard Biener  <rguenther@suse.de>
28303         PR middle-end/83517
28304         * match.pd ((t * 2) / 2) -> t): Add missing :c.
28306 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
28308         PR middle-end/81897
28309         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
28310         basic blocks with a small number of successors.
28311         (convert_control_dep_chain_into_preds): Improve handling of
28312         forwarder blocks.
28313         (dump_predicates): Split apart into...
28314         (dump_pred_chain): ...here...
28315         (dump_pred_info): ...and here.
28316         (can_one_predicate_be_invalidated_p): Add debugging printfs.
28317         (can_chain_union_be_invalidated_p): Improve check for invalidation
28318         of paths.
28319         (uninit_uses_cannot_happen): Avoid unnecessary if
28320         convert_control_dep_chain_into_preds yielded nothing.
28322 2018-01-06  Martin Sebor  <msebor@redhat.com>
28324         PR tree-optimization/83640
28325         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
28326         subtracting negative offset from size.
28327         (builtin_access::overlap): Adjust offset bounds of the access to fall
28328         within the size of the object if possible.
28330 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
28332         PR rtl-optimization/83699
28333         * expmed.c (extract_bit_field_1): Restrict the vector usage of
28334         extract_bit_field_as_subreg to cases in which the extracted
28335         value is also a vector.
28337         * lra-constraints.c (process_alt_operands): Test for the equivalence
28338         substitutions when detecting a possible reload cycle.
28340 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
28342         PR debug/83480
28343         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
28344         by default if flag_selective_schedling{,2}.  Formatting fixes.
28346         PR rtl-optimization/83682
28347         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
28348         if it has non-VECTOR_MODE element mode.
28349         (vec_duplicate_p): Likewise.
28351         PR middle-end/83694
28352         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
28353         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
28355 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28357         PR target/83604
28358         * config/i386/i386-builtin.def
28359         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
28360         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
28361         Require also OPTION_MASK_ISA_AVX512F in addition to
28362         OPTION_MASK_ISA_GFNI.
28363         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
28364         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
28365         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
28366         to OPTION_MASK_ISA_GFNI.
28367         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
28368         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
28369         OPTION_MASK_ISA_AVX512BW.
28370         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
28371         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
28372         addition to OPTION_MASK_ISA_GFNI.
28373         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
28374         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
28375         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
28376         to OPTION_MASK_ISA_GFNI.
28377         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
28378         a requirement for all ISAs rather than any of them with a few
28379         exceptions.
28380         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
28381         processing.
28382         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
28383         bitmasks to be enabled with 3 exceptions, instead of requiring any
28384         enabled ISA with lots of exceptions.
28385         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
28386         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
28387         Change avx512bw in isa attribute to avx512f.
28388         * config/i386/sgxintrin.h: Add license boilerplate.
28389         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
28390         to __AVX512F__ and __AVX512VL to __AVX512VL__.
28391         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
28392         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
28393         defined.
28394         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
28395         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28396         temporarily sse2 rather than sse if not enabled already.
28398         PR target/83604
28399         * config/i386/sse.md (VI248_VLBW): Rename to ...
28400         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
28401         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28402         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28403         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28404         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28405         mode iterator instead of VI248_VLBW.
28407 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
28409         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28410         (record_modified): Skip clobbers; add debug output.
28411         (param_change_prob): Use sreal frequencies.
28413 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28415         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28416         punt for user-aligned variables.
28418 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28420         * tree-chrec.c (chrec_contains_symbols): Return true for
28421         POLY_INT_CST.
28423 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
28425         PR target/82439
28426         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28427         of (x|y) == x for BICS pattern.
28429 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28431         PR tree-optimization/83605
28432         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28433         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28434         can throw.
28436 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28438         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28439         * config/epiphany/rtems.h: New file.
28441 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28442             Uros Bizjak  <ubizjak@gmail.com>
28444         PR target/83554
28445         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28446         QIreg_operand instead of register_operand predicate.
28447         * config/i386/i386.c (ix86_rop_should_change_byte_p,
28448         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28449         comments instead of -fmitigate[-_]rop.
28451 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28453         PR bootstrap/81926
28454         * cgraphunit.c (symbol_table::compile): Switch to text_section
28455         before calling assembly_start debug hook.
28456         * run-rtl-passes.c (run_rtl_passes): Likewise.
28457         Include output.h.
28459 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28461         * tree-vrp.c (extract_range_from_binary_expr_1): Check
28462         range_int_cst_p rather than !symbolic_range_p before calling
28463         extract_range_from_multiplicative_op_1.
28465 2018-01-04  Jeff Law  <law@redhat.com>
28467         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28468         redundant test in assertion.
28470 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28472         * doc/rtl.texi: Document machine_mode wrapper classes.
28474 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28476         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28477         using tree_to_uhwi.
28479 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28481         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28482         the VEC_PERM_EXPR fold to fail.
28484 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28486         PR debug/83585
28487         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28488         to switched_sections.
28490 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28492         PR target/83680
28493         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28494         test for d.testing.
28496 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
28498         PR target/83387
28499         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28500         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28502 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28504         PR debug/83666
28505         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28506         is BLKmode and bitpos not zero or mode change is needed.
28508 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28510         PR target/83675
28511         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28512         TARGET_VIS2.
28514 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
28516         PR target/83628
28517         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28518         instead of MULT rtx.  Update all corresponding splitters.
28519         (*saddl_se): Ditto.
28520         (*ssub<modesuffix>): Ditto.
28521         (*ssubl_se): Ditto.
28522         (*cmp_sadd_di): Update split patterns.
28523         (*cmp_sadd_si): Ditto.
28524         (*cmp_sadd_sidi): Ditto.
28525         (*cmp_ssub_di): Ditto.
28526         (*cmp_ssub_si): Ditto.
28527         (*cmp_ssub_sidi): Ditto.
28528         * config/alpha/predicates.md (const23_operand): New predicate.
28529         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28530         Look for ASHIFT, not MULT inner operand.
28531         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28533 2018-01-04  Martin Liska  <mliska@suse.cz>
28535         PR gcov-profile/83669
28536         * gcov.c (output_intermediate_file): Add version to intermediate
28537         gcov file.
28538         * doc/gcov.texi: Document new field 'version' in intermediate
28539         file format. Fix location of '-k' option of gcov command.
28541 2018-01-04  Martin Liska  <mliska@suse.cz>
28543         PR ipa/82352
28544         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28546 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28548         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28550 2018-01-03  Martin Sebor  <msebor@redhat.com>
28552         PR tree-optimization/83655
28553         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28554         checking calls with invalid arguments.
28556 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28558         * tree-vect-stmts.c (vect_get_store_rhs): New function.
28559         (vectorizable_mask_load_store): Delete.
28560         (vectorizable_call): Return false for masked loads and stores.
28561         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
28562         instead of gimple_assign_rhs1.
28563         (vectorizable_load): Handle IFN_MASK_LOAD.
28564         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28566 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28568         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28569         split out from..,
28570         (vectorizable_mask_load_store): ...here.
28571         (vectorizable_load): ...and here.
28573 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28575         * tree-vect-stmts.c (vect_build_all_ones_mask)
28576         (vect_build_zero_merge_argument): New functions, split out from...
28577         (vectorizable_load): ...here.
28579 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28581         * tree-vect-stmts.c (vect_check_store_rhs): New function,
28582         split out from...
28583         (vectorizable_mask_load_store): ...here.
28584         (vectorizable_store): ...and here.
28586 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28588         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28589         split out from...
28590         (vectorizable_mask_load_store): ...here.
28592 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28594         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28595         (vect_model_store_cost): Take a vec_load_store_type instead of a
28596         vect_def_type.
28597         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28598         (vect_model_store_cost): Take a vec_load_store_type instead of a
28599         vect_def_type.
28600         (vectorizable_mask_load_store): Update accordingly.
28601         (vectorizable_store): Likewise.
28602         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28604 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28606         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28607         IFN_MASK_LOAD calls here rather than...
28608         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28610 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28611             Alan Hayward  <alan.hayward@arm.com>
28612             David Sherwood  <david.sherwood@arm.com>
28614         * expmed.c (extract_bit_field_1): For vector extracts,
28615         fall back to extract_bit_field_as_subreg if vec_extract
28616         isn't available.
28618 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28619             Alan Hayward  <alan.hayward@arm.com>
28620             David Sherwood  <david.sherwood@arm.com>
28622         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28623         they are variable or constant sized.
28624         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28625         slots for constant-sized data.
28627 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28628             Alan Hayward  <alan.hayward@arm.com>
28629             David Sherwood  <david.sherwood@arm.com>
28631         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28632         handling COND_EXPRs with boolean comparisons, try to find a better
28633         basis for the mask type than the boolean itself.
28635 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28637         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28638         is calculated and how it can be overridden.
28639         * genmodes.c (max_bitsize_mode_any_mode): New variable.
28640         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28641         if defined.
28642         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28643         if nonzero.
28645 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28646             Alan Hayward  <alan.hayward@arm.com>
28647             David Sherwood  <david.sherwood@arm.com>
28649         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28650         Remove the mode argument.
28651         (aarch64_simd_valid_immediate): Remove the mode and inverse
28652         arguments.
28653         * config/aarch64/iterators.md (bitsize): New iterator.
28654         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28655         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28656         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28657         aarch64_simd_valid_immediate.
28658         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28659         (aarch64_reg_or_bic_imm): Likewise.
28660         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28661         with an insn_type enum and msl with a modifier_type enum.
28662         Replace element_width with a scalar_mode.  Change the shift
28663         to unsigned int.  Add constructors for scalar_float_mode and
28664         scalar_int_mode elements.
28665         (aarch64_vect_float_const_representable_p): Delete.
28666         (aarch64_can_const_movi_rtx_p)
28667         (aarch64_simd_scalar_immediate_valid_for_move)
28668         (aarch64_simd_make_constant): Update call to
28669         aarch64_simd_valid_immediate.
28670         (aarch64_advsimd_valid_immediate_hs): New function.
28671         (aarch64_advsimd_valid_immediate): Likewise.
28672         (aarch64_simd_valid_immediate): Remove mode and inverse
28673         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
28674         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28675         and aarch64_float_const_representable_p on the result.
28676         (aarch64_output_simd_mov_immediate): Remove mode argument.
28677         Update call to aarch64_simd_valid_immediate and use of
28678         simd_immediate_info.
28679         (aarch64_output_scalar_simd_mov_immediate): Update call
28680         accordingly.
28682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28683             Alan Hayward  <alan.hayward@arm.com>
28684             David Sherwood  <david.sherwood@arm.com>
28686         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28687         (mode_nunits): Likewise CONST_MODE_NUNITS.
28688         * machmode.def (ADJUST_NUNITS): Document.
28689         * genmodes.c (mode_data::need_nunits_adj): New field.
28690         (blank_mode): Update accordingly.
28691         (adj_nunits): New variable.
28692         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28693         parameter.
28694         (emit_mode_size_inline): Set need_bytesize_adj for all modes
28695         listed in adj_nunits.
28696         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28697         listed in adj_nunits.  Don't emit case statements for such modes.
28698         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28699         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
28700         nothing if adj_nunits is nonnull.
28701         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28702         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28703         (emit_mode_fbit): Update use of print_maybe_const_decl.
28704         (emit_move_size): Likewise.  Treat the array as non-const
28705         if adj_nunits.
28706         (emit_mode_adjustments): Handle adj_nunits.
28708 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28710         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28711         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28712         (VECTOR_MODES): Use it.
28713         (make_vector_modes): Take the prefix as an argument.
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         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28720         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28721         for MODE_VECTOR_BOOL.
28722         * machmode.def (VECTOR_BOOL_MODE): Document.
28723         * genmodes.c (VECTOR_BOOL_MODE): New macro.
28724         (make_vector_bool_mode): New function.
28725         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28726         MODE_VECTOR_BOOL.
28727         * lto-streamer-in.c (lto_input_mode_table): Likewise.
28728         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28729         Likewise.
28730         * stor-layout.c (int_mode_for_mode): Likewise.
28731         * tree.c (build_vector_type_for_mode): Likewise.
28732         * varasm.c (output_constant_pool_2): Likewise.
28733         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28734         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
28735         for MODE_VECTOR_BOOL.
28736         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28737         of mode class checks.
28738         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28739         instead of a list of mode class checks.
28740         (expand_vector_scalar_condition): Likewise.
28741         (type_for_widest_vector_mode): Handle BImode as an inner mode.
28743 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28744             Alan Hayward  <alan.hayward@arm.com>
28745             David Sherwood  <david.sherwood@arm.com>
28747         * machmode.h (mode_size): Change from unsigned short to
28748         poly_uint16_pod.
28749         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28750         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28751         or if measurement_type is not polynomial.
28752         (fixed_size_mode::includes_p): Check for constant-sized modes.
28753         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28754         return a poly_uint16 rather than an unsigned short.
28755         (emit_mode_size): Change the type of mode_size from unsigned short
28756         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
28757         (emit_mode_adjustments): Cope with polynomial vector sizes.
28758         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28759         for GET_MODE_SIZE.
28760         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28761         for GET_MODE_SIZE.
28762         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28763         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28764         * caller-save.c (setup_save_areas): Likewise.
28765         (replace_reg_with_saved_mem): Likewise.
28766         * calls.c (emit_library_call_value_1): Likewise.
28767         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28768         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28769         (gen_lowpart_for_combine): Likewise.
28770         * convert.c (convert_to_integer_1): Likewise.
28771         * cse.c (equiv_constant, cse_insn): Likewise.
28772         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28773         (cselib_subst_to_values): Likewise.
28774         * dce.c (word_dce_process_block): Likewise.
28775         * df-problems.c (df_word_lr_mark_ref): Likewise.
28776         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28777         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28778         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
28779         (rtl_for_decl_location): Likewise.
28780         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
28781         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
28782         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
28783         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
28784         (expand_expr_real_1): Likewise.
28785         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
28786         (pad_below): Likewise.
28787         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28788         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
28789         * ira.c (get_subreg_tracking_sizes): Likewise.
28790         * ira-build.c (ira_create_allocno_objects): Likewise.
28791         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
28792         (ira_sort_regnos_for_alter_reg): Likewise.
28793         * ira-costs.c (record_operand_costs): Likewise.
28794         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
28795         (resolve_simple_move): Likewise.
28796         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
28797         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
28798         (lra_constraints): Likewise.
28799         (CONST_POOL_OK_P): Reject variable-sized modes.
28800         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
28801         (add_pseudo_to_slot, lra_spill): Likewise.
28802         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28803         * optabs-query.c (get_best_extraction_insn): Likewise.
28804         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28805         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
28806         (expand_mult_highpart, valid_multiword_target_p): Likewise.
28807         * recog.c (offsettable_address_addr_space_p): Likewise.
28808         * regcprop.c (maybe_mode_change): Likewise.
28809         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
28810         * regrename.c (build_def_use): Likewise.
28811         * regstat.c (dump_reg_info): Likewise.
28812         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
28813         (find_reloads, find_reloads_subreg_address): Likewise.
28814         * reload1.c (eliminate_regs_1): Likewise.
28815         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
28816         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
28817         (simplify_binary_operation_1, simplify_subreg): Likewise.
28818         * targhooks.c (default_function_arg_padding): Likewise.
28819         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
28820         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
28821         (verify_gimple_assign_ternary): Likewise.
28822         * tree-inline.c (estimate_move_cost): Likewise.
28823         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28824         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
28825         (get_address_cost_ainc): Likewise.
28826         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
28827         (vect_supportable_dr_alignment): Likewise.
28828         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28829         (vectorizable_reduction): Likewise.
28830         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
28831         (vectorizable_operation, vectorizable_load): Likewise.
28832         * tree.c (build_same_sized_truth_vector_type): Likewise.
28833         * valtrack.c (cleanup_auto_inc_dec): Likewise.
28834         * var-tracking.c (emit_note_insn_var_location): Likewise.
28835         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
28836         (ADDR_VEC_ALIGN): Likewise.
28838 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28839             Alan Hayward  <alan.hayward@arm.com>
28840             David Sherwood  <david.sherwood@arm.com>
28842         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
28843         unsigned short.
28844         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28845         or if measurement_type is polynomial.
28846         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
28847         * combine.c (make_extraction): Likewise.
28848         * dse.c (find_shift_sequence): Likewise.
28849         * dwarf2out.c (mem_loc_descriptor): Likewise.
28850         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
28851         (extract_bit_field, extract_low_bits): Likewise.
28852         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
28853         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
28854         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
28855         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
28856         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28857         * reload.c (find_reloads): Likewise.
28858         * reload1.c (alter_reg): Likewise.
28859         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
28860         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
28861         * tree-if-conv.c (predicate_mem_writes): Likewise.
28862         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
28863         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
28864         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
28865         * valtrack.c (dead_debug_insert_temp): Likewise.
28866         * varasm.c (mergeable_constant_section): Likewise.
28867         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
28869 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28870             Alan Hayward  <alan.hayward@arm.com>
28871             David Sherwood  <david.sherwood@arm.com>
28873         * expr.c (expand_assignment): Cope with polynomial mode sizes
28874         when assigning to a CONCAT.
28876 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28877             Alan Hayward  <alan.hayward@arm.com>
28878             David Sherwood  <david.sherwood@arm.com>
28880         * machmode.h (mode_precision): Change from unsigned short to
28881         poly_uint16_pod.
28882         (mode_to_precision): Return a poly_uint16 rather than an unsigned
28883         short.
28884         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
28885         or if measurement_type is not polynomial.
28886         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
28887         in which the mode is already known to be a scalar_int_mode.
28888         * genmodes.c (emit_mode_precision): Change the type of mode_precision
28889         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
28890         initializer.
28891         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28892         for GET_MODE_PRECISION.
28893         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28894         for GET_MODE_PRECISION.
28895         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
28896         as polynomial.
28897         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
28898         (expand_field_assignment, make_extraction): Likewise.
28899         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
28900         (get_last_value): Likewise.
28901         * convert.c (convert_to_integer_1): Likewise.
28902         * cse.c (cse_insn): Likewise.
28903         * expr.c (expand_expr_real_1): Likewise.
28904         * lra-constraints.c (simplify_operand_subreg): Likewise.
28905         * optabs-query.c (can_atomic_load_p): Likewise.
28906         * optabs.c (expand_atomic_load): Likewise.
28907         (expand_atomic_store): Likewise.
28908         * ree.c (combine_reaching_defs): Likewise.
28909         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
28910         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
28911         * tree.h (type_has_mode_precision_p): Likewise.
28912         * ubsan.c (instrument_si_overflow): Likewise.
28914 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28915             Alan Hayward  <alan.hayward@arm.com>
28916             David Sherwood  <david.sherwood@arm.com>
28918         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
28919         polynomial numbers of units.
28920         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
28921         (valid_vector_subparts_p): New function.
28922         (build_vector_type): Remove temporary shim and take the number
28923         of units as a poly_uint64 rather than an int.
28924         (build_opaque_vector_type): Take the number of units as a
28925         poly_uint64 rather than an int.
28926         * tree.c (build_vector_from_ctor): Handle polynomial
28927         TYPE_VECTOR_SUBPARTS.
28928         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
28929         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
28930         (build_vector_from_val): If the number of units is variable,
28931         use build_vec_duplicate_cst for constant operands and
28932         VEC_DUPLICATE_EXPR otherwise.
28933         (make_vector_type): Remove temporary is_constant ().
28934         (build_vector_type, build_opaque_vector_type): Take the number of
28935         units as a poly_uint64 rather than an int.
28936         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
28937         VECTOR_CST_NELTS.
28938         * cfgexpand.c (expand_debug_expr): Likewise.
28939         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
28940         (store_constructor, expand_expr_real_1): Likewise.
28941         (const_scalar_mask_from_tree): Likewise.
28942         * fold-const-call.c (fold_const_reduction): Likewise.
28943         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
28944         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
28945         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
28946         (fold_relational_const): Likewise.
28947         (native_interpret_vector): Likewise.  Change the size from an
28948         int to an unsigned int.
28949         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
28950         TYPE_VECTOR_SUBPARTS.
28951         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
28952         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
28953         duplicating a non-constant operand into a variable-length vector.
28954         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
28955         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
28956         * ipa-icf.c (sem_variable::equals): Likewise.
28957         * match.pd: Likewise.
28958         * omp-simd-clone.c (simd_clone_subparts): Likewise.
28959         * print-tree.c (print_node): Likewise.
28960         * stor-layout.c (layout_type): Likewise.
28961         * targhooks.c (default_builtin_vectorization_cost): Likewise.
28962         * tree-cfg.c (verify_gimple_comparison): Likewise.
28963         (verify_gimple_assign_binary): Likewise.
28964         (verify_gimple_assign_ternary): Likewise.
28965         (verify_gimple_assign_single): Likewise.
28966         * tree-pretty-print.c (dump_generic_node): Likewise.
28967         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28968         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
28969         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
28970         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
28971         (vect_shift_permute_load_chain): Likewise.
28972         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
28973         (expand_vector_condition, optimize_vector_constructor): Likewise.
28974         (lower_vec_perm, get_compute_type): Likewise.
28975         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28976         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
28977         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
28978         (vect_recog_mask_conversion_pattern): Likewise.
28979         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
28980         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
28981         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28982         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
28983         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
28984         (vectorizable_shift, vectorizable_operation, vectorizable_store)
28985         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
28986         (supportable_widening_operation): Likewise.
28987         (supportable_narrowing_operation): Likewise.
28988         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
28989         Likewise.
28990         * varasm.c (output_constant): Likewise.
28992 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28993             Alan Hayward  <alan.hayward@arm.com>
28994             David Sherwood  <david.sherwood@arm.com>
28996         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
28997         so that both the length == 3 and length != 3 cases set up their
28998         own permute vectors.  Add comments explaining why we know the
28999         number of elements is constant.
29000         (vect_permute_load_chain): Likewise.
29002 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29003             Alan Hayward  <alan.hayward@arm.com>
29004             David Sherwood  <david.sherwood@arm.com>
29006         * machmode.h (mode_nunits): Change from unsigned char to
29007         poly_uint16_pod.
29008         (ONLY_FIXED_SIZE_MODES): New macro.
29009         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
29010         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
29011         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
29012         New typedefs.
29013         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
29014         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
29015         or if measurement_type is not polynomial.
29016         * genmodes.c (ZERO_COEFFS): New macro.
29017         (emit_mode_nunits_inline): Make mode_nunits_inline return a
29018         poly_uint16.
29019         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
29020         Use ZERO_COEFFS when emitting initializers.
29021         * data-streamer.h (bp_pack_poly_value): New function.
29022         (bp_unpack_poly_value): Likewise.
29023         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29024         for GET_MODE_NUNITS.
29025         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29026         for GET_MODE_NUNITS.
29027         * tree.c (make_vector_type): Remove temporary shim and make
29028         the real function take the number of units as a poly_uint64
29029         rather than an int.
29030         (build_vector_type_for_mode): Handle polynomial nunits.
29031         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
29032         * emit-rtl.c (const_vec_series_p_1): Likewise.
29033         (gen_rtx_CONST_VECTOR): Likewise.
29034         * fold-const.c (test_vec_duplicate_folding): Likewise.
29035         * genrecog.c (validate_pattern): Likewise.
29036         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
29037         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29038         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
29039         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
29040         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
29041         * rtlanal.c (subreg_get_info): Likewise.
29042         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29043         (vect_grouped_load_supported): Likewise.
29044         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
29045         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
29046         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
29047         (simplify_const_unary_operation, simplify_binary_operation_1)
29048         (simplify_const_binary_operation, simplify_ternary_operation)
29049         (test_vector_ops_duplicate, test_vector_ops): Likewise.
29050         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
29051         instead of CONST_VECTOR_NUNITS.
29052         * varasm.c (output_constant_pool_2): Likewise.
29053         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
29054         explicit-encoded elements in the XVEC for variable-length vectors.
29056 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29058         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
29060 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29061             Alan Hayward  <alan.hayward@arm.com>
29062             David Sherwood  <david.sherwood@arm.com>
29064         * coretypes.h (fixed_size_mode): Declare.
29065         (fixed_size_mode_pod): New typedef.
29066         * builtins.h (target_builtins::x_apply_args_mode)
29067         (target_builtins::x_apply_result_mode): Change type to
29068         fixed_size_mode_pod.
29069         * builtins.c (apply_args_size, apply_result_size, result_vector)
29070         (expand_builtin_apply_args_1, expand_builtin_apply)
29071         (expand_builtin_return): Update accordingly.
29073 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29075         * cse.c (hash_rtx_cb): Hash only the encoded elements.
29076         * cselib.c (cselib_hash_rtx): Likewise.
29077         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
29078         CONST_VECTOR encoding.
29080 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29081             Jeff Law  <law@redhat.com>
29083         PR target/83641
29084         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
29085         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
29086         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
29087         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
29089         PR target/83641
29090         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
29091         explicitly probe *sp in a noreturn function if there were any callee
29092         register saves or frame pointer is needed.
29094 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29096         PR debug/83621
29097         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
29098         BLKmode for ternary, binary or unary expressions.
29100         PR debug/83645
29101         * var-tracking.c (delete_vta_debug_insn): New inline function.
29102         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
29103         insns from get_insns () to NULL instead of each bb separately.
29104         Use delete_vta_debug_insn.  No longer static.
29105         (vt_debug_insns_local, variable_tracking_main_1): Adjust
29106         delete_vta_debug_insns callers.
29107         * rtl.h (delete_vta_debug_insns): Declare.
29108         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
29109         instead of variable_tracking_main.
29111 2018-01-03  Martin Sebor  <msebor@redhat.com>
29113         PR tree-optimization/83603
29114         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
29115         arguments past the endof the argument list in functions declared
29116         without a prototype.
29117         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
29118         Avoid checking when arguments are null.
29120 2018-01-03  Martin Sebor  <msebor@redhat.com>
29122         PR c/83559
29123         * doc/extend.texi (attribute const): Fix a typo.
29124         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
29125         issuing -Wsuggest-attribute for void functions.
29127 2018-01-03  Martin Sebor  <msebor@redhat.com>
29129         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
29130         offset_int::from instead of wide_int::to_shwi.
29131         (maybe_diag_overlap): Remove assertion.
29132         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
29133         * gimple-ssa-sprintf.c (format_directive): Same.
29134         (parse_directive): Same.
29135         (sprintf_dom_walker::compute_format_length): Same.
29136         (try_substitute_return_value): Same.
29138 2018-01-03  Jeff Law  <law@redhat.com>
29140         PR middle-end/83654
29141         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
29142         non-constant residual for zero at runtime and avoid probing in
29143         that case.  Reorganize code for trailing problem to mirror handling
29144         of the residual.
29146 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29148         PR tree-optimization/83501
29149         * tree-ssa-strlen.c (get_string_cst): New.
29150         (handle_char_store): Call get_string_cst.
29152 2018-01-03  Martin Liska  <mliska@suse.cz>
29154         PR tree-optimization/83593
29155         * tree-ssa-strlen.c: Include tree-cfg.h.
29156         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
29157         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
29158         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
29159         to false.
29160         (strlen_dom_walker::before_dom_children): Call
29161         gimple_purge_dead_eh_edges. Dump tranformation with details
29162         dump flags.
29163         (strlen_dom_walker::before_dom_children): Update call by adding
29164         new argument cleanup_eh.
29165         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
29167 2018-01-03  Martin Liska  <mliska@suse.cz>
29169         PR ipa/83549
29170         * cif-code.def (VARIADIC_THUNK): New enum value.
29171         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
29172         thunks.
29174 2018-01-03  Jan Beulich  <jbeulich@suse.com>
29176         * sse.md (mov<mode>_internal): Tighten condition for when to use
29177         vmovdqu<ssescalarsize> for TI and OI modes.
29179 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29181         Update copyright years.
29183 2018-01-03  Martin Liska  <mliska@suse.cz>
29185         PR ipa/83594
29186         * ipa-visibility.c (function_and_variable_visibility): Skip
29187         functions with noipa attribure.
29189 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29191         * gcc.c (process_command): Update copyright notice dates.
29192         * gcov-dump.c (print_version): Ditto.
29193         * gcov.c (print_version): Ditto.
29194         * gcov-tool.c (print_version): Ditto.
29195         * gengtype.c (create_file): Ditto.
29196         * doc/cpp.texi: Bump @copying's copyright year.
29197         * doc/cppinternals.texi: Ditto.
29198         * doc/gcc.texi: Ditto.
29199         * doc/gccint.texi: Ditto.
29200         * doc/gcov.texi: Ditto.
29201         * doc/install.texi: Ditto.
29202         * doc/invoke.texi: Ditto.
29204 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29206         * vector-builder.h (vector_builder::m_full_nelts): Change from
29207         unsigned int to poly_uint64.
29208         (vector_builder::full_nelts): Update prototype accordingly.
29209         (vector_builder::new_vector): Likewise.
29210         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
29211         (vector_builder::operator ==): Likewise.
29212         (vector_builder::finalize): Likewise.
29213         * int-vector-builder.h (int_vector_builder::int_vector_builder):
29214         Take the number of elements as a poly_uint64 rather than an
29215         unsigned int.
29216         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
29217         from unsigned int to poly_uint64.
29218         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
29219         (vec_perm_indices::new_vector): Likewise.
29220         (vec_perm_indices::length): Likewise.
29221         (vec_perm_indices::nelts_per_input): Likewise.
29222         (vec_perm_indices::input_nelts): Likewise.
29223         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
29224         number of elements per input as a poly_uint64 rather than an
29225         unsigned int.  Use the original encoding for variable-length
29226         vectors, rather than clamping each individual element.
29227         For the second and subsequent elements in each pattern,
29228         clamp the step and base before clamping their sum.
29229         (vec_perm_indices::series_p): Handle polynomial element counts.
29230         (vec_perm_indices::all_in_range_p): Likewise.
29231         (vec_perm_indices_to_tree): Likewise.
29232         (vec_perm_indices_to_rtx): Likewise.
29233         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
29234         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
29235         (tree_vector_builder::new_binary_operation): Handle polynomial
29236         element counts.  Return false if we need to know the number
29237         of elements at compile time.
29238         * fold-const.c (fold_vec_perm): Punt if the number of elements
29239         isn't known at compile time.
29241 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29243         * vec-perm-indices.h (vec_perm_builder): Change element type
29244         from HOST_WIDE_INT to poly_int64.
29245         (vec_perm_indices::element_type): Update accordingly.
29246         (vec_perm_indices::clamp): Handle polynomial element_types.
29247         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29248         (vec_perm_indices::all_in_range_p): Likewise.
29249         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
29250         than shwi trees.
29251         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
29252         polynomial vec_perm_indices element types.
29253         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
29254         * fold-const.c (fold_vec_perm): Likewise.
29255         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
29256         * tree-vect-generic.c (lower_vec_perm): Likewise.
29257         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29258         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
29259         element type to HOST_WIDE_INT.
29261 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29262             Alan Hayward  <alan.hayward@arm.com>
29263             David Sherwood  <david.sherwood@arm.com>
29265         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
29266         rather than an int.  Use plus_constant.
29267         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
29268         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
29270 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29271             Alan Hayward  <alan.hayward@arm.com>
29272             David Sherwood  <david.sherwood@arm.com>
29274         * calls.c (emit_call_1, expand_call): Change struct_value_size from
29275         a HOST_WIDE_INT to a poly_int64.
29277 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29278             Alan Hayward  <alan.hayward@arm.com>
29279             David Sherwood  <david.sherwood@arm.com>
29281         * calls.c (load_register_parameters): Cope with polynomial
29282         mode sizes.  Require a constant size for BLKmode parameters
29283         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
29284         forces a parameter to be padded at the lsb end in order to
29285         fill a complete number of words, require the parameter size
29286         to be ordered wrt UNITS_PER_WORD.
29288 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29289             Alan Hayward  <alan.hayward@arm.com>
29290             David Sherwood  <david.sherwood@arm.com>
29292         * reload1.c (spill_stack_slot_width): Change element type
29293         from unsigned int to poly_uint64_pod.
29294         (alter_reg): Treat mode sizes as polynomial.
29296 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29297             Alan Hayward  <alan.hayward@arm.com>
29298             David Sherwood  <david.sherwood@arm.com>
29300         * reload.c (complex_word_subreg_p): New function.
29301         (reload_inner_reg_of_subreg, push_reload): Use it.
29303 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29304             Alan Hayward  <alan.hayward@arm.com>
29305             David Sherwood  <david.sherwood@arm.com>
29307         * lra-constraints.c (process_alt_operands): Reject matched
29308         operands whose sizes aren't ordered.
29309         (match_reload): Refer to this check here.
29311 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29312             Alan Hayward  <alan.hayward@arm.com>
29313             David Sherwood  <david.sherwood@arm.com>
29315         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
29316         that the mode size is in the set {1, 2, 4, 8, 16}.
29318 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29319             Alan Hayward  <alan.hayward@arm.com>
29320             David Sherwood  <david.sherwood@arm.com>
29322         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
29323         Use plus_constant instead of gen_rtx_PLUS.
29325 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29326             Alan Hayward  <alan.hayward@arm.com>
29327             David Sherwood  <david.sherwood@arm.com>
29329         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
29330         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
29331         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
29332         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
29333         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
29334         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
29335         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
29336         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
29337         * config/i386/i386.c (ix86_push_rounding): ...this new function.
29338         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29339         a poly_int64.
29340         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
29341         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
29342         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
29343         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
29344         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
29345         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
29346         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
29347         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
29348         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
29349         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
29350         function.
29351         * expr.c (emit_move_resolve_push): Treat the input and result
29352         of PUSH_ROUNDING as a poly_int64.
29353         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
29354         (emit_push_insn): Likewise.
29355         * lra-eliminations.c (mark_not_eliminable): Likewise.
29356         * recog.c (push_operand): Likewise.
29357         * reload1.c (elimination_effects): Likewise.
29358         * rtlanal.c (nonzero_bits1): Likewise.
29359         * calls.c (store_one_arg): Likewise.  Require the padding to be
29360         known at compile time.
29362 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29363             Alan Hayward  <alan.hayward@arm.com>
29364             David Sherwood  <david.sherwood@arm.com>
29366         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
29367         Use plus_constant instead of gen_rtx_PLUS.
29369 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29370             Alan Hayward  <alan.hayward@arm.com>
29371             David Sherwood  <david.sherwood@arm.com>
29373         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
29374         rather than an int.
29376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29377             Alan Hayward  <alan.hayward@arm.com>
29378             David Sherwood  <david.sherwood@arm.com>
29380         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
29381         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
29382         via stack temporaries.  Treat the mode size as polynomial too.
29384 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29385             Alan Hayward  <alan.hayward@arm.com>
29386             David Sherwood  <david.sherwood@arm.com>
29388         * expr.c (expand_expr_real_2): When handling conversions involving
29389         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
29390         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
29391         as a poly_uint64 too.
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         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
29399 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29400             Alan Hayward  <alan.hayward@arm.com>
29401             David Sherwood  <david.sherwood@arm.com>
29403         * combine.c (can_change_dest_mode): Handle polynomial
29404         REGMODE_NATURAL_SIZE.
29405         * expmed.c (store_bit_field_1): Likewise.
29406         * expr.c (store_constructor): Likewise.
29407         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29408         and polynomial REGMODE_NATURAL_SIZE.
29409         (gen_lowpart_common): Likewise.
29410         * reginfo.c (record_subregs_of_mode): Likewise.
29411         * rtlanal.c (read_modify_subreg_p): Likewise.
29413 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29414             Alan Hayward  <alan.hayward@arm.com>
29415             David Sherwood  <david.sherwood@arm.com>
29417         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29418         numbers of elements.
29420 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29421             Alan Hayward  <alan.hayward@arm.com>
29422             David Sherwood  <david.sherwood@arm.com>
29424         * match.pd: Cope with polynomial numbers of vector elements.
29426 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29427             Alan Hayward  <alan.hayward@arm.com>
29428             David Sherwood  <david.sherwood@arm.com>
29430         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29431         in a POINTER_PLUS_EXPR.
29433 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29434             Alan Hayward  <alan.hayward@arm.com>
29435             David Sherwood  <david.sherwood@arm.com>
29437         * omp-simd-clone.c (simd_clone_subparts): New function.
29438         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29439         (ipa_simd_modify_function_body): Likewise.
29441 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29442             Alan Hayward  <alan.hayward@arm.com>
29443             David Sherwood  <david.sherwood@arm.com>
29445         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29446         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29447         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29448         (expand_vector_condition, vector_element): Likewise.
29449         (subparts_gt): New function.
29450         (get_compute_type): Use subparts_gt.
29451         (count_type_subparts): Delete.
29452         (expand_vector_operations_1): Use subparts_gt instead of
29453         count_type_subparts.
29455 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29456             Alan Hayward  <alan.hayward@arm.com>
29457             David Sherwood  <david.sherwood@arm.com>
29459         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29460         (vect_compile_time_alias): ...this new function.  Do the calculation
29461         on poly_ints rather than trees.
29462         (vect_prune_runtime_alias_test_list): Update call accordingly.
29464 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29465             Alan Hayward  <alan.hayward@arm.com>
29466             David Sherwood  <david.sherwood@arm.com>
29468         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29469         numbers of units.
29470         (vect_schedule_slp_instance): Likewise.
29472 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29473             Alan Hayward  <alan.hayward@arm.com>
29474             David Sherwood  <david.sherwood@arm.com>
29476         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29477         constant and extern definitions for variable-length vectors.
29478         (vect_get_constant_vectors): Note that the number of units
29479         is known to be constant.
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         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29486         of units as polynomial.  Choose between WIDE and NARROW based
29487         on multiple_p.
29489 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29490             Alan Hayward  <alan.hayward@arm.com>
29491             David Sherwood  <david.sherwood@arm.com>
29493         * tree-vect-stmts.c (simd_clone_subparts): New function.
29494         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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         * tree-vect-stmts.c (vectorizable_call): Treat the number of
29501         vectors as polynomial.  Use build_index_vector for
29502         IFN_GOMP_SIMD_LANE.
29504 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29505             Alan Hayward  <alan.hayward@arm.com>
29506             David Sherwood  <david.sherwood@arm.com>
29508         * tree-vect-stmts.c (get_load_store_type): Treat the number of
29509         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29510         for variable-length vectors.
29511         (vectorizable_mask_load_store): Treat the number of units as
29512         polynomial, asserting that it is constant if the condition has
29513         already been enforced.
29514         (vectorizable_store, vectorizable_load): Likewise.
29516 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29517             Alan Hayward  <alan.hayward@arm.com>
29518             David Sherwood  <david.sherwood@arm.com>
29520         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29521         of units as polynomial.  Punt if we can't tell at compile time
29522         which vector contains the final result.
29524 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29525             Alan Hayward  <alan.hayward@arm.com>
29526             David Sherwood  <david.sherwood@arm.com>
29528         * tree-vect-loop.c (vectorizable_induction): Treat the number
29529         of units as polynomial.  Punt on SLP inductions.  Use an integer
29530         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
29531         cast of such a series for variable-length floating-point
29532         reductions.
29534 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29535             Alan Hayward  <alan.hayward@arm.com>
29536             David Sherwood  <david.sherwood@arm.com>
29538         * tree.h (build_index_vector): Declare.
29539         * tree.c (build_index_vector): New function.
29540         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29541         of units as polynomial, forcibly converting it to a constant if
29542         vectorizable_reduction has already enforced the condition.
29543         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
29544         to create a {1,2,3,...} vector.
29545         (vectorizable_reduction): Treat the number of units as polynomial.
29546         Choose vectype_in based on the largest scalar element size rather
29547         than the smallest number of units.  Enforce the restrictions
29548         relied on above.
29550 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29551             Alan Hayward  <alan.hayward@arm.com>
29552             David Sherwood  <david.sherwood@arm.com>
29554         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29555         number of units as polynomial.
29557 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29558             Alan Hayward  <alan.hayward@arm.com>
29559             David Sherwood  <david.sherwood@arm.com>
29561         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29562         * target.def (autovectorize_vector_sizes): Return the vector sizes
29563         by pointer, using vector_sizes rather than a bitmask.
29564         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29565         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29566         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29567         Likewise.
29568         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29569         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29570         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29571         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29572         * omp-general.c (omp_max_vf): Likewise.
29573         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29574         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29575         * tree-vect-loop.c (vect_analyze_loop): Likewise.
29576         * tree-vect-slp.c (vect_slp_bb): Likewise.
29577         * doc/tm.texi: Regenerate.
29578         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29579         to a poly_uint64.
29580         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29581         the vector size as a poly_uint64 rather than an unsigned int.
29582         (current_vector_size): Change from an unsigned int to a poly_uint64.
29583         (get_vectype_for_scalar_type): Update accordingly.
29584         * tree.h (build_truth_vector_type): Take the size and number of
29585         units as a poly_uint64 rather than an unsigned int.
29586         (build_vector_type): Add a temporary overload that takes
29587         the number of units as a poly_uint64 rather than an unsigned int.
29588         * tree.c (make_vector_type): Likewise.
29589         (build_truth_vector_type): Take the number of units as a poly_uint64
29590         rather than an unsigned int.
29592 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29593             Alan Hayward  <alan.hayward@arm.com>
29594             David Sherwood  <david.sherwood@arm.com>
29596         * target.def (get_mask_mode): Take the number of units and length
29597         as poly_uint64s rather than unsigned ints.
29598         * targhooks.h (default_get_mask_mode): Update accordingly.
29599         * targhooks.c (default_get_mask_mode): Likewise.
29600         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29601         * doc/tm.texi: Regenerate.
29603 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29604             Alan Hayward  <alan.hayward@arm.com>
29605             David Sherwood  <david.sherwood@arm.com>
29607         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29608         * omp-general.c (omp_max_vf): Likewise.
29609         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29610         (expand_omp_simd): Handle polynomial safelen.
29611         * omp-low.c (omplow_simd_context): Add a default constructor.
29612         (omplow_simd_context::max_vf): Change from int to poly_uint64.
29613         (lower_rec_simd_input_clauses): Update accordingly.
29614         (lower_rec_input_clauses): Likewise.
29616 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29617             Alan Hayward  <alan.hayward@arm.com>
29618             David Sherwood  <david.sherwood@arm.com>
29620         * tree-vectorizer.h (vect_nunits_for_cost): New function.
29621         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29622         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29623         (vect_analyze_slp_cost): Likewise.
29624         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29625         (vect_model_load_cost): Likewise.
29627 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29628             Alan Hayward  <alan.hayward@arm.com>
29629             David Sherwood  <david.sherwood@arm.com>
29631         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29632         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29633         from an unsigned int * to a poly_uint64_pod *.
29634         (calculate_unrolling_factor): New function.
29635         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
29637 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29638             Alan Hayward  <alan.hayward@arm.com>
29639             David Sherwood  <david.sherwood@arm.com>
29641         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29642         from an unsigned int to a poly_uint64.
29643         (_loop_vec_info::slp_unrolling_factor): Likewise.
29644         (_loop_vec_info::vectorization_factor): Change from an int
29645         to a poly_uint64.
29646         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29647         (vect_get_num_vectors): New function.
29648         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29649         (vect_get_num_copies): Use vect_get_num_vectors.
29650         (vect_analyze_data_ref_dependences): Change max_vf from an int *
29651         to an unsigned int *.
29652         (vect_analyze_data_refs): Change min_vf from an int * to a
29653         poly_uint64 *.
29654         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29655         than an unsigned HOST_WIDE_INT.
29656         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29657         (vect_analyze_data_ref_dependence): Change max_vf from an int *
29658         to an unsigned int *.
29659         (vect_analyze_data_ref_dependences): Likewise.
29660         (vect_compute_data_ref_alignment): Handle polynomial vf.
29661         (vect_enhance_data_refs_alignment): Likewise.
29662         (vect_prune_runtime_alias_test_list): Likewise.
29663         (vect_shift_permute_load_chain): Likewise.
29664         (vect_supportable_dr_alignment): Likewise.
29665         (dependence_distance_ge_vf): Take the vectorization factor as a
29666         poly_uint64 rather than an unsigned HOST_WIDE_INT.
29667         (vect_analyze_data_refs): Change min_vf from an int * to a
29668         poly_uint64 *.
29669         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29670         vfm1 as a poly_uint64 rather than an int.  Make the same change
29671         for the returned bound_scalar.
29672         (vect_gen_vector_loop_niters): Handle polynomial vf.
29673         (vect_do_peeling): Likewise.  Update call to
29674         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29675         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29676         be constant.
29677         * tree-vect-loop.c (vect_determine_vectorization_factor)
29678         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29679         (vect_get_known_peeling_cost): Likewise.
29680         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29681         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29682         (vect_transform_loop): Likewise.  Use the lowest possible VF when
29683         updating the upper bounds of the loop.
29684         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
29685         rather than an int.
29686         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29687         polynomial unroll factors.
29688         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29689         (vect_make_slp_decision): Likewise.
29690         (vect_supported_load_permutation_p): Likewise, and polynomial
29691         vf too.
29692         (vect_analyze_slp_cost): Handle polynomial vf.
29693         (vect_slp_analyze_node_operations): Likewise.
29694         (vect_slp_analyze_bb_1): Likewise.
29695         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29696         than an unsigned HOST_WIDE_INT.
29697         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29698         (vectorizable_load): Handle polynomial vf.
29699         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29700         a poly_uint64.
29701         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29703 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29704             Alan Hayward  <alan.hayward@arm.com>
29705             David Sherwood  <david.sherwood@arm.com>
29707         * match.pd: Handle bit operations involving three constants
29708         and try to fold one pair.
29710 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29712         * tree-vect-loop-manip.c: Include gimple-fold.h.
29713         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29714         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
29715         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29716         Add a path that uses a step of VF instead of 1, but disable it
29717         for now.
29718         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29719         and niters_no_overflow parameters.  Update calls to
29720         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29721         Create a new SSA name if the latter choses to use a ste other
29722         than zero, and return it via niters_vector_mult_vf_var.
29723         * tree-vect-loop.c (vect_transform_loop): Update calls to
29724         vect_do_peeling, vect_gen_vector_loop_niters and
29725         slpeel_make_loop_iterate_ntimes.
29726         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29727         (vect_gen_vector_loop_niters): Update declarations after above changes.
29729 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
29731         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29732         128-bit round to integer instructions.
29733         (ceil<mode>2): Likewise.
29734         (btrunc<mode>2): Likewise.
29735         (round<mode>2): Likewise.
29737 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29739         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29740         unaligned VSX load/store on P8/P9.
29741         (expand_block_clear): Allow the use of unaligned VSX
29742         load/store on P8/P9.
29744 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29746         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29747         New function.
29748         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29749         swap associated with both a load and a store.
29751 2018-01-02  Andrew Waterman  <andrew@sifive.com>
29753         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29754         * config/riscv/riscv.md (clear_cache): Use it.
29756 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
29758         * web.c: Remove out-of-date comment.
29760 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29762         * expr.c (fixup_args_size_notes): Check that any existing
29763         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29764         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29765         (emit_single_push_insn): ...here.
29767 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29769         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29770         (const_vector_encoded_nelts): New function.
29771         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29772         (const_vector_int_elt, const_vector_elt): Declare.
29773         * emit-rtl.c (const_vector_int_elt_1): New function.
29774         (const_vector_elt): Likewise.
29775         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29776         of CONST_VECTOR_ELT.
29778 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29780         * expr.c: Include rtx-vector-builder.h.
29781         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
29782         directly on the tree encoding.
29783         (const_vector_from_tree): Likewise.
29784         * optabs.c: Include rtx-vector-builder.h.
29785         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
29786         sequence of "u" values.
29787         * vec-perm-indices.c: Include rtx-vector-builder.h.
29788         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
29789         directly on the vec_perm_indices encoding.
29791 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29793         * doc/rtl.texi (const_vector): Describe new encoding scheme.
29794         * Makefile.in (OBJS): Add rtx-vector-builder.o.
29795         * rtx-vector-builder.h: New file.
29796         * rtx-vector-builder.c: Likewise.
29797         * rtl.h (rtx_def::u2): Add a const_vector field.
29798         (CONST_VECTOR_NPATTERNS): New macro.
29799         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
29800         (CONST_VECTOR_DUPLICATE_P): Likewise.
29801         (CONST_VECTOR_STEPPED_P): Likewise.
29802         (CONST_VECTOR_ENCODED_ELT): Likewise.
29803         (const_vec_duplicate_p): Check for a duplicated vector encoding.
29804         (unwrap_const_vec_duplicate): Likewise.
29805         (const_vec_series_p): Check for a non-duplicated vector encoding.
29806         Say that the function only returns true for integer vectors.
29807         * emit-rtl.c: Include rtx-vector-builder.h.
29808         (gen_const_vec_duplicate_1): Delete.
29809         (gen_const_vector): Call gen_const_vec_duplicate instead of
29810         gen_const_vec_duplicate_1.
29811         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
29812         (gen_const_vec_duplicate): Use rtx_vector_builder.
29813         (gen_const_vec_series): Likewise.
29814         (gen_rtx_CONST_VECTOR): Likewise.
29815         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
29816         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29817         Build a new vector rather than modifying a CONST_VECTOR in-place.
29818         (handle_special_swappables): Update call accordingly.
29819         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
29820         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29821         Build a new vector rather than modifying a CONST_VECTOR in-place.
29822         (handle_special_swappables): Update call accordingly.
29824 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29826         * simplify-rtx.c (simplify_const_binary_operation): Use
29827         CONST_VECTOR_ELT instead of XVECEXP.
29829 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29831         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
29832         the selector elements to be different from the data elements
29833         if the selector is a VECTOR_CST.
29834         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
29835         ssizetype for the selector.
29837 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29839         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
29840         before testing each element individually.
29841         * tree-vect-generic.c (lower_vec_perm): Likewise.
29843 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29845         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
29846         * selftest-run-tests.c (selftest::run_tests): Call it.
29847         * vector-builder.h (vector_builder::operator ==): New function.
29848         (vector_builder::operator !=): Likewise.
29849         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
29850         (vec_perm_indices::all_from_input_p): New function.
29851         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29852         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
29853         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
29854         instead of reading the VECTOR_CST directly.  Detect whether both
29855         vector inputs are the same before constructing the vec_perm_indices,
29856         and update the number of inputs argument accordingly.  Use the
29857         utility functions added above.  Only construct sel2 if we need to.
29859 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29861         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
29862         the broadcast of the low byte.
29863         (expand_mult_highpart): Use an explicit encoding for the permutes.
29864         * optabs-query.c (can_mult_highpart_p): Likewise.
29865         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
29866         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29867         (vectorizable_bswap): Likewise.
29868         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
29869         explicit encoding for the power-of-2 permutes.
29870         (vect_permute_store_chain): Likewise.
29871         (vect_grouped_load_supported): Likewise.
29872         (vect_permute_load_chain): Likewise.
29874 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29876         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
29877         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
29878         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
29879         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29880         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
29881         (vect_gen_perm_mask_any): Likewise.
29883 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29885         * int-vector-builder.h: New file.
29886         * vec-perm-indices.h: Include int-vector-builder.h.
29887         (vec_perm_indices): Redefine as an int_vector_builder.
29888         (auto_vec_perm_indices): Delete.
29889         (vec_perm_builder): Redefine as a stand-alone class.
29890         (vec_perm_indices::vec_perm_indices): New function.
29891         (vec_perm_indices::clamp): Likewise.
29892         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
29893         (vec_perm_indices::new_vector): New function.
29894         (vec_perm_indices::new_expanded_vector): Update for new
29895         vec_perm_indices class.
29896         (vec_perm_indices::rotate_inputs): New function.
29897         (vec_perm_indices::all_in_range_p): Operate directly on the
29898         encoded form, without computing elided elements.
29899         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
29900         encoding.  Update for new vec_perm_indices class.
29901         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
29902         the given vec_perm_builder.
29903         (expand_vec_perm_var): Update vec_perm_builder constructor.
29904         (expand_mult_highpart): Use vec_perm_builder instead of
29905         auto_vec_perm_indices.
29906         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
29907         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
29908         or double series encoding as appropriate.
29909         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
29910         vec_perm_indices instead of auto_vec_perm_indices.
29911         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29912         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29913         (vect_permute_store_chain): Likewise.
29914         (vect_grouped_load_supported): Likewise.
29915         (vect_permute_load_chain): Likewise.
29916         (vect_shift_permute_load_chain): Likewise.
29917         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29918         (vect_transform_slp_perm_load): Likewise.
29919         (vect_schedule_slp_instance): Likewise.
29920         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29921         (vectorizable_mask_load_store): Likewise.
29922         (vectorizable_bswap): Likewise.
29923         (vectorizable_store): Likewise.
29924         (vectorizable_load): Likewise.
29925         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
29926         vec_perm_indices instead of auto_vec_perm_indices.  Use
29927         tree_to_vec_perm_builder to read the vector from a tree.
29928         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
29929         vec_perm_builder instead of a vec_perm_indices.
29930         (have_whole_vector_shift): Use vec_perm_builder and
29931         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
29932         truncation to calc_vec_perm_mask_for_shift.
29933         (vect_create_epilog_for_reduction): Likewise.
29934         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
29935         from auto_vec_perm_indices to vec_perm_indices.
29936         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29937         instead of changing individual elements.
29938         (aarch64_vectorize_vec_perm_const): Use new_vector to install
29939         the vector in d.perm.
29940         * config/arm/arm.c (expand_vec_perm_d::perm): Change
29941         from auto_vec_perm_indices to vec_perm_indices.
29942         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29943         instead of changing individual elements.
29944         (arm_vectorize_vec_perm_const): Use new_vector to install
29945         the vector in d.perm.
29946         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
29947         Update vec_perm_builder constructor.
29948         (rs6000_expand_interleave): Likewise.
29949         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
29950         (rs6000_expand_interleave): Likewise.
29952 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29954         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
29955         to qimode could truncate the indices.
29956         * optabs.c (expand_vec_perm_var): Likewise.
29958 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29960         * Makefile.in (OBJS): Add vec-perm-indices.o.
29961         * vec-perm-indices.h: New file.
29962         * vec-perm-indices.c: Likewise.
29963         * target.h (vec_perm_indices): Replace with a forward class
29964         declaration.
29965         (auto_vec_perm_indices): Move to vec-perm-indices.h.
29966         * optabs.h: Include vec-perm-indices.h.
29967         (expand_vec_perm): Delete.
29968         (selector_fits_mode_p, expand_vec_perm_var): Declare.
29969         (expand_vec_perm_const): Declare.
29970         * target.def (vec_perm_const_ok): Replace with...
29971         (vec_perm_const): ...this new hook.
29972         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
29973         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
29974         * doc/tm.texi: Regenerate.
29975         * optabs.def (vec_perm_const): Delete.
29976         * doc/md.texi (vec_perm_const): Likewise.
29977         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
29978         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
29979         expand_vec_perm for constant permutation vectors.  Assert that
29980         the mode of variable permutation vectors is the integer equivalent
29981         of the mode that is being permuted.
29982         * optabs-query.h (selector_fits_mode_p): Declare.
29983         * optabs-query.c: Include vec-perm-indices.h.
29984         (selector_fits_mode_p): New function.
29985         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
29986         is defined, instead of checking whether the vec_perm_const_optab
29987         exists.  Use targetm.vectorize.vec_perm_const instead of
29988         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
29989         fit in the vector mode before using a variable permute.
29990         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
29991         vec_perm_indices instead of an rtx.
29992         (expand_vec_perm): Replace with...
29993         (expand_vec_perm_const): ...this new function.  Take the selector
29994         as a vec_perm_indices rather than an rtx.  Also take the mode of
29995         the selector.  Update call to shift_amt_for_vec_perm_mask.
29996         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
29997         Use vec_perm_indices::new_expanded_vector to expand the original
29998         selector into bytes.  Check whether the indices fit in the vector
29999         mode before using a variable permute.
30000         (expand_vec_perm_var): Make global.
30001         (expand_mult_highpart): Use expand_vec_perm_const.
30002         * fold-const.c: Includes vec-perm-indices.h.
30003         * tree-ssa-forwprop.c: Likewise.
30004         * tree-vect-data-refs.c: Likewise.
30005         * tree-vect-generic.c: Likewise.
30006         * tree-vect-loop.c: Likewise.
30007         * tree-vect-slp.c: Likewise.
30008         * tree-vect-stmts.c: Likewise.
30009         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
30010         Delete.
30011         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
30012         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
30013         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
30014         (aarch64_vectorize_vec_perm_const): ...this new function.
30015         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30016         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30017         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
30018         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
30019         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30020         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30021         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
30022         into...
30023         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
30024         check for NEON modes.
30025         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
30026         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
30027         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
30028         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
30029         into...
30030         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
30031         the old VEC_PERM_CONST conditions.
30032         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
30033         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
30034         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
30035         (ia64_vectorize_vec_perm_const_ok): Merge into...
30036         (ia64_vectorize_vec_perm_const): ...this new function.
30037         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
30038         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
30039         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
30040         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
30041         * config/mips/mips.c (mips_expand_vec_perm_const)
30042         (mips_vectorize_vec_perm_const_ok): Merge into...
30043         (mips_vectorize_vec_perm_const): ...this new function.
30044         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
30045         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
30046         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
30047         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
30048         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
30049         (rs6000_expand_vec_perm_const): Delete.
30050         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
30051         Delete.
30052         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30053         (altivec_expand_vec_perm_const_le): Take each operand individually.
30054         Operate on constant selectors rather than rtxes.
30055         (altivec_expand_vec_perm_const): Likewise.  Update call to
30056         altivec_expand_vec_perm_const_le.
30057         (rs6000_expand_vec_perm_const): Delete.
30058         (rs6000_vectorize_vec_perm_const_ok): Delete.
30059         (rs6000_vectorize_vec_perm_const): New function.
30060         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30061         an element count and rtx array.
30062         (rs6000_expand_extract_even): Update call accordingly.
30063         (rs6000_expand_interleave): Likewise.
30064         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
30065         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
30066         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
30067         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
30068         (rs6000_expand_vec_perm_const): Delete.
30069         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30070         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30071         (altivec_expand_vec_perm_const_le): Take each operand individually.
30072         Operate on constant selectors rather than rtxes.
30073         (altivec_expand_vec_perm_const): Likewise.  Update call to
30074         altivec_expand_vec_perm_const_le.
30075         (rs6000_expand_vec_perm_const): Delete.
30076         (rs6000_vectorize_vec_perm_const_ok): Delete.
30077         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
30078         reference to the SPE evmerge intructions.
30079         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30080         an element count and rtx array.
30081         (rs6000_expand_extract_even): Update call accordingly.
30082         (rs6000_expand_interleave): Likewise.
30083         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
30084         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
30085         new function.
30086         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30088 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30090         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
30091         vector mode and that that mode matches the mode of the data
30092         being permuted.
30093         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
30094         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
30095         directly using expand_vec_perm_1 when forcing selectors into
30096         registers.
30097         (expand_vec_perm_var): New function, split out from expand_vec_perm.
30099 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30101         * optabs-query.h (can_vec_perm_p): Delete.
30102         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
30103         * optabs-query.c (can_vec_perm_p): Split into...
30104         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
30105         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
30106         particular selector is valid.
30107         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30108         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30109         (vect_grouped_load_supported): Likewise.
30110         (vect_shift_permute_load_chain): Likewise.
30111         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30112         (vect_transform_slp_perm_load): Likewise.
30113         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30114         (vectorizable_bswap): Likewise.
30115         (vect_gen_perm_mask_checked): Likewise.
30116         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
30117         implementations of variable permutation vectors into account
30118         when deciding which selector to use.
30119         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
30120         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
30121         with a false third argument.
30122         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
30123         to test whether the constant selector is valid and can_vec_perm_var_p
30124         to test whether a variable selector is valid.
30126 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30128         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
30129         * optabs-query.c (can_vec_perm_p): Likewise.
30130         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
30131         instead of vec_perm_indices.
30132         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
30133         (vect_gen_perm_mask_checked): Likewise,
30134         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
30135         (vect_gen_perm_mask_checked): Likewise,
30137 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30139         * optabs-query.h (qimode_for_vec_perm): Declare.
30140         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
30141         (qimode_for_vec_perm): ...this new function.
30142         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
30144 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30146         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
30147         does not have a conditional at the top.
30149 2018-01-02  Richard Biener  <rguenther@suse.de>
30151         * ipa-inline.c (big_speedup_p): Fix expression.
30153 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30155         PR target/81616
30156         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
30157         for generic 4->6.
30159 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30161         PR target/81616
30162         Generic tuning.
30163         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
30164         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
30165         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
30166         cond_taken_branch_cost 3->4.
30168 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
30170         PR tree-optimization/83581
30171         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
30172         TODO_cleanup_cfg if any changes have been made.
30174         PR middle-end/83608
30175         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
30176         convert_modes if target mode has the right side, but different mode
30177         class.
30179         PR middle-end/83609
30180         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
30181         last argument when extracting from CONCAT.  If either from_real or
30182         from_imag is NULL, use expansion through memory.  If result is not
30183         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
30184         the parts directly to inner mode, if even that fails, use expansion
30185         through memory.
30187         PR middle-end/83623
30188         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
30189         check for bswap in mode rather than HImode and use that in expand_unop
30190         too.
30192 Copyright (C) 2018 Free Software Foundation, Inc.
30194 Copying and distribution of this file, with or without modification,
30195 are permitted in any medium without royalty provided the copyright
30196 notice and this notice are preserved.