2018-11-07 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
bloba34bf51e7bccf4d492cc3bdc950c9154e159d0ac
1 2018-11-07  Richard Biener  <rguenther@suse.de>
3         PR lto/87906
4         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
5         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
7 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
9         PR rtl-optimization/87874
10         * lra.c (lra_substitute_pseudo): Do not create a subreg for
11         const wide ints.
13 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
15         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
16         if not in indexed or indirect form.
17         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
18         (bswap<mode>2_store): Ditto.
20 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
22         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
23         the UNSPEC.
25 2018-11-06  Richard Biener  <rguenther@suse.de>
27         PR tree-optimization/86850
28         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
29         instead of src.m_vec.
31 2018-11-06  Jan Hubicka  <jh@suse.cz>
33         * tree.c (fld_simplified_type_name): Break out form ...
34         (free_lang_data_in_type): ... here.
35         (fld_type_variant_equal_p): Use it.
37 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
39         * config/default-d.c: Include memmodel.h.
41         * config/sol2-d.c: New file.
42         * config/t-sol2 (sol2-d.o): New rule.
43         * config.gcc <*-*-solaris2*>: Set d_target_objs,
44         target_has_targetdm.
46 2018-11-06  Jan Hubicka  <jh@suse.cz>
48         * tree.c (fld_type_variant): Also copy alignment; be sure that
49         new variant is equal.
51 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
53         PR target/87762
54         * config/s390/s390.md: Add relative_long attribute.
56 2018-11-06  Jan Hubicka  <jh@suse.cz>
58         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
59         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
60         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
61         stream TYPE_NEEDS_CONSTRUCTING.
62         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
63         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
65 2018-11-06  Richard Biener  <rguenther@suse.de>
67         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
68         dump-scope ...
69         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
71 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
73         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
74         REG_CLASS_CONTENTS[GEN_REGS].
75         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
77 2018-11-06  Jan Hubicka  <jh@suse.cz>
79          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
81 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
83         PR sanitizer/80953
84         * config/sol2.h (ASAN_CC1_SPEC): Define.
85         (LD_WHOLE_ARCHIVE_OPTION): Define.
86         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
87         (ASAN_REJECT_SPEC): Provide default.
88         (LIBASAN_EARLY_SPEC): Define.
89         (LIBTSAN_EARLY_SPEC): Define.
90         (LIBLSAN_EARLY_SPEC): Define.
91         * config/i386/sol2.h (CC1_SPEC): Redefine.
92         (ASAN_REJECT_SPEC): Define.
94         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
95         (TARGET_ASAN_SHADOW_OFFSET): Define.
96         (sparc_asan_shadow_offset): New function.
97         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
98         (ASAN_REJECT_SPEC): Define.
100 2018-11-06  Jan Hubicka  <jh@suse.cz>
102         * tree.c (fld_type_variant): Copy canonical type.
103         (fld_incomplete_type_of): Check that canonical types looks sane;
104         copy canonical type.
105         (verify_type): Accept when incomplete type has complete canonical type.
107 2018-11-06  Jan Hubicka  <jh@suse.cz>
109         * tree.c (free_lang_data): Reset overwite_assembler_name,
110         print_xnode, print_decl, print_type and print_identifier of
111         langhooks.
113 2018-11-06  Richard Biener  <rguenther@suse.de>
115         PR tree-optimization/87889
116         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
117         Do nothing if old and new arg are the same
119 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
121         PR target/87723
122         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
123         attributes for operands 3 and 4.
125 2018-11-06  Richard Biener  <rguenther@suse.de>
127         PR middle-end/18041
128         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
129         matching bitfield insertion.
131 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
133         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
134         (reg_next_debug_use): New.
135         (attempt_change): Propagate adjusted expression into affected
136         debug insns.
137         (merge_in_block): Track uses in debug insns.
138         (pass_inc_dec::execute): Allocate and release
139         reg_next_debug_use.
141 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
143         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
144         (_mm512_fixupimm_round_pd): Update parameters and builtin.
145         (_mm512_maskz_fixupimm_round_pd): Ditto.
146         (_mm512_fixupimm_round_ps): Ditto.
147         (_mm512_maskz_fixupimm_round_ps): Ditto.
148         (_mm_fixupimm_round_sd): Ditto.
149         (_mm_maskz_fixupimm_round_sd): Ditto.
150         (_mm_fixupimm_round_ss): Ditto.
151         (_mm_maskz_fixupimm_round_ss): Ditto.
152         (_mm512_fixupimm_pd): Ditto.
153         (_mm512_maskz_fixupimm_pd): Ditto.
154         (_mm512_fixupimm_ps): Ditto.
155         (_mm512_maskz_fixupimm_ps): Ditto.
156         (_mm_fixupimm_sd): Ditto.
157         (_mm_maskz_fixupimm_sd): Ditto.
158         (_mm_fixupimm_ss): Ditto.
159         (_mm_maskz_fixupimm_ss): Ditto.
160         (_mm512_mask_fixupimm_round_pd): Update builtin.
161         (_mm512_mask_fixupimm_round_ps): Ditto.
162         (_mm_mask_fixupimm_round_sd): Ditto.
163         (_mm_mask_fixupimm_round_ss): Ditto.
164         (_mm512_mask_fixupimm_pd): Ditto.
165         (_mm512_mask_fixupimm_ps): Ditto.
166         (_mm_mask_fixupimm_sd): Ditto.
167         (_mm_mask_fixupimm_ss): Ditto.
168         * config/i386/avx512vlintrin.h:
169         (_mm256_fixupimm_pd): Update parameters and builtin.
170         (_mm256_maskz_fixupimm_pd): Ditto.
171         (_mm256_fixupimm_ps): Ditto.
172         (_mm256_maskz_fixupimm_ps): Ditto.
173         (_mm_fixupimm_pd): Ditto.
174         (_mm_maskz_fixupimm_pd): Ditto.
175         (_mm_fixupimm_ps): Ditto.
176         (_mm_maskz_fixupimm_ps): Ditto.
177         (_mm256_mask_fixupimm_pd): Update builtin.
178         (_mm256_mask_fixupimm_ps): Ditto.
179         (_mm_mask_fixupimm_pd): Ditto.
180         (_mm_mask_fixupimm_ps): Ditto.
181         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
182         * config/i386/i386-builtin.def: Update builtin definitions.
183         * config/i386/i386.c: Handle new builtin types and remove useless ones.
184         * config/i386/sse.md: Update VFIXUPIMM* patterns.
185         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
186         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
187         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
188         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
189         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
190         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
191         * config/i386/subst.md:
192         (round_saeonly_sd_mask_operand4): Add new subst_attr.
193         (round_saeonly_sd_mask_op4): Ditto.
194         (round_saeonly_expand_operand5): Ditto.
195         (round_saeonly_expand): Update.
197 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
199         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
201 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
203         PR rtl-optimization/87871
204         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
206 2018-11-05  Paul Koning  <ni1d@arrl.net>
208         * doc/sourcebuild.texi (target attributes): Document new "inf"
209         effective target keyword.
211 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
213         * config/s390/s390.c (s390_register_move_cost): Increase costs for
214         moves involving the CC reg.
216 2018-11-05  Richard Biener  <rguenther@suse.de>
218         PR tree-optimization/87873
219         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
220         argument.
221         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
222         * tree-vect-loop.c (vect_transform_loop): When splitting the
223         loop exit also create forwarder PHIs for constants.
224         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
225         Handle constant to_arg, add extra checking we match up the correct
226         PHIs.
228 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
230         * config/s390/s390.md: QImode and HImode for load on condition.
232 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
234         * config/s390/predicates.md: Fix typo.
235         * config/s390/s390.md: Allow immediates for load on condition.
237 2018-11-05  Martin Liska  <mliska@suse.cz>
239         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
240         * bitmap.h (struct bitmap_usage): Likewise.
241         * ggc-common.c (SCALE): Remove.
242         (LABEL): Likewise.
243         (struct ggc_usage): Use SIZE_AMOUNT. And update
244         compare method.
245         * ggc-page.c (SCALE): Remove.
246         (STAT_LABEL): Remove.
247         (ggc_print_statistics): Use SIZE_AMOUNT.
248         * gimple.h (SCALE): Remove.
249         (LABEL): Likewise.
250         * input.c (ONE_K): Remove.
251         (ONE_M): Likewise.
252         (SCALE): Likewise.
253         (STAT_LABEL): Likewise.
254         (FORMAT_AMOUNT): Likewise.
255         (dump_line_table_statistics): Use SIZE_AMOUNT.
256         * mem-stats.h (struct mem_usage): Likewise.
257         * rtl.c (dump_rtx_statistics): Likewise.
258         (rtx_alloc_counts): Change type to size_t.
259         (rtx_alloc_sizes): Likewise.
260         (rtx_count_cmp): New.
261         (dump_rtx_statistics): Sort first based on counts.
262         * tree.c (tree_nodes_cmp): New.
263         (tree_codes_cmp): New.
264         (dump_tree_statistics): Sort first based on counts.
265         * system.h (ONE_K): New.
266         (ONE_M): Likewise.
267         (SIZE_SCALE): Likewise.
268         (SIZE_LABEL): Likewise.
269         (SIZE_AMOUNT): Likewise.
270         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
271         * tree-dfa.c (dump_dfa_stats): Likewise.
272         * tree-phinodes.c (phinodes_print_statistics): Likewise.
273         * tree-ssanames.c (ssanames_print_statistics): Likewise.
274         * tree.c (dump_tree_statistics): Likewise.
275         * vec.c (struct vec_usage): Likewise.
276         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
277         trigger a -Werror=format-overflow with
278         --enable-gather-detailed-stats.
280 2018-11-05  Martin Liska  <mliska@suse.cz>
282         * mem-stats.h (mem_alloc_description::release_instance_overhead):
283         Return T *.
284         * vec.c (struct vec_usage): Register m_element_size.
285         (vec_prefix::register_overhead): New arguments: elements and
286         element_size.
287         (vec_prefix::release_overhead): Subtract elements.
288         * vec.h (struct vec_prefix): Change signature.
289         (va_heap::reserve): Pass proper arguments.
290         (va_heap::release): Likewise.
292 2018-11-05  Martin Liska  <mliska@suse.cz>
294         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
295         style.
296         * vec.c: Likewise.
298 2018-11-05  Richard Biener  <rguenther@suse.de>
300         * tree-scalar-evolution.h (final_value_replacement_loop): Update
301         prototype.
302         * tree-scalar-evolution.c (final_value_replacement_loop): Return
303         whether anything was done.
304         (scev_const_prop): Remove constant propagation part, fold
305         remains into ...
306         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
307         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
308         conditionally.
310 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
312         PR tree-optimization/87859
313         * gimple-ssa-store-merging.c (struct merged_store_group): Add
314         only_constants and first_nonmergeable_order members.
315         (merged_store_group::merged_store_group): Initialize them.
316         (merged_store_group::do_merge): Clear only_constants member if
317         adding something other than INTEGER_CST store.
318         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
319         stores with order >= first_nonmergeable_order.  Use
320         merged_store->only_constants instead of always recomputing it.
321         Set merged_store->first_nonmergeable_order if we've skipped any
322         stores.  Attempt to merge overlapping INTEGER_CST stores that
323         we would otherwise skip.
325         PR sanitizer/87837
326         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
328 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
330         PR target/87853
331         * config/i386/emmintrin.h (__v16qs): New to cope with option
332         -funsigned-char.
333         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
334         (_mm_cmplt_epi8): Likewise.
335         (_mm_cmpgt_epi8): Likewise.
337 2018-11-05  Richard Biener  <rguenther@suse.de>
339         PR rtl-optimization/87852
340         * fwprop.c (use_killed_between): Only consider single-defs of the
341         use whose definition statement dominates the use.
343 2018-11-05  Martin Liska  <mliska@suse.cz>
345         PR web/87829
346         * doc/invoke.texi: Remove options that are
347         not disabled with -Os.
349 2018-11-05  Martin Liska  <mliska@suse.cz>
351         PR c/87811
352         * doc/extend.texi: Update constrain about the last argument
353         of __builtin_expect_with_probability.
355 2018-11-05  Martin Liska  <mliska@suse.cz>
357         PR c/87811
358         * predict.c (expr_expected_value_1): Verify
359         that last argument is a real constants and emit
360         error.
362 2018-11-05  Martin Liska  <mliska@suse.cz>
364         PR gcov-profile/77698
365         * ipa-profile.c (ipa_profile): Adjust hotness threshold
366         only in LTO mode.
368 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
370         PR tree-optimization/86572
371         * builtins.c (c_strlen): Handle negative offsets in a safe way.
373 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
375         PR tree-optimization/87672
376         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
377         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
379 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
381         PR middle-end/58372
382         * cfgexpand.c (pass_expand::execute): Move the call to
383         finish_eh_generation in front of the call to expand_stack_alignment.
385 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
387         * common/config/i386/i386-common.c (processor_alias_table): Add
388         znver2 entry.
389         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
390         (case ${target}): Add znver2.
391         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
392         -march=native recognize znver2 processors.
393         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
394         * config/i386/i386.c (m_znver2): New definition.
395         (m_ZNVER): New definition.
396         (m_AMD_MULTIPLE): Includes m_znver2.
397         (processor_cost_table): Add znver2 entry.
398         (processor_target_table): Add znver2 entry.
399         (get_builtin_code_for_version): Set priority for
400         PROCESSOR_ZNVER2.
401         (processor_model): Add M_AMDFAM17H_ZNVER2.
402         (arch_names_table): Ditto.
403         (ix86_reassociation_width): Include znver2.
404         * config/i386/i386.h (TARGET_znver2): New definition.
405         (struct ix86_size_cost): Add TARGET_ZNVER2.
406         (enum processor_type): Add PROCESSOR_ZNVER2.
407         * config/i386/i386.md (define_attr "cpu"): Add znver2.
408         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
409         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
410         (ix86_adjust_cost): Add znver2.
411         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
412         * gcc/doc/extend.texi: Add details about znver2.
413         * gcc/doc/invoke.texi: Add details about znver2.
415 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
417         PR target/87079
419         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
420         pattern.
422 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
424         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
425         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
427 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
429         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
430         shrink-wrapping when -mtrack-speculation.
432 2018-11-02  Richard Biener  <rguenther@suse.de>
434         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
435         (pop_cost_one_pair): Do not free pair.
436         (pop_best_coalesce): Likewise.
437         (create_coalesce_list): Initialize obstack.
438         (delete_coalesce_list): Free obstack.
439         (find_coalesce_pair): Obstack-allocate coalesce pairs.
440         (add_cost_one_coalesce): Likewise.
441         (struct live_track): Remove bitmap pointer indirections.
442         (new_live_track): Adjust.
443         (delete_live_track): Likewise.
444         (live_track_remove_partition): Likewise.
445         (live_track_add_partition): Likewise.
446         (live_track_live_p): Likewise.
447         (live_track_process_def): Likewise.
448         (live_track_clear_base_vars): Likewise.
450 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
452         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
453         * configure: Regenerate.
454         * config.in: Regenerate.
455         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
456         HAVE_AS_REGISTER_PSEUDO_OP guard.
457         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
458         !HAVE_AS_REGISTER_PSEUDO_OP support.
459         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
460         guard.
462 2018-11-02  Richard Biener  <rguenther@suse.de>
464         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
465         with is_type_die.
467 2018-11-02  Richard Biener  <rguenther@suse.de>
469         PR tree-optimization/87776
470         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
471         executable when iterating but running into rpo-vn-max-loop-depth
472         and not eliding the iteration.
474 2018-11-30  Jan Hubicka  <jh@suse.cz>
476         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
477         TYPE_DECL.
479 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
481         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
482         prototype.
483         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
484         Combine with rs6000_address_for_fpconvert.
485         (rs6000_address_for_fpconvert): Combine with
486         rs6000_force_indexed_or_indirect_mem.
487         (rs6000_expand_vector_init): Change function call from
488         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
489         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
490         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
491         (floatsi<mode>2_lfiwax_mem): Ditto.
492         (floatunssi<mode>2_lfiwzx): Ditto.
493         (floatunssi<mode>2_lfiwzx_mem): Ditto.
494         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
495         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
496         (fix_trunc<mode>si2_stfiwx): Ditto.
497         (fixuns_trunc<mode>si2_stfiwx): Ditto.
498         (float_<mode>si2_hw): Ditto.
499         (floatuns_<mode>si2_hw): Ditto.
500         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
501         (vsx_splat_<mode>): Ditto.
503 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
505         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
506         at end of assembler input text.
507         * configure: Regenerate.
509 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
511         PR tree-optimization/87826
512         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
513         negative or larger or equal to type's precision.
515 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
517         * opts.c (default_options_table): Do not enable
518         OPT_fdelayed_branch at -Og.
519         * doc/invoke.texi (-fdelayed-branch): Document it.
521 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
523         * optabs-libfuncs.c (build_libfunc_function_visibility):
524         New, split out from...
525         (build_libfunc_function): ... here.
526         (init_one_libfunc_visibility): New, split out from ...
527         (init_one_libfunc): ... here.
529         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
530         scratch register need not be early-clobber.  Document the reason
531         why we cannot use ST<OP>.
533 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
535         PR bootstrap/82856
536         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
537         line for second argument of AC_DEFINE_UNQUOTED.
538         * doc/install.texi (Tools/packages necessary for modifying GCC):
539         Update to autoconf 2.69 and automake 1.15.1.
540         * aclocal.m4, config.in, configure: Regenerate.
542 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
544         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
545         initialization.
547 2018-10-31  Martin Liska  <mliska@suse.cz>
549         PR driver/83193
550         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
551         Add new argument invalid_extension.
552         (aarch64_get_all_extension_candidates): New function.
553         (aarch64_rewrite_selected_cpu): Add NULL to function call.
554         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
555         new argument.
556         (aarch64_get_all_extension_candidates): New function.
557         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
558         argument invalid_extension.
559         (aarch64_parse_cpu): Likewise.
560         (aarch64_print_hint_for_extensions): New function.
561         (aarch64_validate_mcpu): Provide hint about invalid extension.
562         (aarch64_validate_march): Likewise.
563         (aarch64_handle_attr_arch): Pass new argument.
564         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
565         (aarch64_handle_attr_isa_flags): Likewise.
567 2018-10-31  Richard Biener  <rguenther@suse.de>
569         PR middle-end/70359
570         PR middle-end/86270
571         * tree-outof-ssa.c (insert_backedge_copies): Restrict
572         copy generation to useful cases.  Place the copy before
573         the definition of the backedge value when possible.
575 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
577         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
578         * config/arc/arc.c (arc_active_insn): New function.
579         (check_store_cacheline_hazard): Likewise.
580         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
581         (arc_override_options): Disable delay slot scheduler for older
582         A7.
583         (arc_store_addr_hazard_p): New implementation, old one renamed to
584         ...
585         (arc_store_addr_hazard_internal_p): Renamed.
586         (arc_reorg): Don't combine into brcc instructions which are part
587         of hardware hazard solution.
588         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
589         (tune_arc700): Likewise.
590         * config/arc/arc.opt (arc7xx): New tune value.
591         * config/arc/arc700.md: Improve A7 scheduler.
593 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
595         * config/arc/arc.c (arc_override_options): Remove
596         TARGET_COMPACT_CASESI.
597         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
598         (CASE_VECTOR_MODE): Likewise.
599         (CASE_VECTOR_PC_RELATIVE): Likewise.
600         (CASE_VECTOR_SHORTEN_MODE): Likewise.
601         (CASE_VECTOR_SHORTEN_MODE1): Delete.
602         (ADDR_VEC_ALIGN): Update.
603         (ASM_OUTPUT_CASE_LABEL): Undefine.
604         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
605         (TARGET_BI_BIH): Define.
606         (DEFAULT_BRANCH_INDEX): Likewise.
607         * config/arc/arc.md (casesi): Rework to accept BI/BIH
608         instructions, remove compact_casesi use case.
609         (casesi_compact_jump): Remove.
610         (casesi_dispatch): New pattern.
611         * config/arc/arc.opt: Add mbranch-index option. Deprecate
612         compact_casesi option.
613         * doc/invoke.texi: Document mbranch-index option.
615 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
617         * config/arc/arc.c (arc_get_tp): Remove function.
618         (arc_emit_call_tls_get_addr): Likewise.
619         (arc_call_tls_get_addr): New function.
620         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
621         * config/arc/arc.md (tls_load_tp_soft): Remove.
622         (tls_gd_get_addr): Likewise.
624 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
626         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
627         (mulsi3_600_lib): Remove pattern.
628         (umulsi3_highpart_600_lib_le): Likewise.
629         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
630         (umulsidi3): Remove call to umulsidi3_600_lib.
631         (umulsidi3_600_lib): Remove pattern.
632         (peephole2): Remove peephole using the above deprecated patterns.
634 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
636         PR target/87374
637         * config/arm/arm.c (arm_option_check_internal): Disable the combined
638         use of -mslow-flash-data and -mword-relocations.
639         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
640         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
641         flag_pic.
642         * doc/invoke.texi (-mword-relocations): Mention conflict with
643         -mslow-flash-data.
644         (-mslow-flash-data): Reciprocally.
646 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
648         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
649         16-byte modes held in GP registers to use an even regno.
651         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
652         (aarch64_atomic_ldop_supported_p): Remove.
653         (aarch64_gen_atomic_ldop): Remove.
654         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
655         Fully expand LSE operations here.
656         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
657         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
658         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
659         and use ATOMIC_LDOP instead; use register_operand for the input;
660         drop the split and emit insns directly.
661         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
662         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
663         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
665         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
666         (aarch64_gen_atomic_ldop): Don't call it.
667         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
668         Use aarch64_reg_or_zero.
669         (aarch64_atomic_exchange<ALLI>): Likewise.
670         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
671         operand 0; use aarch64_reg_or_zero for input; merge ...
672         (@aarch64_atomic_swp<ALLI>): ... this and remove.
674         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
675         (aarch64_split_compare_and_swap): Use it.
676         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
677         test oldval against the proper predicate.
678         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
679         Use nonmemory_operand for expected.
680         (cas_short_expected_pred): New.
681         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
682         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
683         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
684         (aarch64_plushi_operand): New.
686         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
687         Force oldval into the rval register for TARGET_LSE; emit the compare
688         during initial expansion so that it may be deleted if unused.
689         (aarch64_gen_atomic_cas): Remove.
690         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
691         Change =&r to +r for operand 0; use match_dup for operand 2;
692         remove is_weak and mod_f operands as unused.  Drop the split
693         and merge with...
694         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
695         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
696         (@aarch64_atomic_cas<GPI>): Similarly.
698 2018-10-31  Richard Biener  <rguenther@suse.de>
700         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
701         using ABSU_EXPR.
703 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
705         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
706         * config/aarch64/aarch64.md: Include saphira.md
707         * config/aarch64/saphira.md: New file for pipeline description.
709 2018-10-30  Martin Sebor  <msebor@redhat.com>
711         PR middle-end/87041
712         * gimple-ssa-sprintf.c (format_directive): Use %G to include
713         inlining context.
714         (sprintf_dom_walker::compute_format_length):
715         Avoid setting POSUNDER4K here.
716         (get_destination_size): Handle null argument values.
717         (get_user_idx_format): New function.
718         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
719         functions, including user-defined with attribute format printf.
720         Use %G to include inlining context.
721         Set POSUNDER4K here.
723 2018-10-30  Jan Hubicka  <jh@suse.cz>
725         * params.def (lto-partitions): Bump from 32 to 128.
727 2018-10-30  Jan Hubicka  <jh@suse.cz>
729         * tree.c
730         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
731         head in file.
732         (free_lang_data_in_type): Forward declare.
733         (fld_type_variant_equal_p): New function.
734         (fld_type_variant): New function
735         (fld_incomplete_types): New hash.
736         (fld_incomplete_type_of): New function
737         (fld_simplfied-type): New function.
738         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
739         (free_lang_data): Allocate and free fld_incomplete_type; update call
740         of free_lang_data_in_decl.
742 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
744         * gcov.c (output_lines): Remove duplicate line.
746 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
748         * config/rs6000/rs6000.md (bswapdi2): Force address into register
749         if not in indexed or indirect form.
750         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
751         (bswapdi2_store): Ditto.
752         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
753         helper function.
754         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
755         Prototype for helper function.
757 2018-10-30  Martin Sebor  <msebor@redhat.com>
759         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
760         (target, pragma GCC optimize, pragma GCC target): Ditto.
762 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
764         * doc/extend.texi: Fix prototype and description of
765         __builtin_expect_with_probability.
767 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
769         * cgraph.h (clone_function_name_1): Replaced by new
770           clone_function_name_numbered that takes name as string; for
771           privatize_symbol_name_1 use only.
772           (clone_function_name): Renamed to
773           clone_function_name_numbered to be explicit about numbering.
774           (clone_function_name): New two-argument function that does
775           not number its output.
776           (clone_function_name): New three-argument function that
777           takes a number to append to its output.
778         * cgraphclones.c (duplicate_thunk_for_node):
779           (clone_function_name_1): Renamed.
780           (clone_function_name_numbered): Two new functions.
781           (clone_function_name): Improved documentation.
782           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
783         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
784         * final.c (final_scan_insn_1): Use the new clone_function_name
785           without numbering.
786         * multiple_target.c (create_dispatcher_calls): Ditto.
787           (create_target_clone): Ditto.
788         * omp-expand.c (grid_expand_target_grid_body): Ditto.
789         * omp-low.c (create_omp_child_function_name): Ditto.
790         * omp-simd-clone.c (simd_clone_create): Ditto.
791         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
792           new clone_function_name without numbering.
794 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
796         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
797         Assert that the allocation size is not zero.
799 2018-10-30  Richard Biener  <rguenther@suse.de>
801         PR tree-optimization/87800
802         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
803         non-induction or reduction PHIs.
805 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
807         * config/aarch64/falkor-tag-collision-avoidance.c
808         (execute_tag_collision_avoidance): Call df_note_add_problem.
810 2018-10-30  Martin Liska  <mliska@suse.cz>
812         * doc/extend.texi: Fix typo in documentation
813         of __builtin_expect_with_probability.
815 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
817         PR c++/87721
818         * input.c (get_substring_ranges_for_loc): Detect if
819         linemap_resolve_location gives us a NULL map, and reject
820         this case.
822 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
824         * config.gcc (xstormy16-*-elf): Set tm_d_file.
826 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
827             Martin Sebor  <msebor@redhat.com>
828             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
830         * doc/ux.texi (Quoting): New subsection, adapted from material at
831         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
832         MartinSebor and ManuelLopezIbanez.
833         (Fix-it hints): Note that fix-it hints shouldn't be marked for
834         translation.
836 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
838         PR middle-end/87469
839         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
840         max value.
842 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
844         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
846 2018-10-29  Paul Koning  <ni1d@arrl.net>
848         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
850 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
852         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
853         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
854         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
855         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
856         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
857         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
858         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
859         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
860         Change 'vector' to '__vector'.
861         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
862         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
863         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
864         _mm_avg_pu16): Likewise.  And, whitespace corrections.
866 2018-10-29  Richard Biener  <rguenther@suse.de>
868         PR tree-optimization/87785
869         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
870         internal defs.
872 2018-10-29  Olivier Hainque  <hainque@adacore.com>
874         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
876 2018-10-29  Olivier Hainque  <hainque@adacore.com>
878         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
879         ports configuration macro, defaults to "gnu".
880         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
881         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
883 2018-10-29  Olivier Hainque  <hainque@adacore.com>
885         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
887 2018-10-29  Richard Biener  <rguenther@suse.de>
889         PR tree-optimization/87790
890         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
891         (vect_make_slp_decision): Adjust.
892         (vect_slp_analyze_bb_1): Likewise.
893         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
894         edges.
896 2018-10-29  Richard Biener  <rguenther@suse.de>
898         PR tree-optimization/87785
899         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
900         and processing.
901         (vect_build_slp_tree): Likewise.
902         (vect_gather_slp_loads): New function.
903         (vect_analyze_slp_instance): Gather loads separately from the
904         SLP tree build.
906 2018-10-29  Martin Liska  <mliska@suse.cz>
908         * Makefile.in: Make dependency to json.o.
909         * doc/gcov.texi: Document new JSON format, remove
910         old intermediate format documentation.
911         * gcov.c (struct function_info): Come up with m_name and
912         m_demangled_name.
913         (function_info::function_info): Initialize it.
914         (function_info::~function_info): Release it.
915         (main): Rename flag_intermediate_format to flag_json_format.
916         (print_usage): Describe --json-format.
917         (process_args): Set flag_json_format.
918         (output_intermediate_line): Remove.
919         (output_intermediate_json_line): Likewise.
920         (get_gcov_intermediate_filename): Return new extension
921         ".gcov.json.gz".
922         (output_intermediate_file): Implement JSON emission.
923         (output_json_intermediate_file): Implement JSON emission.
924         (generate_results): Use ::get_name for function name.
925         Handle JSON output file.
926         (read_graph_file): Use ::get_name instead of cplus_demangle.
927         (read_count_file): Likewise.
928         (solve_flow_graph): Likewise.
929         (add_line_counts): Likewise.
930         (accumulate_line_counts): Use new flag_json_format.
931         (output_function_details): Use ::get_name instead of cplus_demangle.
932         (output_lines): Likewise.
933         * json.cc (test_writing_literals): Add new tests.
934         * json.h (class literal): Add new boolean constructor.
936 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
938         PR rtl-optimization/87701
939         PR rtl-optimization/87780
940         * combine.c (make_more_copies): Rewrite.
942 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
944         * doc/generic.texi (ABSU_EXPR): Document.
945         * match.pd (absu(x)*absu(x) -> x*x): Handle.
946         (absu(absu(X)) -> absu(X)): Likewise.
947         (absu(-X) -> absu(X)): Likewise.
948         (absu(X)  where X is nonnegative -> X): Likewise.
950 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
952         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
953         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
954         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
955         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
956         (s-tm-texi): Also check timestamp on d-target.def.
957         (generated_files): Add TM_D_H and d-target-hooks-def.h.
958         (build/genhooks.o): Also depend on D_TARGET_DEF.
959         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
960         variables.
961         * config/aarch64/aarch64-d.c: New file.
962         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
963         Define.
964         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
965         prototype.
966         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
967         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
968         * config/arm/arm-d.c: New file.
969         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
970         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
971         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
972         * config/arm/t-arm (arm-d.o): New rule.
973         * config/default-d.c: New file.
974         * config/glibc-d.c: New file.
975         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
976         * config/i386/i386-d.c: New file.
977         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
978         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
979         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
980         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
981         * config/i386/t-i386 (i386-d.o): New rule.
982         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
983         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
984         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
985         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
986         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
987         * config/mips/mips-d.c: New file.
988         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
989         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
990         * config/mips/t-mips (mips-d.o): New rule.
991         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
992         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
993         * config/powerpcspe/powerpcspe-d.c: New file.
994         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
995         New prototype.
996         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
997         Support GNU D by using 0 as the language type.
998         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
999         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
1000         * config/riscv/riscv-d.c: New file.
1001         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
1002         prototype.
1003         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
1004         * config/riscv/t-riscv (riscv-d.o): New rule.
1005         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1006         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1007         * config/rs6000/rs6000-d.c: New file.
1008         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
1009         prototype.
1010         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1011         Support GNU D by using 0 as the language type.
1012         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
1013         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
1014         * config/s390/s390-d.c: New file.
1015         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
1016         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
1017         * config/s390/t-s390 (s390-d.o): New rule.
1018         * config/sparc/sparc-d.c: New file.
1019         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
1020         prototype.
1021         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
1022         * config/sparc/t-sparc (sparc-d.o): New rule.
1023         * config/t-glibc (glibc-d.o): New rule.
1024         * configure: Regenerated.
1025         * configure.ac (tm_d_file): New variable.
1026         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
1027         * doc/contrib.texi (Contributors): Add self for the D frontend.
1028         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
1029         * doc/install.texi (Configuration): Mention libphobos as an option for
1030         --enable-shared.  Mention d as an option for --enable-languages.
1031         (Testing): Mention check-d as a target.
1032         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
1033         name suffixes.  Mention d as a -x option.
1034         * doc/sourcebuild.texi (Top Level): Mention libphobos.
1035         * doc/standards.texi (Standards): Add section on D language.
1036         * doc/tm.texi: Regenerated.
1037         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
1038         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
1039         * dwarf2out.c (is_dlang): New function.
1040         (gen_compile_unit_die): Use DW_LANG_D for D.
1041         (declare_in_namespace): Return module die for D, instead of adding
1042         extra declarations into the namespace.
1043         (gen_namespace_die): Generate DW_TAG_module for D.
1044         (gen_decl_die): Handle CONST_DECLSs for D.
1045         (dwarf2out_decl): Likewise.
1046         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
1047         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
1048         kinds of aggregates.
1049         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
1050         * genhooks.c: Include d/d-target.def.
1052 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
1054         PR target/85669
1055         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
1056         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
1057         (STACK_DYNAMIC_OFFSET): Likewise.
1059 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
1061         PR target/80024
1062         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
1063         error message.
1065 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
1067         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
1068         return type and other typos.
1070 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
1072         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
1073         a shorter sequence with fewer branches.
1074         (emit_final_str_compare_gpr): Ditto.
1076 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1078         * config/rs6000/tmmintrin.h: New file.
1079         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
1081 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1083         * config/rs6000/mmintrin.h: Enable 32bit compilation.
1084         * config/rs6000/xmmintrin.h: Likewise.
1086 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1088         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
1090 2018-10-26  Richard Biener  <rguenther@suse.de>
1092         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
1093         and wrapper.
1094         (vect_mark_slp_stmts_relevant): Likewise.
1095         (vect_detect_hybrid_slp_stmts): Likewise.
1096         (vect_bb_slp_scalar_cost): Likewise.
1097         (vect_remove_slp_scalar_calls): Likewise.
1099 2018-10-26  Jan Hubicka  <jh@suse.cz>
1101         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
1102         (warn_types_mismatch): Fix walk of DECL_NAME.
1103         (odr_types_equivalent_p): Fix overactive assert.
1105 2018-10-26  Richard Biener  <rguenther@suse.de>
1107         PR tree-optimization/87105
1108         * tree-vectorizer.h (_slp_tree::refcnt): New member.
1109         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
1110         refcnt.
1111         (vect_create_new_slp_node): Initialize refcnt to one.
1112         (bst_traits): Move.
1113         (scalar_stmts_set_t, bst_fail): Remove.
1114         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
1115         (vect_build_slp_tree): Add bst_map argument and lookup
1116         already created SLP nodes.
1117         (vect_print_slp_tree): Handle a SLP graph, print SLP node
1118         addresses.
1119         (vect_slp_rearrange_stmts): Handle a SLP graph.
1120         (vect_analyze_slp_instance): Adjust and free SLP nodes from
1121         the CSE map.  Fix indenting.
1122         (vect_schedule_slp_instance): Add short-cut.
1124 2018-10-26  Martin Liska  <mliska@suse.cz>
1126         PR testsuite/86158
1127         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
1128         addr_expr and not with pointers.
1130 2018-10-26  Jan Hubicka  <jh@suse.cz>
1132         * tree.c (free_lang_data_in_type): Only check main variants.
1133         * ipa-devirt.c (warn_odr): Make static.
1134         (types_same_for_odr): Drop strict variant.
1135         (types_odr_comparable): Likewise.
1136         (odr_or_derived_type_p): Look for main variants.
1137         (odr_name_hasher::equal): Cleanup comment.
1138         (odr_subtypes_equivalent): Add warn and warned arguments; check main
1139         variants.
1140         (type_variants_equivalent_p): break out from ...
1141         (odr_types_equivalent): ... here; go for main variants where needed.
1142         (warn_odr): ... here; turn static.
1143         (warn_types_mismatch): Compare mangled names of main variants.
1144         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
1145         (type_with_linkage_p): Sanity check that we look at main variant.
1146         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
1147         * tree.h (types_same_for_odr): Drop strict argument.
1149 2018-10-26  Richard Biener  <rguenther@suse.de>
1151         PR tree-optimization/87746
1152         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1153         Simplify and fix WRT strided store groups with size not
1154         equal to step in element count.
1155         (vect_analyze_group_access_1): Dump the whole group.
1157 2018-10-25  Carl Love  <cel@us.ibm.com>
1159         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
1160         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
1161         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
1162         precicion entry for each overloaded builtin.
1163         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
1164         VSCEDPUO): Rename overloaded name.
1165         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
1166         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
1167         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
1168         define_expand for xscmpexqp instruction.
1169         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
1171 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1172             Jinsong Ji <jji@us.ibm.com>
1174         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
1175         function with vec_sl.
1176         (_mm_slli_epi32): Likewise.
1177         (_mm_slli_epi64): Likewise.
1178         (_mm_srai_epi16): Replace deprecated function with vec_sra.
1179         (_mm_srai_epi32): Likewise.
1180         (_mm_srli_epi16): Replace deprecated function with vec_sr.
1181         (_mm_srli_epi32): Likewise.
1182         (_mm_srli_epi64): Likewise.
1183         (_mm_sll_epi16): Replace deprecated function with vec_sl.
1184         (_mm_sll_epi32): Likewise.
1185         (_mm_sll_epi64): Likewise.
1186         (_mm_sra_epi16): Replace deprecated function with vec_sra.
1187         (_mm_sra_epi32): Likewise.
1188         (_mm_srl_epi16): Replace deprecated function with vec_sr.
1189         (_mm_srl_epi32): Likewise.
1190         (_mm_srl_epi64): Likewise.
1192 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1193             Jinsong Ji <jji@us.ibm.com>
1195         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
1196         comparison operators with vec_cmp* for compatibility due to
1197         unfortunate history; clean up formatting and use types more
1198         appropriately.
1199         (_mm_sll_epi32): Likewise.
1200         (_mm_sll_epi64): Likewise.
1201         (_mm_srl_epi16): Likewise.
1202         (_mm_srl_epi32): Likewise.
1203         (_mm_srl_epi64): Likewise.
1205 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1206             Jinsong Ji <jji@us.ibm.com>
1208         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
1209         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
1210         __vector __bool int.  Use vec_cmpgt in preference to deprecated
1211         function vec_vcmpgtfp.
1212         (_mm_max_ps): Likewise.
1214 2018-10-25  Jeff Law  <law@redhat.com>
1216         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
1217         if returning false.
1219 2018-10-25  Martin Sebor  <msebor@redhat.com>
1221         * doc/extend.texi (aligned): Expand attribute description.
1222         (Alignment): Rename section.  Discuss function arguments.
1224 2018-10-25  Jan Hubicka  <jh@suse.cz>
1226         * ipa-devirt.c (main_odr_variant): Remove.
1227         (hash_odr_name, types_same_for_odr, types_odr_comparable,
1228         odr_name_hasher::equal, odr_subtypes_equivalent_p):
1229         Drop use of main_odr_variant.
1230         (add_type_duplicate): Silence confused warnings on integer types.
1231         (get_odr_type): Always look for main variant.
1232         (register_odr_type): Simplify.
1234 2018-10-25  Richard Biener  <rguenther@suse.de>
1236         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
1237         Initialize ng to silence error with release checking bootstrap.
1239 2018-10-25  Richard Biener  <rguenther@suse.de>
1241         * tree-if-conv.c: Include tree-ssa-sccvn.h.
1242         (tree_if_conversion): Run CSE on the if-converted loop body.
1244 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1246         * config/s390/constraints.md (ZL): New constraint.
1247         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
1248         operands.
1249         * config/s390/s390.md (movdi_larl): Remove.
1250         (movdi_64): Add the LARL alternative.
1252 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1254         PR bootstrap/87747
1255         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
1256         (RTX_CODE_HWINT_P): New macro.
1257         (rtx_code_size): Use RTX_CODE_HWINT_P ().
1259 2018-10-25  Jan Hubicka  <jh@suse.cz>
1261         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
1262         is anonymous.
1264 2018-10-25  Richard Biener  <rguenther@suse.de>
1266         PR tree-optimization/87665
1267         PR tree-optimization/87745
1268         * tree-vectorizer.h (get_earlier_stmt): Remove.
1269         (get_later_stmt): Pick up UID from the original non-pattern stmt.
1271 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
1273         * options.texi (Deprecated): Move list to Var section.
1275 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
1276             Jinsong Ji <jji@us.ibm.com>
1278         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
1279         __vector long to __vector long long.
1280         (_mm_cvtpd_ps): Likewise.
1281         (_mm_cvttpd_epi32): Likewise.
1282         (_mm_cvtpi32_pd): Likewise.
1283         (_mm_unpackhi_epi64): Likewise.
1284         (_mm_unpacklo_epi64): Likewise.
1286 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
1288         PR rtl-optimization/87720
1289         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
1291 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
1293         * gimple-ssa-isolate-paths.c
1294         (find_implicit_erroneous_behavior): Do not change code if the
1295         pass is running for warnings only.
1296         (find_explicit_erroneous_behavior): Likewise.
1298 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
1300         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
1301         Define as rs6000_mangle_decl_assembler_name.
1302         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
1303         long double to IEEE long double, switch the names of the long
1304         double built-in functions to be <func>f128 instead of <func>l.
1306 2018-10-24  Martin Sebor  <msebor@redhat.com>
1308         * doc/extend.texi (nonnull): List no-argument form.  Reference
1309         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
1311 2018-10-24  Richard Biener  <rguenther@suse.de>
1313         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
1315 2018-10-24  Martin Liska  <mliska@suse.cz>
1317         PR tree-optimization/84436
1318         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
1319         Remove.
1320         (switch_conversion::contains_linear_function_p): New.
1321         (switch_conversion::build_one_array): Support linear
1322         transformation on input.
1323         * tree-switch-conversion.h (struct switch_conversion): Add
1324         contains_linear_function_p declaration.
1326 2018-10-24  Richard Biener  <rguenther@suse.de>
1328         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
1329         if its argument is CONSTANT_CLASS_P.
1331 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1333         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
1334         it is wrong for forward declarations.
1336 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1338         * config/s390/s390.c (s390_check_qrst_address): Add the missing
1339         SYMBOL_REF_P () check.
1341 2018-10-24  Richard Biener  <rguenther@suse.de>
1343         PR tree-optimization/87105
1344         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
1345         dump classification.
1346         (vect_analyze_data_ref_accesses): Handle duplicate loads and
1347         stores by splitting the affected group after the fact.
1348         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
1349         fail the SLP build because of size constraints.
1351 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1353         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
1354         * configure: Regenerate.
1355         * config.in: Regenerate.
1356         * varasm.c (mergeable_string_section): Use readonly_data_section
1357         if linker doesn't support SHF_MERGE with alignment > 8.
1358         (mergeable_constant_section): Likewise.
1360 2018-10-24  Richard Biener  <rguenther@suse.de>
1362         PR tree-optimization/84013
1363         * tree-ssa-structalias.c (struct msdi_data): New struct for
1364         marshalling data to walk_stmt_load_store_ops.
1365         (maybe_set_dependence_info): Refactor as callback for
1366         walk_stmt_load_store_ops.
1367         (compute_dependence_clique): Set restrict info on all stmt kinds.
1369 2018-10-24  Martin Liska  <mliska@suse.cz>
1371         * cgraph.c (cgraph_node::dump):
1372         Remove reduntant dumps and make tp_first_run dump more compact.
1374 2018-10-24  Richard Biener  <rguenther@suse.de>
1376         PR tree-optimization/87665
1377         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
1378         to reflect reality.
1380 2018-10-12  Jeff Law  <law@redhat.com>
1382         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
1383         for H8/S.
1385 2018-10-23  Richard Biener  <rguenther@suse.de>
1387         * tree-vrp.c (add_assert_info): Guard dump_printf with
1388         dump_enabled_p.
1389         * gimple-ssa-evrp-analyze.c
1390         (evrp_range_analyzer::record_ranges_from_incoming_edge):
1391         Use value_range::ignore_equivs_equal_p.
1393 2018-10-23  Richard Biener  <rguenther@suse.de>
1395         PR tree-optimization/87105
1396         PR tree-optimization/87608
1397         * passes.def (pass_all_early_optimizations): Add early phi-opt
1398         after dce.
1399         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
1400         addition to debug stmts.
1401         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
1402         and abs replacement early.
1403         * tree-cfg.c (gimple_empty_block_p): Likewise.
1405 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
1407         PR target/86383
1408         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
1409         specified to configure.
1410         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
1412 2018-10-23  Richard Biener  <rguenther@suse.de>
1414         PR tree-optimization/87700
1415         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
1417 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
1419         PR target/87674
1420         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
1421         second argument from __mmask16 to __mmask8.
1422         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
1423         _mm_mask_packs_epi32): Likewise.
1424         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
1425         Likewise.
1426         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
1428 2018-10-23  Richard Biener  <rguenther@suse.de>
1430         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
1432 2018-10-23  Richard Biener  <rguenther@suse.de>
1434         PR tree-optimization/86144
1435         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
1436         over simd attribute.
1438 2018-10-23  Richard Biener  <rguenther@suse.de>
1440         PR tree-optimization/87693
1441         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
1442         the case we do not find the taken edge.
1444 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
1445             Jinsong Ji  <jji@us.ibm.com>
1447         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
1448         (_mm_store_pd): Use unaligned vector type for pointer cast.
1449         (_mm_maskmoveu_si128): Likewise.
1450         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
1451         (_mm_store_ps): Use unaligned vector type for pointer cast.
1453 2018-10-22  Paul Koning  <ni1d@arrl.net>
1455         * symtab.c (symtab_node::increase_alignment): Correct max
1456         alignment check.
1458 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
1460         PR tree-optimization/87633
1461         * match.pd: Do not generate unordered integer comparisons.
1463 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1465         PR rtl-optimization/87600
1466         * combine.c: Add include of expr.h.
1467         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
1468         register to a pseudo.
1469         (make_more_copies): New function, add a copy to a new pseudo after
1470         the moves from hard registers into pseudos.
1471         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
1472         later.  Call make_more_copies.
1474 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
1476         * lra-constraints.c (process_alt_operands): New local array,
1477         matching_early_clobber.  Check matching_early_clobber before
1478         decrementing reject, and set matching_early_clobber after.
1480 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1482         PR target/87598
1483         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
1484         call output_addr_const and hope for the best.
1486 2018-10-22  Richard Biener  <rguenther@suse.de>
1488         * gimple-ssa-evrp-analyze.c
1489         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
1490         smarter about what ranges to use.
1491         * tree-vrp.c (add_assert_info): Dump here.
1492         (register_edge_assert_for_2): Instead of here at multiple but
1493         not all places.
1495         * gcc.dg/tree-ssa/evrp12.c: New testcase.
1496         * gcc.dg/predict-6.c: Adjust.
1497         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
1498         * gcc.dg/tree-ssa/vrp02.c: Likewise.
1499         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
1501 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
1502         Richard Biener  <rguenther@suse.de>
1504         * bitmap.h: Update data structure documentation, including a
1505         description of bitmap views as either linked-lists or splay trees.
1506         (struct bitmap_element_def): Update comments for splay tree bitmaps.
1507         (struct bitmap_head_def): Likewise.
1508         (bitmap_list_view, bitmap_tree_view): New prototypes.
1509         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
1510         tree_form fields.
1511         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
1512         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
1513         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
1514         released bitmap element here.
1515         (bitmap_element_free): Remove.
1516         (bitmap_elt_clear_from): Work on splay tree bitmaps.
1517         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
1518         this function similar ones such that linked-list bitmap implementation
1519         functions are grouped.
1520         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
1521         and moved for grouping.
1522         (bitmap_list_insert_element_after): Renamed from
1523         bitmap_elt_insert_after, and moved for grouping.
1524         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
1525         (bitmap_tree_link_left, bitmap_tree_link_right,
1526         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
1527         bitmap_tree_link_element, bitmap_tree_unlink_element,
1528         bitmap_tree_find_element): New functions for splay-tree bitmap
1529         implementation.
1530         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
1531         Renamed and moved, see above entries.
1532         (bitmap_tree_listify_from): New function to convert part of a splay
1533         tree bitmap to a linked-list bitmap.
1534         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
1535         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
1536         (bitmap_find_bit): Remove.
1537         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
1538         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
1539         Handle splay tree bitmaps.
1540         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
1541         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
1542         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
1543         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
1544         bitmap_intersect_compl_p, bitmap_ior_and_compl,
1545         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
1546         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
1547         corresponding changes to use linked-list specific bitmap_element
1548         manipulation functions as applicable for efficiency.
1549         (bitmap_tree_to_vec): New function.
1550         (debug_bitmap_elt_file): New function split out from ...
1551         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
1552         (bitmap_print): Likewise.
1554         PR tree-optimization/63155
1555         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
1556         SSA edge worklists.
1557         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
1558         in tree-view.
1560 2018-10-22  Martin Liska  <mliska@suse.cz>
1562         PR tree-optimization/87686
1563         Revert
1564         2018-08-29  Martin Liska  <mliska@suse.cz>
1566         * tree-switch-conversion.c (switch_conversion::expand):
1567         Strenghten assumption about gswitch statements.
1569 2018-10-22  Martin Liska  <mliska@suse.cz>
1571         * ipa-icf.c (sem_item::compare_attributes): Remove.
1572         (sem_item::compare_referenced_symbol_properties): Use
1573         attribute_list_equal instead.
1574         (sem_function::equals_wpa): Likewise.
1575         * ipa-icf.h: Remove compare_attributes.
1577 2018-10-22  Richard Biener  <rguenther@suse.de>
1579         PR middle-end/87682
1580         * mem-stats.h (mem_usage::operator==): Fix pasto.
1582 2018-10-22  Richard Biener  <rguenther@suse.de>
1584         PR tree-optimization/87640
1585         * tree-vrp.c (set_value_range_with_overflow): Decompose
1586         incomplete result.
1587         (extract_range_from_binary_expr_1): Adjust.
1589 2018-10-22  Martin Jambor  <mjambor@suse.cz>
1591         * tree-eh.h (stmt_could_throw_p): Add function parameter.
1592         (stmt_can_throw_external): Likewise.
1593         (stmt_can_throw_internal): Likewise.
1594         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
1595         (lower_eh_constructs_2): Likewise.
1596         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
1597         (stmt_can_throw_external): Likewise.
1598         (stmt_can_throw_internal): Likewise.
1599         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
1600         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
1601         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
1602         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
1603         (pass_lower_eh_dispatch::execute): Pass cfun to
1604         stmt_can_throw_external.
1605         (cleanup_empty_eh): Likewise.
1606         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
1607         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
1608         stmt_can_throw_external instead of pushing it to cfun.
1609         (symbol_table::create_edge): Likewise.
1610         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
1611         stmt_can_throw_internal.
1612         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
1613         to stmt_could_throw_p.
1614         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
1615         stmt_can_throw_internal.
1616         (pass_store_merging::execute): Likewise.
1617         * gimple-ssa-strength-reduction.c
1618         (find_candidates_dom_walker::before_dom_children): Pass cfun to
1619         stmt_could_throw_p.
1620         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
1621         stmt_can_throw_internal.
1622         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1623         to stmt_can_throw_external.
1624         (check_stmt): Pass cfun to stmt_could_throw_p.
1625         (check_stmt): Pass cfun to stmt_can_throw_external.
1626         (pass_nothrow::execute): Likewise.
1627         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1628         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1629         stmt_can_throw_internal.
1630         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1631         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1632         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1633         * tree-complex.c (expand_complex_libcall): Pass cfun to
1634         stmt_could_throw_p and to stmt_can_throw_internal.
1635         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1636         * tree-inline.c (copy_edges_for_bb): Likewise.
1637         (maybe_move_debug_stmts_to_successors): Likewise.
1638         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1639         stmt_could_throw_p.
1640         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1641         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1642         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1643         stmt_can_throw_internal.
1644         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1645         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1646         stmt_could_throw_p.
1647         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1648         stmt_can_throw_internal.
1649         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1650         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1651         stmt_could_throw_p.
1652         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1653         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1654         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1655         (convert_mult_to_fma_1): Likewise.
1656         (convert_to_divmod): Likewise.
1657         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1658         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1659         * tree-ssa-propagate.c
1660         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1661         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1662         (maybe_optimize_range_tests): Likewise.
1663         (linearize_expr_tree): Likewise.
1664         (reassociate_bb): Likewise.
1665         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1666         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1667         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1668         (handle_char_store): Likewise.
1669         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1670         stmt_can_throw_internal.
1671         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1672         stmt_could_throw_p.
1673         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1674         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1675         (vectorizable_simd_clone_call): Likewise.
1676         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1677         (gimple_stringop_fixed_value): Likewise.
1679 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
1681         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1682         literal pool references.
1683         (s390_check_qrst_address): Adapt to the new behavior of
1684         s390_loadrelative_operand_p ().
1686 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1688         PR target/72782
1689         * config/i386/sse.md (*andnot<mode>3_bcst): New.
1691 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1693         PR target/72782
1694         * config/i386/sse.md (*<code><mode>3_bcst): New.
1696 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1698         PR target/72782
1699         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1700         V4DI, V16SI and V8DI.
1701         (*sub<mode>3<mask_name>_bcst): New.
1702         (*add<mode>3<mask_name>_bcst): Likewise.
1704 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
1705             Jinsong Ji  <jji@us.ibm.com>
1707         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1708         __m64 with __vector unsigned long long for compatibility.
1709         (_mm_movemask_epi8): Likewise.
1710         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1711         (_mm_cvttps_pi32): Likewise.
1712         (_mm_cvtpi32_ps): Likewise.
1713         (_mm_cvtps_pi16): Likewise.
1714         (_mm_loadh_pi): Likewise.
1715         (_mm_storeh_pi): Likewise.
1716         (_mm_movehl_ps): Likewise.
1717         (_mm_movelh_ps): Likewise.
1718         (_mm_loadl_pi): Likewise.
1719         (_mm_storel_pi): Likewise.
1720         (_mm_movemask_ps): Likewise.
1721         (_mm_shuffle_pi16): Likewise.
1723 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1725         PR target/72782
1726         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1727         __builtin_ia32_vfnmsubpd512_mask.
1728         (_mm512_mask_fnmsub_round_pd): Likewise.
1729         (_mm512_fnmsub_pd): Likewise.
1730         (_mm512_mask_fnmsub_pd): Likewise.
1731         (_mm512_maskz_fnmsub_round_pd): Use
1732         __builtin_ia32_vfnmsubpd512_maskz.
1733         (_mm512_maskz_fnmsub_pd): Likewise.
1734         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1735         (_mm512_mask_fnmsub_round_ps): Likewise.
1736         (_mm512_fnmsub_ps): Likewise.
1737         (_mm512_mask_fnmsub_ps): Likewise.
1738         (_mm512_maskz_fnmsub_round_ps): Use
1739         __builtin_ia32_vfnmsubps512_maskz.
1740         (_mm512_maskz_fnmsub_ps): Likewise.
1741         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1742         __builtin_ia32_vfnmsubpd256_mask.
1743         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1744         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1745         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1746         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1747         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1748         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1749         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1750         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1751         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1752         __builtin_ia32_vfnmsubpd.
1753         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1754         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1755         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1756         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1757         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1758         * config/i386/i386-builtin.def: Add
1759         __builtin_ia32_vfnmsubpd256_mask,
1760         __builtin_ia32_vfnmsubpd256_maskz,
1761         __builtin_ia32_vfnmsubpd128_mask,
1762         __builtin_ia32_vfnmsubpd128_maskz,
1763         __builtin_ia32_vfnmsubps256_mask,
1764         __builtin_ia32_vfnmsubps256_maskz,
1765         __builtin_ia32_vfnmsubps128_mask,
1766         __builtin_ia32_vfnmsubps128_maskz,
1767         __builtin_ia32_vfnmsubpd512_mask,
1768         __builtin_ia32_vfnmsubpd512_maskz,
1769         __builtin_ia32_vfnmsubps512_mask,
1770         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1771         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1772         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1773         __builtin_ia32_vfnmsubpd256.
1774         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1775         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1776         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1777         Likewise.
1778         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1779         Likewise.
1780         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1781         Likewise.
1782         (fmai_vmfnmsub_<mode><round_name>): Likewise.
1784 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1786         PR target/72782
1787         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1788         __builtin_ia32_vfnmaddpd512_mask.
1789         (_mm512_mask_fnmadd_round_pd): Likewise.
1790         (_mm512_fnmadd_pd): Likewise.
1791         (_mm512_mask_fnmadd_pd): Likewise.
1792         (_mm512_maskz_fnmadd_round_pd): Use
1793         __builtin_ia32_vfnmaddpd512_maskz.
1794         (_mm512_maskz_fnmadd_pd): Likewise.
1795         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1796         (_mm512_mask_fnmadd_round_ps): Likewise.
1797         (_mm512_fnmadd_ps): Likewise.
1798         (_mm512_mask_fnmadd_ps): Likewise.
1799         (_mm512_maskz_fnmadd_round_ps): Use
1800         __builtin_ia32_vfnmaddps512_maskz.
1801         (_mm512_maskz_fnmadd_ps): Likewise.
1802         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1803         __builtin_ia32_vfnmaddpd256_mask.
1804         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1805         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1806         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1807         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1808         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1809         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1810         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1811         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1812         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1813         __builtin_ia32_vfnmaddpd.
1814         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1815         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1816         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1817         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1818         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1819         * config/i386/i386-builtin.def: Add
1820         __builtin_ia32_vfnmaddpd256_mask,
1821         __builtin_ia32_vfnmaddpd256_maskz,
1822         __builtin_ia32_vfnmaddpd128_mask,
1823         __builtin_ia32_vfnmaddpd128_maskz,
1824         __builtin_ia32_vfnmaddps256_mask,
1825         __builtin_ia32_vfnmaddps256_maskz,
1826         __builtin_ia32_vfnmaddps128_mask,
1827         __builtin_ia32_vfnmaddps128_maskz,
1828         __builtin_ia32_vfnmaddpd512_mask,
1829         __builtin_ia32_vfnmaddpd512_maskz,
1830         __builtin_ia32_vfnmaddps512_mask,
1831         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1832         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1833         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1834         __builtin_ia32_vfnmaddpd256.
1835         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1836         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1837         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1838         Likewise.
1839         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1840         Likewise.
1841         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1842         Likewise.
1843         (fmai_vmfnmadd_<mode><round_name>): Likewise.
1845 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1847         PR target/72782
1848         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1849         __builtin_ia32_vfmsubpd512_mask.
1850         (_mm512_mask_fmsub_round_pd): Likewise.
1851         (_mm512_fmsub_pd): Likewise.
1852         (_mm512_mask_fmsub_pd): Likewise.
1853         (_mm512_maskz_fmsub_round_pd): Use
1854         __builtin_ia32_vfmsubpd512_maskz.
1855         (_mm512_maskz_fmsub_pd): Likewise.
1856         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1857         (_mm512_mask_fmsub_round_ps): Likewise.
1858         (_mm512_fmsub_ps): Likewise.
1859         (_mm512_mask_fmsub_ps): Likewise.
1860         (_mm512_maskz_fmsub_round_ps): Use
1861         __builtin_ia32_vfmsubps512_maskz.
1862         (_mm512_maskz_fmsub_ps): Likewise.
1863         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1864         __builtin_ia32_vfmsubpd256_mask.
1865         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1866         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1867         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1868         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1869         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1870         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1871         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1872         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1873         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1874         __builtin_ia32_vfmsubpd.
1875         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1876         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1877         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1878         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1879         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1880         * config/i386/i386-builtin.def: Add
1881         __builtin_ia32_vfmsubpd256_mask,
1882         __builtin_ia32_vfmsubpd256_maskz,
1883         __builtin_ia32_vfmsubpd128_mask,
1884         __builtin_ia32_vfmsubpd128_maskz,
1885         __builtin_ia32_vfmsubps256_mask,
1886         __builtin_ia32_vfmsubps256_maskz,
1887         __builtin_ia32_vfmsubps128_mask,
1888         __builtin_ia32_vfmsubps128_maskz,
1889         __builtin_ia32_vfmsubpd512_mask,
1890         __builtin_ia32_vfmsubpd512_maskz,
1891         __builtin_ia32_vfmsubps512_mask,
1892         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1893         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1894         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1895         __builtin_ia32_vfmsubpd256.
1896         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1897         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1898         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1899         Likewise.
1900         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1901         Likewise.
1902         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1903         Likewise.
1904         (fmai_vmfmsub_<mode><round_name>): Likewise.
1906 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1908         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1909         Remove plus.  Renamed to ...
1910         (*sub<mode>3<mask_name>_bcst): This.
1911         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1912         (*add<mode>3<mask_name>_bcst): This.
1914 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1916         PR target/72782
1917         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1919 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1921         PR target/87662
1922         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1923         (_mm_or_epi32): Likewise.
1924         (_mm256_xor_epi32): Likewise.
1925         (_mm_xor_epi32): Likewise.
1926         (_mm256_or_epi64): Likewise.
1927         (_mm_or_epi64): Likewise.
1928         (_mm256_xor_epi64): Likewise.
1929         (_mm_xor_epi64): Likewise.
1931 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
1933         PR target/72782
1934         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1936 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
1938         PR middle-end/87647
1939         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1941 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
1943         * doc/ux.texi: Move @section directly after @node.
1945 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
1947         PR middle-end/85488
1948         PR middle-end/87649
1949         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1950         depend closely nested inside of loop with ordered clause with
1951         a parameter.
1953 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
1955         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1956         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1957         * doc/ux.texi: New file.
1959 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
1961         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1962         be the first CR field allocated.
1964 2018-10-19  Richard Biener  <rguenther@suse.de>
1966         PR target/87657
1967         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1968         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1970 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1972         PR target/72782
1973         * config/i386/sse.md
1974         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1975         (*add<mode>3<mask_name>_bcst_2): Likewise.
1977 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1979         * config/i386/sse.md
1980         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1981         Replace nonimmediate_operand with register_operand.
1982         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1983         Likewise.
1984         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1985         Likewise.
1987 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
1989         PR rtl-optimization/87596
1990         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1991         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1992         for instructions in FROM..TO range.
1994 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
1996         * cfgexpand.c (expand_one_var): Use specific wording in error message
1997         for non-local frame variables.
1998         * stor-layout.c (layout_decl): Do not issue a warning for them.
2000 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
2002         * haifa-sched.c (priority): Add force_recompute parameter.
2003         (apply_replacement): Call priority () with force_recompute = true.
2004         (restore_pattern): Likewise.
2006 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2008         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
2009         HOST_BITS_PER_WIDE_INT.
2010         (test_vector_ops_duplicate): Likewise.
2012 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2014         PR target/72782
2015         * config/i386/sse.md (VF_AVX512): New.
2016         (avx512bcst): Likewise.
2017         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
2018         Likewise.
2019         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
2020         Likewise.
2021         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
2022         Likewise.
2024 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
2026         * doc/invoke.texi (-dumpversion): Improve grammar.
2027         (-dumpfullversion): Make more consistent with -dumpversion.
2029 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
2031         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
2032         Set JUMP_LABEL to the jump insn.
2033         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
2034         Predict emitted jump and add label to jump insn.
2036 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
2038         PR tree-optimization/87562
2039         * input.c (get_substring_ranges_for_loc): Use
2040         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
2041         getting the linemap for the endpoint.  Verify that it's either
2042         in the same linemap as the start point's spelling location, or
2043         at least in the same file.
2045 2018-10-18  Richard Biener  <rguenther@suse.de>
2047         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
2048         feed width-specific load/store costs through ix86_vec_cost.
2049         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
2050         (k8_cost): Likewise.
2051         (bdver_cost): Likewise.
2052         (znver1_cost): Likewise.
2053         (btver1_cost): Likewise.
2054         (btver2_cost): Likewise.
2056 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2058         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
2059         to simplify subreg of vec_merge.
2061 2018-10-18  Richard Biener  <rguenther@suse.de>
2063         * config/i386/i386.c: Fix costing of vector FMA.
2065 2018-10-18  Richard Biener  <rguenther@suse.de>
2067         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
2068         and argument.
2069         (ix86_builtin_vectorization_cost): For vec_construct properly
2070         cost insertion into SSE regs.
2071         (...): Adjust calls to ix86_vec_cost.
2073 2018-10-18  Richard Biener  <rguenther@suse.de>
2075         PR middle-end/87087
2076         Revert
2077         2018-02-07  Richard Biener  <rguenther@suse.de>
2079         PR tree-optimization/84204
2080         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2081         this place.
2083 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2085         PR target/87537
2086         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
2087         of vec_duplicate.
2088         (test_vector_ops_duplicate): Add test for a scalar subreg of a
2089         VEC_MERGE of a VEC_DUPLICATE.
2091 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
2093         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
2094         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
2095         * doc/standards.texi (C Language): Document C2X.
2096         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
2097         (rl78_option_override): Handle "GNU C2X" language name.
2099 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
2101         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
2102         Document C17 as published in 2018.
2104 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
2106         PR middle-end/87623
2107         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
2108         bail out if both sides do not have the same storage order.
2110 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2112         * bitmap.c (bitmap_head::dump): New.
2113         * bitmap.h (bitmap_head): Add dump().
2114         * gimple-ssa-evrp-analyze.c
2115         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
2116         (evrp_range_analyzer::set_ssa_range_info): Same.
2117         (evrp_range_analyzer::record_ranges_from_phis): Same.
2118         (evrp_range_analyzer::record_ranges_from_stmt): Same.
2119         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
2120         * gimple-ssa-sprintf.c (get_int_range): Same.
2121         (format_integer): Same.
2122         (sprintf_dom_walker::handle_gimple_call): Same.
2123         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
2124         (ipcp_vr_lattice::top_p): Same.
2125         (ipcp_vr_lattice::bottom_p): Same.
2126         (ipcp_vr_lattice::set_to_bottom): Same.
2127         (ipa_vr_operation_and_type_effects): Same.
2128         (propagate_vr_across_jump_function): Same.
2129         (ipcp_store_vr_results): Same.
2130         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
2131         (ipa_print_node_jump_functions_for_edge): Same.
2132         (ipa_get_value_range): Same.
2133         (ipa_compute_jump_functions_for_edge): Same.
2134         (ipa_write_jump_function): Same.
2135         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
2136         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2137         Same.
2138         * vr-values.c (set_value_range_to_nonnegative): Same.
2139         (set_value_range_to_truthvalue): Same.
2140         (vr_values::get_value_range): Same.
2141         (vr_values::set_defs_to_varying): Same.
2142         (vr_values::update_value_range): Same.
2143         (symbolic_range_based_on_p): Same.
2144         (vr_values::op_with_boolean_value_range_p): Same.
2145         (vr_values::extract_range_for_var_from_comparison_expr): Same.
2146         (vr_values::extract_range_from_ssa_name): Same.
2147         (vr_values::extract_range_from_binary_expr): Same.
2148         (vr_values::extract_range_from_unary_expr): Same.
2149         (vr_values::extract_range_from_cond_expr): Same.
2150         (vr_values::extract_range_from_comparison): Same.
2151         (vr_values::check_for_binary_op_overflow): Same.
2152         (vr_values::extract_range_basic): Same.
2153         (vr_values::extract_range_from_assignment): Same.
2154         (compare_ranges): Same.
2155         (compare_range_with_value): Same.
2156         (vr_values::adjust_range_with_scev): Same.
2157         (vrp_valueize): Same.
2158         (vrp_valueize_1): Same.
2159         (vr_values::get_vr_for_comparison): Same.
2160         (vr_values::compare_name_with_value): Same.
2161         (vr_values::compare_names): Same.
2162         (vr_values::vrp_evaluate_conditional): Same.
2163         (find_case_label_ranges): Same.
2164         (vr_values::vrp_visit_switch_stmt): Same.
2165         (vr_values::extract_range_from_phi_node): Same.
2166         (vr_values::simplify_div_or_mod_using_ranges): Same.
2167         (vr_values::simplify_bit_ops_using_ranges): Same.
2168         (test_for_singularity): Same.
2169         (range_fits_type_p): Same.
2170         (vr_values::simplify_cond_using_ranges_1): Same.
2171         (vr_values::simplify_switch_using_ranges): Same.
2172         (vr_values::simplify_float_conversion_using_ranges): Same.
2173         (vr_values::two_valued_val_range_p): Same.
2174         (vr_values::add_equivalence): Move to value_range::equiv_add.
2175         * vr-values.h (vr_values::add_equivalence): Remove.
2176         (VR_INITIALIZER): Remove.
2177         * tree-vrp.c (value_range::set): New.
2178         (value_range::equiv_add): New.
2179         (value_range::value_range): New.
2180         (value_range::deep_copy): New.
2181         (value_range::check): New.
2182         (value_range::equal_p): New.
2183         (value_range::ignore_equivs_equal_p): New.
2184         (value_range::operator==): New.
2185         (value_range::operator!=): New.
2186         (value_range::symbolic_p): New.
2187         (value_range::numeric_p): New.
2188         (value_range::set_undefined): New.
2189         (value_range::set_varying): New.
2190         (value_range::may_contain_p): New.
2191         (value_range::equiv_clear): New.
2192         (value_range::singleton_p): New.
2193         (value_range::intersect): New.
2194         (value_range::dump): New.
2195         (value_range::set_and_canonicalize): New.
2196         (set_value_range): Adjust for value_range API.
2197         (set_value_range_to_undefined): Same.
2198         (set_value_range_to_varying): Same.
2199         (set_and_canonicalize_value_range): Same.
2200         (set_value_range_to_nonnull): Same.
2201         (set_value_range_to_null): Same.
2202         (range_is_null): Same.
2203         (range_is_nonnull): Same.
2204         (range_int_cst_p): Same.
2205         (range_int_cst_singleton_p): Same.
2206         (symbolic_range_p): Same.
2207         (range_includes_zero_p): Same.
2208         (value_range_constant_singleton): Same.
2209         (vrp_set_zero_nonzero_bits): Same.
2210         (ranges_from_anti_range): Same.
2211         (extract_range_into_wide_ints): Same.
2212         (extract_range_from_multiplicative_op): Same.
2213         (set_value_range_with_overflow): Same.
2214         (extract_range_from_binary_expr_1): Same.
2215         (extract_range_from_unary_expr): Same.
2216         (dump_value_range): Same.
2217         (debug_value_range): Same.
2218         (vrp_prop::check_array_ref): Same.
2219         (vrp_prop::check_mem_ref): Same.
2220         (vrp_prop::vrp_initialize): Same.
2221         (vrp_prop::visit_stmt): Same.
2222         (intersect_ranges): Same.
2223         (vrp_prop::visit_phi): Same.
2224         (vrp_prop::vrp_finalize): Same.
2225         (determine_value_range_1): Same.
2226         (determine_value_range): Same.
2227         (vrp_intersect_ranges_1): Rename to...
2228         (vrp_intersect_1): this.
2229         (vrp_intersect_ranges): Rename to...
2230         (value_range::intersect_helper): ...this.
2231         (vrp_meet_1): Rename to...
2232         (value_range::union_helper): ...this.
2233         (vrp_meet): Rename to...
2234         (value_range::union_): ...this.
2235         (copy_value_range): Remove.
2236         * tree-vrp.h (struct value_range): Rewrite into a proper class.
2237         (value_range::vrtype): New.
2238         (value_range::type): New.
2239         (value_range::equiv): New.
2240         (value_range::min): New.
2241         (value_range::max): New.
2242         (value_range::varying_p): New.
2243         (value_range::undefined_p): New.
2244         (value_range::null_p): New.
2245         (value_range::equiv_add): New.
2246         (copy_value_range): Remove.
2248 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
2250         * Makefile.in (SELFTEST_TARGETS): New.
2251         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
2252         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
2253         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
2254         c/Make-lang.in.
2255         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
2256         (selftest-c++-gdb, selftest-c++-valgrind): Move to
2257         cp/Make-lang.in.
2258         * configure: Regenerate.
2259         * configure.ac (selftest_languages): New.
2261 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2263         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
2264         overflow wraps argument.
2265         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
2266         wide_int_range_multiplicative_op.
2267         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
2268         overflow wraps argument.
2269         (wide_int_range_multiplicative_op): Same.
2270         (wide_int_range_lshift): Same.
2271         (wide_int_range_div): Same.
2272         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
2273         (wide_int_range_lshift): Same.
2274         (wide_int_range_div): Same.
2276 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2278         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
2279         use sign as argument.
2280         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
2281         wide_int_range_shift_undefined_p.
2283 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2285         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
2286         Rename to...
2287         (@despeculate_copy<ALLI_TI:mode>): ... This.
2288         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
2289         switch statement.
2291 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2293         * config.gcc: Obsolete *-*-solaris2.10*.
2294         * doc/install.texi (Specific, *-*-solaris2*): Document it.
2296 2018-10-12  Jeff Law  <law@redhat.com>
2298         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
2299         reg + sym +- const_int addressing modes.
2301 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2303         * common.opt (fdiagnostics-minimum-margin-width=): New option.
2304         * diagnostic-show-locus.c (layout::layout): Apply the minimum
2305         margin width.
2306         (layout::start_annotation_line): Only print up to 3 of the
2307         margin character, to avoid touching the left-hand side.
2308         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
2309         minimum margin width, as set by test_diagnostic_context's ctor.
2310         (selftest::test_fixit_insert_containing_newline): Likewise.
2311         (selftest::test_fixit_insert_containing_newline_2): Likewise.
2312         (selftest::test_line_numbers_multiline_range): Clear
2313         dc.min_margin_width.
2314         * diagnostic.c (diagnostic_initialize): Initialize
2315         min_margin_width.
2316         * diagnostic.h (struct diagnostic_context): Add field
2317         "min_margin_width".
2318         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
2319         * opts.c (common_handle_option): Handle
2320         OPT_fdiagnostics_minimum_margin_width_.
2321         * selftest-diagnostic.c
2322         (selftest::test_diagnostic_context::test_diagnostic_context):
2323         Initialize min_margin_width to 6.
2324         * toplev.c (general_init): Initialize global_dc->min_margin_width.
2326 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2328         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2329         Fix usage of "error_at_rich_loc" in the comment.
2331 2018-10-15  Renlin Li  <renlin.li@arm.com>
2333         PR target/87563
2334         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
2335         if-conversioned loop when it contains ifn with types not
2336         supported by backend.
2337         * internal-fn.c (expand_direct_optab_fn): Add an assert.
2338         (direct_internal_fn_supported_p): New helper function.
2339         * internal-fn.h (direct_internal_fn_supported_p): Declare.
2341 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
2343         PR target/87572
2344         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2345         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2346         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2348 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
2350         PR tree-optimization/87022
2351         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
2352         bits in dist vector rather than the first one.
2354 2018-10-15  Richard Biener  <rguenther@suse.de>
2356         PR middle-end/87610
2357         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2358         (visit_loadstore): When a used restrict tag escaped verify that
2359         the points-to solution of "other" pointers do not include
2360         escaped.
2361         (compute_dependence_clique): If a used restrict tag escaped
2362         communicated that down to visit_loadstore.
2364 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
2366         * config/s390/s390.c (s390_expand_vec_init): Force vector element
2367         into reg if it isn't a general operand.
2369 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2371         PR target/87599
2372         * config/i386/sse.md (*vec_dupv2di): Add register source to
2373         movddup.
2375 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2377         PR target/87572
2378         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2379         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
2380         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2381         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2383 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
2385         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
2386         (notice_args_size): Set it in the current trace if no insn that can
2387         throw internally has been seen yet.
2388         (connect_traces): When connecting args_size between traces, allow the
2389         incoming values not to match if there is an insn setting it before the
2390         first insn that can throw internally; in that case, force the creation
2391         of a CFI note on this latter insn.
2393 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
2395         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
2396         the base class, because using the injected-class-name was not clearly
2397         specified until DR 176.
2399 2018-10-12  Paul Koning  <ni1d@arrl.net>
2401         * config/pdp11/pdp11.md (doloop_end): New expander.
2402         (doloop_end_insn): renamed from "doloop_end".
2403         (addqi3): New pattern.
2404         (subqi3): New pattern.
2405         * config/pdp11/predicates.md (incdec_operand): New predicate.
2407 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
2409         PR middle-end/81376
2410         * real.c (format_helper::can_represent_integral_type_p): New function
2411         * real.h (format_helper::can_represent_integral_type_p): Ditto.
2412         * match.pd: New pattern.
2414 2018-10-12  Alexandre Oliva <oliva@adacore.com>
2416         * configure.ac: Introduce --enable-large-address-aware
2417         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
2418         * doc/install.texi: Document it.
2419         * configure, config.in: Rebuilt.
2420         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
2421         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
2422         (LINK_SPEC): Insert it.
2423         * config/i386/mingw-w64.h: Likewise.
2425         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
2427 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
2429         PR rtl-optimization/87600
2430         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
2432 2018-10-12  Paul Koning  <ni1d@arrl.net>
2434         * doc/md.texi (doloop_end): Document that the pattern code may
2435         need to check operand mode.
2437 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
2439         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
2440         to zero-extend between int and floating-point registers.
2441         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
2442         ldp into floating-point registers.  Add type and arch attributes.
2443         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
2444         Use f_loads for type attribute.
2446 2018-10-11  Martin Sebor  <msebor@redhat.com>
2448         * doc/extend.texi (attribute packed): Correct typos.
2450 2018-10-11  Martin Sebor  <msebor@redhat.com>
2452         * doc/extend.texi (attribute flatten): Mention interaction with
2453         noinline.
2455 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
2457         PR target/87156
2458         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2459         Set new_decl virtual flag to zero.
2461 2018-10-11  Martin Sebor  <msebor@redhat.com>
2463         PR middle-end/87593
2464         * doc/extend.texi (attribute format_arg): Discuss using multiple
2465         attributes on a single function.
2467 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
2469         PR tree-optimization/86829
2470         * match.pd (sin (atan (x))): New simplification rules.
2471         (cos (atan (x))): Likewise.
2472         * real.c (build_sinatan_real): New function.
2473         * real.h (build_sinatan_real): Prototype.
2475 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
2477         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
2478         function.
2479         (fold_mergeeo_helper): New helper function.
2480         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
2481         intrinsics.  Correct some whitespace indentation issues.
2483 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
2485         PR target/87511
2486         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2487         Use HOST_WIDE_INT_1U for shift.
2489 2018-10-11  Doug Rupp  <rupp@adacore.com>
2490             Olivier Hainque  <hainque@adacore.com>
2492         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
2493         Pass --relax to the linker for RTPs.
2494         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
2496 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
2497             Jan Hubicka  <jh@suse.cz>
2498             Martin Jambor  <mjambor@suse.cz>
2500         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
2501         the same elements are repeated rather than printing all of them.
2502         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
2503         "repeated" elements.
2504         * read-rtl-function.c (test_loading_repeat): New function.
2505         (read_rtl_function_c_tests): Call test_loading_repeat.
2506         * rtl-tests.c (test_dumping_repeat): New function.
2507         (rtl_tests_c_tests): Call test_dumping_repeat.
2509 2018-10-11  Richard Biener  <rguenther@suse.de>
2511         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
2512         bdver?_cost): Unify to ...
2513         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
2514         * config/i386/i386.c (processor_cost_table): Adjust.
2516 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
2518         PR middle-end/87574
2519         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
2520         the thunk when expanding to GIMPLE.
2522 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2524         * varasm.c (mergeable_string_section): Don't try to move zero-length
2525         strings to the merge section.
2527 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
2529         PR target/87573
2530         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
2532 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
2534         PR target/87550
2535         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2536         to special_args set.
2538 2018-10-10  Richard Biener  <rguenther@suse.de>
2540         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
2541         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
2542         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
2543         and recursing and pattern terminating the recursion on SSE
2544         vector width using ix86_expand_reduc.
2545         (reduc_sminmax_scal_<mode>): Split into part reducing to half
2546         width and recursing and SSE2 vector variant doing the final
2547         reduction with ix86_expand_reduc.
2548         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
2549         with terminating the recursion at AVX level, splitting that
2550         to SSE there.
2552 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
2554         * genmatch.c (error_cb): Rename to...
2555         (diagnostic_cb): ...this, converting int params to enums.
2556         (fatal_at): Update for renaming.
2557         (warning_at): Likewise.
2558         (main): Likewise.
2559         * input.c (selftest::ebcdic_execution_charset::apply):
2560         Update for renaming of...
2561         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
2562         to...
2563         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
2564         converting level and reason to enums.
2565         (class selftest::lexer_error_sink): Rename to...
2566         (class selftest::lexer_test_options): ...this, renaming field
2567         "m_errors" to "m_diagnostics".
2568         (selftest::lexer_test_options::apply): Update for renaming of...
2569         (selftest::lexer_test_options::on_error): ...this, renaming to...
2570         (selftest::lexer_test_options::on_diagnostic): ...this
2571         converting level and reason to enums.
2572         (selftest::test_lexer_string_locations_raw_string_unterminated):
2573         Update for renamings.
2574         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
2575         "reason".
2577 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
2579         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
2580         * config/rs6000/pmmintrin.h: New file.
2582 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
2584         PR tree-optimization/86659
2585         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
2587 2018-10-09  Richard Biener  <rguenther@suse.de>
2589         PR tree-optimization/63155
2590         * tree-ssa-structalias.c: Include tree-ssa.h.
2591         (get_constraint_for_ssa_var): For undefs return nothing_id.
2592         (find_func_aliases): Cleanup PHI handling.
2594 2018-10-09  Richard Biener  <rguenther@suse.de>
2596         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
2597         replacements.
2599 2018-10-09  Martin Liska  <mliska@suse.cz>
2601         * asan.c (asan_emit_stack_protection): If a stack variable
2602         is located in a same file as current function, then emit
2603         line info into variable definition string.
2605 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2607         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
2608         information.
2610 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2612         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
2613         on the thunk.
2615 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2617         PR tree-optimization/86659
2618         * gimple-match.h (struct gimple_match_op): Add reverse field.
2619         (gimple_match_op::set_op): New overloaded method.
2620         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
2621         the REF_REVERSE_STORAGE_ORDER flag on the value.
2622         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2623         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2624         
2625 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
2627         PR middle-end/63155
2628         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2629         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2631 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
2633         PR target/87517
2634         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2635         Defined with __builtin_ia32_vfmaddsubpd512_mask.
2637 2018-10-08  Richard Biener  <rguenther@suse.de>
2639         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2640         cost the same as AVX128 ones.
2642 2018-10-08  Paul Koning  <ni1d@arrl.net>
2644         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2645         (expand_block_move): New function.
2646         * config/pdp11/pdp11.c (output_block_move): Remove.
2647         (expand_block_move): New function.
2648         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2649         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2650         (*movmemhi1): Remove.
2652 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
2654         * config/s390/2827.md: Increase latencies for some FP instructions.
2656 2018-10-08  Richard Biener  <rguenther@suse.de>
2658         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2659         Open a dump scope.
2660         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2661         * tree-vectorizer.h (dump_stmt_cost): Adjust.
2662         (add_stmt_cost): Dump return value of the hook.
2664 2018-10-08  Richard Biener  <rguenther@suse.de>
2666         PR tree-optimization/63155
2667         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2668         (ssa_propagation_engine::ssa_propagate): Remove redundant
2669         bitmap bit clearing.
2671 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
2673         PR rtl-optimization/86939
2674         PR rtl-optimization/87479
2675         * ira.h (non_conflicting_reg_copy_p): New prototype.
2676         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2677         (make_hard_regno_dead): Don't add conflicts for register
2678         ignore_reg_for_conflicts.
2679         (make_object_dead): Likewise.
2680         (non_conflicting_reg_copy_p): New function.
2681         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2682         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2683         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2684         (make_hard_regno_dead): Don't add conflicts for register
2685         ignore_reg_for_conflicts.  Remove special conflict handling of
2686         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
2687         check_pic_pseudo_p and update callers.
2688         (mark_pseudo_dead): Don't add conflicts for register
2689         ignore_reg_for_conflicts.
2690         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2692 2018-10-05  Andrew Waterman  <andrew@sifive.com>
2693             Jim Wilson  <jimw@sifive.com>
2695         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2696         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
2697         new pattern using HONOR_SNANS that emits one extra instruction.
2699 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2701         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2702         patterns): Merge SI and DI patterns to a GPR pattern.
2703         (unnamed define_insn and define_split for record form of that): Merge
2704         to a single define_insn_and_split pattern.
2706 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
2708         PR c++/56856
2709         * input.c (expand_location_to_spelling_point): Add param "aspect"
2710         and use rather than hardcoding LOCATION_ASPECT_CARET.
2711         (get_substring_ranges_for_loc): Handle the case of a single token
2712         within a macro expansion.
2713         * input.h (expand_location_to_spelling_point): Add "aspect" param,
2714         defaulting to LOCATION_ASPECT_CARET.
2716 2018-10-05  Paul Koning  <ni1d@arrl.net>
2718         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2719         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2720         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2721         (pdp11_guard_type): New function.
2723 2018-10-05  Paul Koning  <ni1d@arrl.net>
2725         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2726         * config/pdp11/pdp11.opt (mfloat32): Remove.
2727         (mfloat64): Remove.
2728         * doc/invoke.texi (pdp11 -mfloat32): Remove:
2729         (pdp11 -mfloat64): Remove.
2731 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2733         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2734         (*cmp<mode>_cc_i387): Ditto.
2735         (*cmpu<mode>_cc_i387): Ditto.
2736         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2737         * config/i386/i386.c (ix86_expand_fp_compare): Remove
2738         "scratch" argument.
2739         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2740         Emit x86_sahf_1 pattern.
2741         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2742         (ix86_expand_carry_flag_compare): Ditto.
2744 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2746         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2747         to reg_or_0_operand.  Add "C" constraint.
2748         (*cmpxf_cc_i387): Ditto.
2749         (*cmp<mode>_i387): Change operand 2 predicate
2750         to nonimm_or_0_operand.  Add "C" constraint.
2751         (*cmp<mode>_cc_i387): Ditto.
2752         (*cmp<mode>_0_i387): Remove insn pattern.
2753         (*cmp<mode>_0_cc_i387): Ditto.
2755 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2757         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2758         * config/i386/predicates.md (nonimm_or_0_operand): Rename
2759         from vector_move_operand.  Update all uses.
2761 2018-10-05  Martin Sebor  <msebor@redhat.com>
2763         PR tree-optimization/87490
2764         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2765         consistently.
2767 2018-10-05  Richard Biener  <rguenther@suse.de>
2769         PR tree-optimization/63155
2770         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2771         vertical space in dumpfiles.
2772         * tree-ssa-propagate.h
2773         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2774         * tree-ssa-propagate.c (cfg_blocks_back): New global.
2775         (ssa_edge_worklist_back): Likewise.
2776         (curr_order): Likewise.
2777         (cfg_blocks_get): Remove abstraction.
2778         (cfg_blocks_add): Likewise.
2779         (cfg_blocks_empty_p): Likewise.
2780         (add_ssa_edge): Add to current or next worklist based on
2781         RPO index.
2782         (add_control_edge): Likewise.
2783         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2784         into ...
2785         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
2786         iteration from CFG and SSA edge worklist so we process
2787         everything in RPO order, prioritizing forward progress
2788         over iteration.
2789         (ssa_prop_init): Allocate new worklists, do not dump
2790         immediate uses.
2791         (ssa_prop_fini): Free new worklists.
2793 2018-10-05  Richard Biener  <rguenther@suse.de>
2795         * tree-core.h (tree_block::abstract_flag): Remove.
2796         (tree_block::block_num): Make full 32bits.
2797         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2798         * tree.h (BLOCK_ABSTRACT): Remove.
2799         * dwarf2out.c (gen_lexical_block_die): Remove dead code
2800         resulting from BLOCK_ABSTRACT being always false.
2801         (gen_inlined_subroutine_die): Likewise.
2802         (gen_block_die): Likewise.
2803         * tree.c (block_ultimate_origin): Likewise.
2804         * tree-pretty-print.c (dump_block_node): Remove code dealing
2805         with BLOCK_ABSTRACT.
2806         * tree-ssa-live.c (dump_scope_block): Likewise.
2807         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2808         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2810 2018-10-05   Richard Biener  <rguenther@suse.de>
2812         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2813         is asked for initialize mode to the component mode of the
2814         vector type.
2816 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
2818         PR target/87522
2819         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2820         assembler for -mavx.
2821         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2823 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2825         PR target/87509
2826         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2827         RS6000_BTM_DFP.
2828         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2829         to be DImode.  When using mffscrn, force the operand to a register.
2831 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
2833         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2834         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2835         X87MODEF mode iterator.
2836         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2837         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2838         X87MODEF mode iterator.
2840 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
2842         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2843         -Wno-prio-ctor-dtor.
2845 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2847         * Makefile.in (OBJS): Add opt-problem.o.
2848         * dump-context.h: Include "selftest.h.
2849         (selftest::temp_dump_context): New forward decl.
2850         (class dump_context): Make friend of class
2851         selftest::temp_dump_context.
2852         (dump_context::dump_loc_immediate): New decl.
2853         (class dump_pretty_printer): Move here from dumpfile.c.
2854         (class temp_dump_context): Move to namespace selftest.
2855         (temp_dump_context::temp_dump_context): Add param
2856         "forcibly_enable_dumping".
2857         (selftest::verify_dumped_text):
2858         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2859         (selftest::verify_item):
2860         (ASSERT_IS_TEXT): Move here from dumpfile.c.
2861         (ASSERT_IS_TREE): Likewise.
2862         (ASSERT_IS_GIMPLE): Likewise.
2863         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2864         to...
2865         (dump_context::dump_loc_immediate): ...this new function.
2866         (class dump_pretty_printer): Move to dump-context.h.
2867         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2868         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2869         (temp_dump_context::temp_dump_context): Move to "selftest"
2870         namespace.  Add param "forcibly_enable_dumping", and use it to
2871         conditionalize the use of m_pp;
2872         (selftest::verify_dumped_text): Make non-static.
2873         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2874         (selftest::verify_item): Make non-static.
2875         (ASSERT_IS_TEXT): Move to dump-context.h.
2876         (ASSERT_IS_TREE): Likewise.
2877         (ASSERT_IS_GIMPLE): Likewise.
2878         (selftest::test_capture_of_dump_calls): Pass "true" for new
2879         param of temp_dump_context.
2880         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2881         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
2882         TDF_COMPARE_DEBUG.
2883         * opt-problem.cc: New file.
2884         * opt-problem.h: New file.
2885         * optinfo-emit-json.cc
2886         (selftest::test_building_json_from_dump_calls): Pass "true" for
2887         new param of temp_dump_context.
2888         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2889         (optinfo::emit_for_opt_problem): New function.
2890         (optinfo::emit): Clarity which emit_item is used.
2891         * optinfo.h (optinfo::get_dump_location): New accessor.
2892         (optinfo::emit_for_opt_problem): New decl.
2893         (optinfo::emit): Make const.
2894         * selftest-run-tests.c (selftest::run_tests): Call
2895         selftest::opt_problem_cc_tests.
2896         * selftest.h (selftest::opt_problem_cc_tests): New decl.
2897         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2898         bool to opt_result, converting fprintf messages to
2899         opt_result::failure_at calls.  Add "stmt" param for use by the
2900         failure_at calls.
2901         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2902         (runtime_alias_check_p): Convert return type from bool to
2903         opt_result, converting dump_printf calls to
2904         opt_result::failure_at, using the statement DDR_A for their
2905         location.
2906         (find_data_references_in_stmt): Convert return type from bool to
2907         opt_result, converting "return false" to opt_result::failure_at
2908         with a new message.
2909         * tree-data-ref.h: Include "opt-problem.h".
2910         (dr_analyze_innermost): Convert return type from bool to opt_result,
2911         and add a const gimple * param.
2912         (find_data_references_in_stmt): Convert return type from bool to
2913         opt_result.
2914         (runtime_alias_check_p): Likewise.
2915         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2916         dr_analyze_innermost.
2917         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2918         Convert return type from bool to opt_result, adding a message for
2919         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2920         (vect_analyze_data_ref_dependence): Convert return type from bool
2921         to opt_result.  Change sense of return type from "false"
2922         effectively meaning "no problems" to "false" meaning a problem,
2923         so that "return false" becomes "return opt_result::success".
2924         Convert "return true" calls to opt_result::failure_at, using
2925         the location of statement A rather than vect_location.
2926         (vect_analyze_data_ref_dependences): Convert return type from bool
2927         to opt_result.
2928         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2929         calls to opt_result::failure_at, using the stmt location rather
2930         than vect_location.
2931         (vect_verify_datarefs_alignment): Convert return type from bool
2932         to opt_result.
2933         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
2934         into multiple more-tightly-scoped copies.
2935         (vect_analyze_data_refs_alignment): Convert return type from bool
2936         to opt_result.
2937         (vect_analyze_data_ref_accesses): Likewise, converting a
2938         "return false" to a "return opt_result::failure_at", adding a
2939         new message.
2940         (vect_prune_runtime_alias_test_list): Convert return type from
2941         bool to opt_result, converting dump_printf_loc to
2942         opt_result::failure_at.  Add a %G to show the pertinent statement,
2943         and use the stmt's location rather than vect_location.
2944         (vect_find_stmt_data_reference): Convert return type from
2945         bool to opt_result, converting dump_printf_loc to
2946         opt_result::failure_at, using stmt's location.
2947         (vect_analyze_data_refs):  Convert return type from bool to
2948         opt_result.  Convert "return false" to "return
2949         opt_result::failure_at", adding messages as needed.
2950         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2951         type from bool to opt_result.
2952         (vect_determine_vf_for_stmt): Likewise.
2953         (vect_determine_vectorization_factor): Likewise, converting
2954         dump_printf_loc to opt_result::failure_at, using location of phi
2955         rather than vect_location.
2956         (vect_analyze_loop_form_1): Convert return type from bool to
2957         opt_result, converting dump_printf_loc calls, retaining the use of
2958         vect_location.
2959         (vect_analyze_loop_form): Convert return type from loop_vec_info
2960         to opt_loop_vec_info.
2961         (vect_analyze_loop_operations): Convert return type from bool to
2962         opt_result, converting dump_printf_loc calls, using the location
2963         of phi/stmt rather than vect_location where available.  Convert
2964         various "return false" to "return opt_result::failure_at" with
2965         "unsupported phi" messages.
2966         (vect_get_datarefs_in_loop): Convert return type from bool to
2967         opt_result.  Add a message for the
2968         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2969         (vect_analyze_loop_2): Convert return type from bool to
2970         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
2971         each "goto again;", adding new messages where needed.
2972         Add "unsupported grouped {store|load}" messages.
2973         (vect_analyze_loop): Convert return type from loop_vec_info to
2974         opt_loop_vec_info.
2975         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2976         bool to opt_result.
2977         * tree-vect-stmts.c (process_use): Likewise, converting
2978         dump_printf_loc call and using stmt location, rather than
2979         vect_location.
2980         (vect_mark_stmts_to_be_vectorized): Likeise.
2981         (vect_analyze_stmt): Likewise, adding a %G.
2982         (vect_get_vector_types_for_stmt): Convert return type from bool to
2983         opt_result, converting dump_printf_loc calls and using stmt
2984         location, rather than vect_location.
2985         (vect_get_mask_type_for_stmt): Convert return type from tree to
2986         opt_tree, converting dump_printf_loc calls and using stmt location.
2987         * tree-vectorizer.c: Include "opt-problem.h.
2988         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2989         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
2990         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
2991         enabled, use it to report at the top level "couldn't vectorize
2992         loop" followed by the problem.
2993         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2994         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2995         to opt_result.
2996         (vect_analyze_stmt): Likewise.
2997         (vect_get_vector_types_for_stmt): Likewise.
2998         (tree vect_get_mask_type_for_stmt): Likewise.
2999         (vect_analyze_data_ref_dependences): Likewise.
3000         (vect_enhance_data_refs_alignment): Likewise.
3001         (vect_analyze_data_refs_alignment): Likewise.
3002         (vect_verify_datarefs_alignment): Likewise.
3003         (vect_analyze_data_ref_accesses): Likewise.
3004         (vect_prune_runtime_alias_test_list): Likewise.
3005         (vect_find_stmt_data_reference): Likewise.
3006         (vect_analyze_data_refs): Likewise.
3007         (vect_analyze_loop): Convert return type from loop_vec_info to
3008         opt_loop_vec_info.
3009         (vect_analyze_loop_form): Likewise.
3010         (vect_analyze_slp): Convert return type from bool to opt_result.
3012 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3014         * doc/invoke.texi (-fopt-info): Document new "internals"
3015         sub-option.
3016         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
3017         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
3018         MSG_ALL_KINDS.
3019         (optinfo_verbosity_options): Add "internals".
3020         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
3021         (dump_context::apply_dump_filter_p): New member function.
3022         (dump_context::dump_loc): Use apply_dump_filter_p rather than
3023         explicitly masking the dump_kind.
3024         (dump_context::begin_scope): Increment the scope depth first.  Use
3025         apply_dump_filter_p rather than explicitly masking the dump_kind.
3026         (dump_context::emit_item): Use apply_dump_filter_p rather than
3027         explicitly masking the dump_kind.
3028         (dump_dec): Likewise.
3029         (dump_hex): Likewise.
3030         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
3031         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
3032         (opt_info_switch_p): Update handling of default
3033         MSG_OPTIMIZED_LOCATIONS to cope with default of
3034         MSG_PRIORITY_USER_FACING.
3035         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
3036         masking the dump_kind.
3037         (selftest::test_capture_of_dump_calls): Update test_dump_context
3038         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
3039         than MSG_ALL.  Generalize scope test to be run at all four
3040         combinations of with/without MSG_PRIORITY_USER_FACING and
3041         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
3042         for each of the two values.
3043         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
3044         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
3045         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
3046         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
3047         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
3048         with MSG_PRIORITY_*.
3049         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
3050         dump messages as MSG_PRIORITY_USER_FACING.
3051         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
3052         about the interaction with MSG_PRIORITY_*.
3054 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3056         * varasm.c (output_constant): Add new parameter merge_strings.
3057         Make strings properly zero terminated in merge string sections.
3058         (mergeable_string_section): Don't fail if the last char is non-zero.
3059         (assemble_variable_contents): Handle merge string sections.
3060         (assemble_variable): Likewise.
3061         (assemble_constant_contents): Likewise.
3062         (output_constant_def_contents): Likewise.
3063         (output_constructor_array_range,
3064         output_constructor_regular_field): Adjust call to output_constant.
3065         (output_object_block): Adjust call to assemble_constant_contents
3066         and assemble_variable_contents.
3068 2018-10-04  Martin Liska  <mliska@suse.cz>
3070         PR c/87483
3071         * cgraphunit.c (process_function_and_variable_attributes):
3072         Warn about a function with alias attribute and a body.
3074 2018-10-04  Martin Liska  <mliska@suse.cz>
3076         PR ipa/82625
3077         * multiple_target.c (redirect_to_specific_clone): New function.
3078         (ipa_target_clone): Use it.
3079         * tree-inline.c: Fix comment.
3081 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3083         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
3084         fields.
3085         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
3086         (gcc::dump_manager::register_pass): New member function, adapted
3087         from loop body in gcc::pass_manager::register_pass, adding a
3088         call to update_dfi_for_opt_info.
3089         (gcc::dump_manager::opt_info_enable_passes): Store the
3090         -fopt-info options into the new fields.  Move the loop
3091         bodies into...
3092         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
3093         function.
3094         * dumpfile.h (struct opt_pass): New forward decl.
3095         (gcc::dump_manager::register_pass): New decl.
3096         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
3097         (class gcc::dump_manager): Add fields "m_optgroup_flags",
3098         "m_optinfo_flags", and "m_optinfo_filename".
3099         * passes.c (gcc::pass_manager::register_pass): Move all of the
3100         dump-handling code to gcc::dump_manager::register_pass.
3102 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
3104         PR rtl-optimization/87466
3105         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
3106         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
3107         * doc/tm.texi: Regenerate.
3108         * ira-lives.c (process_bb_node_lives): Use the new target hook.
3109         * lra-lives.c (process_bb_lives): Likewise.
3110         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
3111         Define.
3113 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
3115         * params.c (add_params): Fix initialization.
3117 2018-10-04  Martin Liska  <mliska@suse.cz>
3119         PR gcov-profile/84107
3120         * tree-profile.c (init_ic_make_global_vars):
3121         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
3122         Come up with new ic_tuple* variables.  Emit
3123         __gcov_indirect_call{,_topn} variables.
3124         (gimple_gen_ic_profiler): Access the variable
3125         and emit gimple.
3126         (gimple_gen_ic_func_profiler): Access
3127         __gcov_indirect_call.callee field.
3128         (gimple_init_gcov_profiler): Use ptr_type_node.
3129         * value-prof.c (gimple_ic): Use ptr_type_node.
3131 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3133         PR tree-optimization/85787
3134         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
3135         into this function and add support for detecting multiple phis.
3136         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
3138 2018-10-04  Martin Liska  <mliska@suse.cz>
3140         PR ipa/87491
3141         * ipa-inline.c (inline_to_all_callers_1):
3142         Call ultimate_alias_target for node being inlined.
3144 2018-10-03  Jeff Law  <law@redhat.com>
3146         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
3147         target's wchar_t.
3148         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
3149         * tree.h (get_typenode_from_name): Prototype.
3151 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
3153         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
3154         Change operand 2 predicate to nonimmediate_operand.
3155         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
3157 2018-10-03  Martin Sebor  <msebor@redhat.com>
3158             Jeff Law  <law@redhat.com>
3160         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
3161         initialize it.
3162         (get_string_length): Detect unterminated arrays.
3163         (format_string): Same.
3164         (format_directive): Warn about unterminated arrays.
3165         (handle_gimple_call): Mark statements with no_warning as needed.
3167 2018-10-03  Jim Wilson  <jimw@sifive.com>
3169         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
3170         also define __riscv_abi_rve.  Delete trailing white space.
3172 2018-10-03  Paul Koning  <ni1d@arrl.net>
3174         Enable LRA register allocator for PDP11.
3175         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
3176         (R): Likewise.
3177         (D): Likewise.
3178         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
3179         * config/pdp11/pdp11.opt (-mlra): New option.
3180         * doc/invoke.texi (PDP-11 Options): Document -mlra.
3182 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
3184         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
3185         (*<absneg:code>extend<mode>xf2): Ditto.
3187 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
3189         PR tree-optimization/87415
3190         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
3191         precision fields.
3193 2018-10-02  Jeff Law  <law@redhat.com>
3195         * gimple-fold.c (get_range_strlen): Only set *nonstr when
3196         an unterminated string is discovered.  Bubble up range
3197         even for unterminated strings.
3198         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
3199         indicates the string was not terminated via NONSTR.
3201 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
3203         * tree-vrp.c (extract_range_from_unary_expr): Special case all
3204         pointer conversions.
3205         Do not do anything special for anti-ranges.
3207 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
3209         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
3210         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
3212 2018-10-03  Martin Liska  <mliska@suse.cz>
3214         PR gcov-profile/86109
3215         * coverage.c (coverage_begin_function): Do not
3216         mark lambdas as artificial.
3217         * tree-core.h (struct GTY): Remove tm_clone_flag
3218         and introduce new lambda_function.
3219         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
3221 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3223         PR target/87474
3224         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
3225         P8_VECTOR and VSX are enabled.
3227 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3229         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
3230         0x3907 as CPU model number.
3232 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3234         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
3235         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
3236         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
3237         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
3238         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
3239         * config/s390/s390.md: Likewise. Rename also the cpu attribute
3240         value from arch12 to z14.
3242 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3244         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
3245         (isinfxf2): Ditto.
3246         (isinf<mode>2): Ditto.
3248 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3250         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
3251         before emitting fxam.  Perform calculations in XFmode.
3253 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3255         * match.pd (((X /[ex] A) +- B) * A): New transformation.
3257 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3259         PR middle-end/87319
3260         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
3261         * tree.c (signed_or_unsigned_type_for): Handle complex.
3263 2018-10-02  Jeff Law  <law@redhat.com>
3265         * gimple-fold.c (get_range_strlen): Remove dead code.
3267 2018-10-02  Martin Sebor  <msebor@redhat.com>
3268             Jeff Law  <law@redhat.com>
3270         * builtins.c (unterminated_array): Add new arguments.
3271         If argument is not terminated, bubble up size and exact
3272         state to callers.
3273         (expand_builtin_strnlen): Detect, avoid expanding
3274         and diagnose unterminated arrays.
3275         (c_strlen): Fill in offset of start of unterminated strings.
3276         * builtins.h (unterminated_array): Update prototype.
3278 2018-10-02  Richard Biener  <rguenther@suse.de>
3280         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
3281         of haddv4df, first reduce to SSE width and exploit the fact
3282         that we only need element zero with the reduction result.
3283         (reduc_plus_scal_v2df): Likewise.
3285 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3287         * dojump.h (do_jump): Delete.
3288         (do_jump_1): Likewise.
3289         (split_comparison): Move around.
3290         * dojump.c (do_jump): Make static.
3291         (do_jump_1): Likewise.
3292         (jumpifnot): Move around.
3293         (jumpifnot_1): Likewise.
3294         (jumpif): Likewise.
3295         (jumpif_1): Likewise.
3296         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
3298 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3300         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
3301         insns in the delay slot and add_insn_after for the jump insn.
3303 2018-10-02  Richard Biener  <rguenther@suse.de>
3305         * tree-inline.c (expand_call_inline): Use the location of
3306         the callee declaration for the inline-entry marker.
3307         * final.c (notice_source_line): Remove special-casing of
3308         NOTE_INSN_INLINE_ENTRY.
3310 2018-10-01  Carl Love  <cel@us.ibm.com>
3312         PR 69431
3313         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
3314         (__builtin_mtfsb0): New.
3315         (__builtin_mtfsb1): New.
3316         ( __builtin_set_fpscr_rn): New.
3317         (__builtin_set_fpscr_drn): New.
3318         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
3319         (rs6000_expand_set_fpscr_rn_builtin): Add.
3320         (rs6000_expand_set_fpscr_drn_builtin): Add.
3321         (rs6000_expand_builtin): Add case statement entries for
3322         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
3323         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
3324         RS6000_BUILTIN_MFFSL.
3325         (rs6000_init_builtins): Add ftype initialization and def_builtin
3326         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
3327         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
3328         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
3329         rs6000_mffscdrn): Add define_insn.
3330         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
3331         * doc/extend.texi: Add documentation for the builtins.
3333 2018-10-01  Richard Biener  <rguenther@suse.de>
3335         PR tree-optimization/87465
3336         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
3337         causing branch miscounts.
3339 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3341         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
3342         aarch64_option_default_param):  New.
3343         (params.h): Include.
3344         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
3345         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
3346         stack-clash protection validation code.
3348 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3350         * params.c (validate_param): New.
3351         (add_params): Use it.
3352         (set_param_value): Refactor param validation into validate_param.
3353         (diagnostic.h): Include.
3354         * diagnostic.h (diagnostic_ready_p): New.
3356 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3358         * params.c (set_param_value):
3359         Add index of parameter being validated.
3360         * common/common-target.def (option_validate_param): New.
3361         * common/common-targhooks.h (default_option_validate_param): New.
3362         * common/common-targhooks.c (default_option_validate_param): New.
3363         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
3364         * doc/tm.texi: Regenerate.
3366 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3368         PR target/86486
3369         * config/aarch64/aarch64.c (aarch64_override_options_internal):
3370         Add validation for stack-clash parameters and set defaults.
3372 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3374         PR target/86486
3375         * configure.ac: Add stack-clash-protection-guard-size.
3376         * doc/install.texi: Document it.
3377         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
3378         * params.def: Update comment for guard-size.
3379         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
3380         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
3381         * configure: Regenerate.
3383 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3385         PR target/86486
3386         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
3387         STACK_DYNAMIC_OFFSET): New.
3388         * config/aarch64/aarch64.c (aarch64_layout_frame):
3389         Update outgoing args size.
3390         (aarch64_stack_clash_protection_alloca_probe_range,
3391         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3393 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3395         PR target/86486
3396         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
3397         probe ranges.
3398         * target.def (stack_clash_protection_alloca_probe_range): New.
3399         (stack_clash_protection_final_dynamic_probe): Remove.
3400         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
3401         (default_stack_clash_protection_final_dynamic_probe): Remove.
3402         * targhooks.c: Likewise.
3403         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3404         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
3405         * doc/tm.texi: Regenerate.
3407 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3409         PR target/86486
3410         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
3411         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
3412         aarch64_clamp_to_uimm12_shift): New.
3413         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
3414         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
3416 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3418         PR target/86486
3419         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
3421 2018-10-01  Jeff Law  <law@redhat.com>
3422             Richard Sandiford <richard.sandiford@linaro.org>
3423             Tamar Christina  <tamar.christina@arm.com>
3425         PR target/86486
3426         * config/aarch64/aarch64.md
3427         (probe_stack_range): Add k (SP) constraint.
3428         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
3429         STACK_CLASH_MAX_UNROLL_PAGES): New.
3430         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
3431         stack probes for stack clash.
3432         (aarch64_allocate_and_probe_stack_space): New.
3433         (aarch64_expand_prologue): Use it.
3434         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
3435         (aarch64_sub_sp): Add emit_move_imm optional param.
3437 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
3439         PR tree-optimization/87261
3440         * match.pd: Remove trailing whitespace.
3441         Add (x & y) | ~(x | y) -> ~(x ^ y),
3442         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
3444 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3446         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
3447         constraints.
3448         (ashlsi3_insn): Update instruction constraints.
3449         (ashrsi3_insn): Likewise.
3450         (rotrsi3): Likewise.
3451         (add_shift): Likewise.
3452         * config/arc/constraints.md (Csz): New 32 bit constraint. It
3453         avoids placing in the limm field small constants which, otherwise,
3454         could end into a small instruction.
3456 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3458         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
3459         destination register is not odd-even.
3460         (umaddsidi4_split): Likewise.
3462 2018-10-01  Richard Biener  <rguenther@suse.de>
3464         * tree-inline.c (expand_call_inline): Store origin of fn
3465         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
3466         * tree.c (block_ultimate_origin): Simplify and do some
3467         checking.
3469 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
3471         * config/i386/mmx.md (EMMS): New int iterator.
3472         (emms): New int attribute.
3473         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
3474         EMMS int iterator.  Explicitly declare clobbers.
3475         (mmx_emms): Remove expander.
3476         (mmx_femms): Ditto.
3477         * config/i386/predicates.md (emms_operation): Remove predicate.
3478         (vzeroall_pattern): New predicate.
3479         (vzeroupper_pattern): Rename from vzeroupper_operation.
3480         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
3481         vzeroupper_pattern and vzeroall_pattern predicates.
3483 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
3485         PR rtl-optimization/86939
3486         * ira-lives.c (make_hard_regno_born): Rename from this...
3487         (make_hard_regno_live): ... to this.  Remove update to conflict
3488         information.  Update function comment.
3489         (make_hard_regno_dead): Add conflict information update.  Update
3490         function comment.
3491         (make_object_born): Rename from this...
3492         (make_object_live): ... to this.  Remove update to conflict information.
3493         Update function comment.
3494         (make_object_dead):  Add conflict information update.  Update function
3495         comment.
3496         (mark_pseudo_regno_live): Call make_object_live.
3497         (mark_pseudo_regno_subword_live): Likewise.
3498         (mark_hard_reg_dead): Update function comment.
3499         (mark_hard_reg_live): Call make_hard_regno_live.
3500         (process_bb_node_lives): Likewise.
3501         * lra-lives.c (make_hard_regno_born): Rename from this...
3502         (make_hard_regno_live): ... to this.  Remove update to conflict
3503         information.  Remove now uneeded check_pic_pseudo_p argument.
3504         Update function comment.
3505         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
3506         to conflict information.  Update function comment.
3507         (mark_pseudo_live): Remove update to conflict information.  Update
3508         function comment.
3509         (mark_pseudo_dead): Add conflict information update.
3510         (mark_regno_live): Call make_hard_regno_live.
3511         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
3512         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
3514 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
3516         PR target/87370
3517         * config/i386/i386.c (construct_container): Use TImode for
3518         BLKmode values in 2 integer registers.
3520 2018-09-29  Jeff Law  <law@redhat.com>
3522         * builtins.c (unterminated_array): Pass in c_strlen_data * to
3523         c_strlen rather than just a tree *.
3524         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
3525         Update recursive calls appropriately.  If caller did not provide a
3526         suitable data pointer, create a local one.  When a non-terminated
3527         string is discovered, bubble up information about the string via the
3528         c_strlen_data object.
3529         * builtins.h (c_strlen): Update prototype.
3530         (c_strlen_data): New structure.
3531         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
3532         For a type 2 call, if c_strlen indicates a non-terminated string
3533         use the length of the non-terminated string.
3534         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
3536 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
3538         PR target/87467
3539         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
3540         __m512d type for __A argument rather than __m512.
3542 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
3544         * match.pd (simple_comparison): Don't optimize if either operand is
3545         a function pointer when target needs function pointer canonicalization.
3547 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3549         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
3550         power5 .. power9 to remove indirection.
3551         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
3552         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
3553         ASM_CPU_476_SPEC): Delete.
3554         (ASM_CPU_SPEC): Adjust.
3555         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
3556         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
3558 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3560         * config.in: Delete HAVE_AS_DCI.
3561         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
3562         * config/rs6000/rs6000.h: Ditto.
3563         * configure.ac: Delete HAVE_AS_DCI.
3564         * configure: Regenerate.
3566 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3568         * config.in (HAVE_AS_LWSYNC): Delete.
3569         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3570         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
3571         do it as a .long .
3572         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3573         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
3574         as a .long .
3575         * configure.ac: Delete HAVE_AS_LWSYNC.
3576         * configure: Regenerate.
3578 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3579             Pierre-Marie de Rodat  <derodat@adacore.com>
3581         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
3582         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
3583         (cgraph_node::create_thunk): Add indirect_offset parameter.
3584         (thunk_adjust): Likewise.
3585         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
3586         and initialize the corresponding field with it.
3587         (cgraph_node::dump): Dump indirect_offset field.
3588         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
3589         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
3590         (thunk_adjust): Add indirect_offset parameter and deal with it.
3591         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
3592         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
3593         or if the thunk is external or local.  Fix formatting.  Do not chain
3594         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
3595         if any, in the GIMPLE representation.
3596         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
3597         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
3598         (input_node): Read indirect_offset field.
3599         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
3600         call to thunk_adjust.
3601         * tree-nested.c (struct nesting_info): Add thunk_p field.
3602         (create_nesting_tree): Set it.
3603         (convert_all_function_calls): Copy static chain from targets to thunks.
3604         (finalize_nesting_tree_1): Return early for thunks.
3605         (unnest_nesting_tree_1): Do not finalize thunks.
3606         (gimplify_all_functions): Do not gimplify thunks.
3608 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3610         * opt-suggestions.c (option_proposer::build_option_suggestions):
3611         Release "option_values".
3613 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3615         * coverage.c (get_coverage_counts): Convert problem-reporting dump
3616         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
3617         * dumpfile.c (kind_as_string): New function.
3618         (dump_loc): Rather than a hardcoded prefix of "note: ", use
3619         kind_as_string to vary the prefix based on dump_kind.
3620         (selftest::test_capture_of_dump_calls): Update for above.
3622 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3624         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3625         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
3627 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3629         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3630         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3631         INVALID_REGNUM instead of FPSR_REG.
3632         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3633         * config/i386/i386.md: Update comment of FP compares.
3634         (fldenv): Do not clobber FPSR_REG.
3636 2018-09-28  Richard Biener  <rguenther@suse.de>
3638         * tree.h (BLOCK_ORIGIN): New.
3639         * omp-expand.c (grid_expand_target_grid_body): Assign
3640         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3641         * tree-inline.c (remap_block): Likewise.
3642         * auto-profile.c (get_function_decl_from_block): Simplify
3643         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3644         * langhooks.c (lhd_print_error_function): Likewise.
3645         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3646         Likewise.
3647         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3648         * tree.c (block_nonartificial_location): Likewise.
3649         (block_ultimate_origin): Likewise.
3650         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
3651         no longer needed LTO case.
3653 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
3654             Jan Hubicka  <jh@suse.cz>
3655             Martin Jambor  <mjambor@suse.cz>
3657         * simplify-rtx.c (simplify_merge_mask): New function.
3658         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3659         same masks are used in op1 or op2.
3660         (test_vec_merge): New function.
3661         (test_vector_ops): Call test_vec_merge.
3663 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3665         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3666         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3667         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3668         (cypress_costs): Set it.
3669         (supersparc_costs): Likewise.
3670         (hypersparc_costs): Likewise.
3671         (leon_cost): Likewise.
3672         (leon3_costs): Likewise.
3673         (sparclet_costs): Likewise.
3674         (ultrasparc_costs): Likewise.
3675         (ultrasparc_costs): Likewise.
3676         (niagara_costs): Likewise.
3677         (niagara2_costs): Likewise.
3678         (niagara3_costs): Likewise.
3679         (niagara4_costs): Likewise.
3680         (niagara7_costs): Likewise.
3681         (m8_costs): Likewise.
3682         (TARGET_CAN_FOLLOW_JUMP): Define.
3683         (pass_work_around_errata::gate): Minor tweak.
3684         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3685         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3686         Automaitcally clear MASK_FSMULD mask for V7 processors.
3687         (sparc_can_follow_jump): New static function.
3688         (output_ubranch): Deal with CROSSING_JUMP_P.
3689         (sparc_use_sched_lookahead): Rewrite using switch statement.
3690         (sparc_issue_rate): Reorder.
3691         (sparc_branch_cost): New function.
3693 2018-09-27  Martin Sebor  <msebor@redhat.com>
3695         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3696         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3697         (int_fits_type_p): Same.
3699 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
3701         * config/i386/i386.md (FPCR_REG): Remove.
3702         (UNSPEC_FLDCW): Remove.
3703         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3704         (x86_fldcw_1): Remove insn pattern.
3705         (fnstenv): Do not clobber FPCR_REG.
3706         (fldenv): Ditto.
3707         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3708         (CALL_USED_REGISTERS): Ditto.
3709         (REG_ALLOC_ORDER): Ditto.
3710         (REG_CLASS_CONTENTS): Ditto.
3711         (HI_REGISTER_NAMES): Ditto.
3712         (ADDITIONAL_REGISTER_NAMES): Use defines instead
3713         of numerical constants.
3714         * config/i386/i386.c (regclass_map): Remove fpsr register.
3715         (dbx_register_map): Ditto.
3716         (dbx64_register_map): Ditto.
3717         (svr4_dbx_register_map): Ditto.
3718         (print_reg): Do not handle FPCR_REG.
3720 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
3722         PR target/87149
3723         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3724         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3725         Delete, always treat as true.
3726         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3727         Ditto.  Simplify remaining code.
3728         * config/powerpcspe/powerpcspe.h: Ditto.
3729         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3730         Simplify remaining code.
3731         (rs6000_expand_builtin): Ditto.
3732         * config/rs6000/rs6000.h: Ditto.
3733         * configure.ac: Ditto.
3734         * configure: Regenerate.
3736 2018-09-27  Martin Liska  <mliska@suse.cz>
3738         * coverage.c (get_coverage_counts): Revert the formatting
3739         of missing profile opt info.
3741 2018-09-27  Richard Biener  <rguenther@suse.de>
3743         PR debug/37801
3744         PR debug/87440
3745         * dwarf2out.c (set_block_origin_self): Do not mark outermost
3746         block as we do not output that.
3747         (gen_inlined_subroutine_die): Elide the originally outermost
3748         block, matching what we do for concrete instances.
3749         (decls_for_scope): Add parameter specifying whether to recurse
3750         to subblocks.
3752 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
3753             Tom de Vries  <tom@codesourcery.com>
3755         PR 82089
3757         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3758         STORE_FLAG_VALUE == 1.
3760 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
3762         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3763         constant definitions.
3764         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3765         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3766         ("speculation_barrier"): New expander definition.
3768 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
3770         PR gcov-profile/86957
3771         * common.opt: New warning option -Wmissing-profile.
3772         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3773         * doc/invoke.texi: Document -Wmissing-profile.
3775 2018-09-26  Jim Wilson  <jimw@sifive.com>
3777         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3778         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3780 2018-09-26  Martin Sebor  <msebor@redhat.com>
3782         * tree.c (zerop): Change return type to bool.
3783         (integer_zerop, integer_onep, integer_each_onep): Same.
3784         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3785         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3786         (real_onep, real_minus_onep, chain_index): Same.
3787         (print_type_hash_statistics, type_list_equal): Same.
3788         * tree.h (zerop): Same.
3789         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
3790         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3791         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3792         (real_onep, real_minus_onep, chain_index): Same.
3793         (print_type_hash_statistics, type_list_equal): Same.
3795 2018-09-26  Jim Wilson  <jimw@sifive.com>
3797         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3799 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
3801         PR target/87414
3802         * config/i386/i386.c: Include debug.h and dwarf2out.h.
3803         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3804         call.
3806 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
3808         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3810 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3812         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3813         and FP_SECOND_SSE_REGS.
3814         (REG_CLASS_NAMES): Ditto.
3815         (REG_CLASS_CONTENTS): Ditto.
3816         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3817         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3818         (ix86_preferred_output_reload_class): Ditto.
3819         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3820         clobber constraint to "=&f".
3821         (fix_truncdi_i387): Ditto.
3822         (lrintxfdi2): Ditto.
3823         (fistdi2_<rounding>): Ditto.
3824         (fpremxf4_i387): Change "=u" constraint to "=f".
3825         (fprem1xf4_i387): Ditto.
3826         (sincosxf3): Ditto.
3827         (fptanxf4_i387): Ditto.
3828         (fxtractxf3_i387): Ditto.
3829         (fscalexf4_i387): Ditto.
3830         (atan2xf3): Change "u" constraint to "f".
3831         (fyl2xxf3_i387): Ditto.
3832         (fyl2xp1xf3_i387): Ditto.
3834 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3836         PR target/87439
3837         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3838         for removed I387_MASK_PM entity.
3841 2018-09-26  Jeff Law  <law@redhat.com>
3842         Revert
3843         2018-09-26  Alexey Neyman  <stilor@att.net>
3845         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3846         headers are no longer pulled in by <isl/val.h>.
3848 2018-09-26  Richard Biener  <rguenther@suse.de>
3850         PR debug/87443
3851         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3852         or concrete instance DIE to the tree.  Create abstract origin
3853         attributes also for concrete instances.
3855 2018-09-26  Alexey Neyman  <stilor@att.net>
3857         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3858         headers are no longer pulled in by <isl/val.h>.
3860 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
3862         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3863         Use new helper functions.
3864         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3865         Use new helper functions.
3866         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3867         aarch_mm_needs_release): New declarations.
3868         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3869         aarch_mm_needs_release): New.
3871 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
3873         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3874         (arm32_output_mi_thunk): Deal with long calls.
3876 2018-09-26  Richard Biener  <rguenther@suse.de>
3878         PR debug/87428
3879         PR debug/87362
3880         * tree-inline.c (expand_call_inline): When the location
3881         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3882         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3883         the inserted BLOCK to make inlined_function_outer_scope_p
3884         recognize it.
3885         * dwarf2out.c (add_call_src_coords_attributes): Do not add
3886         coords for reserved locations.
3888 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3890         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3891         (*call_indirect_nonlocal_sysv<mode>): Ditto.
3892         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3893         (*sibcall_nonlocal_sysv<mode>): Ditto.
3894         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3895         (<bd>_<mode>): Ditto.
3896         (<bd>tf_<mode>): Ditto.
3898 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3900         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3901         control string as a list of templates instead of as C code.
3902         (*altivec_movti): Ditto.
3903         * config/rs6000/darwin.md (movdf_low_di): Ditto.
3905 2018-09-25  Jim Wilson  <jimw@sifive.com>
3907         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3908         when target symbol is weak.
3910 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3912         PR c/87387
3913         * builtins.c (unterminated_array): Simplify.
3914         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
3915         where pointer arithmetic is safe.
3917 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3919         PR target/86987
3920         * config/rs6000/altivec.md (altivec_vspltb): Use
3921         const_0_to_15_operand instead of u5bit_cint_operand.
3922         (*altivec_vspltb_internal): Ditto.
3923         (altivec_vspltb_direct): Ditto.
3924         (altivec_vsplth): Use const_0_to_7_operand instead of
3925         u5bit_cint_operand.
3926         (*altivec_vsplth_internal): Ditto.
3927         (altivec_vsplth_direct): Ditto.
3928         (altivec_vspltw): Use const_0_to_3_operand instead of
3929         u5bit_cint_operand.
3930         (*altivec_vspltw_internal): Ditto.
3931         (altivec_vspltw_direct): Ditto.
3932         (altivec_vspltsf): Ditto.
3933         (*altivec_vspltsf_internal): Ditto.
3934         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3935         various splats with the proper size immediate.  Reorder the various
3936         cases by ascending size of immediate, and put all such together.
3938 2018-09-25  Richard Biener  <rguenther@suse.de>
3940         PR debug/83941
3941         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3942         GC-ification.
3943         (maybe_create_die_with_external_ref): Do not create
3944         DW_TAG_imported_unit here.
3945         (add_abstract_origin_attribute): Handle external BLOCK refs.
3946         (dwarf2out_abstract_function): Simplify LTO case.
3947         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3948         rather than using maybe_create_die_with_external_ref.
3950 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
3952         PR target/71278
3953         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3954         (frndintxf2_mask_pm_i387): Ditto.
3955         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3956         Enable for !flag_trapping_math.
3957         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3958         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3959         Change operand 1 predicate to nonimmediate_operand.
3960         (attr "i387_cw"): Remove mask_pm.
3961         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3962         (enum ix86_entity): Remove I387_MASK_PM.
3963         * config/i386/i386.c (ix86_i387_mode_needed): Do not
3964         handle I387_MASK_PM.
3965         (ix86_mode_needed): Ditto.
3966         (ix86_mode_after): Ditto.
3967         (ix86_mode_entry): Ditto.
3968         (ix86_mode_exit): Ditto.
3969         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3971 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
3973         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3974         to_update_switch_stmts to vNULL instead of calling create on them
3975         immediately.
3977 2018-09-25  Richard Biener  <rguenther@suse.de>
3979         PR tree-optimization/87402
3980         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3981         (visit_phi): Re-instantiate handling of supposed to be VARYING
3982         but non-VARYING backedge value.
3984 2018-09-25  Richard Biener  <rguenther@suse.de>
3986         PR debug/83941
3987         * dwarf2out.c (struct sym_off_pair): New.
3988         (external_die_map): New global.
3989         (lookup_decl_die): When in LTO create DIEs lazily from the
3990         external_die_map.
3991         (lookup_block_die): New function, create DIEs lazily in LTO.
3992         (equate_block_to_die): New function.
3993         (dwarf2out_die_ref_for_decl): During WPA get the association
3994         from the external DIE map.
3995         (dwarf2out_register_external_die): Record mapping into the
3996         external DIE map.
3997         (maybe_create_die_with_external_ref): New function split out from
3998         DIE generation part of old dwarf2out_register_external_die.
3999         (add_abstract_origin_attribute): Do not return the DIE.  When
4000         in LTO reference externals directly.
4001         (dwarf2out_abstract_function): When in LTO ignore calls for
4002         decls with external DIEs (already present abstract instances).
4003         (gen_call_site_die): Adjust.
4004         (add_high_low_attributes): Likewise.
4005         (gen_lexical_block_die): Likewise.
4006         (gen_inlined_subroutine_die): Likewie.
4007         (gen_block_die): Likewise.
4008         (dwarf2out_inline_entry): Likewise.
4009         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
4010         DIEs.
4012 2018-09-25  Martin Liska  <mliska@suse.cz>
4014         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
4015         integers and not by a float value.
4017 2018-09-25  Martin Liska  <mliska@suse.cz>
4019         PR fortran/87394
4020         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
4021         instead of NULL.
4022         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
4023         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
4024         * godump.c (go_format_type): Remove extra parenthesis.
4026 2018-09-25  Martin Liska  <mliska@suse.cz>
4028         * alias.c (set_dest_equal_p): Remove unused function.
4029         * config/i386/i386.c (def_builtin_pure2): Likewise.
4030         * diagnostic-show-locus.c (class layout): Remove
4031         unused field.
4032         (layout::layout): Likewise here.
4033         * dump-context.h (class temp_dump_context): Likewise.
4034         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
4035         (add_AT_loclistsptr): Likewise.
4036         (add_AT_offset): Likewise.
4037         (get_AT_hi_pc): Likewise.
4038         (is_comdat_die): Likewise.
4039         (type_is_enum): Likewise.
4040         (ceiling): Likewise.
4041         (add_AT_vms_delta): Likewise.
4042         (is_class_die): Likewise.
4043         * edit-context.c (class line_event): Remove unused field.
4044         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
4045         unused function.
4046         * ipa-cp.c (ipa_get_vr_lat): Likewise.
4047         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
4048         (ok_for_base_p_nonstrict): Likewise.
4049         * tree-chrec.c (is_not_constant_evolution): Likewise.
4050         (chrec_fold_poly_cst): Likewise.
4051         * tree-if-conv.c (has_pred_critical_p): Likewise.
4052         * tree-ssa-coalesce.c (print_exprs): Likewise.
4053         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
4054         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
4055         * tree-vrp.c (value_ranges_intersect_p): Likewise.
4056         (value_range_nonnegative_p): Likewise.
4058 2018-09-25  Martin Liska  <mliska@suse.cz>
4060         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
4061         Do not handle "GNU Pascal".
4062         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
4063         Likewise.
4064         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
4065         from documentation. Likewise.
4066         * dbxout.c (dbxout_range_type): Likewise.
4067         * doc/cpp.texi: Likewise.
4068         * doc/extend.texi: Likewise.
4069         * doc/frontends.texi: Likewise.
4070         * doc/invoke.texi: Remove Pascal entry.
4071         * tree.def (CLEANUP_POINT_EXPR): Likewise.
4072         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
4074 2018-09-25  Martin Liska  <mliska@suse.cz>
4076         PR middle-end/86078
4077         * doc/invoke.texi: Document all parameters and remove default
4078         of the parameters.
4080 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
4082         PR bootstrap/87417
4083         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
4084         contains HOST_WIDE_INTs when computing its size.
4086 2018-09-24  Jim Wilson  <jimw@sifive.com>
4088         PR target/87391
4089         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
4090         not TARGET_RVE.
4091         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
4093 2018-09-24  Andrew Pinski  <apinski@marvell.com>
4095         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
4096         access prev before checking it for NULLness in the
4097         AARCH64_FUSE_CMP_BRANCH case.
4099 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
4101         PR target/82699
4102         * config/i386/i386.c (rest_of_insert_endbranch): Set
4103         endbr_queued_at_entrance to true and don't insert ENDBR if
4104         x86_function_profiler will be called.
4105         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
4106         is true.
4107         * config/i386/i386.h (machine_function): Add
4108         endbr_queued_at_entrance.
4110 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
4112         * genattrtab.c (mk_attr_alt): Use alternative_mask.
4113         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
4114         types.
4115         (check_attr_test): Use alternative_mask.
4116         (get_attr_value): Likewise.
4117         (compute_alternative_mask): Use alternative_mask and XWINT.
4118         (make_alternative_compare): Use alternative_mask.
4119         (attr_alt_subset_p): Use XWINT.
4120         (attr_alt_subset_of_compl_p): Likewise.
4121         (attr_alt_intersection): Use alternative_mask and XWINT.
4122         (attr_alt_union): Likewise.
4123         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
4124         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
4125         (simplify_test_exp): Use alternative_mask and XWINT.
4126         (write_test_expr): Use alternative_mask and XWINT, adjust bit
4127         number calculation to support 64 bits.  Generate code that
4128         checks 64-bit masks.
4129         (main): Use alternative_mask.
4130         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
4132 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
4134         PR target/80080
4135         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
4136         RETURN+USE when returning via %r14.
4138 2018-09-24  Martin Liska  <mliska@suse.cz>
4140         * gcov.c (output_lines): Print colorization legend
4141         for both flag_use_colors and flag_use_hotness_colors.
4142         Reword the help.
4144 2018-09-24  Martin Liska  <mliska@suse.cz>
4146         * coverage.c (get_coverage_counts): Use warning_at
4147         with current_function_decl location. Use %qD in warning
4148         message.
4150 2018-09-24  Martin Liska  <mliska@suse.cz>
4152         * memory-block.h (memory_block_pool::release): Annotate with
4153         valgrind that the memory is not accessible.
4155 2018-09-24  Martin Liska  <mliska@suse.cz>
4157         PR sanitizer/85774
4158         * asan.c: Make asan_handled_variables extern.
4159         * asan.h: Likewise.
4160         * cfgexpand.c (expand_stack_vars): Make sure
4161         a representative is unpoison if another
4162         variable in the partition is handled by
4163         use-after-scope sanitization.
4165 2018-09-24  Richard Biener  <rguenther@suse.de>
4167         PR tree-optimization/63155
4168         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
4169         the worklist when the edge of the respective argument isn't
4170         executable.
4172 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4174         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
4175         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
4176         (MASK_CLASS_P): Update for rename.
4177         (MAYBE_MASK_CLASS_P): Ditto.
4178         (REG_CLASS_NAMES): Update.
4179         (REG_CLASS_CONTENT): Update.
4180         * config/i386/i386.c (regclass_map): Update for MASK_REG
4181         and ALL_MASK_REGS rename.
4182         * config/i386/constraints.md (Yk): Update for rename.
4183         (k): Ditto.
4185 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4187         * config/i386/i386.h (enum reg_class): Remove
4188         EVEX_SSE_REGS and MOD4_SSE_REGS.
4189         (REG_CLASS_NAMES): Update.
4190         (REG_CLASS_CONTENT): Update.
4191         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
4192         registers as ALL_SSE_REGS.
4193         (ix86_additional_allocno_class_p): Remove.
4194         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
4195         (ix86_register_priority): Lower priority of EVEX SSE registers.
4196         Use IN_RANGE macro where appropriate.
4197         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
4198         AVX-5124VNNIW checks.
4199         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
4200         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
4201         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
4202         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
4203         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
4204         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
4205         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
4206         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
4207         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
4208         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
4209         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
4210         * config/i386/constraints.md (Yh): Remove.
4212 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4214         * config/i386/i386.c (regclass_map): Declare integer REX registers
4215         as GENERAL_REGS.
4217 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
4219         * doc/service.texi (Service): Switch the fsf.org link to https.
4221 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4223         PR target/86798
4224         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4225         Define to speculation_safe_value_not_needed.
4227 2018-09-21  Florian Weimer  <fweimer@redhat.com>
4229         PR middle-end/81035
4230         * doc/extend.texi (Common Function Attributes): Mention that
4231         noreturn suppresses tail call optimization.
4233 2018-09-21  Jeff Law  <law@redhat.com>
4235         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
4236         vr_values::cleanup_edges_and_switches.
4237         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
4238         vr_values class.
4239         (identify_jump_threads): Remove EDGE_IGNORE handling.
4240         (execute_vrp): Move handling of to_remove_edges and
4241         to_update_switch_stmts into vr_values class member functions.
4242         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
4243         (to_update_switch_stmts): Likewise.
4244         * vr-values.c: Include cfghooks.h.
4245         (vr_values::vr_values): Initialize to_remove_edges and
4246         to_update_switch_stmts.
4247         (vr_values::~vr_values): Verify to_remove_edges and
4248         to_update_switch_stmts are empty.
4249         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
4250         (vr_values::cleanup_edges_and_switches): New member function.
4251         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
4252         function.  Add new data members.
4254 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
4256         PR tree-optimization/87309
4257         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
4258         calls with pflags and alt_flags.
4259         (selftest::test_capture_of_dump_calls): Add test of interaction of
4260         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
4262 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4264         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
4266 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4268         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
4270 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4272         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
4273         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
4274         (VXWORKS_LIBS_RTP): Minor reordering.
4276 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4278         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
4279         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
4281 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4283         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
4284         (PTRDIFF_TYPE): Likewise.
4286 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4288         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
4289         triplet, similar to support for VxWorks7.
4290         * config/vxworks-dummy.h: Provide a default definition
4291         of TARGET_VXWORKS64 to 0.
4293 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4295         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
4296         * config/vxworks-dummy.h: here.
4298 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4300         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
4302 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
4303             Bo Zhou  <zbo.zhou@hisilicon.com>
4305         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
4306         * config/aarch64/aarch64-tune.md: Regenerated.
4307         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
4308         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
4309         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
4311 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
4312             Julian Brown  <julian@codesourcery.com>
4314         * builtins.c (get_builtin_sync_mem): Handle address spaces.
4316 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4318         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
4319         if the call takes a static chain.
4321 2018-09-21  Martin Liska  <mliska@suse.cz>
4323         * auto-profile.c (autofdo_source_profile::read): Do not
4324         set sum_all.
4325         (read_profile): Do not add working sets.
4326         (read_autofdo_file): Remove sum_all.
4327         (afdo_callsite_hot_enough_for_early_inline): Remove const
4328         qualifier.
4329         * coverage.c (struct counts_entry): Remove gcov_summary.
4330         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
4331         do not support GCOV_TAG_PROGRAM_SUMMARY.
4332         (get_coverage_counts): Remove summary and expected
4333         arguments.
4334         * coverage.h (get_coverage_counts): Likewise.
4335         * doc/gcov-dump.texi: Remove -w option.
4336         * gcov-dump.c (dump_working_sets): Remove.
4337         (main): Do not support '-w' option.
4338         (print_usage): Likewise.
4339         (tag_summary): Likewise.
4340         * gcov-io.c (gcov_write_summary): Do not dump
4341         histogram.
4342         (gcov_read_summary): Likewise.
4343         (gcov_histo_index): Remove.
4344         (gcov_histogram_merge): Likewise.
4345         (compute_working_sets): Likewise.
4346         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
4347         it not obsolete.
4348         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
4349         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
4350         (GCOV_HISTOGRAM_SIZE): Remove.
4351         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
4352         (struct gcov_summary): Simplify rapidly just
4353         to runs and sum_max fields.
4354         (gcov_histo_index): Remove.
4355         (NUM_GCOV_WORKING_SETS): Likewise.
4356         (compute_working_sets): Likewise.
4357         * gcov-tool.c (print_overlap_usage_message): Remove
4358         trailing empty line.
4359         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
4360         (output_lines): Remove program related line.
4361         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
4362         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
4363         histogram.
4364         (input_profile_summary): Do not read it.
4365         (merge_profile_summaries): And do not merge it.
4366         (input_symtab): Do not call removed function.
4367         * modulo-sched.c (sms_schedule): Do not print sum_max.
4368         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
4369         removed when histogram method was invented.
4370         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
4371         mode.
4372         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
4373         GCOV coding style.
4374         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
4375         and dump selected value.
4376         * profile.c (add_working_set): Remove.
4377         (get_working_sets): Likewise.
4378         (find_working_set): Likewise.
4379         (get_exec_counts): Do not work with working sets.
4380         (read_profile_edge_counts): Do not inform as sum_max is removed.
4381         (compute_branch_probabilities): Likewise.
4382         (compute_value_histograms): Remove argument for call of
4383         get_coverage_counts.
4384         * profile.h: Do not make gcov_summary const.
4386 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
4388         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
4390 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4392         PR tree-optimization/86990
4393         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
4394         Check that the entire merged store group is made of constants only for
4395         overlapping stores.
4397 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
4399         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
4400         (VTABLE_VERIFICATION_SPEC): Likewise.
4401         (SANITIZER_EARLY_SPEC): Likewise.
4402         (SANITIZER_SPEC): Likewise.
4403         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
4404         * doc/invoke.texi (Link Options): Document -r.
4406 2018-09-20  Richard Biener <rguenther@suse.de>
4408         PR middle-end/87054
4409         * gimplify.c (gimplify_expr): Retain alignment of
4410         addressable lvalue in dereference.
4412 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
4414         PR bootstrap/87013
4415         * configure.ac: Check for .loc is_stmt support.
4416         * configure, config.in: Rebuilt.
4417         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
4418         if not supported.
4420 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4422         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
4423         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
4424         -misel=no.
4426 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4428         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4429         VECTOR_OTHER.
4430         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
4431         case VECTOR_OTHER.
4433 2018-09-20  Marek Polacek  <polacek@redhat.com>
4435         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
4437 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4439         PR tree-optimization/87288
4440         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
4441         into account when determining PEELING_FOR_NITERS.
4443 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4445         PR tree-optimization/86877
4446         * tree-vect-loop.c (vect_analyze_loop_2): Call
4447         vect_verify_datarefs_alignment.
4449 2018-09-19  Marek Polacek  <polacek@redhat.com>
4451         * doc/invoke.texi: Document -Wclass-conversion.
4453 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
4455         * config/pa/pa.c (pa_adjust_priority): Delete.
4456         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
4458         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
4459         (atomic_storehi): Likewise.
4460         (atomic_storesi): Likewise.
4461         (atomic_loaddi): Restore compare and swap exchange loop code.
4463 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
4465         PR rtl-optimization/86902
4466         * combine.c (try_combine): When changing the CC mode used, don't change
4467         an unrelated mode in other_insn to that new CC mode.
4469 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
4471         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
4472         with %T in place of calls to dump_generic_expr.
4473         (prune_runtime_alias_test_list): Likewise.
4474         (create_runtime_alias_checks): Likewise.
4475         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
4476         (vect_analyze_data_ref_dependence): Likewise.
4477         (vect_slp_analyze_data_ref_dependence): Likewise.
4478         (vect_record_base_alignment): Likewise.  Use %G in place of call
4479         to dump_gimple_stmt.
4480         (vect_compute_data_ref_alignment): Likewise.
4481         (verify_data_ref_alignment): Likewise.
4482         (vect_find_same_alignment_drs): Likewise.
4483         (vect_analyze_group_access_1): Likewise.
4484         (vect_analyze_data_ref_accesses): Likewise.
4485         (dependence_distance_ge_vf): Likewise.
4486         (dump_lower_bound): Likewise.
4487         (vect_prune_runtime_alias_test_list): Likewise.
4488         (vect_find_stmt_data_reference): Likewise.
4489         (vect_analyze_data_refs): Likewise.
4490         (vect_create_addr_base_for_vector_ref): Likewise.
4491         (vect_create_data_ref_ptr): Likewise.
4492         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
4493         (vect_can_advance_ivs_p): Likewise.
4494         (vect_update_ivs_after_vectorizer): Likewise.
4495         (vect_gen_prolog_loop_niters): Likewise.
4496         (vect_prepare_for_masked_peels): Likewise.
4497         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4498         (vect_determine_vectorization_factor): Likewise.
4499         (vect_is_simple_iv_evolution): Likewise.
4500         (vect_analyze_scalar_cycles_1): Likewise.
4501         (vect_analyze_loop_operations): Likewise.
4502         (report_vect_op): Likewise.
4503         (vect_is_slp_reduction): Likewise.
4504         (check_reduction_path): Likewise.
4505         (vect_is_simple_reduction): Likewise.
4506         (vect_create_epilog_for_reduction): Likewise.
4507         (vect_finalize_reduction:): Likewise.
4508         (vectorizable_induction): Likewise.
4509         (vect_transform_loop_stmt): Likewise.
4510         (vect_transform_loop): Likewise.
4511         (optimize_mask_stores): Likewise.
4512         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
4513         (vect_split_statement): Likewise.
4514         (vect_recog_over_widening_pattern): Likewise.
4515         (vect_recog_average_pattern): Likewise.
4516         (vect_determine_min_output_precision_1): Likewise.
4517         (vect_determine_precisions_from_range): Likewise.
4518         (vect_determine_precisions_from_users): Likewise.
4519         (vect_mark_pattern_stmts): Likewise.
4520         (vect_pattern_recog_1): Likewise.
4521         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4522         (vect_record_max_nunits): Likewise.
4523         (vect_build_slp_tree_1): Likewise.
4524         (vect_build_slp_tree_2): Likewise.
4525         (vect_print_slp_tree): Likewise.
4526         (vect_analyze_slp_instance): Likewise.
4527         (vect_detect_hybrid_slp_stmts): Likewise.
4528         (vect_detect_hybrid_slp_1): Likewise.
4529         (vect_slp_analyze_operations): Likewise.
4530         (vect_slp_analyze_bb_1): Likewise.
4531         (vect_transform_slp_perm_load): Likewise.
4532         (vect_schedule_slp_instance): Likewise.
4533         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4534         (vect_mark_stmts_to_be_vectorized): Likewise.
4535         (vect_init_vector_1): Likewise.
4536         (vect_get_vec_def_for_operand): Likewise.
4537         (vect_finish_stmt_generation_1): Likewise.
4538         (vect_check_load_store_mask): Likewise.
4539         (vectorizable_call): Likewise.
4540         (vectorizable_conversion): Likewise.
4541         (vectorizable_operation): Likewise.
4542         (vectorizable_load): Likewise.
4543         (vect_analyze_stmt): Likewise.
4544         (vect_is_simple_use): Likewise.
4545         (vect_get_vector_types_for_stmt): Likewise.
4546         (vect_get_mask_type_for_stmt): Likewise.
4547         * tree-vectorizer.c (increase_alignment): Likewise.
4549 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
4551         * doc/rtl.texi: Adjust vec_select description.
4552         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
4553         non-constant selectors.
4555 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
4557         * config/aarch64/aarch64-protos.h
4558         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
4559         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
4560         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
4561         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
4562         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
4563         (AARCH64_FL_PROFILE): Move index so flags are ordered.
4564         (AARCH64_ISA_RCPC8_4): New flag.
4565         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
4566         to aarch64_offset_9bit_signed_unscaled_p.
4567         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
4568         and use stlur.
4569         * config/aarch64/constraints.md (Ust): New constraint.
4570         * config/aarch64/predicates.md.
4571         (aarch64_9bit_offset_memory_operand): New predicate.
4572         (aarch64_rcpc_memory_operand): New predicate.
4574 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
4576         PR rtl-optimization/87361
4577         * rtlanal.c (nonzero_bits1): Revert accidental change.
4579 2018-09-19  Richard Biener  <rguenther@suse.de>
4581         PR tree-optimization/87349
4582         PR tree-optimization/87342
4583         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
4585 2018-09-18  Marek Polacek  <polacek@redhat.com>
4587         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
4588         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
4590 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4592         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
4594 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4596         PR rtl-optimization/86882
4597         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
4599 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
4601         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
4602         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
4604 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
4606         PR other/87353
4607         * doc/invoke.texi (Link Options): Fix formatting and grammar.
4609 2018-09-18  Richard Biener  <rguenther@suse.de>
4611         PR middle-end/63155
4612         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
4613         (compute_samebase_partition_bases): Likewise.
4614         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
4615         (gimple_can_coalesce_p): Simplify.
4617 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
4619         Handle a library implementation of ffs calling __builtin_ffs.
4620         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
4621         (mmix_init_libfuncs): New function: make __builtin_ffs expand
4622         to __ffsdi2.
4624 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4626         * diagnostic-show-locus.c (class layout_range): Add field
4627         "m_original_idx".
4628         (layout_range::layout_range): Add "original_idx" param and use it
4629         to initialize new field.
4630         (make_range): Use 0 for original_idx.
4631         (layout::layout): Pass in index to calls to
4632         maybe_add_location_range.
4633         (layout::maybe_add_location_range): Add param "original_idx" and
4634         pass it on to layout_range.
4635         (layout::print_any_labels): Pass on range->m_original_idx to
4636         get_text call.
4637         (gcc_rich_location::add_location_if_nearby): Use 0 for
4638         original_idx.
4639         * gcc-rich-location.h (text_range_label::get_text): Update for new
4640         param.
4641         (range_label_for_type_mismatch::get_text): Likewise.
4643 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4645         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4647 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4649         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4650         format_string_diagnostic_t.
4651         (fmtwarn_n): Likewise.
4652         * substring-locations.c
4653         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4654         (format_warning_n_va): Convert to...
4655         (format_string_diagnostic_t::emit_warning_n_va): ...this.
4656         (format_warning_va): Convert to...
4657         (format_string_diagnostic_t::emit_warning_va): ...this.
4658         (format_warning_at_substring): Convert to...
4659         (format_string_diagnostic_t::emit_warning): ...this.
4660         (format_warning_at_substring_n): Convert to...
4661         (format_string_diagnostic_t::emit_warning_n): ...this.
4662         * substring-locations.h (class format_string_diagnostic_t): New
4663         class.
4664         (format_warning_va): Convert to
4665         format_string_diagnostic_t::emit_warning_va.
4666         (format_warning_n_va): Convert to
4667         format_string_diagnostic_t::emit_warning_n_va.
4668         (format_warning_at_substring): Convert to
4669         format_string_diagnostic_t::emit_warning.
4670         (format_warning_at_substring_n): Convert to
4671         format_string_diagnostic_t::emit_warning_n.
4673 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
4674             Bernd Schmidt <bernds_cb1@t-online.de>
4676         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4677         SImode args.
4679 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4681         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4682         operand 0 predicate to nonimmediate operand.
4683         (rint<mode>2_frndint): Remove insn pattern.
4684         (rint<mode>2): Change operand 1 predicate to general_operand.
4685         Extend operand 1 to XFmode and generate rintxf2 insn.
4686         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4687         Do not use X87MODEF mode macro.
4688         (frndintxf2_<rounding>_i387): Rename from
4689         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
4690         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4691         to XFmode and generate significandxf3 insn.
4693 2018-09-17  Richard Biener  <rguenther@suse.de>
4695         PR tree-optimization/87328
4696         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4697         visiting unexecutable backedges when not iterating.
4698         (do_rpo_vn): Mark all edges not executable even when not
4699         iterating.
4701 2018-09-17  Martin Jambor  <mjambor@suse.cz>
4703         PR c/63886
4704         * doc/invoke.texi (Warning Options): Likewise.
4706 2018-09-17  Richard Biener  <rguenther@suse.de>
4708         PR tree-optimization/87301
4709         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4710         clean EH info from leftover copy assignments.
4712 2018-09-17  Martin Liska  <mliska@suse.cz>
4714         PR gcov-profile/85871
4715         * gcov.c (output_intermediate_file): Fix out of bounds
4716         access.
4718 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
4720         * config/arc/arc.c: Object attributes for core4 not reflected
4721         correctly.
4722         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4723         core3).
4725 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
4727         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4729 2018-09-17  Martin Liska  <mliska@suse.cz>
4731         * doc/gcov.texi: Document new option --use-hotness-colors.
4732         * gcov.c (struct source_info): Declare new field.
4733         (source_info::source_info): Set default for maximum_count.
4734         (print_usage): Add new -q option.
4735         (process_args): Process it.
4736         (accumulate_line_info): Save src->maximum_count.
4737         (output_line_beginning): Make color line number if
4738         flag_use_hotness_colors is set.
4739         (output_line_details): Pass default argument value.
4740         (output_lines): Pass src->maximum_count.
4742 2018-09-17  Martin Liska  <mliska@suse.cz>
4744         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4745         Use processor_names table.
4746         * config/i386/i386.c (ix86_default_align): Use
4747         processor_cost_table for alignment values.
4748         (ix86_option_override_internal): Use processor_names.
4749         (ix86_function_specific_print): Likewise.
4750         * config/i386/i386.h (struct processor_costs):
4751         Add alignment values.
4752         (struct ptt): Remove and replace with const char *.
4753         * config/i386/x86-tune-costs.h (struct processor_costs):
4754         Declare default alignments for all costs.
4756 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
4758         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4759         symbolics or VR_VARYING ranges for ABS_EXPR.
4760         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4761         when range will wrap.
4763 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
4765         PR middle-end/86864
4766         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4767         before and after a JUMP_TABLE_DATA.
4769 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
4771         PR middle-end/87188
4772         * dojump.c (do_compare_and_jump): Canonicalize function pointers
4773         when one operand is a function pointer.  Use POINTER_TYPE_P and
4774         FUNC_OR_METHOD_TYPE_P.
4775         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4776         * fold-const.c (build_range_check): Likewise.
4777         * match.pd (simple_comparison): Likewise.
4779 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
4781         PR c/82967
4782         * spellcheck.c (get_edit_distance_cutoff): New function.
4783         (selftest::test_edit_distance_unit_test_oneway): Rename to...
4784         (selftest::test_get_edit_distance_one_way): ...this.
4785         (selftest::test_get_edit_distance_unit): Rename to...
4786         (selftest::test_get_edit_distance_both_ways): ...this.
4787         (selftest::test_edit_distances): Move tests to this new function,
4788         and test some more pairs of strings.  Update for above renaming.
4789         (selftest::get_old_cutoff): New function.
4790         (selftest::test_get_edit_distance_cutoff): New function.
4791         (selftest::assert_suggested_for): New function.
4792         (ASSERT_SUGGESTED_FOR): New macro.
4793         (selftest::assert_not_suggested_for): New function.
4794         (ASSERT_NOT_SUGGESTED_FOR): New macro.
4795         (selftest::test_suggestions): New function.
4796         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4797         tests to selftest::test_edit_distances and call it.  Add calls to
4798         selftest::test_get_edit_distance_cutoff and
4799         selftest::test_suggestions.
4800         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4801         (best_match::consider): Replace hard-coded cutoff calculation with
4802         a call to...
4803         (best_match::get_cutoff): New declaration.
4804         (best_match::get_best_meaningful_candidate): Likewise.
4806 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4808         * builtins.c (fold_builtin_strlen): Remove TODO comment.
4810 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4812         revert:
4813         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4815         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4816         terminated string literal.
4818 2018-09-14  Martin Sebor  <msebor@redhat.com>
4820         * builtins.c (unterminated_array): Handle ARRAY_REF.
4821         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4822         * builtins.h (unterminated_array): Declare extern.
4823         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4824         arrays.
4825         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4826         calls.
4828 2018-09-14  Martin Sebor  <msebor@redhat.com>
4829             Jeff Law  <law@redhat.com>
4831         * builtins.c (unterminated_array): New.
4832         (expand_builtin_strcpy): Adjust.
4833         (expand_builtin_strcpy_args): Detect unterminated arrays.
4834         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
4835         unterminated arrays.
4836         * gimple-fold.h (get_maxval_strlen): Add argument.
4837         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4839         * gimple-fold.c (get_range_strlen): Add argument.
4840         (get_maxval_strlen): Adjust.
4841         * gimple-fold.h (get_range_strlen): Add argument.
4843 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
4845         * config/i386/movdirintrin.h: Fix copyright year.
4847 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
4849         * reg-stack.c: Include regs.h.
4850         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4851         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4852         FIRST_STACK_REG, not DFmode.
4853         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4854         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
4855         (change stack): Default register mode to the reg_raw_mode of
4856         FIRST_STACK_REG, not DFmode.
4857         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4858         (*swapxf): Rename from swapxf.
4860 2018-09-14  Carl Love  <cel@us.ibm.com>
4862         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4863         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4865 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
4867         PR target/87224
4868         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4869         alternatives.
4871 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
4873         PR target/85628
4874         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4876 2018-09-14  Jason Merrill  <jason@redhat.com>
4878         Fix --enable-gather-detailed-mem-stats.
4879         * hash-table.c (hash_table_usage): Change from variable to function.
4880         * hash-table.h: Adjust.
4881         * Makefile.in: Add missing dependencies on hash-table.h.
4883 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4885         PR tree-optimization/87259
4886         PR lto/87283
4887         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4888         execute_cse_reciprocals_1 has tried transforming.
4890 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
4892         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4893         VR_VARYING for PLUS/MINUS_EXPR.
4895 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4897         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4898         formatting.
4900 2018-09-14  Richard Biener  <rguenther@suse.de>
4902         PR middle-end/63155
4903         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4904         bits for the merged partition.
4906 2018-09-13  Martin Sebor  <msebor@redhat.com>
4907             Bernd Edlinger  <bernd.edlinger@hotmail.de>
4909         * builtins.h (c_srlen): Add argument.
4910         * builtins.c (warn_string_no_nul): New function.
4911         (c_strlen): Add argument and use it.  Update recursive calls.
4912         Pass DECL argument to string_constant to get info on non
4913         terminated strings.  Update *NONSTR as needed.
4914         (fold_builtin_strlen): Add argument to calls to c_strlen.
4915         Warn for unterminated arrays.
4916         (warn_string_no_null): Add prototype.
4917         * expr.c (string_constant): Update arguments.  Update recursive
4918         calls appropriately.  Detect missing NUL terminator and outermost
4919         declaration its missing in.
4920         Improve checks for arrays with nonzero lower bound or elements
4921         that are not a single byte.  Simplify offset computation.
4922         Simplify checks for non-NUL terminated strings.
4923         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4924         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4926 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4928         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4929         correctly.
4930         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
4931         argument.  Fix range checks.
4932         * fold-const.h (c_getstr): Adjust protoype.
4933         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4934         string is constant but contains no NUL byte.
4936         * expr.c (string_constant): Adjust function comment.
4937         Remove bogus check for zero termination.
4939         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4941         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4942         (get_constant_size): Don't make STRING_CSTs larger than they are.
4943         (check_string_literal): New check function for STRING_CSTs.
4944         (output_constant): Use it.
4946 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
4948         PR target/86812
4949         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4951 2018-09-13  Richard Biener  <rguenther@suse.de>
4953         PR tree-optimization/87263
4954         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4955         (struct unwind_state): Add max_rpo field.
4956         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4957         Compute max_rpo, the max RPO number a block can be backwards reached
4958         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
4959         separating it from the iterating mode.
4961 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4963         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4964         (rfs_decision): New scheduling decision.
4966 2018-09-13  Richard Biener  <rguenther@suse.de>
4968         PR bootstrap/87134
4969         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4970         (vn_nary_op_insert_pieces_predicated): Do not write useless
4971         valid_dominated_by_p entry outside of the allocated storage.
4973 2018-09-13  Omar Sandoval  <osandov@osandov.com>
4974             Tom de Vries  <tdevries@suse.de>
4976         PR debug/86985
4977         * dwarf2out.c (is_c): New function.
4978         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4980 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
4982         PR target/85628
4983         * config/aarch64/aarch64.md (*aarch64_bfxil):
4984         Define.
4985         * config/aarch64/constraints.md (Ulc): Define.
4986         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4987         Define.
4988         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4989         New function.
4991 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4993         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4994         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4995         aarch64_layout_frame call.
4996         (aarch64_expand_epilogue): Likewise.
4997         (aarch64_initial_elimination_offset): Likewise.
4998         (aarch64_get_separate_components): Likewise.
4999         (aarch64_use_return_insn_p): Likewise.
5000         (aarch64_layout_frame): Remove unneeded check.
5002 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5004         * configure.ac: Only append
5005         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
5006         gcc_config_arguments if it was never reconfigured or last reconfigure
5007         was with different arguments.
5008         * configure: Regenerated.
5010 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5011             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5013         PR middle-end/87290
5014         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
5015         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
5017 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5019         PR tree-optimization/87287
5020         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
5021         X % C == 0 to X % (unsigned) C == 0 optimization to ...
5022         * match.pd (X % C == 0): ... here.  New optimization.
5024 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5026         PR middle-end/82853
5027         * expr.h (maybe_optimize_mod_cmp): Declare.
5028         * expr.c (mod_inv): New function.
5029         (maybe_optimize_mod_cmp): New function.
5030         (do_store_flag): Use it.
5031         * cfgexpand.c (expand_gimple_cond): Likewise.
5033 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
5034             Julian Brown  <julian@codesourcery.com>
5036         PR middle-end/86336
5037         * gimplify.c (gimplify_scan_omp_clauses): Set
5038         target_firstprivatize_array_bases in OpenACC parallel and kernels
5039         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
5040         OpenACC data regions.
5042 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
5044         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
5045         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
5047 2018-09-12  Richard Biener  <rguenther@suse.de>
5049         PR tree-optimization/87280
5050         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
5051         edge but unreachable target.
5052         (do_rpo_vn): For conservatively handling a PHI only mark
5053         the backedge executable but not the block reachable.
5055 2018-09-12  Richard Biener  <rguenther@suse.de>
5057         PR tree-optimization/87266
5058         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
5059         visited blocks.
5061 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
5063         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
5064         constants.
5065         ("trunc<BFP:mode><DFP_ALL:mode>2")
5066         ("trunc<DFP_ALL:mode><BFP:mode>2")
5067         ("extend<BFP:mode><DFP_ALL:mode>2")
5068         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
5069         according to the target operand type.
5071 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5072             Andreas Krebbel  <krebbel@linux.ibm.com>
5074         PR tree-optimization/86844
5075         * gimple-ssa-store-merging.c
5076         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
5077         there are any overlapping stores in between them, make sure they are
5078         also coalesced or we give up completely.
5080 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5082         PR middle-end/87248
5083         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
5084         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
5086 2018-09-12  Tom de Vries  <tdevries@suse.de>
5088         * common.opt (gdescribe-dies): Add option.
5089         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
5090         attribute for artifical and nameless decls.
5091         (dwarf2out_register_external_die): Add description attribute to
5092         external reference die.
5093         (add_desc_attribute): New functions.
5094         (gen_subprogram_die): Add description attribute to
5095         DW_TAG_call_site_parameter.
5096         * tree-pretty-print.c (print_generic_expr_to_str): New function.
5097         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
5098         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
5099         -gno-describe-dies.
5100         (@item -gdescribe-dies): Add.
5102 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
5104         * tree-vrp.c (vrp_shift_undefined_p): Remove.
5105         (extract_range_from_binary_expr_1: Call
5106         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
5107         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
5108         depend on sign.
5110 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
5112         * gimple-ssa-warn-alloca.c
5113         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
5114         field for ALLOCA_BOUND_*_LARGE.
5116 2018-09-11  Nathan Sidwell  <nathan@acm.org>
5118         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
5120 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
5122         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
5123         for clobbers.  Remove obsolete comment.
5125 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
5127         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
5128         mpxchk, mpxld and mpxst types.
5129         (define_attr length_immediate): Remove all processing of mpx types.
5130         (define_attr prefix_0f): Ditto.
5131         (define_attr memory): Ditto.
5133 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
5135         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
5136         (log<mode>2): Change operand 1 predicate to general_operand.
5137         Extend operand 1 to XFmode and generate logxf3 insn.
5138         (log10<mode>2): Change operand 1 predicate to general_operand.
5139         Extend operand 1 to XFmode and generate log10xf3 insn.
5140         (log2<mode>2): Change operand 1 predicate to general_operand.
5141         Extend operand 1 to XFmode and generate log2xf3 insn.
5142         (fyl2xp1_extend<mode>xf3_i387): Remove.
5143         (log1p<mode>2): Change operand 1 predicate to general_operand.
5144         Extend operand 1 to XFmode and generate log1pxf3 insn.
5145         (fxtract_extend<mode>xf3_i387): Remove.
5146         (logb<mode>2): Change operand 1 predicate to general_operand.
5147         Extend operand 1 to XFmode and generate logbxf3 insn.
5148         (ilogb<mode>2): Change operand 1 predicate to general_operand.
5149         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
5150         (significand<mode>2): Change operand 1 predicate to general_operand.
5151         Extend operand 1 to XFmode and generate significandxf3 insn.
5153 2018-09-11  Nathan Sidwell  <nathan@acm.org>
5155         * gcc.c (perror_with_name, pfatal_with_name): Delete.
5156         (load_specs): Use fatal_error.
5157         (DELETE_IF_ORDINARY, process_command): Use error.
5158         (execute, run_attempt): Use fatal_error.
5160 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
5162         * diagnostic-core.h (sorry_at): New prototype.
5163         * diagnostic.c (sorry_at): New function.
5165 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
5167         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
5168         by zero as VR_UNDEFINED.
5170 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
5172         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
5173         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
5174         (<sincos>mode2): New expander.
5175         (sincos_extend<mode>xf3_i387): Remove insn pattern.
5176         (sincos -> sin, cos splitters): Remove splitter patterns.
5177         (sincos<mode>3): Change operand 2 predicate to general_operand.
5178         Extend operand 2 to XFmode and generate sincosxf3 insn.
5179         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
5180         Change operand 3 predicate to const1_operand.
5181         (fptan_extend<mode>xf4_i387): Remove insn pattern.
5182         (tanxf2): Update operands in the call to fptanxf4_i387.
5183         (tan<mode>2): Change operand 1 predicate to general_operand.
5184         Extend operand 1 to XFmode and generate tanxf3 insn.
5185         (atan2xf3): Rename from *fpatanxf3_i387.
5186         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
5187         (atan2xf3): Remove expander.
5188         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
5189         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
5190         (atan<mode>2): Change operand 1 predicate to general_operand.
5191         Extend operand 1 to XFmode and generate atanxf3 insn.
5193 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
5195         * config/i386/i386.md (x87/SSE constant load splitter): Use
5196         memory_operand instead of nonimmediate_operand for input operand
5197         predicate.
5199 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
5201         * config/i386/i386.md (float partial SSE register stall splitter): Move
5202         splitter near its instruction pattern.
5203         (float_extend partial SSE register stall splitter): Ditto.
5204         (float_truncate partial SSE register stall splitter): Ditto.
5206 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
5208         PR target/86794
5209         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5210         to speculation_safe_value_not_needed.
5212         PR target/85666
5213         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
5214         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
5215         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
5216         leaf_function_p, instead use has_hard_reg_initial_val.
5218 2018-09-09  Nathan Sidwell  <nathan@acm.org>
5220         * gcc.h (pfatal_with_name): Don't declare here.
5221         * gcc.c (pfatal_with_name): Make static.
5223 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
5225         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
5226         earlyclobber.
5228 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
5230         PR rtl-optimization/85458
5231         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
5232         priority hook to reduce the priority of EXPR.
5234 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
5236         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
5237         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
5238         with X87_ENABLE_FLOAT.  Remove preparation code.
5239         (*float<SWI48:mode><MODEF:mode>2): Rename from
5240         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
5241         math using "enabled" attribute.
5242         (*floatdi<MODEF:mode>2_i387): Rename from
5243         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
5244         enable for 32bit targets only.
5245         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
5246         splitter.
5247         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
5248         as operand 1 predicate.  Rewrite as define_insn_and_split.
5249         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
5251 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5253         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
5254         to fallthru to FLOAT case.
5256 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5258         PR target/86731
5259         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
5260         around folding of vec_sl to handle out of range shift values.
5262 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5264         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
5265         Update callers to gen_fix_trunc<mode>_i387_fisttp
5266         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
5267         nonimmediate_operand.
5268         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
5269         and corresponding splitters.
5270         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
5271         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
5272         (fix_truncdi_i387_with_temp): Remove insn pattern
5273         and corresponding splitters.
5274         (fix_trunc<mode>_i387): Change operand 0 predicate to
5275         nonimmediate_operand.
5276         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
5277         and corresponding splitters.
5278         (*fistdi2_1): Remove.
5279         (fistdi2): Ditto.
5280         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
5281         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
5282         (*fist<mode>2_1): Remove.
5283         (fist<mode>2): Ditto.
5284         (fist<mode>2_with_temp): Remove insn pattern and corresponding
5285         splitters.
5286         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
5287         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
5288         (fistdi2_<rounding>): Change operand 0 predicate to
5289         nonimmediate_operand.
5290         (fistdi2_<rounding>_with_temp): Remove insn pattern
5291         and corresponding splitters.
5292         (fist<mode>2_<rounding>): Change operand 0 predicate to
5293         nonimmediate_operand.
5294         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
5295         and corresponding splitters.
5297         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
5299 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5301         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
5302         the init value.
5304 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5306         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
5307         early gimple folding of vec_splat().
5308         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
5309         * gimple-fold.h: Add an extern define for tree_vec_extract().
5311 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5313         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
5314         wrappers around TREE_TYPE comparisons.
5316 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5318         PR target/80080
5319         * config/s390/predicates.md: Add nonsym_memory_operand.
5320         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
5321         contains a SYMBOL_REF, load it into an intermediate pseudo.
5322         (s390_emit_compare_and_swap): Legitimize operand.
5323         * config/s390/s390.md: Use the new nonsym_memory_operand
5324         with UNSPECV_CAS patterns.
5326 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5328         PR target/80080
5329         * config/s390/s390-passes.def: New file.
5330         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
5331         declaration.
5332         (make_pass_s390_early_mach): Add declaration.
5333         * config/s390/s390.c (make_pass_s390_early_mach):
5334         (s390_option_override): Remove dynamic registration.
5335         * config/s390/t-s390: Add s390-passes.def.
5337 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5339         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5340         Remove UNSPEC_LTREL_BASE check.
5341         (annotate_constant_pool_refs): Likewise.
5342         (find_constant_pool_ref): Likewise.
5343         (find_ltrel_base): Removed.
5344         (replace_ltrel_base): Removed.
5345         (s390_mainpool_finish): Remove replace_ltrel_base call.
5346         (s390_chunkify_start): Remove pending LTREL_BASE logic.
5347         (s390_chunkify_finish): Remove replace_ltrel_base call.
5348         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
5350 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
5352         PR target/86779
5353         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5354         to speculation_safe_value_not_needed.
5356 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
5357             Bernd Schmidt  <bernds_cb1@t-online.de>
5359         * config/nvptx/nvptx-opts.h: New file.
5360         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
5361         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
5362         (ASM_SPEC): Define.
5363         (TARGET_SM35): New macro.
5364         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
5365         correct predicate.
5366         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
5367         values.
5368         (misa=): New option.
5369         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
5371 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
5373         * config/i386/i386.md (truncdfsf2): Remove expander.
5374         (truncdfsf2_with_temp): Ditto.
5375         (truncxf<mode>2): Ditto.
5376         (*truncdfsf_fast_mixed): Remove insn pattern.
5377         (*truncdfsf_fast_i387): Ditto.
5378         (*truncdfsf_mixed): Ditto.
5379         (*truncdfsf_i387): Ditto.
5380         (*truncdfsf2_i387_1): Ditto.
5381         (*truncxfsf2_mixed): Ditto.
5382         (*truncxfdf2_mixed): Ditto.
5383         (*truncxf<mode>2_i387_noop): Ditto. Update callers
5384         to call gen_truncxf<mode>2 instead.
5385         (*truncxf<mode>2_i387): Remove.
5386         (reg->reg splitters): Remove splitter pattern.
5387         (reg->mem splitters): Ditto.
5389         (truncdfsf2): New insn pattern.
5390         (truncxf<mode>2): Ditto.
5392 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5394         * tree-ssa-math-opts.c (is_mult_by): New function.
5395         (is_square_of): Use the above.
5396         (optimize_recip_sqrt): New function.
5397         (pass_cse_reciprocals::execute): Use the above.
5399 2018-09-05  Richard Biener  <rguenther@suse.de>
5401         PR bootstrap/87134
5402         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
5403         to zero-init the emplaced vec.
5405 2018-09-05  Martin Liska  <mliska@suse.cz>
5407         PR tree-optimization/87205
5408         * tree-switch-conversion.c (pass_lower_switch::execute):
5409         Group cases for switch statements.
5411 2018-09-05  Richard Biener  <rguenther@suse.de>
5413         PR tree-optimization/87217
5414         * tree-ssa-sccvn.c (vuse_valueize): New.
5415         (vn_reference_lookup_pieces): Use it.
5416         (vn_reference_lookup): Likewise.
5418 2018-09-05  Nathan Sidwell  <nathan@acm.org>
5420         PR c++/87137
5421         * stor-layout.c (place_field): Scan forwards to check last
5422         bitfield when ms_bitfield_placement is in effect.
5424 2018-09-05  Richard Biener  <rguenther@suse.de>
5426         PR bootstrap/87225
5427         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
5428         return.
5430 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5431             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
5433         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
5434         * config.gcc (extra_objs): Build it.
5435         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
5436         Likewise.
5437         * config/aarch64/aarch64-passes.def
5438         (pass_tag_collision_avoidance): New pass.
5439         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
5440         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
5441         (aarch64_classify_address): Remove static qualifier.
5442         (aarch64_address_info, aarch64_address_type): Move to...
5443         * config/aarch64/aarch64-protos.h: ... here.
5444         (make_pass_tag_collision_avoidance): New function.
5445         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
5446         New tuning flag.
5448 2018-09-05  Martin Liska  <mliska@suse.cz>
5450         * doc/gcov.texi: Update documentation of humar
5451         readable mode.
5452         * gcov.c (format_count): Print one decimal place, it provides
5453         more fine number of situations like '1G' vs. '1.4G'.
5455 2018-09-05  Martin Liska  <mliska@suse.cz>
5457         PR target/87164
5458         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
5459         * optc-gen.awk: Allow 'Var' for Deprecated options in order
5460         to generate a MASK value.
5462 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
5464         PR debug/86593
5465         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
5466         if frame pointer isn't used.
5467         (compute_frame_pointer_to_fb_displacement): Likewise.
5469 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
5471         PR target/87198
5472         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
5473         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
5474         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
5475         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
5476         and OPTION_MASK_ISA_XSAVEC_UNSET.
5478 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
5480         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
5481         XOR operations in NAND case.
5483 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5485         * wide-int-range.cc (wide_int_range_convert): New.
5486         * wide-int-range.h (wide_int_range_convert): New.
5487         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
5488         code into wide_int_range_convert.
5489         (extract_range_into_wide_ints): Do not munge anti range constants
5490         into the entire domain.  Just return the range back.
5492 2018-09-04  Martin Liska  <mliska@suse.cz>
5494         * genmatch.c (output_line_directive): Add new argument
5495         fnargs.
5496         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
5498 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5500         * doc/invoke.texi (Option Summary): Add whitespace.
5502         * doc/invoke.texi (Option Summary): Add -Waligned-new.
5504 2018-09-04  Richard Biener  <rguenther@suse.de>
5506         PR tree-optimization/87211
5507         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
5508         backedge value we're supposed to treat as VARYING also number
5509         the PHI to VARYING in case it got a different value-number already.
5511 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5513         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
5514         (extract_range_from_binary_expr_1): Do not call
5515         vrp_can_optimize_bit_op.
5516         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
5517         static.
5518         (wide_int_range_get_mask_and_bounds): New.
5519         (wide_int_range_optimize_bit_op): New.
5520         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
5521         (wide_int_range_bit_and): Same.
5522         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
5523         (wide_int_range_optimize_bit_op): New.
5524         (wide_int_range_get_mask_and_bounds): New.
5526 2018-09-04  Richard Biener  <rguenther@suse.de>
5528         PR tree-optimization/87176
5529         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
5530         variable.  When value-numbering a virtual PHI node make sure
5531         to not value-number to the backedge value.
5533 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5535         * doc/extend.texi (Long Long, Hex Floats): Document support for
5536         long long and hex floats in more recent versions of ISO C++.
5538 2018-09-03  Richard Biener  <rguenther@suse.de>
5540         PR tree-optimization/87177
5541         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
5542         cleanup.
5544 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5546         * bb-reorder.c (edge_order): Convert to C-qsort-style
5547         tri-state comparator.
5548         (reorder_basic_blocks_simple): Change std::stable_sort to
5549         gcc_stablesort.
5551 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5553         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
5554         tri-state comparator.
5555         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
5557 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5559         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
5560         (mergesort): ... here as maximum count for using netsort.
5561         (gcc_qsort): Set nlim to 3 if stable sort is requested.
5562         (gcc_stablesort): New.
5563         * system.h (gcc_stablesort): Declare.
5565 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5567         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
5568         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
5569         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
5571 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
5573         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
5574         lxsdx and stxsdx alternatives.
5575         (*mov<mode>_hardfloat64): Ditto.
5576         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
5578 2018-09-03  Richard Biener  <rguenther@suse.de>
5580         PR tree-optimization/87200
5581         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
5582         simplify result.
5584 2018-09-03  Martin Liska  <mliska@suse.cz>
5586         PR tree-optimization/87201
5587         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5588         Fix parenthesis in an expression.
5590 2018-09-03  Richard Biener  <rguenther@suse.de>
5592         PR tree-optimization/87197
5593         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
5594         visited.  CSE the VN_INFO hashtable lookup.
5596         PR tree-optimization/87169
5597         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
5598         iterating make sure there's no extra backedges from irreducible
5599         regions feeding the header.  Mark the destination block
5600         executable.
5602 2018-09-03  Martin Liska  <mliska@suse.cz>
5604         PR driver/83193
5605         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
5606         * common/common-targhooks.c (default_get_valid_option_values):
5607         New function.
5608         * common/common-targhooks.h (default_get_valid_option_values):
5609         Likewise.
5610         * common/config/i386/i386-common.c: Move processor_target_table
5611         from i386.c.
5612         (ix86_get_valid_option_values): New function.
5613         (TARGET_GET_VALID_OPTION_VALUES): New macro.
5614         * config/i386/i386.c (struct ptt): Move to i386-common.c.
5615         (PTA_*): Move all defined masks into i386-common.c.
5616         (ix86_function_specific_restore): Use new processor_cost_table.
5617         * config/i386/i386.h (struct ptt): Moved from i386.c.
5618         (struct pta): Likewise.
5619         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
5620         * doc/tm.texi.in: Likewise.
5621         * opt-suggestions.c (option_proposer::suggest_option):
5622         Pass prefix to build_option_suggestions.
5623         (option_proposer::get_completions): Likewise.
5624         (option_proposer::build_option_suggestions): Use the new target
5625         hook.
5626         * opts.c (struct option_help_tuple): New struct.
5627         (print_filtered_help): Use the new target hook.
5629 2018-09-03  Martin Liska  <mliska@suse.cz>
5631         PR middle-end/59521
5632         * predict.c (set_even_probabilities): Add likely_edges
5633         argument and handle cases where we have precisely one
5634         likely edge.
5635         (combine_predictions_for_bb): Catch also likely_edges.
5636         (tree_predict_by_opcode): Handle gswitch statements.
5637         * tree-cfg.h (find_case_label_for_value): New declaration.
5638         (find_taken_edge_switch_expr): Likewise.
5639         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5640         Find pivot in decision tree based on probabily, not by number of
5641         nodes.
5643 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
5645         * doc/standards.texi (Standards): Update Objective-C reference.
5647 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5649         * doc/install.texi (Prerequisites): Update link for MPC.
5651 2018-09-01  Michael Matz  <matz@suse.de>
5653         PR tree-optimization/87074
5654         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5655         PHIs for outer-loop uses.
5657 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5659         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5660         * doc/invoke.texi (C Dialect Options): Ditto.
5662 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5664         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5666 2018-08-31  Richard Biener  <rguenther@suse.de>
5668         PR tree-optimization/87168
5669         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5670         (rpo_elim::eliminate_avail): When OP was not visited it must
5671         be available.
5673 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
5675         * tree-vrp.c (copy_value_range): Convert param "from" from
5676         "value_range *" to "const value_range *".
5677         (range_is_null): Likewise for param "vr".
5678         (range_int_cst_p): Likewise.
5679         (range_int_cst_singleton_p): Likewise.
5680         (symbolic_range_p): Likewise.
5681         (value_ranges_intersect_p): Likewise for both params.
5682         (value_range_nonnegative_p): Likewise for param "vr".
5683         (value_range_constant_singleton): Likewise.
5684         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5685         (extract_range_into_wide_ints): Likewise for param "vr".
5686         (extract_range_from_multiplicative_op): Likewise for params "vr0"
5687         and "vr1".
5688         (vrp_can_optimize_bit_op): Likewise.
5689         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5690         "vr1_".
5691         (extract_range_from_unary_expr): Likewise.
5692         (debug_value_range): Likewise for param "vr".
5693         (value_range::dump): Add "const" qualifier.
5694         (vrp_prop::check_array_ref): Convert local "vr" from
5695         "value_range *" to "const value_range *".
5696         (vrp_prop::check_mem_ref): Likewise.
5697         (vrp_prop::visit_stmt): Likewise for local "old_vr".
5698         (vrp_intersect_ranges_1): Likewise for param "vr_1".
5699         (vrp_intersect_ranges): Likewise.
5700         (simplify_stmt_for_jump_threading): Likewise for local "vr".
5701         (vrp_prop::vrp_finalize): Likewise.
5702         * tree-vrp.h (value_range::dump): Add "const" qualifier.
5703         (vrp_intersect_ranges): Add "const" qualifier to params as above.
5704         (extract_range_from_unary_expr): Likewise.
5705         (value_range_constant_singleton): Likewise.
5706         (symbolic_range_p): Likewise.
5707         (copy_value_range): Likewise.
5708         (extract_range_from_binary_expr_1): Likewise.
5709         (range_int_cst_p): Likewise.
5710         (vrp_set_zero_nonzero_bits): Likewise.
5711         (range_int_cst_singleton_p): Likewise.
5713 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
5715         * config/aarch64/arm_neon.h (vabsd_s64): New.
5716         (vnegd_s64): Likewise.
5718 2018-08-31  Martin Jambor  <mjambor@suse.cz>
5720         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5722 2018-08-31  Martin Liska  <mliska@suse.cz>
5724         * ipa-icf.c (sem_item::add_type): Use
5725         sem_item::m_type_hash_cache.
5726         * ipa-icf.h: Move the cache from sem_item_optimizer
5727         to sem_item.
5729 2018-08-31  Nathan Sidwell  <nathan@acm.org>
5731         * doc/extend.texi (Backwards Compatibility): Remove implicit
5732         extern C leeway of () being (...).
5734 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5736         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5738 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
5740         PR target/86684
5741         PR target/87149
5742         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5744 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
5746         PR middle-end/87138
5747         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5748         gen_int_mode.  Formatting fixes.
5750 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
5752         * target.def (custom_function_descriptors): Improve documentation.
5753         * doc/tm.texi.in (Trampolines): Expand discussion of function
5754         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5755         beginning of the section.
5756         * doc/tm.texi: Regenerated.
5758 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
5760         * cfg.h (class auto_edge_flag): Spell out the template-id of the
5761         base class in the initializer list.  This is a workaround for
5762         building with older GCC.
5763         (class auto_bb_flag): Likewise.
5765 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5767         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5768         (altivec_vcmpequ<VI_char>_p): Remove star.
5769         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5770         vector load modes.
5771         (expand_strncmp_vec_sequence): New function.
5772         (emit_final_str_compare_vec): New function.
5773         (expand_strn_compare): Add support for vector strncmp.
5774         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5775         length specification to bytes.
5776         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5777         (vcmpnezb_p): New pattern.
5778         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5779         for option -mstring-compare-inline-limit.
5781 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
5783         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5784         (PTA_SKYLAKE): Add PTA_AES.
5785         (PTA_GOLDMONT): Likewise.
5787 2018-08-29  Jan Hubicka  <jh@suse.cz>
5789         PR lto/86517
5790         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5791         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5793 2018-08-29  Jan Hubicka  <jh@suse.cz>
5795         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5796         TYPE_STUB_DECL.
5797         (hash_tree): Do not visit TYPE_STUB_DECL.
5798         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5799         stream TYPE_STUB_DECL.
5800         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5801         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5802         after free_lang_data.
5803         (type_in_anonymous_namespace_p): Likewise.
5805 2018-08-29  Jan Hubicka  <jh@suse.cz>
5807         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5808         comment that it has to be even number.
5809         (class sreal): Change m_sig type to int32_t.
5810         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5811         int64_t for temporary calculations.
5812         (sreal_verify_basics): Drop one bit from minimum and maximum.
5814 2018-08-30  Richard Biener  <rguenther@suse.de>
5816         PR tree-optimization/87147
5817         * tree-ssa-sccvn.c (SSA_VISITED): New function.
5818         (visit_phi): When the degenerate result is from the backedge and
5819         we didn't visit its definition yet drop to VARYING.
5820         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5822 2018-08-29  Jan Hubicka  <jh@suse.cz>
5824         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5825         DECL_VINDEX.
5826         (hash_tree): Likewise.
5828 2018-08-29  Jan Hubicka  <jh@suse.cz>
5830         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5831         and TYPE_NEXT_REF_TO.
5833 2018-08-29  Jan Hubicka  <jh@suse.cz>
5835         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5836         comment that it has to be even number.
5837         (class sreal): Change m_sig type to int32_t.
5838         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5839         int64_t for temporary calculations.
5840         (sreal_verify_basics): Drop one bit from minimum and maximum.
5842 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
5844         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5846 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
5848         PR middle-end/86995
5849         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5850         if to_add is negative.
5852 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5854         PR middle-end/87053
5855         * builtins.c (c_strlen): Improve range checks.
5857 2018-08-29  Martin Sebor  <msebor@redhat.com>
5858             Jeff Law  <law@redhat.com>
5860         PR tree-optimization/86714
5861         PR tree-optimization/86711
5862         * builtins.c (c_strlen): Add arguments to call to string_constant.
5863         * expr.c (string_constant): Add argument.  Detect missing nul
5864         terminator and outermost declaration it's missing in.
5865         * expr.h (string_constant): Add argument.
5866         * fold-const.c (read_from_constant_string): Add arguments to call to
5867         string_constant.
5868         (c_getstr): Likewise.
5869         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5870         to string_constant.
5871         * tree-ssa-strlen.c (get_stridx): Likewise.
5873 2018-08-29  Jan Hubicka  <jh@suse.cz>
5875         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5876         Do not stream DECL_VINDEX.
5877         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5878         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5879         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5881 2018-08-29  Richard Biener  <rguenther@suse.de>
5883         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5884         virtual operands that are not default defs to honor region
5885         boundaries.
5886         (rpo_vn_valueize): Remove ineffective code here.
5888 2018-08-29  Richard Biener  <rguenther@suse.de>
5890         PR tree-optimization/87132
5891         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5892         when skipping defs reachable over backedges.
5894 2018-08-29  Richard Biener  <rguenther@suse.de>
5896         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5897         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5898         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5899         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5900         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5901         (vn_reference_lookup_3): Remove use of const_parms.
5902         (free_rpo_vn): Do not free const_parms.
5903         (do_rpo_vn): Do not call init_const_parms.
5904         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5905         SSA_NAME_POINTS_TO_READONLY_MEMORY.
5906         (call_may_clobber_ref_p_1): Likewise.
5908 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
5910         PR other/86726
5911         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5912         (-O): Ditto.
5913         (-ftree-scev-cprop): Document.
5915 2018-08-29  Jan Hubicka  <jh@suse.cz>
5917         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5918         parameters.
5919         (sreal constructor): Update.
5920         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5921         sreal:operator/): Update.
5923 2018-08-29  Martin Liska  <mliska@suse.cz>
5925         * tree-switch-conversion.c (switch_conversion::expand):
5926         Strenghten assumption about gswitch statements.
5928 2018-08-29  Richard Biener  <rguenther@suse.de>
5930         PR tree-optimization/87117
5931         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5932         re-value-number released SSA VDEFs.
5934 2018-08-29  Richard Biener  <rguenther@suse.de>
5936         PR tree-optimization/87126
5937         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5939 2018-08-28  Jim Wilson  <jimw@sifive.com>
5941         * config/riscv/pic.md: Rewrite.
5942         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5943         invalid address.
5944         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5945         (SOFTF, default_load, softload, softstore): New.
5947 2018-08-28  Jeff Law  <law@redhat.com>
5949         * fold-const.c (fold_binary_loc): Remove recently added assert.
5951 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
5953         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5954         to OP parmeter of generated function.
5956 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
5958         PR tree-optimization/87009
5959         * match.pd: Add boolean optimizations.
5961 2018-08-28  Martin Sebor  <msebor@redhat.com>
5963         PR middle-end/86631
5964         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5965         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5966         (pass_walloca::gate): Use it.
5967         (alloca_call_type): Same.
5968         (pass_walloca::execute): Same.
5969         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5971 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
5973         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5974         GCC_VERSION for usage of "__gcc_dump_printf__" format from
5975         >= 3005 to >= 9000.
5977 2018-08-28  Richard Biener  <rguenther@suse.de>
5979         PR tree-optimization/87124
5980         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5981         constants before looking up avail.
5983 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
5985         PR middle-end/87099
5986         * calls.c (maybe_warn_nonstring_arg): Punt early if
5987         warn_stringop_overflow is zero.  Don't call get_range_strlen
5988         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5989         Swap comparison operands to have constants on rhs.  Only use
5990         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
5991         increment lenrng[0].
5993 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
5995         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5996         use of tree_to_shwi.  Remove duplicated test for the size being
5997         a whole number of bytes.
5999 2018-08-28  Richard Biener  <rguenther@suse.de>
6001         PR tree-optimization/87117
6002         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
6003         Handle removed stmt without LHS (GIMPLE_NOP).
6005 2018-08-28  Richard Biener  <rguenther@suse.de>
6007         PR tree-optimization/87117
6008         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
6009         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
6011 2018-08-28  Richard Biener  <rguenther@suse.de>
6013         PR tree-optimization/87117
6014         * tree-ssa-pre.c (compute_avail): Do not make expressions
6015         with predicated values available.
6016         (get_expr_value_id): Assert we do not run into predicated value
6017         expressions.
6019 2018-08-28  Richard Biener  <rguenther@suse.de>
6021         PR tree-optimization/87117
6022         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
6023         get virtual operands.
6024         (get_expr_operands): Handle STRING_CST like other decls.
6026 2018-08-28  Martin Liska  <mliska@suse.cz>
6028         * tree.h: Update documentation of fndecl_built_in_p
6029         functions.
6032 2018-08-27  Jeff Law  <law@redhat.com>
6033         PR tree-optimization/87110
6034         * tree-ssa-dse.c (compute_trims): Handle non-constant
6035         TYPE_SIZE_UNIT.
6037 2018-08-27  Martin Sebor  <msebor@redhat.com>
6039         PR tree-optimization/86914
6040         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
6042 2018-08-27  Martin Sebor  <msebor@redhat.com>
6044         PR tree-optimization/87112
6045         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
6046         the type of the bound argument.
6048 2018-08-27  Jeff Law  <law@redhat.com>
6050         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
6051         type does not have a TYPE_SIZE_UNIT.
6053 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
6055         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
6056         with include of backend.h.
6058 2018-08-27  Richard Biener  <rguenther@suse.de>
6060         PR tree-optimization/86927
6061         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
6062         use const cond reduction code.
6064 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
6066         PR tree-optimization/85758
6067         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
6069 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
6071         PR c++/87091
6072         * diagnostic-show-locus.c (class layout_range): Update for
6073         conversion of show_caret_p to a tri-state.
6074         (layout_range::layout_range): Likewise.
6075         (make_range): Likewise.
6076         (layout::maybe_add_location_range): Likewise.
6077         (layout::should_print_annotation_line_p): Don't show annotation
6078         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
6079         (layout::get_state_at_point): Update for conversion of
6080         show_caret_p to a tri-state.  Bail out early for
6081         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
6082         underlining or source colorization.
6083         (gcc_rich_location::add_location_if_nearby): Update for conversion
6084         of show_caret_p to a tri-state.
6085         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
6086         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
6087         Likewise.
6088         (selftest::test_one_liner_labels): Likewise.
6089         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
6090         conversion of show_caret_p to a tri-state.
6091         * pretty-print.c (text_info::set_location): Likewise.
6092         * pretty-print.h (text_info::set_location): Likewise.
6093         * substring-locations.c (format_warning_n_va): Likewise.
6094         * tree-diagnostic.c (default_tree_printer): Likewise.
6095         * tree-pretty-print.c (newline_and_indent): Likewise.
6097 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
6099         PR c++/87091
6100         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
6101         line above for line-insertion fix-it hints.
6102         (selftest::test_fixit_insert_containing_newline): Update the
6103         expected results, and add a test with line-numbering enabled.
6105 2018-08-27  Martin Liska  <mliska@suse.cz>
6107         PR sanitizer/86962
6108         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
6109         params with DECL_HAS_VALUE_EXPR_P.
6111 2018-08-27  Martin Liska  <mliska@suse.cz>
6113         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
6114         selected expansion strategy.
6116 2018-08-27  Martin Liska  <mliska@suse.cz>
6118         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
6119         * builtins.c (is_builtin_fn): Likewise.
6120         * attribs.c (diag_attr_exclusions): Use new function
6121         fndecl_built_in_p and remove check for FUNCTION_DECL if
6122         possible.
6123         (builtin_mathfn_code): Likewise.
6124         (fold_builtin_expect): Likewise.
6125         (fold_call_expr): Likewise.
6126         (fold_builtin_call_array): Likewise.
6127         (fold_call_stmt): Likewise.
6128         (set_builtin_user_assembler_name): Likewise.
6129         (is_simple_builtin): Likewise.
6130         * calls.c (gimple_alloca_call_p): Likewise.
6131         (maybe_warn_nonstring_arg): Likewise.
6132         * cfgexpand.c (expand_call_stmt): Likewise.
6133         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
6134         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
6135         (cgraph_node::verify_node): Likewise.
6136         * cgraphclones.c (build_function_decl_skip_args): Likewise.
6137         (cgraph_node::create_clone): Likewise.
6138         * config/arm/arm.c (arm_insert_attributes): Likewise.
6139         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
6140         * dse.c (scan_insn): Likewise.
6141         * expr.c (expand_expr_real_1): Likewise.
6142         * fold-const.c (operand_equal_p): Likewise.
6143         (fold_binary_loc): Likewise.
6144         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
6145         * gimple-low.c (lower_stmt): Likewise.
6146         * gimple-pretty-print.c (dump_gimple_call): Likewise.
6147         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6148         Likewise.
6149         * gimple.c (gimple_build_call_from_tree): Likewise.
6150         (gimple_call_builtin_p): Likewise.
6151         (gimple_call_combined_fn): Likewise.
6152         * gimplify.c (gimplify_call_expr): Likewise.
6153         (gimple_boolify): Likewise.
6154         (gimplify_modify_expr): Likewise.
6155         (gimplify_addr_expr): Likewise.
6156         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
6157         * ipa-cp.c (determine_versionability): Likewise.
6158         * ipa-fnsummary.c (compute_fn_summary): Likewise.
6159         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
6160         * ipa-split.c (visit_bb): Likewise.
6161         (split_function): Likewise.
6162         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6163         * lto-cgraph.c (input_node): Likewise.
6164         * lto-streamer-out.c (write_symbol): Likewise.
6165         * omp-low.c (setjmp_or_longjmp_p): Likewise.
6166         (lower_omp_1): Likewise.
6167         * predict.c (strip_predict_hints): Likewise.
6168         * print-tree.c (print_node): Likewise.
6169         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
6170         * trans-mem.c (is_tm_irrevocable): Likewise.
6171         (is_tm_load): Likewise.
6172         (is_tm_simple_load): Likewise.
6173         (is_tm_store): Likewise.
6174         (is_tm_simple_store): Likewise.
6175         (is_tm_abort): Likewise.
6176         (tm_region_init_1): Likewise.
6177         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
6178         * tree-cfg.c (verify_gimple_call): Likewise.
6179         (move_stmt_r): Likewise.
6180         (stmt_can_terminate_bb_p): Likewise.
6181         * tree-eh.c (lower_eh_constructs_2): Likewise.
6182         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
6183         * tree-inline.c (remap_gimple_stmt): Likewise.
6184         (copy_bb): Likewise.
6185         (estimate_num_insns): Likewise.
6186         (fold_marked_statements): Likewise.
6187         * tree-sra.c (scan_function): Likewise.
6188         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
6189         (optimize_stack_restore): Likewise.
6190         (pass_fold_builtins::execute): Likewise.
6191         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
6192         (mark_all_reaching_defs_necessary_1): Likewise.
6193         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
6194         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
6195         (pass_forwprop::execute): Likewise.
6196         * tree-ssa-loop-im.c (stmt_cost): Likewise.
6197         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
6198         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
6199         * tree-ssa-strlen.c (get_string_length): Likewise.
6200         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
6201         (find_func_aliases_for_call): Likewise.
6202         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
6203         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
6204         * tree-tailcall.c (find_tail_calls): Likewise.
6205         * tree.c (need_assembler_name_p): Likewise.
6206         (free_lang_data_in_decl): Likewise.
6207         (get_call_combined_fn): Likewise.
6208         * ubsan.c (is_ubsan_builtin_p): Likewise.
6209         * varasm.c (incorporeal_function_p): Likewise.
6210         * tree.h (DECL_BUILT_IN): Remove and replace with
6211         fndecl_built_in_p.
6212         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
6213         (fndecl_built_in_p): New.
6215 2018-08-27  Martin Liska  <mliska@suse.cz>
6217         PR tree-optimization/86847
6218         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
6219         Dump also subtree probability.
6220         (switch_decision_tree::do_jump_if_equal): New function.
6221         (switch_decision_tree::emit_case_nodes): Handle special
6222         situations in balanced tree that can be emitted much simpler.
6223         Fix calculation of probabilities that happen in tree expansion.
6224         * tree-switch-conversion.h (struct cluster): Add
6225         is_single_value_p.
6226         (struct simple_cluster): Likewise.
6227         (struct case_tree_node): Add new function has_child.
6228         (do_jump_if_equal): New.
6230 2018-08-27  Martin Liska  <mliska@suse.cz>
6232         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
6233         Add new argument to bit_test_cluster constructor.
6234         (bit_test_cluster::emit): Set bits really number of values
6235         handlel by a test.
6236         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
6237         probability argument.
6238         * tree-switch-conversion.h (struct bit_test_cluster):
6239         Add m_handles_entire_switch.
6241 2018-08-27  Martin Liska  <mliska@suse.cz>
6243         PR tree-optimization/86702
6244         * tree-switch-conversion.c (jump_table_cluster::emit):
6245         Make probabilities even for values in jump table
6246         according to number of cases handled.
6247         (switch_decision_tree::compute_cases_per_edge): Pass
6248         argument to reset_out_edges_aux function.
6249         (switch_decision_tree::analyze_switch_statement): Likewise.
6250         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
6251         Make it static.
6253 2018-08-27  Martin Liska  <mliska@suse.cz>
6255         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
6256         cfun argument explicitly.
6257         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
6258         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
6259         function gimple_switch_default_bb.
6260         (convert_switch_statements):
6261         (expand_builtins):
6262         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
6263         * stmt.c (label_to_block_fn): Use label_to_block and pass
6264         cfun argument explicitly and use gimple_switch_label_bb.
6265         (expand_case): Likewise.
6266         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
6267         cfun argument explicitly. Likewise.
6268         (make_edges_bb): Likewise.
6269         (make_cond_expr_edges): Likewise.
6270         (get_cases_for_edge): Likewise.
6271         (make_gimple_switch_edges): Likewise.
6272         (label_to_block_fn): Likewise.
6273         (label_to_block): Likewise.
6274         (make_goto_expr_edges): Likewise.
6275         (make_gimple_asm_edges): Likewise.
6276         (main_block_label): Likewise.
6277         (group_case_labels_stmt): Likewise.
6278         (find_taken_edge_computed_goto): Likewise.
6279         (find_taken_edge_switch_expr): Likewise.
6280         (gimple_verify_flow_info): Likewise.
6281         (gimple_redirect_edge_and_branch): Likewise.
6282         (gimple_switch_label_bb): New function.
6283         (gimple_switch_default_bb): Likewise.
6284         (gimple_switch_edge): Likewise.
6285         (gimple_switch_default_edge): Likewise.
6286         * tree-cfg.h (label_to_block_fn): Remove and replace ...
6287         (label_to_block): ... with this.
6288         (gimple_switch_label_bb): New.
6289         (gimple_switch_default_bb): Likewise.
6290         (gimple_switch_edge): Likewise.
6291         (gimple_switch_default_edge): Likewise.
6292         * tree-cfgcleanup.c (convert_single_case_switch): Use
6293         new gimple functions and pass new argument to label_to_block.
6294         (cleanup_control_flow_bb):
6295         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
6296         cfun argument explicitly.
6297         (make_eh_edges): Likewise.
6298         (redirect_eh_dispatch_edge): Likewise.
6299         (lower_resx): Likewise.
6300         (lower_eh_dispatch): Likewise.
6301         (maybe_remove_unreachable_handlers): Likewise.
6302         (unsplit_eh): Likewise.
6303         (cleanup_empty_eh): Likewise.
6304         (verify_eh_edges): Likewise.
6305         (verify_eh_dispatch_edge): Likewise.
6306         * tree-ssa-dom.c (record_edge_info): Likewise.
6307         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
6308         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
6309         (thread_through_normal_block): Likewise.
6310         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
6311         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
6312         * tree-switch-conversion.c (switch_conversion::collect): Use new
6313         gimple functions.
6314         (switch_conversion::check_final_bb): Likewise.
6315         (switch_conversion::gather_default_values): Pass new argument
6316         to label_to_block.
6317         (switch_conversion::build_constructors): Likewise.
6318         (switch_decision_tree::compute_cases_per_edge): Use new
6319         gimple_switch_edge function.
6320         (switch_decision_tree::analyze_switch_statement): Pass new argument
6321         to label_to_block.
6322         (switch_decision_tree::try_switch_expansion): Use
6323         gimple_switch_default_edge.
6324         * tree-vrp.c (find_switch_asserts): Pass new argument
6325         to label_to_block.
6326         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
6327         (vr_values::simplify_switch_using_ranges): Likewise.
6329 2018-08-27  Richard Biener  <rguenther@suse.de>
6331         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
6332         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
6334         * tree-ssa-sccvn.h (struct vn_pval): New structure.
6335         (struct vn_nary_op_s): Add unwind_to member.  Add
6336         predicated_values flag and put result into a union together
6337         with a linked list of vn_pval.
6338         (struct vn_ssa_aux): Add name member to make maintaining
6339         a map of SSA name to vn_ssa_aux possible.  Remove no longer
6340         needed info, dfsnum, low, visited, on_sccstack, use_processed
6341         and range_info_anti_range_p members.
6342         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
6343         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
6344         New functions.
6345         (vn_valueize): New global.
6346         (vn_context_bb): Likewise.
6347         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
6348         VN_INFO_PTR_INFO): Remove.
6349         * tree-ssa-sccvn.c: ... (rewrite)
6350         (pass_fre::execute): For -O2+ initialize loops and run
6351         RPO VN in optimistic mode (iterating).  For -O1 and -Og
6352         run RPO VN in non-optimistic mode.
6353         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
6354         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
6355         * doc/invoke.texi (sccvn-max-scc-size): Remove.
6356         (rpo-vn-max-loop-depth): Document.
6357         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
6358         when valuezing the VUSE signals we walked out of the region.
6359         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
6360         (phi_translate): Set VN context block to use for availability
6361         lookup.
6362         (compute_avail): Likewise.
6363         (pre_valueize): New function.
6364         (pass_pre::execute): Adjust to the RPO VN API.
6366         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
6367         (propagate_constants_for_unrolling): Remove.
6368         (tree_unroll_loops_completely): Perform value-numbering
6369         on the unrolled bodies loop parent.
6371 2018-08-27  Richard Biener  <rguenther@suse.de>
6373         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
6374         for partial antic compute.
6376 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
6378         PR rtl-optimization/87065
6379         * combine.c (simplify_if_then_else): Formatting fix.
6380         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
6381         check.
6382         (known_cond): Don't return const_true_rtx for vector modes.  Use
6383         CONST0_RTX instead of const0_rtx.  Formatting fixes.
6385 2018-08-27  Martin Liska  <mliska@suse.cz>
6387         PR gcov-profile/87069
6388         * gcov.c (process_file): Record files already processed
6389         and warn about a file being processed multiple times.
6391 2018-08-27  Martin Liska  <mliska@suse.cz>
6393         PR driver/83193
6394         * config/aarch64/aarch64.c (aarch64_override_options_internal):
6395         Set default values for x_aarch64_*_string strings.
6396         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
6397         prefix.  For -mabi do not print '=ABI' in help and use
6398         <option_value> format for -msve-vector-bits and -moverride
6399         options.
6401 2018-08-26  Jeff Law  <law@redhat.com>
6403         * config/mips/frame-header-opt.c: Include "backend.h" rather than
6404         "cfg.h"
6406 2018-08-26  Marek Polacek  <polacek@redhat.com>
6408         PR c++/87029, Implement -Wredundant-move.
6409         * doc/invoke.texi: Document -Wredundant-move.
6411 2018-08-25  Martin Sebor  <msebor@redhat.com>
6413         PR tree-optimization/87059
6414         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
6415         to the same type as the other.
6416         * fold-const.c (fold_binary_loc): Assert expectation.
6418 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
6420         * config/darwin.c (machopic_legitimize_pic_address): Clean up
6421         extraneous parentheses, dead code section and formatting.
6423 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
6425         PR c++/87091
6426         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
6427         wide enough for jumps in the line-numbering to be visible.
6428         (layout::print_gap_in_line_numbering): New member function.
6429         (layout::calculate_line_spans): When using line numbering, merge
6430         line spans that are only 1 line apart.
6431         (diagnostic_show_locus): When printing line numbers, show gaps in
6432         line numbering directly, rather than printing headers.
6433         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
6434         line-numbering with multiple line spans.
6435         (selftest::test_fixit_insert_containing_newline_2): Add test of
6436         line-numbering, in which the spans are close enough to be merged.
6438 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6440         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
6441         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
6442         * tree-vrp.c (range_is_nonnull): Remove.
6443         (range_includes_zero_p): Accept value_range instead of min/max.
6444         (extract_range_from_binary_expr_1): Do not early bail on
6445         POINTER_PLUS_EXPR.
6446         Use range_includes_zero_p instead of range_is_nonnull.
6447         (extract_range_from_unary_expr): Use range_includes_zero_p instead
6448         of range_is_nonnull.
6449         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
6450         special case VR_ANTI_RANGE.
6451         (vrp_finalize): Same.
6452         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
6453         instead of min/max.
6454         (range_is_nonnull): Remove.
6455         * vr-values.c (vrp_stmt_computes_nonzero): Use
6456         range_includes_zero_p instead of range_is_nonnull.
6457         (extract_range_basic): Pass value_range to range_includes_zero_p
6458         instead of range_is_nonnull.
6460 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
6462         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
6463         * emit-rtl.h (rtl_data): Remove return_bnd.
6464         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
6465         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
6466         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
6467         (POINTER_BOUNDS_MODE): Remove definition.
6468         (make_pointer_bounds_mode): Remove.
6469         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
6470         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
6471         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
6472         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
6473         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
6474         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
6475         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
6477         * config/i386/i386-modes.def (BND32, BND64): Remove.
6478         * config/i386/i386.c (dbx_register_map): Remove bound registers.
6479         (dbx64_register_map): Ditto.
6480         (svr4_dbx_register_map): Ditto.
6481         (indirect_thunk_bnd_needed): Remove.
6482         (indirect_thunks_bnd_used): Ditto.
6483         (indirect_return_bnd_needed): Ditto.
6484         (indirect_return_via_cx_bnd): Ditto.
6485         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
6486         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
6487         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
6488         (output_indirect_thunk_function): Remove handling of
6489         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
6490         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
6491         (ix86_save_reg): Remove handling of crtl->return_bnd.
6492         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
6493         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
6494         and UNSPEC_BNDLX_ADDR.
6495         (ix86_output_indirect_branch_via_reg): Remove handling of
6496         indirect_thunk_prefix_bnd.
6497         (ix86_output_indirect_branch_via_push): Ditto.
6498         (ix86_output_function_return): Ditto.
6499         (ix86_output_indirect_function_return): Ditto.
6500         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
6501         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
6502         (CALL_USED_REGISTERS): Ditto.
6503         (REG_ALLOC_ORDER): Update for removal of bound registers.
6504         (HI_REGISTER_NAMES): Ditto.
6505         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
6506         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
6507         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
6508         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
6509         (FIRST_PSEUDO_REG): Update.
6510         (BND): Remove mode iterator.
6511         * config/i386/predicates.md (bnd_mem_operator): Remove.
6513 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6515         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
6516         vectors.
6518 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6520         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
6521         the case in which the permute needs only a single element and
6522         repeats for every vector of the result.  Extend that case to
6523         handle variable-length vectors.
6524         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6526 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
6528         PR debug/79342
6529         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
6530         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
6532 2018-08-24  Richard Biener  <rguenther@suse.de>
6534         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
6535         bb_flags_allocated members.
6536         (auto_flag): New RAII class for allocating flags.
6537         (auto_edge_flag): New RAII class for allocating edge flags.
6538         (auto_bb_flag): New RAII class for allocating bb flags.
6539         * cfgloop.c (verify_loop_structure): Allocate temporary edge
6540         flag dynamically.
6541         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
6542         in favor of temporarily allocated BB flag.
6543         * hsa-brig.c: Re-order includes.
6544         * hsa-dump.c: Likewise.
6545         * hsa-regalloc.c: Likewise.
6546         * print-rtl.c: Likewise.
6547         * profile-count.c: Likewise.
6549 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
6551         PR target/86989
6552         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
6553         the TOC register.
6555 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6557         PR 87073/bootstrap
6558         * wide-int-range.cc (wide_int_range_div): Do not ignore result
6559         from wide_int_range_multiplicative_op.
6561 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6563         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
6564         "permutaion".
6566 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
6568         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
6569         to 'expanded'.
6571 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
6573         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
6574         full GENERIC expression used for replacement.
6576 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
6578         * tree-vrp.c (abs_extent_range): Remove.
6579         (extract_range_into_wide_ints): Pass wide ints by reference.
6580         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
6581         Pass wide ints by reference in all calls to
6582         extract_range_into_wide_ints.
6583         * wide-int-range.cc (wide_int_range_div): New.
6584         * wide-int-range.h (wide_int_range_div): New.
6585         (wide_int_range_includes_zero_p): New.
6586         (wide_int_range_zero_p): New.
6588 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
6590         * config/aarch64/aarch64.md (arches): New enum.
6591         (arch): New enum attr.
6592         (arch_enabled): New attr.
6593         (enabled): Now uses arch_enabled only.
6594         (simd, sve, fp16): Removed attribute.
6595         (fp): Attr now defined in terms of 'arch'.
6596         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
6597         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
6598         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
6599         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
6600         attributes into 'arch'.
6601         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
6602         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
6603         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
6604         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
6605         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
6606         'simd' attribute into 'arch'.
6607         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
6608         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
6609         Convert use of 'fp' attribute to 'arch'.
6610         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
6611         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
6612         into 'arch'.
6613         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
6614         (different modes) Merge 'fp' and 'simd' into 'arch'.
6615         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
6616         'simd' into 'arch'.
6618 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
6620         PR rtl-optimization/87026
6621         * expmed.c (canonicalize_comparison): If we can no longer create
6622         pseudoregisters, don't.
6624 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
6626         PR target/86951
6627         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6628         prototype.
6629         * config/arm/arm.c (speculation_barrier_libfunc): New static
6630         variable.
6631         (arm_init_libfuncs): Initialize it.
6632         (arm_emit_speculation_barrier): New function.
6633         * config/arm/arm.md (speculation_barrier): Call
6634         arm_emit_speculation_barrier for architectures that do not have
6635         DSB or ISB.
6636         (speculation_barrier_insn): Only match on Armv7 or later.
6638 2018-08-23  Richard Biener  <rguenther@suse.de>
6640         PR middle-end/87024
6641         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6642         calls.
6644 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6646         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6647         of single-vector TBLs.
6648         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6649         one input is given.
6651 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6653         PR target/85910
6654         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6655         aarch64_evpc_tbl guard.
6657 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6659         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6660         behaviour.
6662 2018-08-22  Martin Sebor  <msebor@redhat.com>
6664         PR middle-end/87052
6665         * tree-pretty-print.c (pretty_print_string): Add argument.
6666         (dump_generic_node): Call to pretty_print_string with string size.
6668 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
6670         PR rtl-optimization/86771
6671         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6672         of two SETs into those two SETs, one to be placed at i2, if that SETs
6673         destination is modified between i2 and i3.
6675 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6677         PR tree-optimization/86725
6678         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6679         function.
6680         (vect_analyze_scalar_cycles_1): Check it.
6682 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6684         PR tree-optimization/86725
6685         * tree-vect-loop.c (vect_is_simple_reduction): When treating
6686         an outer loop phi as a double reduction, make sure that the
6687         single user of the phi result is an inner loop phi.
6689 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6691         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6692         grouped stores with gaps to a strided group.
6694 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6696         * tree-vect-stmts.c (get_group_load_store_type)
6697         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6698         first statement in a group.
6700 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6702         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6703         the sequence used in gcc/gcc.c.
6705 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6707         PR other/704
6708         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6709         building it.
6711 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6713         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6714         Darwin10-specific unwinder-shim.
6715         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6716         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6717         New to cater for Darwin10 Rosetta.
6719 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6721         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6722         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6724 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6726         PR bootstrap/81033
6727         PR target/81733
6728         PR target/52795
6729         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6730         (dwarf2out_switch_text_section): Generate a local label for the
6731         second function sub-section and apply it as the second FDE start
6732         label.
6733         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6734         second sub-section start.
6736 2018-08-22  Richard Biener  <rguenther@suse.de>
6738         PR tree-optimization/86988
6739         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6741 2018-08-22  Richard Biener  <rguenther@suse.de>
6743         PR tree-optimization/86945
6744         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6746 2018-08-22  Alexandre Oliva <oliva@adacore.com>
6748         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6749         a comment about how uses of r2 for .sdata2 come about.
6751 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
6753         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6755 2018-08-21  Marek Polacek  <polacek@redhat.com>
6757         PR c++/86981, Implement -Wpessimizing-move.
6758         * doc/invoke.texi: Document -Wpessimizing-move.
6760 2018-08-21  Jan Hubicka  <jh@suse.cz>
6762         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6763         * tree.h (is_redundant_typedef): Remove.
6764         * dwarf2out.c (is_redundant_typedef): Turn into static function.
6766 2018-08-21  Jan Hubicka  <jh@suse.cz>
6768         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6769         when possible.
6771 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
6773         * expmed.c (extract_low_bits): Reject invalid subregs early.
6775 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6777         PR middle-end/86121
6778         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6779         behaviour.
6781 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6783         * config/vxworks.h: Guard vxworks_asm_out_constructor and
6784         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6785         * config/vxworks.c: Likewise.
6787 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6789         * config/vxworks.c: Set targetm.have_ctors_dtors
6790         if HAVE_INITFINI_ARRAY_SUPPORT.
6791         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6792         if HAVE_INITFINI_ARRAY_SUPPORT.
6794 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6796         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6797         default search path for VxWorks < 7.
6799 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6801         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6802         (get_format_string): Refer to c_getstr.
6804 2018-08-21  Tom de Vries  <tdevries@suse.de>
6806         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6807         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6808         (debuginfo_early_stop): Declare.
6809         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6810         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6811         (debuginfo_early_stop): New function.
6812         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6813         and debuginfo_early_stop.
6814         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6815         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6816         (general_init): Call debuginfo_early_init.
6817         (finalize): Call debuginfo_fini.
6818         (do_compile): Call debuginfo_init.
6819         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6820         -fdump-early-debug.
6821         (@item -fdump-debug, @item -fdump-earlydebug): Add.
6823 2018-08-21  Tom de Vries  <tdevries@suse.de>
6825         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6826         flag_dump_noaddr and flag_dump_unnumbered.
6828 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
6830         * wide-int-range.cc (wide_int_range_abs): New.
6831         (wide_int_range_order_set): Rename from wide_int_range_min_max.
6832         * wide-int-range.h (wide_int_range_abs): New.
6833         (wide_int_range_min_max): New.
6834         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6835         case to call wide_int_range_abs.
6836         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6837         (extract_range_from_abs_expr): Delete.
6839 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
6841         PR target/87033
6842         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6843         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6844         for indexed loads.
6846 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6847             Jeff Law <law@redhat.com>
6849         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6850         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6852 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
6854         PR other/84889
6855         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6856         (decl_attributes): Likewise.
6857         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6858         instance.
6859         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6860         * diagnostic-core.h (class auto_diagnostic_group): New class.
6861         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6862         (diagnostic_report_diagnostic): Handle the first diagnostics within
6863         a group.
6864         (emit_diagnostic): Add auto_diagnostic_group instance.
6865         (inform): Likewise.
6866         (inform_n): Likewise.
6867         (warning): Likewise.
6868         (warning_at): Likewise.
6869         (warning_n): Likewise.
6870         (pedwarn): Likewise.
6871         (permerror): Likewise.
6872         (error): Likewise.
6873         (error_n): Likewise.
6874         (error_at): Likewise.
6875         (sorry): Likewise.
6876         (fatal_error): Likewise.
6877         (internal_error): Likewise.
6878         (internal_error_no_backtrace): Likewise.
6879         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6880         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6881         * diagnostic.h (struct diagnostic_context): Add fields
6882         "diagnostic_group_nesting_depth",
6883         "diagnostic_group_emission_count", "begin_group_cb",
6884         "end_group_cb".
6885         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6886         Add auto_diagnostic_group instance(s).
6887         (find_explicit_erroneous_behavior): Likewise.
6888         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6889         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6890         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6891         (gimplify_va_arg_expr): Likewise.
6892         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6893         (HSA_SORRY_AT): Likewise.
6894         * ipa-devirt.c (compare_virtual_tables): Likewise.
6895         (warn_odr): Likewise.
6896         * multiple_target.c (expand_target_clones): Likewise.
6897         * opts-common.c (cmdline_handle_error): Likewise.
6898         * reginfo.c (globalize_reg): Likewise.
6899         * substring-locations.c (format_warning_n_va): Likewise.
6900         * tree-inline.c (expand_call_inline): Likewise.
6901         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6902         * tree-ssa-loop-niter.c
6903         (do_warn_aggressive_loop_optimizations): Likewise.
6904         * tree-ssa-uninit.c (warn_uninit): Likewise.
6905         * tree.c (warn_deprecated_use): Likewise.
6907 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
6909         PR target/87014
6910         * config/i386/i386.md (eh_return): Always update EH return
6911         address in word_mode.
6913 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
6915         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6916         TARGET_SPLIT_COMPLEX_ARG is defined.
6918 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6920         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6922 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6924         PR target/86984
6925         * expr.c (expand_assignment): Assert that bitpos is positive.
6926         (store_field): Likewise
6927         (expand_expr_real_1): Make sure that bitpos is positive.
6928         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6929         integer overflow.
6931 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6933         * Makefile.in (CPP_ID_DATA_H): Delete.
6934         (CPP_INTERNAL_H): Don't add it.
6935         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6936         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6938 2018-08-20  Richard Biener  <rguenther@suse.de>
6940         PR tree-optimization/78655
6941         * tree-vrp.c (extract_range_from_binary_expr_1): Make
6942         pointer + offset nonnull if either operand is nonnull work.
6944 2018-08-20  Tom de Vries  <tdevries@suse.de>
6946         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6947         unless the referenced die describes the added property using
6948         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
6949         Otherwise, add a DW_AT_location to the referenced die.
6951 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
6953         PR target/86994
6954         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6955         register_operand when calling ix86_set_reg_reg_cost.
6956         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6957         Set *total to 0 for operands that satisfy x86_64_immediate_operand
6958         predicate and to 1 otherwise.
6960 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
6962         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6963         emit a diagnostic that it is not supported and reset the option.
6964         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6965         supported and consume the option.  (ASM_FINAL_SPEC): New.
6967 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6969         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6970         a sentence.
6972 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
6974         C-SKY port: Documentation
6976         * doc/extend.texi (C-SKY Function Attributes): New section.
6977         * doc/invoke.texi (Option Summary): Add C-SKY options.
6978         (C-SKY Options): New section.
6979         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6981 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6982             Huibin Wang  <huibin_wang@c-sky.com>
6983             Sandra Loosemore  <sandra@codesourcery.com>
6984             Chung-Lin Tang  <cltang@codesourcery.com>
6986         C-SKY port: Backend implementation
6988         * config/csky/*: New.
6989         * common/config/csky/*: New.
6991 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6992             Huibin Wang  <huibin_wang@c-sky.com>
6993             Sandra Loosemore  <sandra@codesourcery.com>
6994             Chung-Lin Tang  <cltang@codesourcery.com>
6995             Andrew Jenner  <andrew@codesourcery.com>
6997         C-SKY port: Configury
6999         * config.gcc (csky-*-*): New.
7000         * configure.ac: Add csky to targets for dwarf2 debug_line support.
7001         * configure: Regenerated.
7003 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
7005         * dump-context.h: Include "dumpfile.h".
7006         (dump_context::dump_printf_va): Convert final param from va_list
7007         to va_list *.  Convert from ATTRIBUTE_PRINTF to
7008         ATTRIBUTE_GCC_DUMP_PRINTF.
7009         (dump_context::dump_printf_loc_va): Likewise.
7010         * dumpfile.c: Include "stringpool.h".
7011         (make_item_for_dump_printf_va): Delete.
7012         (make_item_for_dump_printf): Delete.
7013         (class dump_pretty_printer): New class.
7014         (dump_pretty_printer::dump_pretty_printer): New ctor.
7015         (dump_pretty_printer::emit_items): New member function.
7016         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
7017         function.
7018         (dump_pretty_printer::emit_item): New member function.
7019         (dump_pretty_printer::stash_item): New member function.
7020         (dump_pretty_printer::format_decoder_cb): New member function.
7021         (dump_pretty_printer::decode_format): New member function.
7022         (dump_context::dump_printf_va): Reimplement in terms of
7023         dump_pretty_printer.
7024         (dump_context::dump_printf_loc_va): Convert final param from va_list
7025         to va_list *.
7026         (dump_context::begin_scope): Reimplement call to
7027         make_item_for_dump_printf.
7028         (dump_printf): Update for change to dump_printf_va.
7029         (dump_printf_loc): Likewise.
7030         (selftest::test_capture_of_dump_calls): Convert "stmt" from
7031         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
7032         with %T, %E, and %G.
7033         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
7034         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
7035         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
7036         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
7037         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
7038         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
7039         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
7040         within a dump_printf_loc call to "%wu".
7041         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
7042         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
7043         missing space after "=".
7044         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
7045         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
7046         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
7047         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
7048         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
7049         duplicate "vectorized" from message.
7051 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
7053         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
7054         polyNxK_t element's TYPE_STRING_FLAG.
7056 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
7058         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
7059         (they were unnamed before).  Fix comments.
7061 2018-08-17  Nathan Sidwell  <nathan@acm.org>
7063         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
7065 2018-08-17  Richard Biener  <rguenther@suse.de>
7067         PR tree-optimization/86841
7068         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
7070 2018-08-17  Martin Liska  <mliska@suse.cz>
7072         * common.opt: Remove Warn, Init and Report for options with
7073         Ignore/Deprecated flag. Warning is done automatically for
7074         Deprecated flags.
7075         * config/i386/i386.opt: Likewise.
7076         * config/ia64/ia64.opt: Likewise.
7077         * config/rs6000/rs6000.opt: Likewise.
7078         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
7079         Remove usage of flag_check_pointer_bounds.
7080         * lto-wrapper.c (merge_and_complain): Do not handle
7081         OPT_fcheck_pointer_bounds.
7082         (append_compiler_options): Likewise.
7083         * opt-functions.awk: Do not handle Deprecated.
7084         * optc-gen.awk: Check that Var, Report and Init are not
7085         used for an option with Ignore/Deprecated flag.
7086         * opts-common.c (decode_cmdline_option): Do not report
7087         CL_ERR_DEPRECATED.
7088         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
7089         options.
7090         * opts.h (struct cl_option): Remove cl_deprecated flag.
7091         (CL_ERR_DEPRECATED): Remove error enum value.
7093 2018-08-17  Richard Biener  <rguenther@suse.de>
7095         PR middle-end/86505
7096         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
7097         across a va-arg-pack using call adjust its return value accordingly.
7099 2018-08-16  Martin Sebor  <msebor@redhat.com>
7101         PR tree-optimization/86853
7102         * gimple-ssa-sprintf.c (struct format_result): Rename member.
7103         (struct fmtresult): Add member and initialize it in ctors.
7104         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
7105         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
7106         (format_directive): Set POSUNDER4K when MAYFAIL is set.
7107         (parse_directive): Handle %C same as %c.
7108         (sprintf_dom_walker::compute_format_length): Adjust.
7109         (is_call_safe): Adjust.
7111 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7113         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
7114         for determining how to count the elements.
7115         * builtins.h (c_strlen): Adjust prototype.
7116         * expr.c (string_constant): Add new parameter mem_size.
7117         Set *mem_size appropriately.
7118         * expr.h (string_constant): Adjust protoype.
7119         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
7120         * gimple-fold.h (get_range_strlen): Adjust prototype.
7121         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
7122         (format_string): Call get_string_length with eltsize.
7124 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
7126         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
7127         to emit the span, rather than setting it as the prefix.
7129 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
7131         * diagnostic-show-locus.c (layout::start_annotation_line): Add
7132         "margin_char" parameter, defaulting to space.  Use it in place
7133         of pp_space for the initial part of the margin.
7134         (layout::print_leading_fixits): Use '+' when filling the margin
7135         of line-insertion fix-it hints.
7137 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7139         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
7140         Delete.
7142 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7144         * config/rs6000/altivec.md: Don't set length attribute to the default
7145         value.
7146         * config/rs6000/darwin.md: Ditto.
7147         * config/rs6000/dfp.md: Ditto.
7148         * config/rs6000/htm.md: Ditto.
7149         * config/rs6000/rs6000.md: Ditto.
7150         * config/rs6000/sync.md: Ditto.
7151         * config/rs6000/vsx.md: Ditto.
7153 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7155         * config/rs6000/altivec.md: Don't set length attribute to the default
7156         value, for branch instructions.
7157         * config/rs6000/darwin.md: Ditto.
7158         * config/rs6000/rs6000.md: Ditto.
7160 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7162         * config/rs6000/rs6000.md (length): Always define as const_int 4.
7163         (unnamed conditional branch define_insn): Set length to 4 or 8
7164         depending on offset.
7165         (<bd>_<mode>): Similar, for alternative 0.
7166         (<bd>tf_<mode>): Ditto.
7168 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7170         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
7172 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
7174         * doc/rtl.texi: Replace old RTX class names with new names.
7177 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
7179         * expmed.h (canonicalize_comparison): New declaration.
7180         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
7181         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
7182         * optabs.c (prepare_cmp_insn): Likewise.
7183         * rtl.h (unsigned_condition_p): New function which checks if a
7184         comparison operator is unsigned.
7186 2018-08-16  Nathan Sidwell  <nathan@acm.org>
7188         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
7189         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
7191 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7193         PR target/84711
7194         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
7195         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
7196         (mov<mov>): ..this and enable unconditionally.
7198 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7200         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
7202 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
7204         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
7205         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
7206         "Common" with "Target".
7208 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
7210         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
7211         * doc/invoke.texi (mmitigate-rop): Remove.
7212         * config/i386/i386.c: Do not include "regrename.h".
7213         (ix86_rop_should_change_byte_p, reg_encoded_number)
7214         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
7215         Remove.
7216         (ix86_reorg): Remove call to ix86_mitigate_rop.
7217         * config/i386/i386.md (attr "modrm_class"): Remove.
7218         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
7219         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
7220         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
7222 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7224         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
7225         allow folding of mergeh() and mergel() for the float and double types.
7226         (fold_mergehl_helper): Rework to handle building a permute tree
7227         for float vectors.
7229 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
7231         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
7232         for TARGET_SSE.
7234 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
7236         * common.opt (fdiagnostics-show-labels): New option.
7237         * diagnostic-show-locus.c (class layout_range): Add field
7238         "m_label".
7239         (class layout): Add field "m_show_labels_p".
7240         (layout_range::layout_range): Add param "label" and use it to
7241         initialize m_label.
7242         (make_range): Pass in NULL for new "label" param of layout_range's
7243         ctor.
7244         (layout::layout): Initialize m_show_labels_p.
7245         (layout::maybe_add_location_range): Pass in loc_range->m_label
7246         when constructing layout_range instances.
7247         (struct line_label): New struct.
7248         (layout::print_any_labels): New member function.
7249         (layout::print_line): Call it if label-printing is enabled.
7250         (selftest::test_one_liner_labels): New test.
7251         (selftest::test_diagnostic_show_locus_one_liner): Call it.
7252         * diagnostic.c (diagnostic_initialize): Initialize
7253         context->show_labels_p.
7254         * diagnostic.h (struct diagnostic_context): Add field
7255         "show_labels_p".
7256         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7257         -fno-diagnostics-show-labels.
7258         * dwarf2out.c (gen_producer_string): Add
7259         OPT_fdiagnostics_show_labels to the ignored options.
7260         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
7261         param.
7262         (gcc_rich_location::maybe_add_expr): Likewise.
7263         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
7264         label" param, defaulting to NULL.
7265         (gcc_rich_location::add_expr): Add "label" param.
7266         (gcc_rich_location::maybe_add_expr): Likewise.
7267         (class text_range_label): New class.
7268         (class range_label_for_type_mismatch): New class.
7269         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
7270         of format_warning_va.
7271         (fmtwarn_n): Likewise for new params of format_warning_n_va.
7272         * lto-wrapper.c (merge_and_complain): Add
7273         OPT_fdiagnostics_show_labels to the "pick one setting" options.
7274         (append_compiler_options): Likewise to the dropped options.
7275         (append_diag_options): Likewise to the passed-on options.
7276         * opts.c (common_handle_option): Handle the new option.
7277         * selftest-diagnostic.c
7278         (test_diagnostic_context::test_diagnostic_context): Enable
7279         show_labels_p.
7280         * substring-locations.c: Include "gcc-rich-location.h".
7281         (format_warning_n_va): Add "fmt_label" and "param_label" params
7282         and use them as appropriate.
7283         (format_warning_va): Add "fmt_label" and "param_label" params,
7284         passing them on to format_warning_n_va.
7285         (format_warning_at_substring): Likewise.
7286         (format_warning_at_substring_n): Likewise.
7287         * substring-locations.h (format_warning_va): Add "fmt_label" and
7288         "param_label" params.
7289         (format_warning_n_va): Likewise.
7290         (format_warning_at_substring): Likewise.
7291         (format_warning_at_substring_n): Likewise.
7292         * toplev.c (general_init): Initialize global_dc->show_labels_p.
7294 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
7296         PR testsuite/86519
7297         * builtins.c (expand_builtin_memcmp): Do not expand the call
7298         when overflow is detected.
7300 2018-08-15  Martin Sebor  <msebor@redhat.com>
7302         PR tree-optimization/71625
7303         * config/aarch64/aarch64-builtins.c
7304         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
7306 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
7308         * config/s390/s390.c (s390_reorg): Remove loop.
7310 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7312         * config/darwin.c
7313          (darwin_function_switched_text_sections): Delete.
7314         * gcc/config/darwin.h
7315          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
7317 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7319         PR target/81685
7320         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
7321         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
7322         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
7324 2018-08-15  Martin Liska  <mliska@suse.cz>
7326         PR tree-optimization/86925
7327         * predict.c (expr_expected_value_1): When taking
7328         later predictor, assign also probability.
7329         Use fold_build2_initializer_loc in order to fold
7330         the expression in -frounding-math.
7332 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
7334         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
7335         patterns.
7336         (expand_vec_perm_1): Try the new method.
7338 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7340         PR target/86547
7341         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
7342         Check whether lra_live_max_point is 0 before dividing.
7344 2018-08-14  Martin Sebor  <msebor@redhat.com>
7346         PR tree-optimization/86650
7347         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
7348         (vrp_prop::check_mem_ref): Same.
7350 2018-08-13  Liu Hao <lh_mouse@126.com>
7352         * pretty-print.c (eat_esc_sequence): Swap the foreground and
7353         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
7354         and clear it thereafter, as it only works for DBCS.
7356 2018-08-13  Liu Hao <lh_mouse@126.com>
7358         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
7359         handle returned by _get_osf_handle().
7361 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7363         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
7364         for folding vec_perm.
7366 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7368         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
7369         Add support for gimple-folding of vec_pack() and vec_unpack()
7370         intrinsics.
7372 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
7374         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
7375         vec_xst variants to the list.
7376         (rs6000_gimple_fold_builtin): Add support for folding unaligned
7377         vector loads and stores.
7379 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
7381         * config.gcc (rs6000-ibm-aix4.x): Delete.
7382         (rs6000-ibm-aix5.1): Delete.
7383         (rs6000-ibm-aix5.2): Delete.
7384         (rs6000-ibm-aix5.3): Delete.
7385         * config/rs6000/aix43.h: Delete.
7386         * config/rs6000/aix51.h: Delete.
7387         * config/rs6000/aix52.h: Delete.
7388         * config/rs6000/t-aix43: Delete.
7390 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
7392         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7393         New function.
7394         (s390_decompose_address): Factor out constant pool ref
7395         decomposition.
7397 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7399         * config/nds32/nds32-predicates.c
7400         (nds32_can_use_bclr_p): Change return type as bool.
7401         (nds32_can_use_bset_p): Ditto.
7402         (nds32_can_use_btgl_p): Ditto.
7403         (nds32_can_use_bitci_p): Ditto.
7404         * config/nds32/nds32-protos.h
7405         (nds32_can_use_bclr_p): Change declaration.
7406         (nds32_can_use_bset_p): Ditto.
7407         (nds32_can_use_btgl_p): Ditto.
7408         (nds32_can_use_bitci_p): Ditto.
7410 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7412         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
7413         Support -msched-prolog-epilog option.
7414         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
7416 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7418         * common/config/nds32/nds32-common.c
7419         (nds32_option_optimization_table): Enalbe -malways-align.
7421 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7423         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
7424         extra_headers.
7425         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
7426         OPT_misr_secure_ case.
7427         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
7428         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
7429         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
7430         secure attribute.
7431         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
7432         (nds32_isr_info): New field security_level.
7433         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
7434         * config/nds32/nds32.md (return_internal): Consider critical attribute.
7435         * config/nds32/nds32.opt (misr-secure): New option.
7436         * config/nds32/nds32_init.inc: New file.
7437         * config/nds32/nds32_isr.h: New file.
7439 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
7441         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
7442         Update comment for atomic instructions.
7443         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
7444         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
7445         Remove.
7446         (atomic_loaddi): Revise fence expansion to only emit fence prior to
7447         load for __ATOMIC_SEQ_CST model.
7448         (atomic_loaddi_1): Remove float register target.
7449         (atomic_storedi): Handle CONST_INT values.
7450         (atomic_storedi_1): Remove float register source.  Add special case
7451         for zero value.
7452         (memory_barrier): New expander and insn.
7454 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
7456         PR tree-optimization/86835
7457         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
7458         new_stmt after def_gsi, make sure to insert new_square_stmt after
7459         that stmt, not 2 stmts before it.
7461 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
7463         PR target/82418
7464         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
7465         instead of SWI48.
7467 2018-08-10  Martin Liska  <mliska@suse.cz>
7469         PR target/83610
7470         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
7471         function type.
7472         * builtins.c (expand_builtin_expect_with_probability):
7473         New function.
7474         (expand_builtin_expect_with_probability): New function.
7475         (build_builtin_expect_predicate): Add new argumnet probability
7476         for BUILT_IN_EXPECT_WITH_PROBABILITY.
7477         (fold_builtin_expect):
7478         (fold_builtin_2):
7479         (fold_builtin_3):
7480         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
7481         * builtins.h (fold_builtin_expect): Set new argument.
7482         * doc/extend.texi: Document __builtin_expect_with_probability.
7483         * doc/invoke.texi: Likewise.
7484         * gimple-fold.c (gimple_fold_call): Pass new argument.
7485         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
7486         also BUILT_IN_EXPECT_WITH_PROBABILITY.
7487         * predict.c (get_predictor_value): New function.
7488         (expr_expected_value): Add new argument probability. Assume
7489         that predictor and probability are always non-null.
7490         (expr_expected_value_1): Likewise.  For __builtin_expect and
7491         __builtin_expect_with_probability set probability.  Handle
7492         combination in binary expressions.
7493         (tree_predict_by_opcode): Simplify code by simply calling
7494         get_predictor_value.
7495         (pass_strip_predict_hints::execute): Add handling of
7496         BUILT_IN_EXPECT_WITH_PROBABILITY.
7497         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
7498         new predictor.
7499         * tree.h (DECL_BUILT_IN_P): New function.
7501 2018-08-10  Martin Liska  <mliska@suse.cz>
7503         PR tree-optimization/85799
7504         * passes.def: Add argument for pass_strip_predict_hints.
7505         * predict.c (class pass_strip_predict_hints): Add new argument
7506         early_p.
7507         (strip_predictor_early): New function.
7508         (pass_strip_predict_hints::execute): Call the function to
7509         strip predictors.
7510         (strip_predict_hints): New function.
7511         * predict.def: Fix comment.
7513 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
7515         * Makefile.in: Clarify which tm.texi to copy over to assert the
7516         right to grant a GFDL license for all.
7518 2018-08-09  Jeff Law  <law@redhat.com>
7520         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
7521         unused variable.
7523 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
7525         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
7526         prototype.
7528 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7530         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
7531         reductions for variable-length vectors.
7533 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
7535         PR other/84889
7536         * common.opt (fdiagnostics-show-line-numbers): New option.
7537         * diagnostic-show-locus.c (class layout): Add fields
7538         "m_show_line_numbers_p" and "m_linenum_width";
7539         (num_digits): New function.
7540         (test_num_digits): New function.
7541         (layout::layout): Initialize new fields.  Update m_x_offset
7542         logic to handle any left margin.
7543         (layout::print_source_line): Print line number when requested.
7544         (layout::start_annotation_line): New member function.
7545         (layout::print_annotation_line): Call it.
7546         (layout::print_leading_fixits): Likewise.
7547         (layout::print_trailing_fixits): Likewise.  Update calls to
7548         move_to_column for new parameter.
7549         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
7550         it to potentially call start_annotation_line.
7551         (layout::show_ruler): Call start_annotation_line.
7552         (selftest::test_line_numbers_multiline_range): New selftest.
7553         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
7554         and selftest::test_line_numbers_multiline_range.
7555         * diagnostic.c (diagnostic_initialize): Initialize
7556         show_line_numbers_p.
7557         * diagnostic.h (struct diagnostic_context): Add field
7558         "show_line_numbers_p".
7559         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7560         -fno-diagnostics-show-line-numbers.
7561         * dwarf2out.c (gen_producer_string): Add
7562         OPT_fdiagnostics_show_line_numbers to the ignored options.
7563         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
7564         one setting" options.
7565         (append_compiler_options): Likewise to the dropped options.
7566         (append_diag_options): Likewise to the passed-on options.
7567         * opts.c (common_handle_option): Handle the new option.
7568         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
7570 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7572         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
7573         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
7574         third argument of type "const signed char" to descriptions of
7575         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
7576         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
7577         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
7578         __builtin_bcdsub_ov functions.
7580 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7582         PR tree-optimization/86858
7583         * tree-vect-loop.c (vect_is_simple_reduction): Restore
7584         flow_bb_inside_loop_p calls.
7586 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7588         PR tree-optimization/86871
7589         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
7590         instead of gimple_assign_lhs.
7592 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
7594         PR target/86887
7595         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
7596         register constraint to operand 0.
7597         (add<mode>3_carryinC): Likewise.
7598         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
7600 2018-08-09  Martin Liska  <mliska@suse.cz>
7602         PR c/86895
7603         * common.opt: Remove extra line.
7605 2018-08-09  Martin Liska  <mliska@suse.cz>
7607         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
7608         at the end of a line, make first letter capital and end up
7609         a sentence with a dot.
7610         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
7611         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
7612         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
7613         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
7614         (PARAM_MAX_ISL_OPERATIONS): Likewise.
7615         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
7616         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
7617         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
7618         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
7619         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
7620         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
7621         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
7622         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7623         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7624         (PARAM_TREE_REASSOC_WIDTH): Likewise.
7625         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7626         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7627         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7629 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
7631         PR target/84332
7632         * config/s390/s390.c (s390_option_override_internal): Reduce the
7633         stack-clash-protection-probe-interval param if it would be too big
7634         for z900.
7636 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
7638         PR target/46179
7639         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7640         * config/m68k/m68k.c (handle_move_double): Don't call
7641         m68k_final_prescan_insn.
7642         (m68k_adjust_decorated_operand): Renamed from
7643         m68k_final_prescan_insn, remove first and third operand and
7644         simplify.
7645         (print_operand): Call it.
7646         (print_operand_address): Call it.
7648 2018-08-08  Nathan Sidwell  <nathan@acm.org>
7650         * diagnostic.c (diagnostic_report_current_module): Use
7651         linemap_included_from & linemap_included_from_linemap.
7653 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
7655         * config/aarch64/aarch64-cores.def: Add phecda core.
7656         * config/aarch64/aarch64-tune.md: Regenerate.
7657         * doc/invoke.texi: Add phecda core.
7659 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7661         PR target/85295
7662         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7663         definitions.
7664         * config/s390/s390.md ("movti"): Add more alternatives for
7665         constant to GPR copies.
7667 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7669         * config/s390/s390.c: Fix whitespace damage throughout the file.
7670         * config/s390/s390.h: Likewise.
7671         * config/s390/tpf.h: Likewise.
7673 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7675         * config/s390/s390.c (s390_loadrelative_operand_p):
7676         Remove TARGET_CPU_ZARCH usages.
7677         (s390_rtx_costs): Likewise.
7678         (s390_legitimate_constant_p): Likewise.
7679         (s390_cannot_force_const_mem): Likewise.
7680         (legitimate_reload_constant_p): Likewise.
7681         (s390_preferred_reload_class): Likewise.
7682         (legitimize_pic_address): Likewise.
7683         (legitimize_tls_address): Likewise.
7684         (s390_split_branches): Removed.
7685         (s390_add_execute): Removed.
7686         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7687         (s390_mainpool_start): Likewise.
7688         (s390_mainpool_finish): Likewise.
7689         (s390_mainpool_cancel): Removed.
7690         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7691         (s390_chunkify_cancel): Likewise.
7692         (s390_return_addr_rtx): Likewise.
7693         (s390_register_info): Remove split_branches_pending_p uages.
7694         (s390_optimize_register_info): Likewise.
7695         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7696         split_branches_pending_p usages.
7697         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7698         (s390_load_got): Likewise.
7699         (s390_expand_split_stack_prologue): Likewise.
7700         (output_asm_nops): Likewise.
7701         (s390_function_profiler): Likewise.
7702         (s390_emit_call): Likewise.
7703         (s390_conditional_register_usage): Likewise.
7704         (s390_optimize_prologue): Likewise.
7705         (s390_reorg): Remove TARGET_CPU_ZARCH and
7706         split_branches_pending_p usages.
7707         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7708         usages.
7709         (s390_output_indirect_thunk_function): Likewise.
7710         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7711         (TARGET_CPU_ZARCH_P): Removed.
7712         (struct machine_function): Remove split_branches_pending_p.
7713         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7715 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7717         * common/config/s390/s390-common.c (processor_flags_table):
7718         Remove flags.
7719         * config.gcc: Remove with_arch/with_tune support.
7720         * config/s390/2064.md: Remove cpu attribute comparisons.
7721         * config/s390/driver-native.c (s390_host_detect_local_cpu):
7722         Remove MTN.
7723         * config/s390/linux.h (ASM_SPEC):
7724         Remove -march support.
7725         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7726         Use a table to get an arch level.
7727         * config/s390/s390-opts.h (enum processor_type):
7728         Remove enum values.
7729         * config/s390/s390.c
7730         (processor_table): Remove entries, add arch_level values.
7731         (s390_issue_rate): Remove cases.
7732         (s390_option_override): Adjust
7733         s390_option_override_internal() call.
7734         (s390_option_override_internal): Remove deprecation warning.
7735         (s390_valid_target_attribute_tree): Adjust
7736         s390_option_override_internal() call.
7737         * config/s390/s390.h (struct s390_processor):
7738         Share with s390-c.c, add arch_level field.
7739         * config/s390/s390.md:
7740         Remove occurrences in cpu attribute.
7741         * config/s390/s390.opt: Remove -march/-mtune support.
7742         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7743         * doc/invoke.texi: Remove deprecation warning.
7745 2018-08-08  Luis Machado  <luis.machado@linaro.org>
7747         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7748         global.
7749         (qdf24xx_tunings): Set vector cost structure to
7750         qdf24xx_vector_cost.
7752         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7753         <register_sextend>: Set to 3.
7755 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
7757         PR target/86838
7758         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7759         * config/aarch64/aarch64-simd.md
7760         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7761         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7762         (aarch64_frecpx<mode>): ...this new pattern.
7763         * config/aarch64/aarch64-simd-builtins.def: Remove comment
7764         about aarch64_frecp<FRECP:frecp_suffix><mode>.
7766 2018-08-07  Martin Liska  <mliska@suse.cz>
7768         PR middle-end/83023
7769         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7770         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7771         * predict.def (PRED_MALLOC_NONNULL): New predictor.
7772         * doc/extend.texi: Document that malloc attribute adds
7773         hit to compiler.
7775 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
7777         PR target/86785
7778         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7779         Define to speculation_safe_value_not_needed.
7781 2018-08-06  Jeff Law  <law@redhat.com>
7783         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7784         the vr_values instance to cprop_into_stmt.
7785         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7786         (cprop_operand): Also query EVRP to determine if OP is a constant.
7788 2018-08-06  Nathan Sidwell  <nathan@acm.org>
7790         * diagnostic.c (diagnostic_report_current_module): Reroll
7791         included-at loop.  Translate text.
7793 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
7795         * function-tests.c (selftest::test_expansion_to_rtl): Call
7796         free_after_compilation.
7798 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7800         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7802 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
7804         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7805         loops with memory block operations from getting unrolled.
7807 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
7809         PR target/86807
7810         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7811         Define to speculation_safe_value_not_needed.
7813 2018-08-06  Jeff Law  <law@redhat.com>
7815         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7816         assert.
7818 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7820         PR target/86662
7821         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7822         with all enabled __intN types.
7824         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7826 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7828         * alias.c (record_set): Check for clobber high.
7829         * cfgexpand.c (expand_gimple_stmt): Likewise.
7830         * combine-stack-adj.c (single_set_for_csa): Likewise.
7831         * combine.c (find_single_use_1): Likewise.
7832         (set_nonzero_bits_and_sign_copies): Likewise.
7833         (get_combine_src_dest): Likewise.
7834         (is_parallel_of_n_reg_sets): Likewise.
7835         (try_combine): Likewise.
7836         (record_dead_and_set_regs_1): Likewise.
7837         (reg_dead_at_p_1): Likewise.
7838         (reg_dead_at_p): Likewise.
7839         * dce.c (deletable_insn_p): Likewise.
7840         (mark_nonreg_stores_1): Likewise.
7841         (mark_nonreg_stores_2): Likewise.
7842         * df-scan.c (df_find_hard_reg_defs): Likewise.
7843         (df_uses_record): Likewise.
7844         (df_get_call_refs): Likewise.
7845         * dwarf2out.c (mem_loc_descriptor): Likewise.
7846         * haifa-sched.c (haifa_classify_rtx): Likewise.
7847         * ira-build.c (create_insn_allocnos): Likewise.
7848         * ira-costs.c (scan_one_insn): Likewise.
7849         * ira.c (equiv_init_movable_p): Likewise.
7850         (rtx_moveable_p): Likewise.
7851         (interesting_dest_for_shprep): Likewise.
7852         * jump.c (mark_jump_label_1): Likewise.
7853         * postreload-gcse.c (record_opr_changes): Likewise.
7854         * postreload.c (reload_cse_simplify): Likewise.
7855         (struct reg_use): Add source expr.
7856         (reload_combine): Check for clobber high.
7857         (reload_combine_note_use): Likewise.
7858         (reload_cse_move2add): Likewise.
7859         (move2add_note_store): Likewise.
7860         * print-rtl.c (print_pattern): Likewise.
7861         * recog.c (decode_asm_operands): Likewise.
7862         (store_data_bypass_p): Likewise.
7863         (if_test_bypass_p): Likewise.
7864         * regcprop.c (kill_clobbered_value): Likewise.
7865         (kill_set_value): Likewise.
7866         * reginfo.c (reg_scan_mark_refs): Likewise.
7867         * reload1.c (maybe_fix_stack_asms): Likewise.
7868         (eliminate_regs_1): Likewise.
7869         (elimination_effects): Likewise.
7870         (mark_not_eliminable): Likewise.
7871         (scan_paradoxical_subregs): Likewise.
7872         (forget_old_reloads_1): Likewise.
7873         * reorg.c (find_end_label): Likewise.
7874         (try_merge_delay_insns): Likewise.
7875         (redundant_insn): Likewise.
7876         (own_thread_p): Likewise.
7877         (fill_simple_delay_slots): Likewise.
7878         (fill_slots_from_thread): Likewise.
7879         (dbr_schedule): Likewise.
7880         * resource.c (update_live_status): Likewise.
7881         (mark_referenced_resources): Likewise.
7882         (mark_set_resources): Likewise.
7883         * rtl.c (copy_rtx): Likewise.
7884         * rtlanal.c (reg_referenced_p): Likewise.
7885         (single_set_2): Likewise.
7886         (noop_move_p): Likewise.
7887         (note_stores): Likewise.
7888         * sched-deps.c (sched_analyze_reg): Likewise.
7889         (sched_analyze_insn): Likewise.
7891 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7893         * cse.c (invalidate_reg): New function extracted from...
7894         (invalidate): ...here.
7895         (canonicalize_insn): Check for clobber high.
7896         (invalidate_from_clobbers): invalidate clobber highs.
7897         (invalidate_from_sets_and_clobbers): Likewise.
7898         (count_reg_usage): Check for clobber high.
7899         (insn_live_p): Likewise.
7900         * cselib.c (cselib_expand_value_rtx_1):Likewise.
7901         (cselib_invalidate_regno): Check for clobber in setter.
7902         (cselib_invalidate_rtx): Pass through setter.
7903         (cselib_invalidate_rtx_note_stores):
7904         (cselib_process_insn): Check for clobber high.
7905         * cselib.h (cselib_invalidate_rtx): Add operand.
7907 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7909         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7910         (mark_not_eliminable): Likewise.
7911         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7912         * lra-lives.c (process_bb_lives): Check for clobber high.
7913         * lra.c (new_insn_reg): Remember clobber highs.
7914         (collect_non_operand_hard_regs): Check for clobber high.
7915         (lra_set_insn_recog_data): Likewise.
7916         (add_regs_to_insn_regno_info): Likewise.
7917         (lra_update_insn_regno_info): Likewise.
7919 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7921         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7922         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7924 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7926         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7927         (copy_insn_1): Likewise.
7928         (gen_hard_reg_clobber_high): New gen function.
7929         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7930         * genemit.c (gen_exp): Likewise.
7931         (gen_emit_seq): Pass through info.
7932         (gen_insn): Check for CLOBBER_HIGH.
7933         (gen_expand): Pass through info.
7934         (gen_split): Likewise.
7935         (output_add_clobbers): Likewise.
7936         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7937         (remove_clobbers): Likewise.
7938         * rtl.h (gen_hard_reg_clobber_high): New declaration.
7940 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7942         * doc/rtl.texi (clobber_high): Add.
7943         (parallel): Add in clobber high
7944         * rtl.c (rtl_check_failed_code3): Add function.
7945         * rtl.def (CLOBBER_HIGH): Add expression.
7946         * rtl.h (RTL_CHECKC3): Add macro.
7947         (rtl_check_failed_code3): Add declaration.
7948         (XC3EXP): Add macro.
7950 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
7952         PR target/86386
7953         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7954         cfun->machine->max_used_stack_alignment if needed.
7956 2018-08-04  Martin Sebor  <msebor@redhat.com>
7958         PR tree-optimization/86571
7959         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7960         NaN output to 4.
7962 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
7964         PR target/86799
7965         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7966         Define.
7968 2018-08-03  Jeff Law  <law@redhat.com>
7970         PR target/86795
7971         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7972         Define to speculation_safe_value_not_needed.
7974 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
7976         * doc/gcov.texi (-x): Remove duplicate "to".
7977         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7978         (-Wif-not-aligned): Remove duplicate "is".
7979         (-flto): Remove duplicate "the".
7980         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7981         duplicate "v5.00.b".
7982         (MSP430 Options): Remove duplicate "and" from the description
7983         of "-mgprel-sec=regexp".
7984         (x86 Options): Remove duplicate copies of "vmldLog102" and
7985         vmlsLog104 from description of "-mveclibabi=type".
7987 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
7989         * internal-fn.h (first_commutative_argument): Declare.
7990         * internal-fn.c (first_commutative_argument): New function.
7991         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7992         restrictions for pattern statements.  Use first_commutative_argument
7993         to look for commutative operands in calls to internal functions.
7995 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
7997         * Makefile.in (wide-int-range.o): New.
7998         * tree-vrp.c: Move all the wide_int_* functions to...
7999         * wide-int-range.cc: ...here.
8000         * tree-vrp.h: Move all the wide_int_* prototypes to...
8001         * wide-int-range.h: ...here.
8003 2018-08-03  Tom de Vries  <tdevries@suse.de>
8005         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
8006         UI_NONE.
8007         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
8008         * except.c (output_function_exception_table): Do early exit if
8009         targetm_common.except_unwind_info (&global_options) == UI_NONE.
8011 2018-08-03  Martin Liska  <mliska@suse.cz>
8013         * predict.c (dump_prediction): Change to 2 digits
8014         in fraction part.
8016 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
8018         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
8019         neon_dup_q to...
8020         (falkor_am_1_gtov_gtov): ... a new insn reservation.
8022 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
8024         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
8025         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
8026         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
8027         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
8028         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
8030 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
8032         * diagnostic-show-locus.c (diagnostic_show_locus): Use
8033         pp_take_prefix when saving the existing prefix.
8034         * diagnostic.c (diagnostic_append_note): Likewise.
8035         * langhooks.c (lhd_print_error_function): Likewise.
8036         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
8037         param's type.  Free the existing prefix.
8038         (pp_take_prefix): New function.
8039         (pretty_printer::pretty_printer): Drop the prefix parameter.
8040         Rename the length parameter to match the comment.
8041         (pretty_printer::~pretty_printer): Free the prefix.
8042         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
8043         parameter.
8044         (struct pretty_printer): Drop the "const" from "prefix" field's
8045         type and clarify memory management.
8046         (pp_set_prefix): Drop the "const" from the 2nd param.
8047         (pp_take_prefix): New decl.
8049 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
8051         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
8052         for word_mode_ok here instead of passing as argument.
8053         (expand_block_compare): Change select_block_compare_mode() call.
8054         (expand_strncmp_gpr_sequence): New function.
8055         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
8057 2018-08-02  Jeff Law  <law@redhat.com>
8059         PR target/86790
8060         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8061         Define to speculation_safe_value_not_needed.
8063         PR target/86784
8064         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8065         Define to speculation_safe_value_not_needed.
8067 2018-08-02  Tom de Vries  <tdevries@suse.de>
8069         PR target/86660
8070         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
8071         function.  Return UI_TARGET unconditionally.
8072         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
8073         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
8075 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8077         * genemit.c (print_overload_test): Fix typo.
8079 2018-08-02  Richard Biener  <rguenther@suse.de>
8081         PR tree-optimization/86816
8082         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
8083         which checks for value availability before querying it.
8084         (gvn_uses_equal): Use it.
8085         (same_succ_hash): Likewise.
8086         (gimple_equal_p): Likewise.
8088 2018-08-02  Nick Clifton  <nickc@redhat.com>
8090         PR target/86813
8091         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8092         Define to speculation_safe_value_not_needed.
8094         PR target/86810
8095         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8096         Define to speculation_safe_value_not_needed.
8098         PR target/86810
8099         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8100         Define to speculation_safe_value_not_needed.
8102         PR target/86803
8103         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8104         Define to speculation_safe_value_not_needed.
8106         PR target/86797
8107         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8108         Define to speculation_safe_value_not_needed.
8110         PR target/86791
8111         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8112         Define to speculation_safe_value_not_needed.
8114         PR target/86789
8115         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8116         Define to speculation_safe_value_not_needed.
8118         PR target/86787
8119         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8120         Define to speculation_safe_value_not_needed.
8122         PR target/86782
8123         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8124         speculation_safe_value_not_needed.
8126         PR target/86781
8127         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
8128         to speculation_safe_value_not_needed.
8130 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8132         * doc/md.texi: Expand the documentation of instruction names
8133         to mention port-local uses.  Document '@' in pattern names.
8134         * read-md.h (overloaded_instance, overloaded_name): New structs.
8135         (mapping): Declare.
8136         (md_reader::handle_overloaded_name): New member function.
8137         (md_reader::get_overloads): Likewise.
8138         (md_reader::m_first_overload): New member variable.
8139         (md_reader::m_next_overload_ptr): Likewise.
8140         (md_reader::m_overloads_htab): Likewise.
8141         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
8142         m_next_overload_ptr and m_overloads_htab.
8143         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
8144         (get_mode_token, get_code_token, get_int_token): New functions.
8145         (map_attr_string): Add an optional argument that passes back
8146         the associated iterator.
8147         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
8148         (md_reader::handle_overloaded_name, add_overload_instance): New
8149         functions.
8150         (apply_iterators): Handle '@' names.  Report an error if '@'
8151         is used without iterators.
8152         (initialize_iterators): Initialize the new iterator_group fields.
8153         * genopinit.c (handle_overloaded_code_for)
8154         (handle_overloaded_gen): New functions.
8155         (main): Use them to print declarations of maybe_code_for_* and
8156         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
8157         * genemit.c (print_overload_arguments, print_overload_test)
8158         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
8159         (main): Use it to print definitions of maybe_code_for_* and
8160         maybe_gen_* functions.
8161         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
8162         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
8163         instead of explicit mode checks.
8164         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
8165         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
8166         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
8167         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
8168         (aarch64_expand_compare_and_swap): Likewise
8169         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
8170         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
8171         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
8172         (aarch64_constant_pool_reload_icode): Delete.
8173         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
8174         instead of aarch64_constant_pool_reload_icode.  Use
8175         code_for_aarch64_reload_mov instead of explicit mode checks.
8176         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
8177         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
8178         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
8179         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
8180         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
8181         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
8182         (aarch64_atomic_load_op_code): Delete.
8183         (aarch64_emit_atomic_load_op): Likewise.
8184         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
8185         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
8186         instead of aarch64_emit_atomic_load_op.
8187         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
8188         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
8189         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
8190         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
8191         character before the pattern name.
8192         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8193         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
8194         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
8195         (aarch64_frecps<mode>): Likewise.
8196         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
8197         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
8198         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
8199         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
8200         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
8202 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8204         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
8205         Allow HFmode constants if TARGET_FP_F16INST.
8207 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
8209         PR target/86014
8210         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
8211         No longer check last store for clobber of address register.
8213 2018-08-02  Martin Liska  <mliska@suse.cz>
8215         PR gcov-profile/86817
8216         * gcov.c (process_all_functions): New function.
8217         (main): Call it.
8218         (process_file): Move functions processing to
8219         process_all_functions.
8221 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
8223         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
8224         "const" to the "gimple *" and "rtx_insn *" parameters.
8225         * dumpfile.h (dump_user_location_t::dump_user_location_t):
8226         Likewise.
8227         (dump_location_t::dump_location_t): Likewise.
8229 2018-08-01  Martin Sebor  <msebor@redhat.com>
8231         PR tree-optimization/86650
8232         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
8233         rather than a "gcall *".  Directly pass the data of interest
8234         to percent_K_format, rather than building a temporary CALL_EXPR
8235         to hold it.
8236         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
8237         (gimple_fold_builtin_strncat): Adjust.
8238         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
8239         gcall* argument with gimple*.
8240         * gimple-ssa-warn-restrict.c (check_call): Same.
8241         (wrestrict_dom_walker::before_dom_children): Same.
8242         (builtin_access::builtin_access): Same.
8243         (check_bounds_or_overlap): Same
8244         (maybe_diag_overlap): Same.
8245         (maybe_diag_offset_bounds): Same.
8246         * tree-diagnostic.c (default_tree_printer): Move usage of
8247         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
8248         to this callsite.
8249         * tree-pretty-print.c (percent_K_format): Add argument.
8250         * tree-pretty-print.h: Add argument.
8251         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
8252         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
8253         (maybe_diag_stxncpy_trunc): Same.
8254         (handle_builtin_stxncpy): Same.
8255         (handle_builtin_strcat): Same.
8257 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8259         * match.pd: Optimise pointer range checks.
8261 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8263         PR tree-optimization/86758
8264         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
8265         to remove pattern statements.
8267 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8269         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
8270         result of dfs_enumerate_from when constructing stmt_vec_infos,
8271         instead of additionally calling get_loop_body.
8273 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8275         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
8276         parameter.
8277         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
8278         When creating an iv, assert that the step is not known to be zero.
8279         (vect_setup_realignment): Update call accordingly.
8280         * tree-vect-stmts.c (vectorizable_store): Likewise.
8281         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
8283 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8285         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
8286         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
8287         (vectorizable_reduction): Likewise.
8288         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8289         (vect_detect_hybrid_slp_stmts): Likewise.
8290         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8292 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
8294         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
8295         (wide_int_set_zero_nonzero_bits): ...this.
8296         (zero_nonzero_bits_from_vr): Rename to...
8297         (vrp_set_zero_nonzero_bits): ...this.
8298         (extract_range_from_multiplicative_op_1): Abstract wide int
8299         code...
8300         (wide_int_range_multiplicative_op): ...here.
8301         (extract_range_from_binary_expr_1): Extract wide int binary
8302         operations into their own functions.
8303         (wide_int_range_lshift): New.
8304         (wide_int_range_can_optimize_bit_op): New.
8305         (wide_int_range_shift_undefined_p): New.
8306         (wide_int_range_bit_xor): New.
8307         (wide_int_range_bit_ior): New.
8308         (wide_int_range_bit_and): New.
8309         (wide_int_range_trunc_mod): New.
8310         (extract_range_into_wide_ints): New.
8311         (vrp_shift_undefined_p): New.
8312         (extract_range_from_multiplicative_op): New.
8313         (vrp_can_optimize_bit_op): New.
8314         * tree-vrp.h (value_range::dump): New.
8315         (wide_int_range_multiplicative_op): New.
8316         (wide_int_range_lshift):New.
8317         (wide_int_range_shift_undefined_p): New.
8318         (wide_int_range_bit_xor): New.
8319         (wide_int_range_bit_ior): New.
8320         (wide_int_range_bit_and): New.
8321         (wide_int_range_trunc_mod): New.
8322         (zero_nonzero_bits_from_bounds): Rename to...
8323         (wide_int_set_zero_nonzero_bits): ...this.
8324         (zero_nonzero_bits_from_vr): Rename to...
8325         (vrp_set_zero_nonzero_bits): ...this.
8326         (range_easy_mask_min_max): Rename to...
8327         (wide_int_range_can_optimize_bit_op): this.
8328         * vr-values.c (simplify_bit_ops_using_ranges): Rename
8329         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
8331 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8333         * tree-vectorizer.h (vect_orig_stmt): New function.
8334         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
8335         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
8336         (vect_create_epilog_for_reduction): Likewise.
8337         (vectorizable_live_operation): Likewise.
8338         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8339         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
8340         * tree-vect-stmts.c (vectorizable_call): Likewise.
8341         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
8343 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8345         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
8346         argument.
8347         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
8348         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
8349         (vect_transform_loop): Likewise.
8350         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8352 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8354         * tree-vectorizer.h (vect_schedule_slp): Return void.
8355         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8356         (vect_schedule_slp): Likewise.
8358 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8360         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
8361         argument.
8362         (vect_transform_loop): Update calls accordingly.  Schedule SLP
8363         instances before the main loop, if any exist.
8365 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8367         PR tree-optimization/86749
8368         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
8369         If the lhs is used in a COND_EXPR, check that it is being used
8370         as the "then" or "else" value.
8372 2018-08-01  Tom de Vries  <tdevries@suse.de>
8374         PR target/86800
8375         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8376         speculation_safe_value_not_needed.
8378 2018-08-01  Richard Biener  <rguenther@suse.de>
8380         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
8381         as base and offset.
8383 2018-08-01  Martin Liska  <mliska@suse.cz>
8385         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
8386         format how successful transformation is dumped.
8387         (gimple_mod_pow2_value_transform): Likewise.
8388         (gimple_mod_subtract_transform): Likewise.
8389         (gimple_stringops_transform): Likewise.
8391 2018-08-01  Martin Liska  <mliska@suse.cz>
8393         PR value-prof/35543
8394         * value-prof.c (interesting_stringop_to_profile_p):
8395         Simplify the code and add BUILT_IN_MEMMOVE.
8396         (gimple_stringops_transform): Likewise.
8398 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
8400         * config/aarch64/aarch64-simd.md
8401         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
8402         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
8403         use GPI iterator instead of SI mode.
8405 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8407         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
8408         rs6000_speculation_barrier.
8409         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
8410         new barrier pattern name.
8412 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8414         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
8415         (speculation_barrier): New insn.
8417 2018-08-01  Richard Biener  <rguenther@suse.de>
8419         PR bootstrap/86724
8420         * graphite.h: Include isl/id.h and isl/space.h to allow build
8421         with ISL 0.20.
8423 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
8425         PR target/86651
8426         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8427         mode for COFF targets.
8428         * defaults.h (TARGET_COFF): Define.
8429         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
8430         TARGET_COFF): Define.
8431         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
8432         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
8433         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
8435 2018-07-31  Alexandre Oliva <oliva@adacore.com>
8437         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
8438         * gimple-streamer-out.c (output_bb): Save it.
8439         * lto-streamer-in.c (input_struct_function_base): Restore
8440         instance discriminator if available.  Create map on demand.
8441         * lto-streamer-out.c (output_struct_function_base): Save it if
8442         available.
8443         * final.c (decl_to_instance_map): Document LTO strategy.
8445 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
8446             Olivier Hainque  <hainque@adacore.com>
8448         * debug.h (decl_to_instance_map_t): New type.
8449         (decl_to_instance_map): Declare.
8450         (maybe_create_decl_to_instance_map): New inline function.
8451         * final.c (bb_discriminator, last_bb_discriminator): New statics,
8452         to track basic block discriminators.
8453         (final_start_function_1): Initialize them.
8454         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
8455         bb_discriminator.
8456         (decl_to_instance_map): New variable.
8457         (map_decl_to_instance, maybe_set_discriminator): New functions.
8458         (notice_source_line): Set discriminator.
8460 2018-07-31  Ian Lance Taylor  <iant@golang.org>
8462         * targhooks.c (default_have_speculation_safe_value): Add
8463         ATTRIBUTE_UNUSED.
8465 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8467         * dump-context.h: Include "pretty-print.h".
8468         (dump_context::refresh_dumps_are_enabled): New decl.
8469         (dump_context::emit_item): New decl.
8470         (class dump_context): Add fields "m_test_pp" and
8471         "m_test_pp_flags".
8472         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8473         (temp_dump_context::get_dumped_text): New decl.
8474         (class temp_dump_context): Add field "m_pp".
8475         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
8476         (dump_context::refresh_dumps_are_enabled): ...and add a test for
8477         m_test_pp.
8478         (set_dump_file): Update for above change.
8479         (set_alt_dump_file): Likewise.
8480         (dump_loc): New overload, taking a pretty_printer *.
8481         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
8482         to any test pretty-printer.
8483         (make_item_for_dump_gimple_stmt): New function, adapted from
8484         optinfo::add_gimple_stmt.
8485         (dump_context::dump_gimple_stmt): Call it, and use the result,
8486         eliminating the direct usage of dump_file and alt_dump_file in
8487         favor of indirectly using them via emit_item.
8488         (make_item_for_dump_gimple_expr): New function, adapted from
8489         optinfo::add_gimple_expr.
8490         (dump_context::dump_gimple_expr): Call it, and use the result,
8491         eliminating the direct usage of dump_file and alt_dump_file in
8492         favor of indirectly using them via emit_item.
8493         (make_item_for_dump_generic_expr): New function, adapted from
8494         optinfo::add_tree.
8495         (dump_context::dump_generic_expr): Call it, and use the result,
8496         eliminating the direct usage of dump_file and alt_dump_file in
8497         favor of indirectly using them via emit_item.
8498         (make_item_for_dump_printf_va): New function, adapted from
8499         optinfo::add_printf_va.
8500         (make_item_for_dump_printf): New function.
8501         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
8502         and use the result, eliminating the direct usage of dump_file and
8503         alt_dump_file in favor of indirectly using them via emit_item.
8504         (make_item_for_dump_dec): New function.
8505         (dump_context::dump_dec): Call it, and use the result,
8506         eliminating the direct usage of dump_file and alt_dump_file in
8507         favor of indirectly using them via emit_item.
8508         (make_item_for_dump_symtab_node): New function, adapted from
8509         optinfo::add_symtab_node.
8510         (dump_context::dump_symtab_node): Call it, and use the result,
8511         eliminating the direct usage of dump_file and alt_dump_file in
8512         favor of indirectly using them via emit_item.
8513         (dump_context::begin_scope): Reimplement, avoiding direct usage
8514         of dump_file and alt_dump_file in favor of indirectly using them
8515         via emit_item.
8516         (dump_context::emit_item): New member function.
8517         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8518         Set up test pretty-printer on the underlying context.  Call
8519         refresh_dumps_are_enabled.
8520         (temp_dump_context::~temp_dump_context): Call
8521         refresh_dumps_are_enabled.
8522         (temp_dump_context::get_dumped_text): New member function.
8523         (selftest::verify_dumped_text): New function.
8524         (ASSERT_DUMPED_TEXT_EQ): New macro.
8525         (selftest::test_capture_of_dump_calls): Run all tests twice, with
8526         and then without optinfo enabled.  Add uses of
8527         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
8528         * dumpfile.h: Update comment for the dump_* API.
8529         * optinfo-emit-json.cc
8530         (selftest::test_building_json_from_dump_calls): Update for new
8531         param for temp_dump_context ctor.
8532         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
8533         and "m_owned" field.
8534         (optinfo_item::~optinfo_item): Likewise.
8535         (optinfo::add_item): New member function.
8536         (optinfo::emit): Update comment.
8537         (optinfo::add_string): Delete.
8538         (optinfo::add_printf): Delete.
8539         (optinfo::add_printf_va): Delete.
8540         (optinfo::add_gimple_stmt): Delete.
8541         (optinfo::add_gimple_expr): Delete.
8542         (optinfo::add_tree): Delete.
8543         (optinfo::add_symtab_node): Delete.
8544         (optinfo::add_dec): Delete.
8545         * optinfo.h (class dump_context): New forward decl.
8546         (optinfo::add_item): New decl.
8547         (optinfo::add_string): Delete.
8548         (optinfo::add_printf): Delete.
8549         (optinfo::add_printf_va): Delete.
8550         (optinfo::add_gimple_stmt): Delete.
8551         (optinfo::add_gimple_expr): Delete.
8552         (optinfo::add_tree): Delete.
8553         (optinfo::add_symtab_node): Delete.
8554         (optinfo::add_dec): Delete.
8555         (optinfo::add_poly_int): Delete.
8556         (optinfo_item::optinfo_item): Remove "owned" param.
8557         (class optinfo_item): Remove field "m_owned".
8559 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
8561         PR middle-end/86705
8562         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
8563         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
8564         requested variable alignment.
8565         (expand_one_ssa_partition): Likewise.
8566         (expand_one_var): Likewise.
8568 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8570         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8571         to speculation_safe_value_not_needed.
8573 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8575         * targhooks.h (speculation_safe_value_not_needed): New prototype.
8576         * targhooks.c (speculation_safe_value_not_needed): New function.
8577         * target.def (have_speculation_safe_value): Update documentation.
8578         * doc/tm.texi: Regenerated.
8580 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8582         * config/aarch64/iterators.md (ALLI_TI): New iterator.
8583         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
8584         expand.
8585         (despeculate_copy<ALLI:mode>_insn): New insn.
8586         (despeculate_copyti_insn): New insn.
8587         (despeculate_simple<ALLI:mode>): New insn
8588         (despeculate_simpleti): New insn.
8589         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
8590         function.
8591         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
8592         aarch64_speculation_safe_value.
8593         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
8595 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8597         * config/aarch64/aarch64-speculation.cc: New file.
8598         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
8599         before pass_reorder_blocks.
8600         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
8601         prototype.
8602         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
8603         X14 and X15 when tracking speculation.
8604         * config/aarch64/aarch64.md (register name constants): Add
8605         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
8606         (unspec): Add UNSPEC_SPECULATION_TRACKER.
8607         (speculation_barrier): New insn attribute.
8608         (cmp<mode>): Allow SP in comparisons.
8609         (speculation_tracker): New insn.
8610         (speculation_barrier): Add speculation_barrier attribute.
8611         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
8612         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
8613         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
8615 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8617         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
8618         aarch64_track_speculation is true.
8619         (tb<optab><mode>1): Likewise.
8620         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
8621         generate CB[N]Z when tracking speculation.
8622         (aarch64_split_compare_and_swap): Likewise.
8623         (aarch64_split_atomic_op): Likewise.
8625 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8627         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8629 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8631         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8632         (speculation_barrier): New insn.
8634 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8636         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8637         * config/arm/arm.md (speculation_barrier): New expand.
8638         (speculation_barrier_insn): New pattern.
8640 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8642         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8643         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8644         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8645         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8646         list.
8647         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8648         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8649         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8650         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8651         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8652         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8653         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8654         * builtins.c (expand_speculation_safe_value): New function.
8655         (expand_builtin): Call it.
8656         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8657         * doc/extend.texi: Document __builtin_speculation_safe_value.
8658         * doc/md.texi: Document "speculation_barrier" pattern.
8659         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8660         TARGET_HAVE_SPECULATION_SAFE_VALUE.
8661         * doc/tm.texi: Regenerated.
8662         * target.def (have_speculation_safe_value, speculation_safe_value): New
8663         hooks.
8664         * targhooks.c (default_have_speculation_safe_value): New function.
8665         (default_speculation_safe_value): New function.
8666         * targhooks.h (default_have_speculation_safe_value): Add prototype.
8667         (default_speculation_safe_value): Add prototype.
8669 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8671         * dump-context.h (dump_context::dump_loc): New decl.
8672         * dumpfile.c (dump_context::dump_loc): New member function.
8673         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8674         and dump_gimple_stmt.
8675         (dump_context::dump_gimple_expr_loc): Likewise, using
8676         dump_gimple_expr.
8677         (dump_context::dump_generic_expr_loc): Likewise, using
8678         dump_generic_expr.
8679         (dump_context::dump_printf_loc_va): Likewise, using
8680         dump_printf_va.
8681         (dump_context::begin_scope): Explicitly using the global function
8682         "dump_loc", rather than the member function.
8684 2018-07-31  Martin Sebor  <msebor@redhat.com>
8686         PR tree-optimization/86741
8687         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8689 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
8691         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8692         depend on whether prefetch instructions will be emitted or not.
8693         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8694         will be emitted or not.
8695         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8696         (TARGET_SETMEM_PFD): New macros.
8698 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8700         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8701         (NULL_STMT_VEC_INFO): Delete.
8702         (stmt_vec_info::operator*): Likewise.
8703         (stmt_vec_info::operator gimple *): Likewise.
8704         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8705         of NULL_STMT_VEC_INFO.
8706         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8707         (vect_reassociating_reduction_p): Likewise.
8708         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8709         (vectorizable_store): Likewise.
8710         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8711         (vec_info::free_stmt_vec_infos): Likewise.
8713 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8715         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8716         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8718 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8720         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8721         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8722         (vec_info::free_stmt_vec_info): New private member functions.
8723         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8724         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8725         * tree-parloops.c (gather_scalar_reductions): Remove calls to
8726         set_stmt_vec_info_vec and free_stmt_vec_infos.
8727         * tree-vect-loop.c (_loop_vec_info): Remove call to
8728         set_stmt_vec_info_vec.
8729         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8730         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8731         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8732         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8733         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
8734         assignments in {vec_info::,}new_stmt_vec_info that are redundant
8735         with the clearing in the xcalloc.
8736         (stmt_vec_info_vec): Delete.
8737         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8738         (vectorize_loops): Likewise.
8739         (vec_info::~vec_info): Remove argument from call to
8740         free_stmt_vec_infos.
8741         (vec_info::add_stmt): Remove vinfo argument from call to
8742         new_stmt_vec_info.
8744 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8746         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8747         rather than a gimple stmt.
8748         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
8749         information for pattern statements when passed the original
8750         statement; instead wait to be passed the pattern statement itself.
8751         Don't call set_vinfo_for_stmt here.
8752         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8753         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8754         stmt_vec_infos here.
8755         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8756         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8757         stmt_vec_infos entry.
8759 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8761         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8762         * tree-vectorizer.c (vec_info::replace_stmt): New function.
8763         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8764         * tree-vect-stmts.c (vectorizable_call): Likewise.
8765         (vectorizable_simd_clone_call): Likewise.
8767 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8769         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8770         * tree-vectorizer.c (vec_info::remove_stmt): New function.
8771         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8772         * tree-vect-loop.c (vect_transform_loop): Likewise.
8773         * tree-vect-slp.c (vect_schedule_slp): Likewise.
8774         * tree-vect-stmts.c (vect_remove_stores): Likewise.
8776 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8778         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8779         (vect_dr_stmt): Delete.
8780         * tree-vectorizer.c (vec_info::lookup_dr): New function.
8781         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8782         of DR_VECT_AUX.
8783         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8784         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8785         (vect_verify_datarefs_alignment, vect_peeling_supportable)
8786         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8787         (vect_analyze_data_refs): Likewise.
8788         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
8789         argument.
8790         (vect_find_same_alignment_drs): Likewise.
8791         (vect_slp_analyze_node_dependences): Update calls accordingly.
8792         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8793         instead of DR_VECT_AUX.
8794         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8795         of a vector data references.  Use vec_info::lookup_dr instead of
8796         DR_VECT_AUX.
8797         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8798         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8799         instead of DR_VECT_AUX.
8801 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8803         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8804         dr_vec_info.
8805         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8806         accordingly.
8807         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8808         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8809         (vect_gen_prolog_loop_niters): Likewise.
8811 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8813         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8814         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8815         (vect_known_alignment_in_bytes, vect_dr_behavior)
8816         (vect_get_scalar_dr_size): Take references as dr_vec_infos
8817         instead of data_references.  Update calls to other routines for
8818         which the same change has been made.
8819         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8820         dr_vec_infos instead of stmt_vec_infos.
8821         (vect_analyze_data_ref_dependence): Update call accordingly.
8822         (vect_slp_analyze_data_ref_dependence)
8823         (vect_record_base_alignments): Use DR_VECT_AUX.
8824         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8825         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8826         (vector_alignment_reachable_p, vect_get_data_access_cost)
8827         (vect_peeling_supportable, vect_analyze_group_access_1)
8828         (vect_analyze_group_access, vect_analyze_data_ref_access)
8829         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8830         (vect_compile_time_alias, vect_small_gap_p)
8831         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8832         (vect_supportable_dr_alignment): Take references as dr_vec_infos
8833         instead of data_references.  Update calls to other routines for
8834         which the same change has been made.
8835         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8836         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8837         (vect_slp_analyze_and_verify_node_alignment)
8838         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8839         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8840         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
8841         above changes.
8842         (_vect_peel_info::dr): Replace with...
8843         (_vect_peel_info::dr_info): ...this new field.
8844         (vect_peeling_hash_get_most_frequent)
8845         (vect_peeling_hash_choose_best_peeling): Update accordingly.
8846         (vect_peeling_hash_get_lowest_cost):
8847         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
8848         routines for which the same change has been made.
8849         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
8850         data_reference.
8851         * tree-vect-loop-manip.c (get_misalign_in_elems)
8852         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
8853         above changes.
8854         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8855         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8856         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8857         (get_group_load_store_type, get_negative_load_store_type)
8858         (vect_get_data_ptr_increment, vectorizable_store)
8859         (vectorizable_load): Likewise.
8860         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8861         Update calls to other routines for which the same change has been made.
8863 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8865         * tree-vectorizer.h (vec_info::move_dr): New member function.
8866         (dataref_aux): Rename to...
8867         (dr_vec_info): ...this and add "dr" and "stmt" fields.
8868         (_stmt_vec_info::dr_aux): Update accordingly.
8869         (_stmt_vec_info::data_ref_info): Delete.
8870         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8871         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8872         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8873         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8874         of data_ref.
8875         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
8876         (STMT_VINFO_DR_INFO): New macro.
8877         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8878         (set_dr_misalignment): Update after rename of dataref_aux.
8879         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
8880         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8881         initialization of STMT_VINFO_DATA_REF.
8882         * tree-vectorizer.c (vec_info::move_dr): New function.
8883         * tree-vect-patterns.c (vect_recog_bool_pattern)
8884         (vect_recog_mask_conversion_pattern)
8885         (vect_recog_gather_scatter_pattern): Use it.
8886         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8887         the "dr" and "stmt" fields of dr_vec_info instead of
8888         STMT_VINFO_DATA_REF.
8890 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8892         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8893         (is_pattern_stmt_p): Use it.
8894         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8895         on pattern statements.
8897 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8899         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8900         original stmt as a stmt_vec_info rather than a gimple stmt.
8901         (vect_pattern_recog_1): Take the statement directly as a
8902         stmt_vec_info, rather than via a gimple_stmt_iterator.
8903         Update call to vect_mark_pattern_stmts.
8904         (vect_pattern_recog): Update calls accordingly.
8906 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8908         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8909         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8910         a vect_def_type for the first argument.
8911         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8912         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
8913         operand if it isn't defined by a vectorized statement.
8914         (vect_build_gather_load_calls): Remove the mask_dt argument and
8915         update calls to vect_get_vec_def_for_stmt_copy.
8916         (vectorizable_bswap): Likewise the dt argument.
8917         (vectorizable_call): Update calls to vectorizable_bswap and
8918         vect_get_vec_def_for_stmt_copy.
8919         (vectorizable_simd_clone_call, vectorizable_assignment)
8920         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8921         (vectorizable_comparison): Update calls to
8922         vect_get_vec_def_for_stmt_copy.
8923         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
8924         vect_is_simple_use.
8925         (vect_get_loop_based_defs): Remove dt argument and update call
8926         to vect_get_vec_def_for_stmt_copy.
8927         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8928         and vect_get_vec_def_for_stmt_copy.
8929         (vectorizable_load): Update calls to vect_build_gather_load_calls
8930         and vect_get_vec_def_for_stmt_copy.
8931         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8932         (vectorizable_reduction, vectorizable_live_operation): Update calls
8933         to vect_get_vec_def_for_stmt_copy.
8935 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8937         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8938         and gimple stmt arguments with a stmt_vec_info.
8939         (vect_record_base_alignments): Update calls accordingly.
8940         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8941         and gimple stmt arguments with a stmt_vec_info.
8942         (vect_build_slp_tree_1): Remove vinfo argument and update call
8943         to vect_record_max_nunits.
8944         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8945         and vect_record_max_nunits.
8947 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8949         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8950         file and take a stmt_vec_info instead of a gimple stmt.
8951         (supportable_widening_operation, vect_finish_replace_stmt)
8952         (vect_finish_stmt_generation, vect_get_store_rhs)
8953         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8954         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8955         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8956         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8957         (vect_create_data_ref_ptr, bump_vector_ptr)
8958         (vect_permute_store_chain, vect_setup_realignment)
8959         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8960         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8961         (vectorizable_reduction, vectorizable_induction)
8962         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8963         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8964         than gimple stmts as arguments.
8965         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8966         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8967         (can_group_stmts_p, vect_check_gather_scatter)
8968         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8969         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8970         (vect_permute_load_chain, vect_shift_permute_load_chain)
8971         (vect_transform_grouped_load)
8972         (vect_record_grouped_load_vectors): Likewise.
8973         * tree-vect-loop.c (vect_fixup_reduc_chain)
8974         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8975         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8976         (vectorizable_reduction, vectorizable_induction)
8977         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8978         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8979         (vect_get_load_store_mask): Likewise.
8980         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8981         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8982         * tree-vect-stmts.c (vect_mark_relevant)
8983         (is_simple_and_all_uses_invariant)
8984         (exist_non_indexing_operands_for_use_p, process_use)
8985         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8986         (vect_get_vec_def_for_operand, vect_get_vec_defs)
8987         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8988         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8989         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8990         (get_negative_load_store_type, get_load_store_type)
8991         (vect_check_load_store_mask, vect_check_store_rhs)
8992         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8993         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8994         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8995         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8996         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8997         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8998         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8999         (supportable_widening_operation): Likewise.
9001 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9003         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
9004         a stmt_vec_info instead of a gcall.
9005         (vect_check_gather_scatter): Update call accordingly.
9006         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
9007         of a gphi.
9008         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
9009         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
9010         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
9011         instead of a gimple stmt.
9012         (vect_transform_loop): Update calls accordingly.
9013         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
9014         stmt_vec_infos instead of gimple stmts.
9015         (vect_analyze_slp_instance): Update use accordingly.
9016         * tree-vect-stmts.c (read_vector_array, write_vector_array)
9017         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
9018         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
9019         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
9020         (vect_gen_widened_results_half, vect_get_loop_based_defs)
9021         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
9022         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
9023         down to subroutines.
9025 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9027         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
9028         of the worklist from a vector of gimple stmts to a vector of
9029         stmt_vec_infos.
9030         * tree-vect-stmts.c (vect_mark_relevant, process_use)
9031         (vect_mark_stmts_to_be_vectorized): Likewise
9033 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9035         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
9036         statement before passing it to vect_analyze_stmt.
9037         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
9038         the chain of phi vector definitions.  Track the exit phi via its
9039         stmt_vec_info.
9040         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
9041         STMT_VINFO_REDUC_DEF.
9042         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
9043         stmt_vec_infos to handle the statement chains.
9044         (vect_get_slp_defs): Record the first statement in the node
9045         using a stmt_vec_info.
9046         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
9047         statements here and pass their stmt_vec_info down to subroutines.
9048         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
9049         down to vect_finish_stmt_generation.
9050         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
9051         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
9052         stmt_vec_infos to subroutines.
9053         (vect_remove_stores): Use stmt_vec_infos to handle the statement
9054         chains.
9056 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9058         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
9059         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
9060         (vect_permute_store_chain, vect_setup_realignment)
9061         (vect_permute_load_chain, vect_shift_permute_load_chain)
9062         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
9063         stmts internally, and when passing values to other vectorizer routines.
9064         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
9065         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
9066         (vect_analyze_loop_operations, get_initial_def_for_reduction)
9067         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
9068         (vectorizable_reduction, vectorizable_induction)
9069         (vectorizable_live_operation, vect_transform_loop_stmt)
9070         (vect_transform_loop): Likewise.
9071         * tree-vect-patterns.c (vect_reassociating_reduction_p)
9072         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
9073         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
9074         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9075         (vect_slp_analyze_node_operations_1): Likewise.
9076         * tree-vect-stmts.c (vect_mark_relevant, process_use)
9077         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
9078         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
9079         (vect_finish_stmt_generation_1, get_group_load_store_type)
9080         (get_load_store_type, vect_build_gather_load_calls)
9081         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9082         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9083         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9084         (vectorizable_store, vectorizable_load, vectorizable_condition)
9085         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
9086         (supportable_widening_operation): Likewise.
9087         (vect_get_vector_types_for_stmt): Likewise.
9088         * tree-vectorizer.h (vect_dr_behavior): Likewise.
9090 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9092         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9093         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
9094         (vect_permute_store_chain, vect_permute_load_chain)
9095         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
9096         repeated stmt_vec_info lookups.
9097         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
9098         (vect_update_ivs_after_vectorizer): Likewise.
9099         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
9100         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9101         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
9102         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9103         (vect_bb_slp_scalar_cost): Likewise.
9104         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
9106 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9108         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
9109         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
9110         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
9111         to get gassigns and gcalls, rather than operating on generc gimple
9112         stmts.
9113         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
9114         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
9115         (vectorizable_load, vect_analyze_stmt): Likewise.
9116         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
9118 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9120         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
9121         return stmt_vec_infos rather than gimple stmts.  Do not accept
9122         null arguments.
9123         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
9124         of a gimple stmt.
9125         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9126         Update use of get_later_stmt.
9127         (vect_get_constant_vectors): Update call accordingly.
9128         (vect_schedule_slp_instance): Likewise
9129         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
9130         (vect_slp_analyze_instance_dependence): Likewise.
9131         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
9133 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9135         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
9136         (stmt_info_for_cost::stmt_info): ...this new field.
9137         (add_stmt_costs): Update accordingly.
9138         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9139         (vect_get_known_peeling_cost): Likewise.
9140         (vect_estimate_min_profitable_iters): Likewise.
9141         * tree-vect-stmts.c (record_stmt_cost): Likewise.
9143 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9145         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
9146         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9147         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
9148         accordingly.
9149         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
9151 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9153         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
9154         a gimple stmt to a stmt_vec_info.
9155         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
9157 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9159         * tree-vectorizer.h (vec_info::grouped_stores): Change from
9160         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9161         (_loop_vec_info::reduction_chains): Likewise.
9162         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
9163         accordingly.
9164         * tree-vect-slp.c (vect_analyze_slp): Likewise.
9166 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9168         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
9169         a gimple stmt to a stmt_vec_info.
9170         (_stmt_vec_info::next_element): Likewise.
9171         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
9172         (vect_slp_analyze_and_verify_node_alignment)
9173         (vect_analyze_group_access_1, vect_analyze_group_access)
9174         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
9175         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
9176         (vect_supportable_dr_alignment): Update accordingly.
9177         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
9178         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
9179         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
9180         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9181         * tree-vect-slp.c (vect_build_slp_tree_1)
9182         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
9183         (vect_split_slp_store_group, vect_analyze_slp_instance)
9184         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
9185         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
9186         (get_group_load_store_type, get_load_store_type)
9187         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
9188         (vect_transform_stmt, vect_remove_stores): Likewise.
9190 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9192         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
9193         than a gimple stmt.
9194         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9195         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
9196         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
9197         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
9198         (vector_alignment_reachable_p, vect_get_data_access_cost)
9199         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
9200         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
9201         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
9202         (vect_analyze_group_access_1, vect_analyze_group_access)
9203         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
9204         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
9205         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
9206         result of vect_dr_stmt and use the stmt_vec_info instead of
9207         the associated gimple stmt.
9208         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9209         (vect_gen_prolog_loop_niters): Likewise.
9210         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9212 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9214         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
9215         to a vec<stmt_vec_info>.
9216         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
9217         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
9218         vec<stmt_vec_info>.
9219         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
9220         to a vec<stmt_vec_info>.
9221         (bst_traits::value_type, bst_traits::value_type): Likewise.
9222         (bst_traits::hash): Update accordingly.
9223         (vect_get_and_check_slp_defs): Change the stmts parameter from
9224         a vec<gimple *> to a vec<stmt_vec_info>.
9225         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
9226         (vect_build_slp_tree): Likewise.
9227         (vect_build_slp_tree_2): Likewise.  Update uses of
9228         SLP_TREE_SCALAR_STMTS.
9229         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
9230         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
9231         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
9232         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
9233         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
9234         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
9235         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
9236         (vect_get_constant_vectors, vect_get_slp_defs)
9237         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
9238         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
9239         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
9240         instead of gimple stmts.
9241         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
9242         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
9243         (vect_slp_analyze_instance_dependence): Update uses of
9244         SLP_TREE_SCALAR_STMTS.
9245         (vect_slp_analyze_and_verify_node_alignment): Likewise.
9246         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
9247         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
9248         (get_initial_defs_for_reduction): Likewise.
9249         (vect_create_epilog_for_reduction): Likewise.
9250         (vectorize_fold_left_reduction): Likewise.
9251         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
9252         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
9253         (can_vectorize_live_stmts): Likewise.
9255 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9257         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
9258         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9259         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
9260         than gimple stmts.
9261         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
9262         of a gimple stmt.
9263         (gather_scalar_reductions): Update after above interface changes.
9264         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
9265         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
9266         than gimple stmts.
9267         (vect_force_simple_reduction): Likewise.
9268         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
9269         LOOP_VINFO_REDUCTIONS.
9270         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9272 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9274         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
9275         a gimple stmt to a stmt_vec_info.
9276         * tree-vect-loop.c (vect_active_double_reduction_p)
9277         (vect_force_simple_reduction, vectorizable_reduction): Update
9278         accordingly.
9280 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9282         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
9283         vec<gimple *> to a vec<stmt_vec_info>.
9284         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
9285         the reduction_phis argument from a vec<gimple *> to a
9286         vec<stmt_vec_info>.
9287         (vectorizable_reduction): Likewise the phis local variable that
9288         is passed to vect_create_epilog_for_reduction.  Update for new type
9289         of SLP_TREE_VEC_STMTS.
9290         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
9291         (vectorizable_live_operation): Likewise.
9292         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
9293         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
9295 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9297         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
9298         a gimple stmt to a stmt_vec_info.
9299         (vectorizable_condition, vectorizable_live_operation)
9300         (vectorizable_reduction, vectorizable_induction): Pass back the
9301         vectorized statement as a stmt_vec_info.
9302         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
9303         use of STMT_VINFO_VEC_STMT.
9304         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
9305         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
9306         as stmt_vec_infos rather than gimple stmts.
9307         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
9308         to a stmt_vec_info.
9309         (vectorizable_live_operation): Likewise.
9310         (vectorizable_reduction, vectorizable_induction): Likewise,
9311         updating use of STMT_VINFO_VEC_STMT.
9312         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
9313         of STMT_VINFO_VEC_STMT.
9314         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
9315         (vectorizable_simd_clone_call, vectorizable_conversion)
9316         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9317         (vectorizable_store, vectorizable_load, vectorizable_condition)
9318         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
9319         from a gimple stmt to a stmt_vec_info.
9320         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
9321         pointer to a stmt_vec_info to the vectorizable_* routines.
9323 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9325         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
9326         a gimple stmt to a stmt_vec_info.
9327         (is_pattern_stmt_p): Update accordingly.
9328         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
9329         (vect_record_grouped_load_vectors): Likewise.
9330         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
9331         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
9332         (vect_model_reduction_cost): Likewise.
9333         (vect_create_epilog_for_reduction): Likewise.
9334         (vectorizable_reduction, vectorizable_induction): Likewise.
9335         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9336         Return the stmt_vec_info for the pattern statement.
9337         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
9338         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
9339         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9340         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
9341         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
9342         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
9343         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
9344         (free_stmt_vec_info, vect_is_simple_use): Likewise.
9346 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9348         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
9349         (vect_finish_stmt_generation): Likewise.
9350         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9351         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
9352         (vect_build_gather_load_calls): Use the return value of the above
9353         functions instead of a separate call to vinfo_for_stmt.  Use narrow
9354         scopes for the input gimple stmt and wider scopes for the associated
9355         stmt_vec_info.  Use vec_info::lookup_def when setting these
9356         stmt_vec_infos from an SSA_NAME definition.
9357         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9358         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9359         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9360         (vectorizable_store, vectorizable_load, vectorizable_condition)
9361         (vectorizable_comparison): Likewise.
9362         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
9363         (vectorizable_reduction): Likewise.
9365 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9367         * tree-vectorizer.h (vect_is_simple_use): Add an optional
9368         stmt_vec_info * parameter before the optional gimple **.
9369         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9370         (process_use, vect_get_vec_def_for_operand_1): Update callers.
9371         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
9372         * tree-vect-loop.c (vectorizable_reduction): Likewise.
9373         (vectorizable_live_operation): Likewise.
9374         * tree-vect-patterns.c (type_conversion_p): Likewise.
9375         (vect_look_through_possible_promotion): Likewise.
9376         (vect_recog_rotate_pattern): Likewise.
9377         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9379 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9381         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
9382         a typedef to a wrapper class.
9383         (NULL_STMT_VEC_INFO): New macro.
9384         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
9385         (stmt_vec_info::operator*): New function.
9386         (stmt_vec_info::operator gimple *): Likewise.
9387         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
9388         (add_stmt_costs): Likewise.
9389         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
9390         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9391         (vect_get_known_peeling_cost): Likewise.
9392         (vect_estimate_min_profitable_iters): Likewise.
9393         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9394         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
9395         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9396         (vectorizable_store, free_stmt_vec_infos): Likewise.
9397         (new_stmt_vec_info): Change return type of xcalloc to
9398         _stmt_vec_info *.
9400 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9402         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
9403         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
9404         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
9405         a single_imm_use-based sequence.
9406         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
9408 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9410         * tree-vectorizer.h (vec_info::lookup_def): Declare.
9411         * tree-vectorizer.c (vec_info::lookup_def): New function.
9412         * tree-vect-patterns.c (vect_get_internal_def): Use it.
9413         (vect_widened_op_tree): Likewise.
9414         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9415         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
9416         (vectorizable_reduction): Likewise.
9417         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
9418         of a gimple *.
9419         (vect_is_slp_reduction): Update calls accordingly.  Use
9420         vec_info::lookup_def.
9421         (vect_is_simple_reduction): Likewise
9422         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
9424 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9426         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
9427         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
9428         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
9429         of vinfo_for_stmt.
9430         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
9431         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
9432         (vect_update_vf_for_slp, vect_analyze_loop_operations)
9433         (vect_is_slp_reduction, vectorizable_induction)
9434         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
9435         * tree-vect-patterns.c (vect_init_pattern_stmt):
9436         (vect_determine_min_output_precision_1, vect_determine_precisions)
9437         (vect_pattern_recog): Likewise.
9438         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
9439         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
9440         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
9441         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9442         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
9443         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
9444         info field from a loop to a loop_vec_info.
9446 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9448         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
9449         (vec_info::add_stmt): Declare.
9450         * tree-vectorizer.c (vec_info::add_stmt): New function.
9451         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
9452         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
9453         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9454         (vectorizable_induction): Likewise.
9455         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
9456         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9457         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
9458         (vectorizable_load): Likewise.
9459         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9460         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
9461         (vect_recog_gather_scatter_pattern): Likewise.
9462         (append_pattern_def_seq): Likewise.  Remove a check that is
9463         performed by add_stmt itself.
9465 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9467         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
9468         which make_ssa_name was called with new_stmt before new_stmt
9469         had been created.
9471 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9473         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
9474         split out from...
9475         (vect_is_slp_reduction): ...here...
9476         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
9477         that are already known to be false.
9479 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9481         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
9482         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
9483         STMT_VINFO_NUM_SLP_USES when it's true.
9484         (vect_free_slp_instance): Add a final_p parameter and pass it to
9485         vect_free_slp_tree.
9486         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
9487         (vect_analyze_slp_instance): Likewise.
9488         (vect_slp_analyze_operations): Likewise.
9489         (vect_slp_analyze_bb_1): Likewise.
9490         * tree-vectorizer.c (vec_info): Likewise.
9491         * tree-vect-loop.c (vect_transform_loop): Likewise.
9493 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9495         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9496         function is not called for second and subsequent members of
9497         a reduction group.
9499 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9501         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
9502         cases for nested loops from here to ...
9503         (vect_create_epilog_for_reduction): ...here.  Only call
9504         vect_is_simple_use for inner-loop reductions.
9506 2018-07-31  Martin Liska  <mliska@suse.cz>
9508         PR gcov-profile/85338
9509         PR gcov-profile/85350
9510         PR gcov-profile/85372
9511         * profile.c (struct location_triplet): New.
9512         (struct location_triplet_hash): Likewise.
9513         (output_location): Do not output a BB that
9514         is already recorded for a line.
9515         (branch_prob): Use streamed_locations.
9517 2018-07-31  Martin Liska  <mliska@suse.cz>
9519         PR gcov-profile/85370
9520         * coverage.c (coverage_begin_function): Do not mark target
9521         clones as artificial functions.
9523 2018-07-31  Martin Liska  <mliska@suse.cz>
9525         PR gcov-profile/83813
9526         PR gcov-profile/84758
9527         PR gcov-profile/85217
9528         PR gcov-profile/85332
9529         * profile.c (branch_prob): Do not record GOTO expressions
9530         for GIMPLE statements which locations are already streamed.
9532 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9534         * gcc.c (handle_spec_function): Accept a soft_matched_part
9535         argument, as do_spec_1.  Pass it down to ...
9536         (eval_spec_function): Accept a soft_matched_part argument,
9537         and pass it down to ...
9538         (do_spec_2): Accept a soft_matched_part argument, and pass
9539         it down to do_spec_1.
9540         (do_spec_1): Pass soft_matched_part to handle_spec_function.
9541         (handle_braces): Update call to handle_spec_function.
9542         (driver::set_up_specs): Update calls to do_spec_2.
9543         (compare_debug_dump_opt_spec_function): Likewise.
9544         (compare_debug_self_opt_spec_function): Likewise.
9546 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9548         * common.opt (nolibc): New option.
9549         * doc/invoke.texi (Link Options): Document it.
9550         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
9551         * config/alpha/linux.h: Likewise.
9552         * config/arc/elf.h: Likewise.
9553         * config/arm/uclinux-elf.h: Likewise.
9554         * config/arm/unknown-elf.h: Likewise.
9555         * config/avr/avrlibc.h: Likewise.
9556         * config/bfin/bfin.h: Likewise.
9557         * config/bfin/linux.h: Likewise.
9558         * config/bfin/uclinux.h: Likewise.
9559         * config/darwin.h: Likewise.
9560         * config/darwin10.h: Likewise.
9561         * config/darwin12.h: Likewise.
9562         * config/gnu-user.h: Likewise.
9563         * config/lm32/uclinux-elf.h: Likewise.
9564         * config/pa/pa-hpux11.h: Likewise.
9565         * config/pa/pa64-hpux.h: Likewise.
9566         * config/sparc/sparc.h: Likewise.
9568 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9570         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
9571         undefined variables.
9573 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9575         PR target/86640
9576         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
9577         instead of GEN_INT.
9579 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9581         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9582         terminated string literal.
9584 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9586         PR rtl-optimization/85160
9587         * combine.c (is_just_move): New function.
9588         (try_combine): Allow combining two instructions into two if neither of
9589         the original instructions was a move.
9591 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
9593         PR target/86673
9594         * doc/extend.texi (Global Register Variables): Discourage use of type
9595         qualifiers.
9596         (Local Register Variables): Likewise.
9598 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
9600         PR tree-optimization/86506
9601         * hwint.h (ceil_log2): Resync with hwint.c implementation.
9603 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
9605         PR target/86547
9606         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
9607         hard_regno, make sure no insn between `from` and `to` clobbers it.
9609 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
9610             Tom de Vries  <tdevries@suse.de>
9612         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
9613         (PTX_DEFAULT_RUNTIME_DIM): ... this.
9614         (nvptx_goacc_validate_dims): Set default worker and gang dims to
9615         PTX_DEFAULT_RUNTIME_DIM.
9616         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
9618 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
9620         * config/pa/pa.c (pa_output_addr_vec): Align address table.
9621         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
9622         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9624 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
9626         * config/rs6000/constraints.md (wG constraint): Delete, no longer
9627         used.
9628         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9629         predicate to reflect toc fusion has been deleted.
9630         (toc_fusion_mem_raw): Delete, no longer used.
9631         (toc_fusion_mem_wrapped): Likewise.
9632         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9633         fusion mask bit.
9634         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9635         Delete, no longer used.
9636         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9637         meant to be used for toc fusion.
9638         (rs6000_debug_print_mode): Delete toc fusion debugging.
9639         (rs6000_debug_reg_global): Likewise.
9640         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9641         fusion and secondary reload support that were never used.
9642         (rs6000_option_override_internal): Delete TOC fusion, that was only
9643         partially defined, and it did not work unless you also used the
9644         -mcmodel= switch.
9645         (rs6000_legitimate_address_p): Delete TOC fusion support.
9646         (rs6000_opt_masks): Likewise.
9647         (fusion_wrap_memory_address): Delete function, no longer used.
9648         (fusion_split_address); Delete TOC fusion support.
9649         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9650         longer used with toc fusion being deleted.
9651         (TARGET_TOC_FUSION_FP): Likewise.
9652         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9653         UNSPEC.
9654         (toc fusion spliter): Delete TOC fusion support.
9655         (toc_fusionload_<mode>): Likewise.
9656         (toc_fusionload_di): Likewise.
9657         (fusion_gpr_load_<mode>): Delete generator function, this insn no
9658         longer needs to be named.  Rename predicate to delete TOC fusion.
9659         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9660         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9661         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9662         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9663         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9665 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9667         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9668         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
9669         __int128 in built-in function prototypes.
9670         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9671         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9673 2018-07-27  Martin Sebor  <msebor@redhat.com>
9675         PR tree-optimization/86696
9676         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9677         types, including enums.
9678         (handle_char_store): Be prepared for the above function to fail.
9680 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
9682         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9683         when optimization level is lower than 2 or optimize for size.
9685 2018-07-26  Martin Sebor  <msebor@redhat.com>
9687         PR tree-optimization/86043
9688         PR tree-optimization/86042
9689         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9690         (get_string_cst_length): Rename...
9691         (get_min_string_length): ...to this.  Add argument.
9692         (handle_char_store): Extend to handle multi-character stores by
9693         MEM_REF.
9694         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
9695         * tree.h (initializer_zerop): Add argument.
9697 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
9699         PR middle-end/86660
9700         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9701         declare target to variables if they have always,{to,from,tofrom} map
9702         kinds.
9704 2018-07-26  Martin Liska  <mliska@suse.cz>
9706         PR lto/86548
9707         * lto-wrapper.c: Add linker_output as prefix
9708         for ltrans_output_file.
9710 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
9712         PR rtl-optimization/85805
9713         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9714         value for hard registers if that was written in the same mode.
9716 2018-07-26  Martin Liska  <mliska@suse.cz>
9718         PR gcov-profile/86536
9719         * gcov.c (format_gcov): Use printf format %.*f directly
9720         and do not handle special values.
9722 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9724         * common/config/arc/arc-common.c (arc_option_optimization_table):
9725         Update default optimizations for size.
9727 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9729         * config/arc/arc.md (movsf_insn): Add short instruction selection.
9730         * config/arc/constraints.md (CfZ): New constraint.
9731         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9732         (subsf3_fpu): Likewise.
9733         (cmpsf_fpu): Likewise.
9734         (cmpsf_fpu_uneq): Likewise.
9736 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9738         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9739         accesses as well.
9740         (arc_is_uncached_mem_p): uncached applies to both the variable and
9741         the pointer.
9743 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9745         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9746         register names.
9748 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
9750         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9751         field "m_scopes" from vec to auto_vec.
9753 2018-07-25  Martin Liska  <mliska@suse.cz>
9755         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9756         return type.
9758 2018-07-25  Richard Biener  <rguenther@suse.de>
9760         PR debug/86654
9761         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9762         special wrt context_die late.
9763         (gen_subprogram_die): Re-use DIEs in local scope.
9765 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
9767         PR tree-optimization/86644
9768         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
9770 2018-07-25  Martin Liska  <mliska@suse.cz>
9772         PR middle-end/86645
9773         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9774         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9776 2018-07-25  Martin Liska  <mliska@suse.cz>
9778         PR sanitizer/79635
9779         * params.def: Explain ASan abbreviation and provide
9780         a documentation link.
9782 2018-07-24  Martin Sebor  <msebor@redhat.com>
9784         PR tree-optimization/86622
9785         PR tree-optimization/86532
9786         * builtins.h (string_length): Declare.
9787         * builtins.c (c_strlen): Correct handling of non-constant offsets.
9788         (check_access): Be prepared for non-constant length ranges.
9789         (string_length): Make extern.
9790         * expr.c (string_constant): Only handle the minor non-constant
9791         array index.  Use string_constant to compute the length of
9792         a generic string constant.
9794 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
9796         PR tree-optimization/86618
9797         * tree-vect-stmts.c (vectorizable_call): Don't take the address
9798         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9800 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
9802         PR tree-optimization/86636
9803         * json.cc (json::object::set): Fix comment.  Add assertions.
9804         (json::array::append): Move here from json.h.  Add comment and an
9805         assertion.
9806         (json::string::string): Likewise.
9807         * json.h (json::array::append): Move to json.cc.
9808         (json::string::string): Likewise.
9809         * optinfo-emit-json.cc
9810         (optrecord_json_writer::impl_location_to_json): Assert that we
9811         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9812         wrapper around it.  Expand the location once, rather than three
9813         times.
9814         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9815         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9816         wrappers.
9817         (optrecord_json_writer::optinfo_to_json): Likewise, in four
9818         places.  Fix some overlong lines.
9820 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
9822         * config/aarch64/aarch64-simd.md
9823         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9824         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9825         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9826         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9827         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9828         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9829         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9830         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9831         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9833 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
9835         PR middle-end/86627
9836         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9837         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
9838         and abs_d == d, do the power of two handling if profitable.
9840 2018-07-24  Richard Biener  <rguenther@suse.de>
9842         * match.pd: Add BIT_FIELD_REF canonicalizations.
9844 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9846         PR c/86617
9847         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9849 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9851         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9852         terminated STRING_CST object.
9854 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9856         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9858 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9860         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9861         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9862         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9863         the elements into a register.
9864         (rs6000_split_v4si_init_di_reg): Delete.
9865         (rs6000_split_v4si_init): Delete.
9866         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9867         (vsx_init_v4si): Rewrite as a define_expand.
9869 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9871         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9872         zero_extend argument from memory): New.
9874 2018-07-22  Martin Sebor  <msebor@redhat.com>
9876         PR bootstrap/86621
9877         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9878         diagnosing calls with unknown arguments unless -Walloca-larger-than
9879         is restricted to less than PTRDIFF_MAX bytes.
9881 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
9883         * doc/gcov.texi (Invoking Gcov): Editorial changes.
9885 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9887         * pretty-print.c (text_info::set_location): Remove redundant
9888         "line_table" parameter from call to rich_location::set_range.
9890 2018-07-20  Martin Sebor  <msebor@redhat.com>
9892         PR middle-end/82063
9893         * builtins.c (expand_builtin_alloca): Adjust.
9894         * calls.c (alloc_max_size): Simplify.
9895         * cgraphunit.c (cgraph_node::expand): Adjust.
9896         * common.opt (larger_than_size, warn_frame_larger_than): Remove
9897         variables.
9898         (frame_larger_than_size): Same.
9899         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9900         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9901         Initialize.
9902         * doc/invoke.texi (GCC Command Options): Document option arguments.
9903         Explain byte-size arguments and suffixes.
9904         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9905         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9906         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9907         * doc/options.texi (UInteger): Expand.
9908         (Host_Wide_Int, ByteSize): Document new properties.
9909         * final.c (final_start_function_1): Include sizes in an error message.
9910         * function.c (frame_offset_overflow): Same.
9911         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9912         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9913         Diagnose unbounded alloca calls only for limits of less than
9914         PTRDIFF_MAX.
9915         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
9916         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
9917         for alloca(0).
9918         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
9919         only for limits of less than PTRDIFF_MAX.
9920         * langhooks-def.h (lhd_handle_option): Change function argument
9921         to HOST_WIDE_INT.
9922         * langhooks.c (lhd_handle_option): Same.
9923         * langhooks.h (handle_option): Same.
9924         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9925         ByteSize flags.
9926         (var_type, var_type_struct): Same.
9927         (var_set): Handle ByteSize flag.
9928         * optc-gen.awk: Add comments to output to ease debugging.  Make
9929         use of HOST_WIDE_INT where appropriate.
9930         * opts-gen-save.awk:  Use %lx to format unsigned long.
9931         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9932         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9933         arguments.  Parse bytes-size suffixes.
9934         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9935         (enum_value_to_arg): Same.
9936         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
9937         (handle_option): Adjust.
9938         (generate_option): Change function argument to HOST_WIDE_INT.
9939         (cmdline_handle_error): Adjust.
9940         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9941         (set_option): Change function argument to HOST_WIDE_INT.
9942         (option_enabled): Handle cl_host_wide_int.
9943         (get_option_state): Handle CLVC_SIZE.
9944         (control_warning_option): Same.
9945         * opts.c (common_handle_option): Change function argument to
9946         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
9947         OPT_Wvla_larger_than_.
9948         * opts.h (enum cl_var_type): Add an enumerator.
9949         * stor-layout.c (layout_decl): Print a more meaningful warning.
9950         * toplev.c (output_stack_usage): Adjust.
9952 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
9954         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9955         call to inline_expand_builtin_string_cmp.
9956         (expand_builtin_strcmp): Likewise.
9957         (expand_builtin_strncmp): Likewise.
9958         (inline_string_cmp): Delete the last parameter, change char_type_node
9959         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9960         two operands.
9961         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9962         the inlining expansion on target where the type of the call has same or
9963         narrower precision than unsigned char.
9965 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9967         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9968         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9969         * common.opt (fsave-optimization-record): New option.
9970         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9971         * doc/invoke.texi (-fsave-optimization-record): New option.
9972         * dumpfile.c: Include "optinfo-emit-json.h".
9973         (struct kv_pair): Move to coretypes.h.
9974         (optgroup_options): Make non-static.
9975         (dump_context::end_scope): Call
9976         optimization_records_maybe_pop_dump_scope.
9977         * dumpfile.h (optgroup_options): New decl.
9978         * json.cc: New file.
9979         * json.h: New file.
9980         * optinfo-emit-json.cc: New file.
9981         * optinfo-emit-json.h: New file.
9982         * optinfo.cc: Include "optinfo-emit-json.h".
9983         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9984         (optinfo_enabled_p): Check optimization_records_enabled_p.
9985         (optinfo_wants_inlining_info_p): Likewise.
9986         * optinfo.h: Update comment.
9987         * profile-count.c (profile_quality_as_string): New function.
9988         * profile-count.h (profile_quality_as_string): New decl.
9989         (profile_count::quality): New accessor.
9990         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9991         and optinfo_emit_json_cc_tests.
9992         * selftest.h (selftest::json_cc_tests): New decl.
9993         (selftest::optinfo_emit_json_cc_tests): New decl.
9994         * toplev.c: Include "optinfo-emit-json.h".
9995         (compile_file): Call optimization_records_finish.
9996         (do_compile): Call optimization_records_start.
9997         * tree-ssa-live.c: Include optinfo.h.
9998         (remove_unused_scope_block_p): Retain inlining information if
9999         optinfo_wants_inlining_info_p returns true.
10001 2018-07-20  Richard Biener  <rguenther@suse.de>
10003         PR debug/86585
10004         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
10005         to cover -flto-partition=none.
10007 2018-07-20  Martin Liska  <mliska@suse.cz>
10009         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
10010         (get_decl_source_range): Remove unused function.
10012 2018-07-20  Richard Biener  <rguenther@suse.de>
10014         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
10015         (struct vn_phi_s): Likewise.
10016         (struct vn_reference_s): Likewise.
10017         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
10018         for searching the slot of an entry known to be in the hash itself.
10019         (vn_phi_hasher::equal): Likewise.
10020         (vn_reference_hasher::equal): Likewise.
10021         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
10022         globals.
10023         (optimistic_info, current_info): Remove, keeping only valid_info.
10024         (vn_reference_lookup_1): Remove fallback lookup.
10025         (vn_reference_lookup_2): Likewise.
10026         (vn_nary_op_lookup_1): Likewise.
10027         (vn_phi_lookup): Likewise.
10028         (vn_nary_build_or_lookup_1): Make sure to not chain the built
10029         hash element.
10030         (vn_reference_insert): Adjust, chain the inserted hash element
10031         at last_inserted_ref.
10032         (vn_reference_insert_pieces): Likewise.
10033         (visit_reference_op_call): Likewise.
10034         (vn_nary_op_insert_into): Chain the inserted hash element at
10035         last_inserted_nary.
10036         (vn_nary_op_insert_pieces): Adjust.
10037         (vn_nary_op_insert): Likewise.
10038         (vn_nary_op_insert_stmt): Likewise.
10039         (vn_phi_insert): Adjust, chain the inserted hash element at
10040         last_inserted_phi.
10041         (process_scc): Remove clearing and copying the optimistic
10042         table.  Instead remove elements inserted during an optimistic
10043         iteration from the single table we maintain.
10044         (init_scc_vn): Adjust.
10045         (free_scc_vn): Likewise.
10046         (sccvn_dom_walker::record_cond): Likewise.
10047         (sccvn_dom_walker::after_dom_children): Likewise.
10049 2018-07-19  Martin Sebor  <msebor@redhat.com>
10051         PR tree-optimization/84047
10052         PR tree-optimization/83776
10053         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
10054         (check_array_bounds): Call it.
10056 2018-07-19  Martin Sebor  <msebor@redhat.com>
10058         * align.h (align_flags): Use member initialization.
10060 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
10062         * Makefile.in (OBJS): Add optinfo.o.
10063         * coretypes.h (class symtab_node): New forward decl.
10064         (struct cgraph_node): New forward decl.
10065         (class varpool_node): New forward decl.
10066         * dump-context.h: New file.
10067         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
10068         "tree-pass.h".
10069         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
10070         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
10071         (set_alt_dump_file): Likewise.
10072         (dump_context::~dump_context): New dtor.
10073         (dump_gimple_stmt): Move implementation to...
10074         (dump_context::dump_gimple_stmt): ...this new member function.
10075         Add the stmt to any pending optinfo, creating one if need be.
10076         (dump_gimple_stmt_loc): Move implementation to...
10077         (dump_context::dump_gimple_stmt_loc): ...this new member function.
10078         Start a new optinfo and add the stmt to it.
10079         (dump_gimple_expr): Move implementation to...
10080         (dump_context::dump_gimple_expr): ...this new member function.
10081         Add the stmt to any pending optinfo, creating one if need be.
10082         (dump_gimple_expr_loc): Move implementation to...
10083         (dump_context::dump_gimple_expr_loc): ...this new member function.
10084         Start a new optinfo and add the stmt to it.
10085         (dump_generic_expr): Move implementation to...
10086         (dump_context::dump_generic_expr): ...this new member function.
10087         Add the tree to any pending optinfo, creating one if need be.
10088         (dump_generic_expr_loc): Move implementation to...
10089         (dump_context::dump_generic_expr_loc): ...this new member
10090         function.  Add the tree to any pending optinfo, creating one if
10091         need be.
10092         (dump_printf): Move implementation to...
10093         (dump_context::dump_printf_va): ...this new member function.  Add
10094         the text to any pending optinfo, creating one if need be.
10095         (dump_printf_loc): Move implementation to...
10096         (dump_context::dump_printf_loc_va): ...this new member function.
10097         Start a new optinfo and add the stmt to it.
10098         (dump_dec): Move implementation to...
10099         (dump_context::dump_dec): ...this new member function.  Add the
10100         value to any pending optinfo, creating one if need be.
10101         (dump_context::dump_symtab_node): New member function.
10102         (dump_context::get_scope_depth): New member function.
10103         (dump_context::begin_scope): New member function.
10104         (dump_context::end_scope): New member function.
10105         (dump_context::ensure_pending_optinfo): New member function.
10106         (dump_context::begin_next_optinfo): New member function.
10107         (dump_context::end_any_optinfo): New member function.
10108         (dump_context::s_current): New global.
10109         (dump_context::s_default): New global.
10110         (dump_scope_depth): Delete global.
10111         (dumpfile_ensure_any_optinfo_are_flushed): New function.
10112         (dump_symtab_node): New function.
10113         (get_dump_scope_depth): Reimplement in terms of dump_context.
10114         (dump_begin_scope): Likewise.
10115         (dump_end_scope): Likewise.
10116         (selftest::temp_dump_context::temp_dump_context): New ctor.
10117         (selftest::temp_dump_context::~temp_dump_context): New dtor.
10118         (selftest::verify_item): New function.
10119         (ASSERT_IS_TEXT): New macro.
10120         (ASSERT_IS_TREE): New macro.
10121         (ASSERT_IS_GIMPLE): New macro.
10122         (selftest::test_capture_of_dump_calls): New test.
10123         (selftest::dumpfile_c_tests): Call it.
10124         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
10125         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
10126         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
10127         descriptive comment.
10128         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
10129         (dump_node, dump_bb): Move these unrelated decls.
10130         (class dump_manager): Add leading comment.
10131         * optinfo.cc: New file.
10132         * optinfo.h: New file.
10134 2018-07-19  Michael Collison  <michael.collison@arm.com>
10135             Richard Henderson <rth@redhat.com>
10137         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
10138         (subti): Handle op1 zero.
10139         (subvti4, usub4ti4): New.
10140         (*sub<GPI>3_compare1_imm): New.
10141         (sub<GPI>3_carryinCV): New.
10142         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
10143         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
10145 2018-07-19  Michael Collison  <michael.collison@arm.com>
10146             Richard Henderson <rth@redhat.com>
10148         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
10149         (addti3): Create simpler code if low part is already known to be 0.
10150         (addvti4, uaddvti4): New.
10151         (*add<GPI>3_compareC_cconly_imm): New.
10152         (*add<GPI>3_compareC_cconly): New.
10153         (*add<GPI>3_compareC_imm): New.
10154         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
10155         handle constants within this pattern..
10156         (*add<GPI>3_compareV_cconly_imm): New.
10157         (*add<GPI>3_compareV_cconly): New.
10158         (*add<GPI>3_compareV_imm): New.
10159         (add<GPI>3_compareV): New.
10160         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
10161         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
10162         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
10163         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
10164         with 'comparison' operator.
10165         (*add<GPI>3_compareV_cconly_imm): Ditto.
10166         (*add<GPI>3_compareV_cconly): Ditto.
10167         (*add<GPI>3_compareV_imm): Ditto.
10168         (add<GPI>3_compareV): Ditto.
10169         (add<mode>3_carryinC): Ditto.
10170         (*add<mode>3_carryinC_zero): Ditto.
10171         (*add<mode>3_carryinC): Ditto.
10172         (add<mode>3_carryinV): Ditto.
10173         (*add<mode>3_carryinV_zero): Ditto.
10174         (*add<mode>3_carryinV): Ditto.
10176 2018-07-19  Michael Collison  <michael.collison@arm.com>
10177             Richard Henderson <rth@redhat.com>
10179         * config/aarch64/aarch64-modes.def (CC_V): New.
10180         * config/aarch64/aarch64-protos.h
10181         (aarch64_addti_scratch_regs): Declare
10182         (aarch64_subvti_scratch_regs): Declare.
10183         (aarch64_expand_subvti): Declare.
10184         (aarch64_gen_unlikely_cbranch): Declare
10185         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
10186         for signed overflow using CC_Vmode.
10187         (aarch64_get_condition_code_1): Handle CC_Vmode.
10188         (aarch64_gen_unlikely_cbranch): New function.
10189         (aarch64_addti_scratch_regs): New function.
10190         (aarch64_subvti_scratch_regs): New function.
10191         (aarch64_expand_subvti): New function.
10193 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10195         * config/aarch64/aarch64-option-extensions.def: New entry for profile
10196         extension.
10197         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
10198         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
10199         extension.
10201 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10203         PR target/83009
10204         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10205         address check not strict.
10207 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10209         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
10210         Umq with Umn.
10211         (store_pair_lanes<mode>): Likewise.
10212         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
10213         enum value 'ADDR_QUERY_LDP_STP_N'.
10214         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
10215         (aarch64_print_address_internal): Add declaration.
10216         (aarch64_print_ldpstp_address): Remove.
10217         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
10218         (aarch64_print_operand): Change printing of 'y'.
10219         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
10220         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
10221         'true' rather than '1'.
10222         * config/aarch64/constraints.md (Uml): Likewise.
10223         (Uml): Rename to Umn.
10224         (Umq): Remove.
10226 2018-07-19  Richard Biener  <rguenther@suse.de>
10228         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
10229         a trailing array.
10230         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
10231         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
10232         (vn_reference_hasher): Likewise.
10233         (struct vn_tables_s): Remove obstack and alloc-pool members.
10234         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
10235         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
10236         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
10237         (vn_reference_insert_pieces): Likewise.
10238         (alloc_vn_nary_op_noinit): Adjust.
10239         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
10240         (vn_phi_eq): Adjust.
10241         (shared_lookup_phiargs): Remove.
10242         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
10243         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
10244         (visit_reference_op_call): Likewise.
10245         (copy_nary, copy_phi, copy_reference): Remove.
10246         (process_scc): Rewind the obstack when iterating.  Do not
10247         copy the elements to valid_info but just move them from one
10248         hashtable to the other.
10249         (allocate_vn_table): Adjust.
10250         (free_vn_table): Likewise.
10251         (init_scc_vn): Likewise.
10252         (free_scc_vn): Likewise.
10254 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
10256         PR target/86560
10257         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
10258         indirect_return as function type attribute.
10259         (ix86_attribute_table): Change indirect_return to function
10260         type attribute.
10261         * doc/extend.texi: Update indirect_return attribute.
10263 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
10265         * wide-int.h (widest2_int): New.
10266         * gimple-fold.c (arith_overflowed_p): Use it.
10267         * tree.h (widest2_int_cst): New.
10268         * tree-vrp.c (wide_int_binop_overflow): Rename from
10269         vrp_int_const_binop.
10270         Rewrite to work on trees.
10271         (extract_range_from_multiplicative_op_1): Abstract code to...
10272         (wide_int_range_min_max): ...here.
10273         (wide_int_range_cross_product): ...and here.
10274         (extract_range_from_binary_expr_1): Abstract overflow code to...
10275         (wide_int_range_mult_wrapping): ...here.
10276         * tree-vrp.h (wide_int_range_cross_product): New.
10277         (wide_int_range_mult_wrapping): New.
10279 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
10280             Julia Koval  <julia.koval@intel.com>
10282         * config/i386/x86-tune-costs.h (skylake_memcpy,
10283         skylake_memset): Replace rep_prefix with unrolling for size 512.
10285 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
10287         PR middle-end/86544
10288         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
10289         comparision with EQ_EXPR in last stmt.
10291 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10293         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
10294         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
10295         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
10296         previously known as "PowerPC AltiVec Built-in Functions".  Move
10297         some material to new subsubsections "PowerPC AltiVec Built-in
10298         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
10299         ISA 2.07".
10300         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
10301         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
10302         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
10303         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
10305 2018-07-18  Richard Biener  <rguenther@suse.de>
10307         PR tree-optimization/86557
10308         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
10309         EXACT_DIV_EXPR.
10311 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
10313         * config/s390/s390.c (s390_function_profiler): Generate CFI.
10315 2018-07-17  Jeff Law  <law@redhat.com>
10317         * config/arm/arm.c (get_label_padding): Update for recent
10318         changes to label_to_alignment.
10320         PR tree-optimization/86010
10321         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
10323         * config/mips/mips.c (vr4130_align_insns): Update for recent
10324         changes to label_to_alignment.
10326         * config/frv/frv.c (frv_label_align): Update for recent changes
10327         to label_to_alignment.
10329         * config/nios2/nios2.c (nios2_label_align): Update for recent
10330         changes which dropped ALIGN_LABELS_LOG.
10332 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
10334         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
10335         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
10337 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
10339         * config/arc/arc.c (arc_label_align): Use align_labels instead of
10340         deprecated align_labels_log.
10342 2018-07-17  Richard Biener  <rguenther@suse.de>
10344         PR lto/86456
10345         * dwarf2out.c (init_sections_and_labels): Always generate
10346         a debug_line_str_section for early LTO debug.
10347         (dwarf2out_finish): Reset debug_line_str_hash output early.
10348         Bump counter for extra dwarf5 .debug_loc labels to not conflict
10349         with fat LTO part.
10350         (dwarf2out_early_finish): Output debug_line_str.
10352 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10354         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
10355         index register on z196 or later.
10357 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10359         * config/s390/s390.c (s390_default_align): Set default function
10360         alignment to 16.
10361         (s390_override_options_after_change): Call s390_default align.
10362         (s390_option_override_internal): Call s390_default_align.
10363         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
10365 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10367         PR middle-end/86542
10368         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
10369         to _looptemp_ clauses, other than the first two.
10371 2018-07-17  Martin Liska  <mliska@suse.cz>
10373         * opts.c: Do not enable OPT_falign_* for -Os.
10375 2018-07-17  Martin Liska  <mliska@suse.cz>
10377         * align.h (MAX_CODE_ALIGN): New.
10378         (MAX_CODE_ALIGN_VALUE): New.
10379         * common/config/i386/i386-common.c (ix86_handle_option):
10380         (MAX_CODE_ALIGN): Moved to align.h.
10381         * final.c (MAX_CODE_ALIGN): Likewise.
10382         * opts.c (parse_and_check_align_values):
10383         (MAX_CODE_ALIGN): Likewise.
10384         (MAX_CODE_ALIGN_VALUE): Likewise.
10386 2018-07-17  Martin Liska  <mliska@suse.cz>
10388         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
10389         in order to fulfil coding style.
10390         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
10391         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10392         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10393         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
10394         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
10395         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
10396         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
10397         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
10398         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10400 2018-07-17  Martin Liska  <mliska@suse.cz>
10402         * align.h: New file.
10403         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
10404         directly.
10405         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
10406         align_flags of label_to_alignment.
10407         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
10408         align_flags class.
10409         * config/m68k/m68k.c: Do not use removed align_labels_value and
10410         align_loops_value.
10411         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
10412         (LOOP_ALIGN): Likewise.
10413         (LABEL_ALIGN): Likewise.
10414         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10415         Remove not used macro.
10416         (rs6000_loop_align): Change return type to align_flags.
10417         (rs6000_loop_align_max_skip): Remove.
10418         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
10419         Change return type to align_flags.
10420         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10421         Remove not used macro.
10422         (rs6000_loop_align):  Change return type to align_flags.
10423         (rs6000_loop_align_max_skip): Remove.
10424         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
10425         * config/rx/rx-protos.h (rx_align_for_label): Make it
10426         static function.
10427         * config/rx/rx.c (rx_align_for_label): Change return type
10428         to align_flags.
10429         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
10430         macro definitions.
10431         into align_flags class.
10432         (LABEL_ALIGN): Likewise.
10433         (LOOP_ALIGN): Likewise.
10434         * config/s390/s390.c (s390_label_align): Use align_flags
10435         class member.
10436         (s390_asm_output_function_label): Likewise.
10437         * config/sh/sh.c (sh_override_options_after_change):
10438         Use align_flags class directly without macros.
10439         (find_barrier): Likewise.
10440         (barrier_align): Likewise.
10441         (sh_loop_align): Likewise.
10442         * config/spu/spu.c (spu_option_override):
10443         Use align_flags_tuple::get_value instead of removed macros.
10444         (spu_sched_init): Likewise.
10445         * config/spu/spu.h (GTY): Likewise.
10446         * config/visium/visium.c (visium_option_override):
10447         Set "8" as default secondary alignment.
10448         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
10449         in order to guarantee secondary alignment of 8.
10450         * coretypes.h: Include align.h header file.
10451         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
10452         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
10453         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
10454         * doc/tm.texi.in: Likewise.
10455         * final.c (struct label_alignment): Remove not used structure.
10456         (LABEL_ALIGN): Change type to align_flags.
10457         (LOOP_ALIGN): Likewise.
10458         (JUMP_ALIGN): Likewise.
10459         (default_loop_align_max_skip): Remove.
10460         (default_label_align_max_skip): Likewise.
10461         (default_jump_align_max_skip): Likewise.
10462         (default_label_align_after_barrier_max_skip):
10463         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
10464         (LABEL_TO_MAX_SKIP): Remove.
10465         (label_to_alignment): Return align_flags type instead of integer.
10466         (label_to_max_skip): Remove.
10467         (align_fuzz): Use align_flags type.
10468         (compute_alignments): Use align_flags type and use align_flags::max
10469         to combine multiple alignments.
10470         (grow_label_align): Grow vec instead of C array.
10471         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
10472         (shorten_branches):  Use align_flags type and use align_flags::max
10473         to combine multiple alignments.
10474         (final_scan_insn_1): Remove usage of secondary alignment that comes
10475         from label alignment, but instead use proper secondary alignment
10476         which is computed in grow_label_align.
10477         * flags.h (struct align_flags_tuple): Move to align.h.
10478         (struct align_flags): Likewise.
10479         (state_align_loops): Rename to align_loops.
10480         (state_align_jumps): Rename to align_jumps.
10481         (state_align_labels): Rename to align_labels.
10482         (state_align_functions): Rename to align_functions.
10483         (align_loops_log): Remove.
10484         (align_jumps_log): Remove.
10485         (align_labels_log): Remove.
10486         (align_functions_log): Remove.
10487         (align_loops_max_skip): Remove.
10488         (align_jumps_max_skip): Remove.
10489         (align_labels_max_skip): Remove.
10490         (align_functions_max_skip): Remove.
10491         (align_loops_value): Remove.
10492         (align_jumps_value): Remove.
10493         (align_labels_value): Remove.
10494         (align_functions_value): Remove.
10495         * output.h (label_to_alignment): Change return type to align_flags.
10496         (label_to_max_skip): Remove.
10497         * target.def: Remove loop_align_max_skip, label_align_max_skip,
10498         jump_align_max_skip macros.
10499         * targhooks.h (default_loop_align_max_skip): Remove.
10500         (default_label_align_max_skip): Likewise.
10501         (default_jump_align_max_skip): Likewise.
10502         (default_label_align_after_barrier_max_skip): Remove.
10503         * toplev.c (read_log_maxskip): Use ::normalize function.
10504         (parse_N_M): Remove not used argument and also call ::normalize.
10505         (parse_alignment_opts): Do not pass unused arguments.
10506         * varasm.c (assemble_start_function): Use directly align_functions
10507         instead of removed macros.
10508         * system.h: Do not poison removed macros.
10510 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10512         PR middle-end/86539
10513         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
10514         and cond temporaries don't have reference type if iterator has
10515         pointer type.  For init use &for_pre_body instead of pre_p if
10516         for_pre_body is non-empty.
10518 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10520         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
10521         double-double modes to SFmode directly directly.
10522         (trunc<mode>sf2_fprs): Delete.
10524 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10526         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
10527         for conversions between IFmode and the decimal floating point modes.
10528         (init_float128_ieee): Use the correct names for conversions between
10529         KFmode and the decimal floating point modes.
10531 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10533         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
10534         for the conversions between TDmode and IFmode.
10535         (init_float128_ieee): Use more correct names for the conversions
10536         between TDmode and KFmode.
10538 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
10540         PR tree-optimization/86526
10541         * builtins.c (expand_builtin_memcmp): Formatting fixes.
10542         (inline_expand_builtin_string_cmp): Likewise.
10543         (inline_string_cmp): Likewise.  Use c_readstr instead of
10544         builtin_memcpy_read_str.  Add unit_mode temporary.
10546 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10548         PR middle-end/86528
10549         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
10550         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
10552 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10554         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
10555         Alphabetize prototypes of built-in functions, separating out
10556         built-in functions that are listed in this section but should be
10557         described elsewhere.
10559 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
10561         PR target/86511
10562         * expmed.c (emit_store_flag): Do not emit setcc followed by a
10563         conditional move when trapping comparison was split to a
10564         non-trapping one (and vice versa).
10566 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10568         * config/s390/s390.c (s390_function_profiler): Generate nops
10569         instead of profiler call sequences.
10570         * config/s390/s390.opt: Add the new option.
10572 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10574         * config/s390/s390.c (s390_function_profiler): Generate
10575         __mcount_loc section.
10576         * config/s390/s390.opt: Add the new option.
10578 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10580         * common.opt: Add the new warning.
10581         * config/s390/s390.c (s390_function_profiler): Emit "brasl
10582         %r0,__fentry__" when -mfentry is specified.
10583         (s390_option_override_internal): Disallow -mfentry for 31-bit
10584         CPUs.
10585         * config/s390/s390.opt: Add the new option.
10587 2018-07-16  Richard Biener  <rguenther@suse.de>
10589         PR lto/86523
10590         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
10591         for function-local FUNCTION_DECL and RESULT_DECL immediately.
10593 2018-07-16  Martin Liska  <mliska@suse.cz>
10595         PR ipa/86529
10596         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
10597         to ::get_create.
10599 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10601         * config/arc/arcHS.md: Update ARCHS scheduling rules.
10603 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10605         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
10606         for ARCHS4x.
10607         * config/arc/arc-cpus.def (hs4x): New cpu.
10608         (hs4xd): Likewise.
10609         * config/arc/arc-tables.opt: Regenerate.
10610         * config/arc/arc.c (arc_sched_issue_rate): New function.
10611         (TARGET_SCHED_ISSUE_RATE): Define.
10612         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
10613         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
10614         fpu_cvt.
10615         (attr tune): Add ARCHS4x tune values.
10616         (attr tune_dspmpy): Define.
10617         (*tst): Correct instruction type.
10618         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
10619         * config/arc/arcHS4x.md: New file.
10620         * config/arc/fpu.md: Update instruction type attributes.
10621         * config/arc/t-multilib: Regenerate.
10623 2018-07-16  Tom de Vries  <tdevries@suse.de>
10625         PR debug/86455
10626         * var-tracking.c (vt_initialize): Fix pre_dec handling.
10628 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
10630         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10631         early clobber.
10633 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
10635         PR tree-optimization/86514
10636         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10637         conversion to a boolean type from a type with greater precision.
10639 2018-07-16  Tom de Vries  <tdevries@suse.de>
10641         * var-tracking.c (vt_initialize): Print adjusted insn slim if
10642         dump_flags request TDF_SLIM.
10644 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
10646         * fold-const.c (int_const_binop_1): Abstract...
10647         (wide_int_binop): ...wide int code here.
10648         (poly_int_binop): ...poly int code here.
10649         Abstract the rest of int_const_binop_1 into int_const_binop.
10650         * fold-const.h (wide_int_binop): New.
10651         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10652         Remove useless PLUS/MINUS_EXPR case.
10653         (zero_nonzero_bits_from_vr): Move wide int code...
10654         (zero_nonzero_bits_from_bounds): ...here.
10655         (extract_range_from_binary_expr_1): Move mask optimization code...
10656         (range_easy_mask_min_max): ...here.
10657         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10658         (range_easy_mask_min_max): New.
10660 2018-07-15  Jeff Law  <law@redhat.com>
10662         PR target/85993
10663         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10664         block.
10666 2018-07-14  Jim Wilson  <jimw@sifive.com>
10668         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10670 2018-07-14  Paul Koning  <ni1d@arrl.net>
10672         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10674 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
10676         * lto-streamer-out.c (copy_function_or_variable): Dump info about
10677         copying section.
10679 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
10680             Steve Munroe  <munroesj52@gmail.com>
10682         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10683         (_mm_andnot_si128): Likewise.
10684         (_mm_or_si128): Likewise.
10685         (_mm_xor_si128): Likewise.
10687 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
10689         PR middle-end/78809
10690         * builtins.c (expand_builtin_memcmp): Inline the calls first
10691         when result_eq is false.
10692         (expand_builtin_strcmp): Inline the calls first.
10693         (expand_builtin_strncmp): Likewise.
10694         (inline_string_cmp): New routine. Expand a string compare
10695         call by using a sequence of char comparison.
10696         (inline_expand_builtin_string_cmp): New routine. Inline expansion
10697         a call to str(n)cmp/memcmp.
10698         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10699         New option.
10700         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10702 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
10704         * config/arm/driver-arm.c: Include arm-native.h.
10705         (host_detect_local_cpu): Use auto-generated data tables.
10706         (vendors, arm_cpu_table): Delete.  Move part information to ...
10707         * config/arm/arm-cpus.in: ... here.
10708         * config/arm/parsecpu.awk (gen_native): New function.
10709         (vendor, part): New CPU fields.
10710         (END): Add support for building the native CPU detection tables.
10711         * config/arm/t-arm (arm-native.h): Add build rule.
10712         (driver-arm.o): Add dependency on arm-native.h.
10714 2018-07-13  Richard Biener  <rguenther@suse.de>
10716         PR middle-end/85974
10717         * match.pd (addr1 - addr2): Allow either of the operand to
10718         have a conversion.
10720 2018-07-13  Tom de Vries  <tdevries@suse.de>
10722         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10723         in remap_ssa_name.
10725 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
10727         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10728         arrays instead of numbered variables.
10730 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
10732         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10733         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10735 2018-07-13  Richard Biener  <rguenther@suse.de>
10737         PR debug/86452
10738         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10739         instead of get_context_die.
10741 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
10742             Richard Biener  <rguenther@suse.de>
10744         PR middle-end/86489
10745         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10746         that the loop latch destination where phi is defined.
10748 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
10750         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10751         (riscv_expand_epilogue): Add assertion to check interrupt mode.
10752         (riscv_set_current_function): Extract getting interrupt type to new
10753         function.
10754         (riscv_get_interrupt_type): New function.
10755         (riscv_merge_decl_attributes): New function, checking interrupt type is
10756         same.
10757         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10759 2018-07-12  Paul Koning  <ni1d@arrl.net>
10761         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10762         directive.
10764 2018-07-12  Paul Koning  <ni1d@arrl.net>
10766         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10767         zero reference, add doloop_end instead.
10768         * doc/md.texi (decrement_and_branch_until_zero): Remove.
10769         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
10770         detail for doloop_end.
10772 2018-07-12  Martin Sebor  <msebor@redhat.com>
10774         PR c/86453
10775         * attribs.c (decl_attributes): Reject conflicting attributes before
10776         calling attribute handlers.
10778 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10780         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10781          parameter.
10782         (gcc::dump_manager::get_dump_file_name): likewise.
10783         (dump_begin): Likewise.
10784         * dumpfile.h (dump_begin): Update prototype.
10785         (gcc::dump_manager::get_dump_file_name,
10786         gcc::dump_manager::get_dump_file_name): Update prototype.
10788 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10790         * internal-fn.h (vectorizable_internal_fn_p): New function.
10791         * tree-vect-slp.c (compatible_calls_p): Likewise.
10792         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
10793         to internal functions.
10794         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10796 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10798         * fold-const.h (inverse_conditions_p): Declare.
10799         * fold-const.c (inverse_conditions_p): New function.
10800         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
10801         that test the inverse condition of a conditional internal function.
10802         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10803         * internal-fn.c (internal_fn_mask_index): Handle conditional
10804         internal functions.
10805         (vectorized_internal_fn_supported_p): New function.
10806         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10807         (any_pred_load_store): Replace with...
10808         (need_to_predicate): ...this new variable.
10809         (redundant_ssa_names): New variable.
10810         (ifcvt_can_use_mask_load_store): Move initial checks to...
10811         (ifcvt_can_predicate): ...this new function.  Handle tree codes
10812         for which a conditional internal function exists.
10813         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10814         instead of ifcvt_can_use_mask_load_store.  Update after variable
10815         name change.
10816         (predicate_load_or_store): New function, split out from
10817         predicate_mem_writes.
10818         (check_redundant_cond_expr): New function.
10819         (value_available_p): Likewise.
10820         (predicate_rhs_code): Likewise.
10821         (predicate_mem_writes): Rename to...
10822         (predicate_statements): ...this.  Use predicate_load_or_store
10823         and predicate_rhs_code.
10824         (combine_blocks, tree_if_conversion): Update after above name changes.
10825         (ifcvt_local_dce): Handle redundant_ssa_names.
10826         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10827         general conditional functions.
10828         * tree-vect-stmts.c (vectorizable_call): Likewise.
10830 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10831             Alan Hayward  <alan.hayward@arm.com>
10832             David Sherwood  <david.sherwood@arm.com>
10834         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10835         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10836         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10837         plus and minus and convert them into IFN_COND_FMA-based sequences.
10838         (convert_mult_to_fma): Handle conditional plus and minus.
10840 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10842         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10843         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10844         (cond_fnms_optab): New optabs.
10845         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10846         internal functions.
10847         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10848         * internal-fn.h (get_conditional_internal_fn): Declare.
10849         (get_unconditional_internal_fn): Likewise.
10850         * internal-fn.c (cond_ternary_direct): New macro.
10851         (expand_cond_ternary_optab_fn): Likewise.
10852         (direct_cond_ternary_optab_supported_p): Likewise.
10853         (FOR_EACH_COND_FN_PAIR): Likewise.
10854         (get_conditional_internal_fn): New function.
10855         (get_unconditional_internal_fn): Likewise.
10856         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10857         (gimple_match_op::gimple_match_op): Add a new overload for 5
10858         operands.
10859         (gimple_match_op::set_op): Likewise.
10860         (gimple_resimplify5): Declare.
10861         * genmatch.c (decision_tree::gen): Generate simplifications for
10862         5 operands.
10863         * gimple-match-head.c (gimple_simplify): Define an overload for
10864         5 operands.  Handle calls with 5 arguments in the top-level overload.
10865         (convert_conditional_op): Handle conversions from unconditional
10866         internal functions to conditional ones.
10867         (gimple_resimplify5): New function.
10868         (build_call_internal): Pass a fifth operand.
10869         (maybe_push_res_to_seq): Likewise.
10870         (try_conditional_simplification): Try converting conditional
10871         internal functions to unconditional internal functions.
10872         Handle 3-operand unconditional forms.
10873         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10874         Define ternary equivalents of the current rules for binary conditional
10875         internal functions.
10876         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10877         ternary operations.
10878         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10879         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10880         (optab): Handle them.
10881         (SVE_COND_FP_TERNARY): New int iterator.
10882         (sve_fmla_op, sve_fmad_op): New int attributes.
10883         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10884         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10885         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10887 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10889         * target.def (preferred_else_value): New target hook.
10890         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10891         * doc/tm.texi: Regenerate.
10892         * targhooks.h (default_preferred_else_value): Declare.
10893         * targhooks.c (default_preferred_else_value): New function.
10894         * internal-fn.h (conditional_internal_fn_code): Declare.
10895         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10896         (get_conditional_internal_fn): Use it.
10897         (conditional_internal_fn_code): New function.
10898         * gimple-match.h (gimple_match_cond): New struct.
10899         (gimple_match_op): Add a cond member function.
10900         (gimple_match_op::gimple_match_op): Update all forms to take a
10901         gimple_match_cond.
10902         * genmatch.c (expr::gen_transform): Use the same condition as res_op
10903         for the suboperation, but don't specify a particular else_value.
10904         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10905         (visit_nary_op, visit_reference_op_load): Pass
10906         gimple_match_cond::UNCOND to the gimple_match_op constructor.
10907         * gimple-match-head.c: Include tree-eh.h
10908         (convert_conditional_op): New function.
10909         (maybe_resimplify_conditional_op): Likewise.
10910         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10911         (gimple_resimplify2): Likewise.
10912         (gimple_resimplify3): Likewise.
10913         (gimple_resimplify4): Likewise.
10914         (maybe_push_res_to_seq): Return null for conditional operations.
10915         (try_conditional_simplification): New function.
10916         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
10917         constructor.
10918         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10919         IFN_COND_* call.
10920         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10921         function.
10922         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10924 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10926         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10927         DECL_FCONTEXT
10928         (hash_tree): Do not hash DECL_FCONTEXT
10929         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10930         Do not stream DECL_FCONTEXT.
10931         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10932         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10934 2018-07-12  Richard Biener  <rguenther@suse.de>
10936         PR debug/86462
10937         * dwarf2out.c (gen_block_die): Only output blocks when they have
10938         at least one !DECL_IGNORED_P variable.
10940 2018-07-12  Richard Biener  <rguenther@suse.de>
10942         PR target/84829
10943         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10944         Remove -mieee-fp handling.
10946 2018-07-12  Richard Biener  <rguenther@suse.de>
10948         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10949         left-over from last patch.
10951 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
10953         PR tree-optimization/86492
10954         * gimple-ssa-store-merging.c
10955         (imm_store_chain_info::coalesce_immediate_stores): Call
10956         check_no_overlap even for the merge_overlapping case.  Formatting fix.
10958 2018-07-12  Richard Biener  <rguenther@suse.de>
10960         PR middle-end/86479
10961         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10962         move possibly trapping operations into the conditional.
10964 2018-07-12  Richard Biener  <rguenther@suse.de>
10966         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10967         (vn_lookup_simplify_result): Remove recursion limit applied
10968         here.
10969         (vn_nary_build_or_lookup_1): Adjust.
10970         (try_to_simplify): Likewise.
10971         * gimple-match-head.c (gimple_resimplify1): Instead apply one
10972         here.
10973         (gimple_resimplify2): Likewise.
10974         (gimple_resimplify3): Likewise.
10975         (gimple_resimplify4): Likewise.
10977 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
10979         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10980         Use __mmask64 type instead of __mmask8 for __M argument.
10981         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10982         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10983         __U argument.
10984         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10985         __mmask16 for __M argument.
10986         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10987         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10988         to __mmask16 instead of __mmask8.
10989         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10990         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10991         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10992         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10993         instead of __mmask16 for __U argument.
10994         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10995         __mmask16 instead of __mmask8 for __U argument.
10996         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10997         __U argument.
10998         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10999         __mmask16.
11000         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
11001         argument.
11002         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
11003         __U argument.
11004         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
11005         __mmask16.
11006         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
11007         of __mmask16.
11008         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
11009         __U argument.
11010         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
11011         __U argument.
11012         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
11013         __U argument.
11014         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
11015         __U argument.
11016         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
11017         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
11018         return type as well as __M argument type and all casts from __mmask8
11019         to __mmask32.
11020         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
11021         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
11022         return type as well as __M argument type and all casts from __mmask8
11023         to __mmask16.
11024         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
11025         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
11026         return type as well as __M argument type and all casts from __mmask8
11027         to __mmask32.
11028         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
11029         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
11030         return type as well as __M argument type and all casts from __mmask8
11031         to __mmask16.
11032         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
11033         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
11034         __mmask16.
11036 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
11038         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
11039         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
11040         for __U argument.
11042 2018-07-11  Paul Koning  <ni1d@arrl.net>
11044         * doc/md.texi (define_subst): Document how multiple occurrences of
11045         the same argument in the replacement pattern are handled.
11047 2018-07-11  Paul Koning  <ni1d@arrl.net>
11049         * doc/extend.texi (Common Variable Attributes): Move "mode" into
11050         alphabetical order.
11051         (Common Type Attributes): Add "mode" attribute.
11053 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
11055         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
11056         stream DECL_ORIGINAL_TYPE.
11057         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
11058         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
11059         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
11060         Do not walk original type.
11061         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
11062         external decls.
11063         (write_ts_decl_non_common_tree_pointers): Do not stream
11064         DECL_ORIGINAL_TYPE
11065         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
11066         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
11068 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
11070         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
11071         thread twice from the same starting edge.
11073 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
11075         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
11076         * gimple.c (gimple_call_nonnull_result_p): ...here...
11077         (gimple_call_nonnull_arg): ...and here.
11078         * gimple.h (gimple_call_nonnull_result_p): New.
11079         (gimple_call_nonnull_arg): New.
11081 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
11083         * config/arm/arm-cpus.in: Move information from fpu field of each
11084         cpu definition to the isa field.
11085         * config/arm/parsecpu.awk (fpu): Delete match rule.
11086         (gen_comm_data): Don't add bits from the CPU's FPU entry.
11088 2018-07-11  Richard Biener  <rguenther@suse.de>
11090         PR debug/86457
11091         * dwarf2out.c (init_sections_and_labels): Use
11092         output_asm_line_debug_info consistently.
11093         (dwarf2out_early_finish): Likewise.
11094         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
11095         type units.
11097 2018-07-11  Richard Biener  <rguenther@suse.de>
11099         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
11100         Rework father_bb setting in a way to avoid propagating constants
11101         multiple times on a loop body.
11103 2018-07-10  Mark Wielaard  <mark@klomp.org>
11105         PR debug/86459
11106         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
11108 2018-07-10  Richard Biener  <rguenther@suse.de>
11110         * hash-map.h (hash_map::iterator::operator*): Return
11111         references to key and value.
11113 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
11115         PR c++/86443
11116         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
11117         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
11118         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
11119         (gimplify_omp_for): For composite loops, move outer
11120         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
11121         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
11122         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
11123         TREE_LIST for both the original class iterator and the "last" helper
11124         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
11125         loop, remember has_decl_expr from outer composite loops for the
11126         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
11128 2018-07-09  Martin Sebor  <msebor@redhat.com>
11130         PR middle-end/77357
11131         PR middle-end/86428
11132         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
11133         accessing implicitly initialized array elements.
11134         * expr.c (string_constant): Handle string initializers of
11135         character arrays within aggregates.
11136         * gimple-fold.c (fold_array_ctor_reference): Add argument.
11137         Store element offset.  As a special case, handle zero size.
11138         (fold_nonarray_ctor_reference): Same.
11139         (fold_ctor_reference): Add argument.  Store subobject offset.
11140         * gimple-fold.h (fold_ctor_reference): Add argument.
11142 2018-07-09  Paul Koning  <ni1d@arrl.net>
11144         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
11145         (pdp11_insn_cost): New function.
11146         (pdp11_md_asm_adjust): New function.
11147         (TARGET_INVALID_WITHIN_DOLOOP): Define.
11148         (pdp11_rtx_costs): Update to match machine better.
11149         (output_addr_const_pdp11): Correct format mismatch warnings.
11150         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
11151         * config/pdp11/pdp11.md: General change to add base_cost and/or
11152         length attributes for use by new pdp11_insn_cost function.
11153         (MIN_BRANCH): Correct definition.
11154         (MIN_SOB): Ditto.
11155         (doloop_end): Use standard pattern name for looping pattern.
11156         (doloop_end_nocc): New.
11157         (movsf): Add another constraint alternative.
11158         (zero_extendqihi2): Add constraint alternatives for not in place
11159         extend.
11160         (zero_extendhisi2): Remove.
11161         (shift patterns): Add CC handling variants.
11162         (bswaphi2): New.
11163         (bswapsi2): New.
11164         (rothi3): New.
11165         (define_peephole2): New peephole to recognize mov that sets CC for
11166         subsequent test.
11168 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11170         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
11171         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
11172         wi::add.
11174 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
11176         PR c/86420
11177         * real.c (real_nextafter): Return true if result is denormal.
11179 2018-07-09  Martin Liska  <mliska@suse.cz>
11181         * common.opt: Add back wrongly removed attribute.
11183 2018-07-09  Richard Biener  <rguenther@suse.de>
11185         PR debug/86413
11186         * dwarf2out.c (gen_block_die): For an early generated DIE
11187         always output high/low PC attributes.
11189 2018-07-09  Tom de Vries  <tdevries@suse.de>
11191         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
11192         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
11193         onto VAR_DECL with abstract origin.
11195 2018-07-07  Jim Wilson  <jimw@sifive.com>
11197         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
11199 2018-07-07  Tom de Vries  <tdevries@suse.de>
11201         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
11203 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
11205         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
11206         overflow_type.
11207         (combine_bound): Use wide-int overflow calculation instead of
11208         rolling our own.
11209         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
11210         overflow_type.
11211         * fold-const.c (int_const_binop_2): Same.
11212         (extract_muldiv_1): Same.
11213         (fold_div_compare): Same.
11214         (fold_abs_const): Same.
11215         * match.pd: Same.
11216         * poly-int.h (add): Same.
11217         (sub): Same.
11218         (neg): Same.
11219         (mul): Same.
11220         * predict.c (predict_iv_comparison): Same.
11221         * profile-count.c (slow_safe_scale_64bit): Same.
11222         * simplify-rtx.c (simplify_const_binary_operation): Same.
11223         * tree-chrec.c (tree_fold_binomial): Same.
11224         * tree-data-ref.c (split_constant_offset_1): Same.
11225         * tree-if-conv.c (idx_within_array_bound): Same.
11226         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
11227         * tree-ssa-phiopt.c (minmax_replacement): Same.
11228         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
11229         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
11230         * vr-values.c (vr_values::adjust_range_with_scev): Same.
11231         * wide-int.cc (wi::add_large): Same.
11232         (wi::mul_internal): Same.
11233         (wi::sub_large): Same.
11234         (wi::divmod_internal): Same.
11235         * wide-int.h: Change overflow type to overflow_type for neg, add,
11236         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
11237         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
11238         mul_internal, divmod_internal.
11239         (overflow_type): New enum.
11240         (accumulate_overflow): New.
11242 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
11244         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
11245         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
11247 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11249         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
11250         argument is checked for zero before entering loop, avoid checking again.
11252 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11254         * gimplify.h (generic_expr_could_trap_p): Set as global function.
11255         * gimplify.h (generic_expr_could_trap_p): Likwise.
11256         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
11258 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
11260         PR tree-optimization/86401
11261         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
11262         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
11263         (fold_bit_and_mask): ... here.  New helper function for match.pd.
11264         * fold-const.h (fold_bit_and_mask): Declare.
11265         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
11267 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
11269         PR target/86324
11270         * target.def (translate_mode_attribute): New hook.
11271         * targhooks.h (default_translate_mode_attribute): Declare.
11272         * targhooks.c (default_translate_mode_attribute): New function.
11273         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
11274         * doc/tm.texi: Regenerate.
11275         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
11276         (rs6000_translate_mode_attribute): New function.
11278 2018-07-06  Paul Koning  <ni1d@arrl.net>
11280         * doc/md.texi (define_split): Document DONE and FAIL.
11281         (define_peephole2): Ditto.
11283 2018-07-05  Jeff Law  <law@redhat.com>
11285         PR tree-optimization/86010
11286         * tree-ssa-dse.c (compute_trims): More aggressively trim at
11287         both the head and tail of mem* and str* calls.
11289 2018-07-05  Jim Wilson  <jimw@sifive.com>
11291         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
11293 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
11295         * config/aarch64/aarch64-simd.md: correct flags text for
11296         MIN_EXPR replacement.
11298 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
11300         * configure: Regenerated.
11302 2018-07-05  Carl Love  <cel@us.ibm.com>
11304         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
11305         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
11306         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
11307         VSX_BUILTIN_DOUBLEL_V4SF.
11309 2018-07-05  Martin Sebor  <msebor@redhat.com>
11311         PR c++/86400
11312         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
11313         than its domain to compute its the upper bound of a char array.
11315 2018-07-05  Nathan Sidwell  <nathan@acm.org>
11317         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
11318         * doc/cpp.texi: Update comment.
11319         * doc/tm.texi: Rebuilt.
11320         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
11321         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
11322         * doc/extend.texi (Backwards Compatibility): Clarify it is system
11323         headers affected by extern "C".
11324         * system.h: Poison NO_IMPLICIT_EXTERN_C.
11325         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
11326         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
11327         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
11328         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
11329         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
11330         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
11331         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
11332         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
11333         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
11334         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
11335         config/sparc/sp64-elf.h, config/spu/spu.h,
11336         config/stormy16/stormy16.h, config/v850/v850.h,
11337         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
11338         define NO_IMPLICIT_EXTERN_C.
11339         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
11341 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11343         PR target/84711
11344         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11345         instead of GET_MODE_SIZE when comparing Units.
11347 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11349         PR target/84711
11350         * rtlanal.c (set_noop_p): Constrain on mode change,
11351         include hard-reg-set.h
11353 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11355         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
11357 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
11359         Revert
11360         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11362         PR sanitizer/84250
11363         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11364         libasan.
11365         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11367 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11369         PR sanitizer/84250
11370         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11371         libasan.
11372         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11374 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
11376         PR middle-end/86380
11377         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
11379 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
11381         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
11382         neg_*_op* variables.
11384 2018-07-04  Martin Liska  <mliska@suse.cz>
11386         * tree-switch-conversion.c: Define
11387         max_ratio_for_speed and max_ratio_for_size constants.
11389 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11390             Martin Liska  <mliska@suse.cz>
11392         PR middle-end/66240
11393         PR target/45996
11394         PR c/84100
11395         * common.opt: Rename align options with 'str_' prefix.
11396         * common/config/i386/i386-common.c (set_malign_value): New
11397         function.
11398         (ix86_handle_option): Use it to set -falign-* options/
11399         * config/aarch64/aarch64-protos.h (struct tune_params): Change
11400         type from int to string.
11401         * config/aarch64/aarch64.c: Update default values from int
11402         to string.
11403         * config/alpha/alpha.c (alpha_override_options_after_change):
11404         Likewise.
11405         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
11406         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11407         max skip conditionally.
11408         * config/i386/freebsd.h (SUBALIGN_LOG): New.
11409         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11410         max skip conditionally.
11411         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11412         max skip conditionally.
11413         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
11414         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11415         max skip conditionally.
11416         * config/i386/i386.c (struct ptt): Change type from int to
11417         string.
11418         (ix86_default_align): Set default values.
11419         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
11420         max skip conditionally.
11421         * config/i386/iamcu.h (SUBALIGN_LOG): New.
11422         (ASM_OUTPUT_MAX_SKIP_ALIGN):
11423         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
11424         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11425         max skip conditionally.
11426         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
11427         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
11428         * config/i386/x86-64.h (SUBALIGN_LOG): New.
11429         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11430         max skip conditionally.
11431         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11432         * config/ia64/ia64.c (ia64_option_override): Set default values
11433         for alignment options.
11434         * config/m68k/m68k.c: Handle new str_align_* options.
11435         * config/mips/mips.c (mips_set_compression_mode): Change
11436         type of constants.
11437         (mips_option_override): Set default values for options.
11438         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
11439         Likewise.
11440         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11441         Likewise.
11442         * config/rx/rx.c (rx_option_override): Likewise.
11443         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
11444         (LABEL_ALIGN): Use align_labels_log.
11445         (LOOP_ALIGN): Use align_loops_align.
11446         * config/s390/s390.c (s390_asm_output_function_label): Use new
11447         macros.
11448         * config/sh/sh.c (sh_override_options_after_change):
11449         Change type of constants.
11450         * config/spu/spu.c (spu_sched_init): Likewise.
11451         * config/sparc/sparc.c (sparc_option_override): Set default
11452         values for options.
11453         * config/visium/visium.c (visium_option_override): Likewise.
11454         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
11455         emit p2align format with last argument if it's not needed.
11456         * doc/invoke.texi: Document extended format of -falign-*.
11457         * final.c: Use align_labels alignment.
11458         * flags.h (struct target_flag_state): Change type to use
11459         align_flags.
11460         (struct align_flags_tuple): New.
11461         (struct align_flags): Likewise.
11462         (align_loops_log): Redefine macro to use new types.
11463         (align_loops_max_skip): Redefine macro to use new types.
11464         (align_jumps_log): Redefine macro to use new types.
11465         (align_jumps_max_skip): Redefine macro to use new types.
11466         (align_labels_log): Redefine macro to use new types.
11467         (align_labels_max_skip): Redefine macro to use new types.
11468         (align_functions_log): Redefine macro to use new types.
11469         (align_loops): Redefine macro to use new types.
11470         (align_jumps): Redefine macro to use new types.
11471         (align_labels): Redefine macro to use new types.
11472         (align_functions): Redefine macro to use new types.
11473         (align_functions_max_skip): Redefine macro to use new types.
11474         (align_loops_value): New macro.
11475         (align_jumps_value): New macro.
11476         (align_labels_value): New macro.
11477         (align_functions_value): New macro.
11478         * function.c (invoke_set_current_function_hook): Propagate
11479         alignment values from flags to global variables default in
11480         topleev.h.
11481         * ipa-icf.c (sem_function::equals_wpa): Use
11482         cl_optimization_option_eq instead of memcmp.
11483         * lto-streamer.h (cl_optimization_stream_out): Support streaming
11484         of string types.
11485         (cl_optimization_stream_in): Likewise.
11486         * optc-save-gen.awk: Support strings in cl_optimization.
11487         * opth-gen.awk: Likewise.
11488         * opts.c (finish_options): Remove error checking of invalid
11489         value ranges.
11490         (MAX_CODE_ALIGN): Remove.
11491         (MAX_CODE_ALIGN_VALUE): Likewise.
11492         (parse_and_check_align_values): New function.
11493         (check_alignment_argument): Likewise.
11494         (common_handle_option): Use check_alignment_argument.
11495         * opts.h (parse_and_check_align_values): Declare.
11496         * toplev.c (init_alignments): Remove.
11497         (read_log_maxskip): New.
11498         (parse_N_M): Likewise.
11499         (parse_alignment_opts): Likewise.
11500         (backend_init_target): Remove usage of init_alignments.
11501         * toplev.h (parse_alignment_opts): Declare.
11502         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
11503         argument.
11504         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
11505         * tree.c (cl_option_hasher::equal): New.
11506         * varasm.c: Use new global macros.
11508 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11510         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
11511         Use a simpler align directive also if MAXSKIP = ALIGN-1.
11512         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11513         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11514         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11515         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11516         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
11517         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
11518         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
11519         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11520         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11521         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11522         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11524 2018-07-04  Martin Liska  <mliska@suse.cz>
11525             Jonathan Wakely  <jwakely@redhat.com>
11527         * coverage.c: Use correct type.
11528         * doc/invoke.texi: Language correction.
11530 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
11532         PR target/85620
11533         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
11534         ENDBRANCH for non-tail call which may return via indirect branch.
11535         * doc/extend.texi: Document indirect_return attribute.
11537 2018-07-03  Martin Sebor  <msebor@redhat.com>
11539         PR tree-optimization/86274
11540         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
11541         precondition.
11542         (format_floating): Correct handling of infinities and NaNs.
11544 2018-07-03  Martin Sebor  <msebor@redhat.com>
11546         * print-tree.c (print_real_cst): New function.
11547         (print_node_brief): Call it.
11548         (print_node): Ditto.
11550 2018-07-03  Jeff Law  <law@redhat.com>
11552         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
11553         into a single pattern.
11555         * config/h8300/h8300.md (ors code_iterator): New.
11556         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
11557         a single pattern and single splitter.
11558         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
11559         (iorqi3_1, xorqi3_1): Likewise.
11560         (iorqi3, xorqi3 expanders): Similarly.
11562         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
11563         (movmd_internal) into a single pattern using the P mode iterator.
11564         (movmd splitters): Similarly.
11565         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
11566         (movsd splitters): Similarly.
11568         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
11569         ADDB, ADDW and ADDL into a single ADD attribute which selects the
11570         right table based on the size of the operand.
11571         * config/h8300/h8300.md (length_table): Corresponding changes. All
11572         references to "addb", "addw" and "addl" changed to "add".
11573         (btst patterns): Merge two variants into a single pattern.
11574         (tstqi, tsthi): Likewise.
11575         (addhi3_incdec, addsi3_incdec): Likewise.
11576         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
11577         (mulhi3, mulsi3): Likewise.
11578         (udivhi3, udivsi3): Likewise.
11579         (divhi3, divsi3): Likewise.
11580         (andorqi3, andorhi3, andorsi3): Likewise.
11582 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
11584         PR target/85694
11585         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
11586         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
11588 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11590         PR tree-optimization/85694
11591         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
11592         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
11593         UNSPEC_URHADD.
11594         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
11595         (<u>avg<mode>3_ceil): New patterns.
11597 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
11599         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
11600         scan-tree-dump directive.
11601         * gcc.dg/vect/slp-perm-2.c: Likewise.
11602         * gcc.dg/vect/slp-perm-3.c: Likewise.
11603         * gcc.dg/vect/slp-perm-5.c: Likewise.
11604         * gcc.dg/vect/slp-perm-6.c: Likewise.
11605         * gcc.dg/vect/slp-perm-7.c: Likewise.
11606         * gcc.dg/vect/slp-perm-8.c: Likewise.
11608 2018-07-03  Marek Polacek  <polacek@redhat.com>
11610         PR middle-end/86202
11611         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
11613 2018-07-03  Richard Biener  <rguenther@suse.de>
11615         PR ipa/86389
11616         * tree-ssa-structalias.c (find_func_clobbers): Properly
11617         handle indirect calls.
11619 2018-07-03  Jeff Law  <law@redhat.com>
11621         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
11622         (shifts): New code iterator.
11623         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11624         expander.  Fix HImode handling on H8/SX.
11625         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11626         (subqi3, subhi3, subsi3 expanders): Likewise.
11627         (andqi3, andhi3, andsi3 expanders): Likewise.
11628         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11629         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11630         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11631         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11632         (zero_extendqihi2, zero_extendqisi2): Likewise.
11633         (extendqihi2, extendqisi2): Likewise.
11634         (rotlqi3, rotlhi3, rotlsi3): Likewise.
11635         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11636         (rotlqi3_1, rotlhi3_1): Likewise.
11637         (logicalhi3_sn, logicalsi3_sn): Likewise.
11638         (logicalhi3, logicalsi3): Likewise.
11640 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11642         * tree-vect-patterns.c (vect_recog_rotate_pattern)
11643         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11644         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11645         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11646         type to append_pattern_def_seq instead of creating a stmt_vec_info
11647         directly.
11648         (build_mask_conversion): Likewise.  Remove vinfo argument.
11649         (vect_add_conversion_to_patterm): Likewise, renaming to...
11650         (vect_add_conversion_to_pattern): ...this.
11651         (vect_recog_mask_conversion_pattern): Update call to
11652         build_mask_conversion.  Pass the vector type to
11653         append_pattern_def_seq here too.
11654         (vect_recog_gather_scatter_pattern): Update call to
11655         vect_add_conversion_to_pattern.
11657 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11659         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11660         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11661         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11662         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11663         STMT_VINFO_PATTERN_DEF_SEQ to null here.
11664         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11665         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11666         append_pattern_def_seq instead of new_pattern_def_seq.
11667         (vect_recog_divmod_pattern): Do both of the above.
11668         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11669         is null.
11671 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11673         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11674         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11675         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11676         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11677         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11678         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11679         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11680         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11681         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11682         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11683         parameter with a single stmt_vec_info.
11684         (vect_recog_func_ptr): Likewise.
11685         (vect_recog_gather_scatter_pattern): Likewise, folding in...
11686         (vect_try_gather_scatter_pattern): ...this.
11687         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11688         the stmt_vec_info of the statement to be matched.  Don't clear
11689         STMT_VINFO_RELATED_STMT.
11690         (vect_pattern_recog): Update call accordingly.
11692 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11694         PR tree-optimization/85694
11695         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11696         (uavgM3_ceil): Document new optabs.
11697         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11698         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11699         functions.
11700         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11701         (savg_ceil_optab): New optabs.
11702         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11703         (vect_vect_recog_func_ptrs): Add it.
11704         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11705         constant directly from the associated lhs.
11707 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11709         * tree-vect-patterns.c (vect_split_statement): New function.
11710         (vect_convert_input): Use it to try to split an existing cast.
11712 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11714         * poly-int.h (print_hex): New function.
11715         * dumpfile.h (dump_dec, dump_hex): Declare.
11716         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11717         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11718         min_input_precision, operation_precision and operation_sign.
11719         * tree-vect-patterns.c (vect_get_range_info): New function.
11720         (vect_same_loop_or_bb_p, vect_single_imm_use)
11721         (vect_operation_fits_smaller_type): Delete.
11722         (vect_look_through_possible_promotion): Add an optional
11723         single_use_p parameter.
11724         (vect_recog_over_widening_pattern): Rewrite to use new
11725         stmt_vec_info infomration.  Handle one operation at a time.
11726         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11727         (vect_truncatable_operation_p, vect_set_operation_type)
11728         (vect_set_min_input_precision): New functions.
11729         (vect_determine_min_output_precision_1): Likewise.
11730         (vect_determine_min_output_precision): Likewise.
11731         (vect_determine_precisions_from_range): Likewise.
11732         (vect_determine_precisions_from_users): Likewise.
11733         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11734         (vect_vect_recog_func_ptrs): Put over_widening first.
11735         Add cast_forwprop.
11736         (vect_pattern_recog): Call vect_determine_precisions.
11738 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11740         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11741         statements that have been replaced by further pattern statements.
11742         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11744 2018-07-03  Richard Biener  <rguenther@suse.de>
11746         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11747         always set *dt.  Dump vectype in vectype overload.
11748         * dumpfile.h (dump_gimple_expr): New function.
11749         (dump_gimple_expr_loc): Likewise.
11750         * dumpfile.c (dump_gimple_expr): New function.
11751         (dump_gimple_expr_loc): Likewise.
11753 2018-07-02  Jeff Law  <law@redhat.com>
11755         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11756         the H8/300, H8/300H and H8/S variants into a single pattern.
11757         (movhi_h8300, movqi_h8300hs): Similarly.
11758         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11759         (QHI mode iterator): New.
11761         * config/h8300/h8300.md: Remove trailing whitespace.
11763 2018-07-02  Jim Wilson  <jimw@sifive.com>
11765         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11766         instead of emit_insn for interrupt returns.
11767         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11768         (riscv_sret, riscv_uret): Likewise.
11770 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11772         * pretty-print.c (selftest::test_pp_format): Move save and restore
11773         of quotes to class auto_fix_quotes, and add an instance.
11774         * selftest.c: Include "intl.h".
11775         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11776         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11777         * selftest.h (selftest::auto_fix_quotes): New class.
11779 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
11781         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11782         (aarch64_sve_prepare_conditional_op): Remove.
11783         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11784         Allow aarch64_simd_reg_or_zero as select operand; remove
11785         the aarch64_sve_prepare_conditional_op call.
11786         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11787         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11788         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11789         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11790         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11791         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11792         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11793         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11794         and a splitters to match all of the *_any patterns.
11795         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11797         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11798         (SVE_COND_FP_BINARY_REV): Remove.
11799         (sve_int_op_rev, sve_fp_op_rev): New.
11800         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11801         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11802         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11803         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11804         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11805         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11806         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11807         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11808         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11810         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11811         Remove match_dup 1 from the inner unspec.
11812         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11814         * config/aarch64/aarch64.md (movprfx): New attr.
11815         (length): Default movprfx to 8.
11816         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11817         (*madd<SVE_I>, *msub<SVE_I): Likewise.
11818         (*<su>mul<SVE_I>3_highpart): Likewise.
11819         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11820         (*v<ASHIFT><SVE_I>3): Likewise.
11821         (*<su><MAXMIN><SVE_I>3): Likewise.
11822         (*<su><MAXMIN><SVE_F>3): Likewise.
11823         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11824         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11825         (*div<SVE_F>4): Likewise.
11827 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
11829         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11830         in dump string.
11832 2018-07-02  Richard Biener  <rguenther@suse.de>
11834         PR tree-optimization/86363
11835         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11836         memset argument refers to a non-variable address.
11838 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11840         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11841         {PLUS,MINUS}_EXPR code to...
11842         (adjust_symbolic_bound): ...here,
11843         (combine_bound): ...here,
11844         (set_value_range_with_overflow): ...and here.
11846 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11848         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11849         code...
11850         (extract_range_from_abs_expr): ...here.
11852 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
11854         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11855         -fno-omit-frame-pointer when not optimizing.
11857 2018-07-02  Martin Liska  <mliska@suse.cz>
11859         PR ipa/86279
11860         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11861         (propagate_nothrow): Likewise.
11863 2018-07-02  Martin Liska  <mliska@suse.cz>
11865         PR ipa/86323
11866         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11868 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11870         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11871         function in r262149, changing "loc" param from source_location to
11872         const dump_location_t &.
11873         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11874         declaration, as above.
11876 2018-07-01  Paul Koning  <ni1d@arrl.net>
11878         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11879         -munit-asm, -mgnu-asm, -mdec-asm.
11880         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11881         (pdp11_output_labelref): New.
11882         (pdp11_output_def): New.
11883         (pdp11_output_addr_vec_elt): New.
11884         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
11885         %# and %@ format codes.
11886         (pdp11_option_override): New.
11887         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11888         (pdp11_output_ident): New.
11889         (pdp11_asm_named_section): New.
11890         (pdp11_asm_init_sections): New.
11891         (pdp11_file_start): New.
11892         (pdp11_file_end): New.
11893         (output_ascii): Use .ascii/.asciz for -mdec-asm.
11894         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
11895         %o, like %c but octal.
11896         (pdp11_option_override): New.
11897         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11898         -mdec-asm.
11899         (DATA_SECTION_ASM_OP): Ditto.
11900         (READONLY_DATA_SECTION_ASM_OP): New.
11901         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11902         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11903         (ASM_OUTPUT_LABELREF): Ditto.
11904         (ASM_OUTPUT_DEF): Ditto.
11905         (ASM_OUTPUT_EXTERNAL): New.
11906         (ASM_OUTPUT_SOURCE_FILENAME): New.
11907         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11908         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11909         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
11910         %# and %@ format codes.
11911         * config/pdp11/pdp11.opt (mgnu-asm): New.
11912         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11913         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11914         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11916 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
11918         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11919         dereferencing path[] beyond its length.
11920         (debug_path): New.
11921         (debug_all_paths): New.
11922         (rewire_first_differing_edge): New.
11923         (adjust_paths_after_duplication): New.
11924         (duplicate_thread_path): Call adjust_paths_after_duplication.
11925         Add new argument.
11926         (thread_through_all_blocks): Add new argument to
11927         duplicate_thread_path.
11929 2018-06-30  Jim Wilson  <jimw@sifive.com>
11931         * config/riscv/predicates.md (p2m1_shift_operand): New.
11932         (high_mask_shift_operand): New.
11933         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11934         pattern using p2m1_shift_operand.
11935         (lshsi3_zero_extend_3+2): New combiner pattern using
11936         high_mask_shift_operand.
11938 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11940         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11941         split out from...
11942         (vect_recog_rotate_pattern): ...here.
11943         (vect_convert_input): Try to insert casts of invariants in the
11944         preheader.
11945         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11946         preheader to be empty.
11948 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11950         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11951         vector type.  If given, install it in the new statement's
11952         STMT_VINFO_VECTYPE.
11953         (vect_element_precision): New function.
11954         (vect_unpromoted_value): New struct.
11955         (vect_unpromoted_value::vect_unpromoted_value): New function.
11956         (vect_unpromoted_value::set_op): Likewise.
11957         (vect_look_through_possible_promotion): Likewise.
11958         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11959         (vect_widened_op_tree, vect_convert_input): Likewise.
11960         (vect_convert_inputs, vect_convert_output): Likewise.
11961         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11962         to handle the optional cast of the multiplication result and
11963         vect_widened_op_tree to detect the widened multiplication itself.
11964         Do not require the input and output of promotion casts to have
11965         the same sign, but base the signedness of the operation on the
11966         input rather than the result.  If the pattern includes two
11967         promotions, check that those promotions have the same sign.
11968         Do not restrict the MULT_EXPR handling to a double-width result;
11969         handle quadruple-width results and wider.  Use vect_convert_inputs
11970         to convert the inputs to the common type.
11971         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
11972         to handle the optional cast of the ABS result.  Also allow a sign
11973         change or a sign extension between the ABS and MINUS.
11974         Use vect_widened_op_tree to detect the widened subtraction and use
11975         vect_convert_inputs to convert the inputs to the common type.
11976         (vect_handle_widen_op_by_const): Delete.
11977         (vect_recog_widen_op_pattern): New function.
11978         (vect_recog_widen_mult_pattern): Use it.
11979         (vect_recog_widen_shift_pattern): Likewise.
11980         (vect_recog_widen_sum_pattern): Use
11981         vect_look_through_possible_promotion to handle the promoted
11982         PLUS_EXPR operand.
11984 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11986         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11987         the containing gimple_seq *.
11988         * gimple-iterator.h (gsi_for_stmt): Declare it.
11989         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11990         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11991         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11992         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11993         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11994         checks.
11995         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11996         split out from...
11997         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
11998         statement being replaced is part of an existing pattern
11999         definition sequence, inserting the new pattern statements before
12000         the original one.
12001         (vect_pattern_recog_1): Don't return a bool.  If the statement
12002         is already part of a pattern, instead apply pattern matching
12003         to the pattern definition statements.  Don't clear the
12004         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
12005         (vect_pattern_recog): Don't break after the first match;
12006         continue processing the pattern definition statements instead.
12007         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
12009 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12011         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12012         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12013         (vect_recog_widen_sum_pattern): Use it.
12015 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12017         * tree-vect-loop.c (vectorizable_reduction): Assert that the
12018         phi is not a pattern statement and has not been replaced by
12019         a pattern statement.
12020         * tree-vect-patterns.c (type_conversion_p): Don't check
12021         STMT_VINFO_IN_PATTERN_P.
12022         (vect_recog_vector_vector_shift_pattern): Likewise.
12023         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
12024         the pattern statement rather than the original statement; check
12025         directly for a WIDEN_MULT_EXPR here.
12026         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
12027         vect_is_simple_use to return the pattern statement rather
12028         than the original statement; use is_pattern_stmt_p to check
12029         for such a pattern statement.
12030         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
12031         to return the pattern statement rather than the original statement;
12032         don't do the same transformation here.
12033         (vect_is_simple_use): If the defining statement has been replaced
12034         by a pattern statement, return the pattern statement instead.
12035         Remove the corresponding (local) transformation from the vectype
12036         overload.
12038 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12040         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
12041         end and default to null.
12042         * tree-vect-loop.c (vect_create_epilog_for_reduction)
12043         (vectorizable_reduction): Update calls accordingly, dropping the
12044         gimple ** argument if the passed-back statement isn't needed.
12045         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
12046         (vect_recog_rotate_pattern): Likewise.
12047         (vect_recog_mask_conversion_pattern): Likewise.
12048         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
12049         (vect_mask_constant_operand_p): Likewise.
12050         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
12051         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
12052         (get_group_load_store_type, get_load_store_type): Likewise.
12053         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
12054         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
12055         (vectorizable_conversion, vectorizable_assignment): Likewise.
12056         (vectorizable_shift, vectorizable_operation): Likewise.
12057         (vectorizable_store, vect_is_simple_cond): Likewise.
12058         (vectorizable_condition, vectorizable_comparison): Likewise.
12059         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
12060         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
12061         and move it to the end.  Cope with null def_stmt_outs.
12063 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12065         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
12067 2018-06-29  Jeff Law  <law@redhat.com>
12069         * config/v850/v850.c (v850_legitimate_address_p): Handle large
12070         displacements for TARGET_V850E2V3 and newer.
12071         (TARGET_LRA_P): Remove.  Defaults to LRA now.
12072         * config/v850/v850.md (sign23byte_load): Remove.
12073         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
12074         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
12076 2018-06-29  Martin Liska  <mliska@suse.cz>
12078         PR lto/85759
12079         * coverage.c (coverage_init): Mangle full path name.
12080         * doc/invoke.texi: Document the change.
12081         * gcov-io.c (mangle_path): New.
12082         * gcov-io.h (mangle_path): Likewise.
12083         * gcov.c (mangle_name): Use mangle_path for path mangling.
12085 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12087         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
12088         if starting source register is not even.
12090 2018-06-29  Martin Liska  <mliska@suse.cz>
12092         PR tree-optimization/86263
12093         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
12094         Make edge redirection.
12096 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
12098         * dumpfile.c (dump_loc): Add indentation based on scope depth.
12099         (dump_scope_depth): New variable.
12100         (get_dump_scope_depth): New function.
12101         (dump_begin_scope): New function.
12102         (dump_end_scope): New function.
12103         * dumpfile.h (get_dump_scope_depth): New declaration.
12104         (dump_begin_scope): New declaration.
12105         (dump_end_scope): New declaration.
12106         (class auto_dump_scope): New class.
12107         (AUTO_DUMP_SCOPE): New macro.
12108         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
12109         AUTO_DUMP_SCOPE.
12111 2018-06-29  Richard Biener  <rguenther@suse.de>
12113         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
12114         compute_all_dependences succeeds.
12115         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
12116         exceed --param loop-max-datarefs-for-datadeps.
12118 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
12120         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
12122 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
12124         PR target/86348
12125         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
12126         alternative 0 in preferred_for_speed attribute.
12128 2018-06-28  Paul Koning  <ni1d@arrl.net>
12130         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
12131         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
12132         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
12133         * config/pdp11/pdp11.md: Correct "length" attribute calculation
12134         for shift insn patterns.
12136 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
12138         * cgraph.c (cgraph_node::get_body): Replace assignments to
12139         "dump_file" with calls to set_dump_file.
12140         * dumpfile.c (alt_dump_file): Make static, and group with...
12141         (alt_flags): ...this definition.
12142         (dumps_are_enabled): New variable.
12143         (refresh_dumps_are_enabled): New function.
12144         (set_dump_file): New function.
12145         (set_alt_dump_file): New function.
12146         (gcc::dump_manager::dump_start): Replace assignments to
12147         "dump_file" and "alt_dump_file" with calls to set_dump_file and
12148         set_alt_dump_file.
12149         (gcc::dump_manager::dump_finish): Likewise.
12150         * dumpfile.h (alt_dump_file): Delete decl.
12151         (dumps_are_enabled): New variable decl.
12152         (set_dump_file): New function decl.
12153         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
12154         global.
12155         * tree-nested.c (lower_nested_functions): Replace assignments to
12156         "dump_file" with calls to set_dump_file.
12158 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
12160         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
12161         goto_locus of each outgoing edge of each basic block.
12163 2018-06-28  Richard Biener  <rguenther@suse.de>
12165         * dwarf2out.c (decl_scope_table): Remove.
12166         (push_decl_scope): Likewise.
12167         (pop_decl_scope): Likewise.
12168         (gen_type_die_for_member): Do not call push/pop_decl_scope.
12169         (gen_struct_or_union_type_die): Likewise.
12170         (gen_tagged_type_die): Likewise.
12171         (dwarf2out_init): Do not initialize decl_scope_table.
12172         (dwarf2out_c_finalize): Do not free it.
12174 2018-06-28  Richard Biener  <rguenther@suse.de>
12176         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
12177         deciding whether to not re-use a DIE.
12179 2018-06-28  Richard Biener  <rguenther@suse.de>
12181         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
12182         DW_AT_abstract_origin attribute.
12184 2018-06-28  Martin Liska  <mliska@suse.cz>
12186         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12187         Use newly introduced constants.
12188         * tree-switch-conversion.h (struct jump_table_cluster):
12189         Define max_ratio_for_size and max_ratio_for_speed.
12191 2018-06-28  Martin Liska  <mliska@suse.cz>
12193         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12194         Add new checking assert to catch invalid state.
12195         (jump_table_cluster::can_be_handled): Handle single case
12196         clusters.
12197         (jump_table_cluster::is_beneficial): Bail out for such case.
12198         (bit_test_cluster::find_bit_tests):
12199         Add new checking assert to catch invalid state.
12200         (bit_test_cluster::can_be_handled): Handle single case
12201         clusters.
12202         (bit_test_cluster::is_beneficial): Bail out for such case.
12203         (switch_decision_tree::analyze_switch_statement):
12204         Fix comment.
12206 2018-06-28  Martin Liska  <mliska@suse.cz>
12208         * common.opt: Introduce -completion option.
12209         * gcc.c (driver_handle_option): Handle it.
12210         (driver::main): Print completions if completion
12211         is set.
12212         * opt-suggestions.c (option_proposer::get_completions):
12213         New function.
12214         (option_proposer::suggest_completion): Likewise.
12215         (option_proposer::find_param_completions): Likewise.
12216         (verify_autocompletions): Likewise.
12217         (test_completion_valid_options): Likewise.
12218         (test_completion_valid_params): Likewise.
12219         (in_completion_p): Likewise.
12220         (empty_completion_p): Likewise.
12221         (test_completion_partial_match): Likewise.
12222         (test_completion_garbage): Likewise.
12223         (opt_proposer_c_tests): Likewise.
12224         * opt-suggestions.h: Declare new functions.
12225         * opts.c (common_handle_option): Handle OPT__completion_.
12226         * selftest-run-tests.c (selftest::run_tests): Add
12227         opt_proposer_c_tests.
12228         * selftest.c (assert_str_startswith): New.
12229         * selftest.h (assert_str_startswith): Likewise.
12230         (opt_proposer_c_tests): New.
12231         (ASSERT_STR_STARTSWITH): Likewise.
12233 2018-06-28  Martin Liska  <mliska@suse.cz>
12235         * Makefile.in: Add opt-suggestions.o.
12236         * gcc-main.c: Include opt-suggestions.h.
12237         * gcc.c (driver::driver): Likewise.
12238         (driver::~driver): Remove m_option_suggestions.
12239         (driver::build_option_suggestions): Moved to option_proposer.
12240         (driver::suggest_option): Likewise.
12241         (driver::handle_unrecognized_options): Use option_proposer.
12242         * gcc.h (class driver): Add new memver m_option_proposer.
12243         * opt-suggestions.c: New file.
12244         * opt-suggestions.h: New file.
12246 2018-06-28  Martin Liska  <mliska@suse.cz>
12248         * vec.h (class auto_string_vec): New (moved from auto_argvec).
12249         (auto_string_vec::~auto_string_vec): Likewise.
12251 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
12253         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
12254         prevent_decl_creation_for_types fields up and add reset_location field.
12255         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
12256         statement if id->reset_location is true.
12257         (copy_edges_for_bb): Do not set goto_locus on the new edges if
12258         id->reset_location is true.
12259         (copy_phis_for_bb): Force input_location on the arguments if
12260         id->reset_location is true.
12261         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
12262         is set on the function to be inlined.
12264 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
12266         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
12268 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
12270         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
12271         registers for Pmode.
12272         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
12273         hard registers for the clobbered pseudo.
12275 2018-06-27  Paul Koning  <ni1d@arrl.net>
12277         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12278         mutually exclusive options.
12279         * config/pdp11/constraints.md (h): New constraint.
12280         (O): Update definition to match shift code generation.
12281         (D): New constraint.
12282         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
12283         (CCFP): Remove.
12284         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
12285         function.
12286         (output_jump): Change arguments.
12287         (pdp11_fixed_cc_regs): New function.
12288         (pdp11_cc_mode): Ditto.
12289         (pdp11_expand_shift): Ditto.
12290         (pdp11_assemble_shift): Ditto.
12291         (pdp11_small_shift): Ditto.
12292         (pdp11_branch_cost): Remove.
12293         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
12294         from output.
12295         (pdp11_register_move_cost): Update for CC registers.
12296         (pdp11_rtx_costs): Add case for LSHIFTRT.
12297         (pdp11_output_jump): Add CCNZ mode conditional branches.
12298         (notice_update_cc_on_set): Remove.
12299         (pdp11_cc_mode): New function.
12300         (simple_memory_operand): Correct pre/post decrement case.
12301         (no_side_effect_operand): New function.
12302         (pdp11_regno_reg_class): Add CC_REGS class.
12303         (pdp11_fixed_cc_regs): New function.
12304         (pdp11_small_shift): New function.
12305         (pdp11_expand_shift): New function to expand shift insns.
12306         (pdp11_assemble_shift): New function to output shifts.
12307         (pdp11_branch_cost): Remove.
12308         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
12309         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
12310         (WCHAR_TYPE): Ditto.
12311         (PTRDIFF_TYPE): Ditto.
12312         (ADJUST_INSN_LENGTH): New macro.
12313         (FIXED_REGISTERS): Add CC registers.
12314         (CALL_USED_REGISTERS): Ditto.
12315         (reg_class): Ditto.
12316         (REG_CLASS_NAMES): Ditto.
12317         (REG_CLASS_CONTENTS): Ditto.
12318         (SELECT_CC_MODE): Use new function.
12319         (TARGET_FLAGS_REGNUM): New macro.
12320         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
12321         (cc0_reg_rtx): Remove.
12322         (CC_STATUS_MDEP): Remove.
12323         (CC_STATUS_MDEFP_INIT): Remove.
12324         (CC_IN_FPU): Remove.
12325         (NOTICE_UPDATE_CC): Remove.
12326         (REGISTER_NAMES): Add CC registers.
12327         (BRANCH_COST): Change to constant 1.
12328         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
12329         handling.
12330         * config/pdp11/pdp11.opt (mbcopy): Remove.
12331         (mbcopy-builtin): Remove.
12332         (mbranch-cheap): Remove.
12333         (mbranch-expensive): Remove.
12334         * config/pdp11/predicates.md (expand_shift_operand): Update to
12335         match shift code generation.
12336         (ccnz_operator): New predicate.
12337         * doc/invoke.texi (PDP-11 Options): Remove deleted options
12338         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
12339         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
12340         exclusive options.
12341         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
12342         description of O constraint.
12344 2018-06-27  Jeff Law  <law@redhat.com>
12345             Austin Law  <austinklaw@gmail.com>
12347         * config/v850/v850.md (addsi3_set_flags): New pattern.
12348         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
12349         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
12350         (zero_extendhisi2_v850_set_flags): Likewise.
12351         (zero_extendqisi2_v850_set_flags): Likewise.
12352         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
12353         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
12354         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
12356         * config/v850/v850-protos.h (notice_update_cc): Remove.
12357         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
12358         (v850_print_operand): Handle 'D' and "d".
12359         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
12360         Add handling of arithmetic/logical operations compared against zero.
12361         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
12362         Do not look at v850_compare_op, instead get mode from last argument.
12363         (v850_gen_compare): Remove
12364         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
12365         after reload for prologue insns.
12366         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
12367         patterns.
12368         (construct_save_jarl): Likewise.
12369         (TARGET_FLAGS_REGNUM): Define.
12370         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
12371         (NOTICE_UPDATE_CC): Remove.
12372         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
12373         than cc0.  Conditionalize on reload_completed.
12374         (cmpsi_insn, setfcc_insn): Likewise.
12375         (tst1 splitter): Turn into define_and_split which sets the flags
12376         after reload.
12377         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
12378         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
12379         (cstoresf4, cstoredf4): Clobber the flags.
12380         (cmpsi, cmpsf, cmpdf): Remove expanders.
12381         (setf_insn): Remove pattern.
12382         (addsi3): Turn into define_and_split which clobbers the flags after
12383         reload and a suitable pattern (addsi3_clobber_flags) for use after
12384         reload.
12385         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
12386         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
12387         (ashrsi3, ashrsi3_v850e2): Likewise.
12388         (bins): Clobber the flags.
12389         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
12390         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
12391         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
12392         (call_value_internal_short, call_value_internal_long): Likewise.
12393         (callt_save_interrupt, callt_return_interrupt): Likewise.
12394         (save_interrupt, return_interrupt): Likewise.
12395         (callt_save_all_interrupt, save_all_interrupt): Likewise.
12396         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
12397         (restore_all_interrupt, _restore_all_interrupt): Likewise.
12398         (All FP comparisons): Only allow after reload has completed.
12399         (trfsr): Likewise.
12400         (divh, divhu): Tweak output template.
12401         (branch_z_normal, branch_z_invert): Remove
12402         (branch_nz_normal, branch_nz_invert): Likewise.
12403         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
12405         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
12406         * config/v850/v850.c (notice_update_cc): Remove.
12407         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
12408         (CC_NO_CARRY): Likewise.
12409         (NOTICE_UPDATE_CC): Define to nothing.
12410         * config/v850/v850.md: Remove block comment on cc0 handling
12411         Remove "cc" attribute from all patterns.  Remove cc_status handling
12412         from all patterns.  Minor formatting fixes.
12414 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12416         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
12417         (cortex-a76.cortex-a55): Likewise.
12418         * config/aarch64/aarch64-tune.md: Regenerate.
12419         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
12420         cortex-a76.cortex-a55.
12422 2018-06-27  Jeff Law  <law@redhat.com>
12424         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
12425         (MULTILIB_DIRNAMES): Similarly.
12427 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
12429         * gimple.h (gimple_return_retbnd): Delete.
12430         (gimple_return_set_retbnd): Likewise.
12431         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
12432         gimple_return_set_retbnd.
12433         * gimple-pretty-print.c (dump_gimple_return): Remove call to
12434         gimple_return_retbnd and adjust.
12435         * tree-inline.h (struct copy_body_data): Remove retbnd field.
12436         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
12437         Explicitly return NULL in a couple more cases.  Move assertion
12438         on debug statements and remove unreachable code.
12439         (reset_debug_binding): Do not test id->retbnd.
12440         (expand_call_inline): Do not set it.
12442 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
12444         * configure.ac: Add --disable-gcov option.
12445         * configure: Regenerate.
12446         * Makefile.in: Honour @enable_gcov@.
12447         * doc/install.texi: Document --disable-gcov.
12449 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12451         * config/arm/arm-cpus.in (cortex-a76): New entry.
12452         (cortex-a76.cortex-a55): Likewise.
12453         * config/arm/arm-tables.opt: Regenerate.
12454         * config/arm/arm-tune.md: Likewise.
12455         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
12456         * doc/invoke.texi (ARM Options): Document cortex-a76 and
12457         cortex-a76.cortex-a55.
12459 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
12461         PR target/85769
12462         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
12464 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
12466         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
12467         comment.
12468         (EPILOGUE_USES): Likewise.
12470 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
12472         * tree-inline.c (remap_location): New function extracted from...
12473         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
12474         (copy_phis_for_bb): ...here.  Call remap_location.
12475         (copy_cfg_body): Adjust call to copy_edges_for_bb.
12477 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
12479         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
12480         unaligned vsx for 16B memset.
12482 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
12484         PR target/86285
12485         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
12486         ieee128_float_type_node to long_double_type_node unless
12487         TARGET_LONG_DOUBLE_128 is set.
12489 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
12491         * cfgloop.c (get_loop_location): Convert return type from
12492         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
12493         by implicit construction from rtx_insn *, and using
12494         dump_user_location_t::from_function_decl for the fallback case.
12495         * cfgloop.h (get_loop_location): Convert return type from
12496         location_t to dump_user_location_t.
12497         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
12498         dump_printf_loc to pass in a dump_location_t rather than a
12499         location_t, via the gimple stmt.
12500         * coverage.c (get_coverage_counts): Update calls to
12501         dump_printf_loc to pass in dump_location_t rather than a
12502         location_t.
12503         * doc/optinfo.texi (Dump types): Convert example of
12504         dump_printf_loc from taking "locus" to taking "insn".  Update
12505         description of the "_loc" calls to cover dump_location_t.
12506         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
12507         "selftest.h".
12508         (dump_user_location_t::dump_user_location_t): New constructors,
12509         from gimple *stmt and rtx_insn *.
12510         (dump_user_location_t::from_function_decl): New function.
12511         (dump_loc): Make static.
12512         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
12513         const dump_location_t &.
12514         (dump_generic_expr_loc): Delete.
12515         (dump_printf_loc): Convert param "loc" from location_t to
12516         const dump_location_t &.
12517         (selftest::test_impl_location): New function.
12518         (selftest::dumpfile_c_tests): New function.
12519         * dumpfile.h: Include "profile-count.h".
12520         (class dump_user_location_t): New class.
12521         (struct dump_impl_location_t): New struct.
12522         (class dump_location_t): New class.
12523         (dump_printf_loc): Convert 2nd param from source_location to
12524         const dump_location_t &.
12525         (dump_generic_expr_loc): Delete.
12526         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
12527         const dump_location_t &.
12528         * gimple-fold.c (fold_gimple_assign): Update call to
12529         dump_printf_loc to pass in a dump_location_t rather than a
12530         location_t, via the gimple stmt.
12531         (gimple_fold_call): Likewise.
12532         * gimple-loop-interchange.cc
12533         (loop_cand::analyze_iloop_reduction_var): Update for change to
12534         check_reduction_path.
12535         (tree_loop_interchange::interchange): Update for change to
12536         find_loop_location.
12537         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
12538         change in return-type of find_loop_location.
12539         (graphite_regenerate_ast_isl): Likewise.
12540         * graphite-optimize-isl.c (optimize_isl): Likewise.
12541         * graphite.c (graphite_transform_loops): Likewise.
12542         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
12543         pass in a dump_location_t rather than a location_t, via the
12544         gimple stmt.
12545         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12546         * ipa.c (walk_polymorphic_call_targets): Likewise.
12547         * loop-unroll.c (report_unroll): Convert "locus" param from
12548         location_t to dump_location_t.
12549         (decide_unrolling): Update for change to get_loop_location's
12550         return type.
12551         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
12552         location_t to dump_user_location_t.
12553         (grid_find_single_omp_among_assignments_1): Updates calls to
12554         dump_printf_loc to pass in a dump_location_t rather than a
12555         location_t, via the gimple stmt.
12556         (grid_parallel_clauses_gridifiable): Convert "tloc" from
12557         location_t to dump_location_t.  Updates calls to dump_printf_loc
12558         to pass in a dump_location_t rather than a location_t, via the
12559         gimple stmt.
12560         (grid_inner_loop_gridifiable_p): Likewise.
12561         (grid_dist_follows_simple_pattern): Likewise.
12562         (grid_gfor_follows_tiling_pattern): Likewise.
12563         (grid_target_follows_gridifiable_pattern): Likewise.
12564         (grid_attempt_target_gridification): Convert initialization
12565         of local "grid" from memset to zero-initialization; FIXME: does
12566         this require C++11?  Update call to dump_printf_loc to pass in a
12567         optinfo_location rather than a location_t, via the gimple stmt.
12568         * profile.c (read_profile_edge_counts): Updates call to
12569         dump_printf_loc to pass in a dump_location_t rather than a
12570         location_t
12571         (compute_branch_probabilities): Likewise.
12572         * selftest-run-tests.c (selftest::run_tests): Call
12573         dumpfile_c_tests.
12574         * selftest.h (dumpfile_c_tests): New decl.
12575         * tree-loop-distribution.c (pass_loop_distribution::execute):
12576         Update for change in return type of find_loop_location.
12577         * tree-parloops.c (parallelize_loops): Likewise.
12578         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
12579         "locus" from location_t to dump_user_location_t.
12580         (canonicalize_loop_induction_variables): Likewise.
12581         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
12582         for change in return type of find_loop_location.
12583         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
12584         to dump_printf_loc to pass in a dump_location_t rather than a
12585         location_t, via the stmt.
12586         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
12587         Likewise.
12588         * tree-vect-loop-manip.c (find_loop_location): Convert return
12589         type from source_location to dump_user_location_t.
12590         (vect_do_peeling): Update for above change.
12591         (vect_loop_versioning): Update for change in type of
12592         vect_location.
12593         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
12594         from location_t to dump_user_location_t.
12595         (vect_estimate_min_profitable_iters): Update for change in type
12596         of vect_location.
12597         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
12598         location_t to dump_location_t.
12599         (vect_slp_bb): Update for change in type of vect_location.
12600         * tree-vectorizer.c (vect_location): Convert from source_location
12601         to dump_user_location_t.
12602         (try_vectorize_loop_1): Update for change in vect_location's type.
12603         (vectorize_loops): Likewise.
12604         (increase_alignment): Likewise.
12605         * tree-vectorizer.h (vect_location): Convert from source_location
12606         to dump_user_location_t.
12607         (find_loop_location): Convert return type from source_location to
12608         dump_user_location_t.
12609         (check_reduction_path): Convert 1st param from location_t to
12610         dump_user_location_t.
12611         * value-prof.c (check_counter): Update call to dump_printf_loc to
12612         pass in a dump_user_location_t rather than a location_t; update
12613         call to error_at for change in type of "locus".
12614         (check_ic_target): Update call to dump_printf_loc to
12615         pass in a dump_user_location_t rather than a location_t, via the
12616         call_stmt.
12618 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12620         * config/s390/s390.h (enum processor_flags): Do not use
12621         default tune parameter when -march was specified.
12623 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
12625         PR target/86314
12626         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12627         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12628         operands.
12630 2018-06-26  Richard Biener  <rguenther@suse.de>
12632         PR tree-optimization/86287
12633         PR bootstrap/86316
12634         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12635         (vect_analyze_loop): Initialize n_stmts.
12637 2018-06-26  Richard Biener  <rguenther@suse.de>
12639         PR middle-end/86271
12640         * fold-const.c (fold_convertible_p): Pointer extension
12641         isn't valid.
12643 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
12645         PR debug/86064
12646         * dwarf2out.c (loc_list_has_views): Adjust comments.
12647         (dw_loc_list): Split single cross-partition range with
12648         nonzero locview.
12650 2018-06-25  Jeff Law  <law@redhat.com>
12652         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12653         on -mbig-switch by default.
12655         * config/v850/predicates.md (const_float_1_operand): Fix match_code
12656         test.
12657         (const_float_0_operand): Remove unused predicate.
12658         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12659         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12660         (recipsf2): New expander.  Original pattern now called
12661         (recipsf2_insn).
12662         (recipdf2, recipdf2_insn): Similarly.
12663         (rsqrtsf2, rsqrtsf2_insn): Similarly
12664         (rsqrtdf2, rsqrtdf2_insn): Similarly
12666 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
12668         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12669         Simplify logic for FreeBSD (twice).
12671 2018-06-25  Martin Sebor  <msebor@redhat.com>
12673         PR tree-optimization/86204
12674         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12675         a strnlen result if it's less than the length of the string.
12677 2018-06-25  Martin Sebor  <msebor@redhat.com>
12679         PR tree-optimization/85700
12680         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12681         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12682         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12684 2018-06-25  Martin Sebor  <msebor@redhat.com>
12686         * doc/extend.texi (Zero-length arrays): Update and clarify.
12688 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
12690         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12691         added IEEE/IBM long double multilib support on PowerPC little
12692         endian Linux systems.
12693         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12694         (MULTILIB_DEFAULTS): Likewise.
12695         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12696         Likewise.
12697         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12698         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12699         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12701 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
12703         PR middle-end/86311
12704         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12705         (REORDER_45): Likewise.
12707 2018-06-25  Jeff Law  <law@redhat.com>
12709         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12710         dividend to 32 bits.  Adjust length.
12711         (udivmodhi4): Cleanup output template.  Fix length.
12713 2018-06-25  Carl Love  <cel@us.ibm.com>
12715         * config/rs6000/vsx.md: Change word selector to prefered location.
12717 2018-06-25  Richard Biener  <rguenther@suse.de>
12719         PR tree-optimization/86304
12720         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12721         epilogue-if-converted loops as well.
12723 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
12725         * lto-section-out.c (lto_begin_section): Do not print section
12726         name for noaddr and unnumbered dumps.
12728 2018-06-25  Richard Biener  <rguenther@suse.de>
12730         * tree-vectorizer.h (struct vec_info_shared): New structure
12731         with parts split out from struct vec_info and loop_nest from
12732         struct _loop_vec_info.
12733         (struct vec_info): Adjust accordingly.
12734         (struct _loop_vec_info): Likewise.
12735         (LOOP_VINFO_LOOP_NEST): Adjust.
12736         (LOOP_VINFO_DATAREFS): Likewise.
12737         (LOOP_VINFO_DDRS): Likewise.
12738         (struct _bb_vec_info): Likewise.
12739         (BB_VINFO_DATAREFS): Likewise.
12740         (BB_VINFO_DDRS): Likewise.
12741         (struct _stmt_vec_info): Add dr_aux member.
12742         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12743         (DR_MISALIGNMENT_UNINITIALIZED): New.
12744         (set_dr_misalignment): Adjust.
12745         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12746         (vect_analyze_loop): Adjust prototype.
12747         (vect_analyze_loop_form): Likewise.
12748         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12749         Compute dependences lazily.
12750         (vect_record_base_alignments): Use shared datarefs/ddrs.
12751         (vect_verify_datarefs_alignment): Likewise.
12752         (vect_analyze_data_refs_alignment): Likewise.
12753         (vect_analyze_data_ref_accesses): Likewise.
12754         (vect_analyze_data_refs): Likewise.
12755         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12756         constructor parameter for shared part.
12757         (vect_analyze_loop_form): Pass in shared part and adjust.
12758         (vect_analyze_loop_2): Pass in storage for the number of
12759         stmts.  Move loop nest finding to the caller.  Compute
12760         datarefs lazily.
12761         (vect_analyze_loop): Pass in shared part.
12762         (vect_transform_loop): Verify shared datarefs are unchanged.
12763         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12764         constructor parameter for shared part.
12765         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12766         (vect_slp_bb): Verify shared datarefs are unchanged before
12767         transform.
12768         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12769         change.
12770         (new_stmt_vec_info): Initialize DR_AUX misalignment to
12771         DR_MISALIGNMENT_UNINITIALIZED.
12772         * tree-vectorizer.c (vec_info::vec_info): Add constructor
12773         parameter for shared part.
12774         (vec_info::~vec_info): Adjust.
12775         (vec_info_shared::vec_info_shared): New.
12776         (vec_info_shared::~vec_info_shared): Likewise.
12777         (vec_info_shared::save_datarefs): Likewise.
12778         (vec_info_shared::check_datarefs): Likewise.
12779         (try_vectorize_loop_1): Construct shared part live for analyses
12780         of a single loop for multiple vector sizes.
12781         * tree-parloops.c (gather_scalar_reductions): Adjust.
12783 2018-06-25  Richard Biener  <rguenther@suse.de>
12785         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12786         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12787         (vect_analyze_data_refs): Remove similar code from here and
12788         simplify accordingly.
12790 2018-06-25  Richard Biener  <rguenther@suse.de>
12792         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12793         for reverse storage order accesses rather than asserting
12794         they cannot happen here.
12796 2018-06-25  Tom de Vries  <tdevries@suse.de>
12798         PR debug/86257
12799         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12800         Use data16 instead of .byte for insn prefix.
12802 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
12804         PR C++/86082
12805         * parser.c (make_char_string_pack): Pass this literal chars
12806         through cpp_interpret_string.
12807         (cp_parser_userdef_numeric_literal): Check the result of
12808         make_char_string_pack.
12810 2018-06-24  Maya Rashish  <coypu@sdf.org>
12812         * ginclude/stddef.h: Simplify conditions around avoiding
12813         re-definition of __size_t.
12815 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12817         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12818         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12820 2018-06-22  Maya Rashish  <coypu@sdf.org>
12822         * doc/invoke.texi (mno-fancy-math-387): Update for changes
12823         made to OpenBSD and NetBSD through the years.
12825 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12827         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12828         behavior of vec_pack (vector double, vector double) to match
12829         behavior of vec_float2 (vector double, vector double).
12831 2018-06-22  Olivier Hainque  <hainque@adacore.com>
12833         * gimplify.c (gimplify_function_tree): Prevent creation
12834         of a trampoline for the address of the current function
12835         passed to entry/exit instrumentation hooks.
12837 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
12839         PR target/86222
12840         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12841         correctly.
12843 2018-06-22  Martin Liska  <mliska@suse.cz>
12845         PR tree-optimization/86263
12846         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12847         Bail out if is_enabled is false.
12848         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12849         New declaration.
12850         (jump_table_cluster::is_enabled): New function.
12852 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12854         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12855         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12856         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12857         (lto_input_ts_binfo_tree_pointers): Likewise.
12858         * tree-streamer-out.c (streamer_write_tree_bitfields,
12859         write_ts_binfo_tree_pointers): Likewise.
12860         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12862 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12864         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12866 2018-06-22  Martin Liska  <mliska@suse.cz>
12868         * symbol-summary.h (get): Make it pure and inline move
12869         functionality from ::get function.
12870         (get): Remove and inline into ::get and ::get_create.
12871         (get_create): Move code from ::get function.
12873 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12875         PR target/85994
12876         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12877         -x assembler-with-cpp.
12879 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12881         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12882         _FILE_OFFSET_BITS=64 for C++.
12884 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
12886         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12887         conversion insn that shows up when pr85657-3.c is compiled using
12888         IEEE 128-bit long double.
12889         (neg<mode>2_internal): Use the correct mode to check whether the
12890         mode is IBM extended.
12891         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12892         multiply and divide external functions from being created more
12893         than once.
12895 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12897         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12898         functions.
12899         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
12900         the edge can be forwarded.
12901         (cfg_layout_merge_blocks): Likewise.
12903 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12905         * except.c (finish_eh_generation): Commit edge insertions only after
12906         the EH edges have been redirected from post-landing to landing pads.
12908 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12910         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12911         create_tmp_var_for to create the FRAME decl.
12912         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12914 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12916         * tree-inline.c (copy_edges_for_bb): Minor tweak.
12917         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12918         debug statement when resetting its value.
12919         (expand_call_inline): Copy the locus of the call onto the assignment
12920         of the return value, if any.  Use local variable in more cases.
12922 2018-06-21  Martin Liska  <mliska@suse.cz>
12924         * ipa-pure-const.c (propagate_nothrow): Use
12925         funct_state_summaries->get.
12926         (dump_malloc_lattice): Likewise.
12927         (propagate_malloc): Likewise.
12929 2018-06-21  Richard Biener  <rguenther@suse.de>
12931         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12932         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12933         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12934         comment.
12935         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12936         BLOCK_ABSTRACT_ORIGIN unconditionally.
12938 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
12940         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12941         deleting it.
12942         * ipa-reference.c (ipa_reference_c_finalize): Delete
12943         ipa_ref_opt_sum_summaries and set it to NULL.
12945 2018-06-21  Tom de Vries  <tdevries@suse.de>
12947         PR tree-optimization/85859
12948         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12949         test with comment from bb_no_side_effects_p.
12951 2018-06-21  Richard Biener  <rguenther@suse.de>
12953         PR tree-optimization/86232
12954         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12955         max for constant niter.
12957 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12959         * config/aarch64/aarch64-simd.md
12960         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12962 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12964         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12965         Make opernads of the unspec commutative.
12967 2018-06-21  Richard Biener  <rguenther@suse.de>
12969         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12970         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12971         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12972         (vect_analyze_data_ref_dependence): Re-order checks to deal with
12973         NULL DR_STEP.
12974         (vect_record_base_alignments): Do not record base alignment
12975         for gathers or scatters.
12976         (vect_compute_data_ref_alignment): Drop return value that is always
12977         true.  Bail out early for gathers or scatters.
12978         (vect_enhance_data_refs_alignment): Bail out early for gathers
12979         or scatters.
12980         (vect_find_same_alignment_drs): Likewise.
12981         (vect_analyze_data_refs_alignment): Remove dead code.
12982         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12983         (vect_analyze_data_refs): For possible gathers or scatters do
12984         not create an alternate DR, just check their possible validity
12985         and mark them.  Adjust DECL_NONALIASED handling to not rely
12986         on DR_BASE_ADDRESS.
12987         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12988         update inits of gathers or scatters.
12989         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12990         Also copy gather/scatter flag to pattern vinfo.
12992 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12994         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12995         behavior of vec_packsu (vector unsigned long long, vector unsigned
12996         long long) to match behavior of vec_packs with same signature.
12998 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
12999             Thomas Schwinge <thomas@codesourcery.com>
13000             Cesar Philippidis  <cesar@codesourcery.com>
13002         * gimplify.c (gimplify_scan_omp_clauses): Add support for
13003         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13004         (gimplify_adjust_omp_clauses): Likewise.
13005         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
13006         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
13007         (gimplify_omp_target_update): Update handling of acc update and
13008         enter/exit data.
13009         * omp-low.c (install_var_field): Remove unused parameter
13010         base_pointers_restrict.
13011         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
13012         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
13013         FINALIZE}
13014         (omp_target_base_pointers_restrict_p): Delete.
13015         (scan_omp_target): Update call to scan_sharing_clauses.
13016         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
13017         FINALIZE}.
13018         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
13019         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13020         (convert_local_omp_clauses): Likewise.
13021         * tree-pretty-print.c (dump_omp_clause): Likewise.
13022         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
13023         FINALIZE}.
13024         (omp_clause_code_name): Likewise.
13026 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
13028         PR debug/86194
13029         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
13030         be narrowed.
13032         PR tree-optimization/86231
13033         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
13034         anti-range don't overwrite *vr0min before using it to compute *vr0max.
13036 2018-06-20  Tom de Vries  <tdevries@suse.de>
13038         PR tree-optimization/86097
13039         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
13040         iv type if signedness of iv type is not the same as that of *nit.
13042 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
13044         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
13045         EDGE_EH edges, verify they are all EDGE_EH.
13047 2018-06-20  Maya Rashish  <coypu@sdf.org>
13049         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
13051 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13053         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
13054         * config/aarch64/aarch64.c (xgene1_tunings): Add
13055         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
13056         (aarch64_mode_valid_for_sched_fusion_p):
13057         Allow 16-byte modes.
13058         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
13059         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
13060         128-bit modes.
13061         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
13062         New pattern.
13063         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
13064         * config/aarch64/iterators.md (VQ2): New mode iterator.
13066 2018-06-20  Martin Liska  <mliska@suse.cz>
13068         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13069         Change default ratio from 10 to 8.
13071 2018-06-20  Martin Liska  <mliska@suse.cz>
13073         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13074         New.
13075         (bit_test_cluster::find_bit_tests): Likewise.
13076         (switch_decision_tree::analyze_switch_statement): Find clusters.
13077         * tree-switch-conversion.h (struct jump_table_cluster): Document
13078         hierarchy.
13080 2018-06-20  Martin Liska  <mliska@suse.cz>
13082         * tree-switch-conversion.c (switch_conversion::collect):
13083         Record m_uniq property.
13084         (switch_conversion::expand): Bail out for special conditions.
13085         (group_cluster::~group_cluster): New.
13086         (group_cluster::group_cluster): Likewise.
13087         (group_cluster::dump): Likewise.
13088         (jump_table_cluster::emit): New.
13089         (switch_decision_tree::fix_phi_operands_for_edges): New.
13090         (struct case_node): Remove struct.
13091         (jump_table_cluster::can_be_handled): New.
13092         (case_values_threshold): Moved to header.
13093         (reset_out_edges_aux): Likewise.
13094         (jump_table_cluster::is_beneficial): New.
13095         (bit_test_cluster::can_be_handled): Likewise.
13096         (add_case_node): Remove.
13097         (bit_test_cluster::is_beneficial): New.
13098         (case_bit_test::cmp): New.
13099         (bit_test_cluster::emit): New.
13100         (expand_switch_as_decision_tree_p): Remove.
13101         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
13102         (fix_phi_operands_for_edge): Likewise.
13103         (switch_decision_tree::analyze_switch_statement): New.
13104         (compute_cases_per_edge): Move ...
13105         (switch_decision_tree::compute_cases_per_edge): ... here.
13106         (try_switch_expansion): Likewise.
13107         (switch_decision_tree::try_switch_expansion): Likewise.
13108         (record_phi_operand_mapping): Likewise.
13109         (switch_decision_tree::record_phi_operand_mapping): Likewise.
13110         (emit_case_decision_tree): Likewise.
13111         (switch_decision_tree::emit): Likewise.
13112         (balance_case_nodes): Likewise.
13113         (switch_decision_tree::balance_case_nodes): Likewise.
13114         (dump_case_nodes): Likewise.
13115         (switch_decision_tree::dump_case_nodes): Likewise.
13116         (emit_jump): Likewise.
13117         (switch_decision_tree::emit_jump): Likewise.
13118         (emit_cmp_and_jump_insns): Likewise.
13119         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
13120         (emit_case_nodes): Likewise.
13121         (switch_decision_tree::emit_case_nodes): Likewise.
13122         (conditional_probability): Remove.
13123         * tree-switch-conversion.h (enum cluster_type): New.
13124         (PRINT_CASE): New.
13125         (struct cluster): Likewise.
13126         (cluster::cluster): Likewise.
13127         (struct simple_cluster): Likewise.
13128         (simple_cluster::simple_cluster): Likewise.
13129         (struct group_cluster): Likewise.
13130         (struct jump_table_cluster): Likewise.
13131         (struct bit_test_cluster): Likewise.
13132         (struct min_cluster_item): Likewise.
13133         (struct case_tree_node): Likewise.
13134         (case_tree_node::case_tree_node): Likewise.
13135         (jump_table_cluster::case_values_threshold): Likewise.
13136         (struct case_bit_test): Likewise.
13137         (struct switch_decision_tree): Likewise.
13138         (struct switch_conversion): Likewise.
13139         (switch_decision_tree::reset_out_edges_aux): Likewise.
13141 2018-06-20  Martin Liska  <mliska@suse.cz>
13143         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
13144         (hoist_edge_and_branch_if_true): Likewise.
13145         (expand_switch_using_bit_tests_p): Likewise.
13146         (struct case_bit_test): Likewise.
13147         (case_bit_test_cmp): Likewise.
13148         (emit_case_bit_tests): Likewise.
13149         (switch_conversion::switch_conversion): New class.
13150         (struct switch_conv_info): Remove old struct.
13151         (collect_switch_conv_info): More to ...
13152         (switch_conversion::collect): ... this.
13153         (check_range): Likewise.
13154         (switch_conversion::check_range): Likewise.
13155         (check_all_empty_except_final): Likewise.
13156         (switch_conversion::check_all_empty_except_final): Likewise.
13157         (check_final_bb): Likewise.
13158         (switch_conversion::check_final_bb): Likewise.
13159         (create_temp_arrays): Likewise.
13160         (switch_conversion::create_temp_arrays): Likewise.
13161         (free_temp_arrays): Likewise.
13162         (gather_default_values): Likewise.
13163         (switch_conversion::gather_default_values): Likewise.
13164         (build_constructors): Likewise.
13165         (switch_conversion::build_constructors): Likewise.
13166         (constructor_contains_same_values_p): Likewise.
13167         (switch_conversion::contains_same_values_p): Likewise.
13168         (array_value_type): Likewise.
13169         (switch_conversion::array_value_type): Likewise.
13170         (build_one_array): Likewise.
13171         (switch_conversion::build_one_array): Likewise.
13172         (build_arrays): Likewise.
13173         (switch_conversion::build_arrays): Likewise.
13174         (gen_def_assigns): Likewise.
13175         (switch_conversion::gen_def_assigns): Likewise.
13176         (prune_bbs): Likewise.
13177         (switch_conversion::prune_bbs): Likewise.
13178         (fix_phi_nodes): Likewise.
13179         (switch_conversion::fix_phi_nodes): Likewise.
13180         (gen_inbound_check): Likewise.
13181         (switch_conversion::gen_inbound_check): Likewise.
13182         (process_switch): Use the newly created class.
13183         (switch_conversion::expand): New.
13184         (switch_conversion::~switch_conversion): New.
13185         * tree-switch-conversion.h: New file.
13187 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13189         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
13190         tree-vect-patterns.c.
13191         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
13192         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
13193         (vect_recog_sad_pattern): Likewise.
13194         (vect_recog_widen_sum_pattern): Likewise.
13195         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
13196         (vect_recog_widen_shift_pattern): Remove the type_in argument.
13197         (vect_recog_rotate_pattern): Likewise.
13198         (vect_recog_mult_pattern): Likewise.
13199         (vect_recog_vector_vector_shift_pattern): Likewise.
13200         (vect_recog_divmod_pattern): Likewise.
13201         (vect_recog_mixed_size_cond_pattern): Likewise.
13202         (vect_recog_bool_pattern): Likewise.
13203         (vect_recog_mask_conversion_pattern): Likewise.
13204         (vect_try_gather_scatter_pattern): Likewise.
13205         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
13206         (vect_recog_over_widening_pattern): Likewise.
13207         (vect_recog_gather_scatter_pattern): Likewise.
13208         (vect_recog_func_ptr): Move from tree-vectorizer.h
13209         (vect_vect_recog_func_ptrs): Move further down the file.
13210         (vect_recog_func): Likewise.  Remove the third argument.
13211         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
13212         (vect_pattern_recog_1): Expect the pattern function to do any
13213         necessary target tests.  Also expect it to provide a vector type.
13214         Remove the type_in handling.
13216 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13218         * tree-vect-patterns.c (vect_pattern_detected): New function.
13219         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
13220         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
13221         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
13222         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
13223         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13224         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13225         (vect_recog_mask_conversion_pattern)
13226         (vect_try_gather_scatter_pattern): Likewise.
13228 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13230         * tree-vect-patterns.c (vect_get_internal_def): New function.
13231         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13232         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
13233         (search_type_for_mask_1): Use it.
13235 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13237         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13238         redundant WIDEN_SUM_EXPR handling.
13239         (vect_recog_sad_pattern): Likewise.
13241 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13243         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13244         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
13245         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
13246         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
13247         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
13249 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13251         * tree-vect-stmts.c (vectorizable_call): Make sure that we
13252         use the stmt_vec_info of the original bb statement for the
13253         new zero assignment, even if the call is part of a pattern.
13255 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13257         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
13258         that the sequence is attached to the original statement rather
13259         than the pattern statement.
13260         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
13261         PATTERN_DEF_SEQ from the original statement rather than
13262         the main pattern statement.
13263         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
13264         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
13265         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
13267 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13269         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
13270         definition statements before the early exit for statements that aren't
13271         live or relevant.
13272         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
13273         split out from...
13274         (vect_transform_loop): ...here.  Process pattern definition
13275         statements without first checking whether the main pattern
13276         statement is live or relevant.
13278 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13280         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
13281         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
13283 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
13285         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
13286         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
13287         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
13288         (expand_block_compare): Change select_block_compare_mode call.
13289         (expand_strncmp_align_check): Use new functions, fix comment.
13290         (emit_final_str_compare_gpr): New function.
13291         (expand_strn_compare): Refactor and clean up code.
13292         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
13294 2018-06-19  Tony Reix  <tony.reix@atos.com>
13295             Damien Bergamini  <damien.bergamini@atos.com>
13296             David Edelsohn  <dje.gcc@gmail.com>
13298         * collect2.c (static_obj): New variable.
13299         (static_libs): New variable.
13300         (is_in_list): Uncomment declaration.
13301         (main): Track AIX libraries linked statically.
13302         (is_in_list): Uncomment definition.
13303         (scan_prog_file): Don't add AIX shared libraries initializer
13304         to constructor list if linking statically.
13306 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
13308         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
13309         constant.
13310         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
13312 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
13314         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
13315         blocks.
13317 2018-06-19  Martin Liska  <mliska@suse.cz>
13319         * config/i386/i386.c (ix86_can_inline_p): Do not use
13320         ipa_fn_summaries::get_create.
13321         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
13322         get.
13323         (devirtualization_time_bonus): Likewise.
13324         (ipcp_propagate_stage): Likewise.
13325         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13326         (edge_set_predicate): Likewise.
13327         (evaluate_conditions_for_known_args): Likewise.
13328         (evaluate_properties_for_edge): Likewise.
13329         (ipa_call_summary::reset): Tranform to ...
13330         (ipa_call_summary::~ipa_call_summary): ... this.
13331         (ipa_fn_summary::reset): Transform to ...
13332         (ipa_fn_summary::~ipa_fn_summary): ... this.
13333         (ipa_fn_summary_t::remove): Rename to ...
13334         (ipa_fn_summary_t::remove_callees): ... this.
13335         (ipa_fn_summary_t::duplicate): Use placement new
13336         instead of memory copy.
13337         (ipa_call_summary_t::duplicate): Likewise.
13338         (ipa_call_summary_t::remove): Remove.
13339         (dump_ipa_call_summary): Change get_create to get.
13340         (ipa_dump_fn_summary): Dump only when summary exists.
13341         (analyze_function_body): Use symbol_summary::get instead
13342         of get_create.
13343         (compute_fn_summary): Likewise.
13344         (estimate_edge_devirt_benefit): Likewise.
13345         (estimate_edge_size_and_time): Likewise.
13346         (inline_update_callee_summaries): Likewise.
13347         (remap_edge_change_prob): Likewise.
13348         (remap_edge_summaries): Likewise.
13349         (ipa_merge_fn_summary_after_inlining): Likewise.
13350         (write_ipa_call_summary): Likewise.
13351         (ipa_fn_summary_write): Likewise.
13352         (ipa_free_fn_summary): Likewise.
13353         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
13354         (struct ipa_call_summary): Likewise.
13355         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
13356         of get_create.
13357         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
13358         (estimate_size_after_inlining): Likewise.
13359         (estimate_growth): Likewise.
13360         (growth_likely_positive): Likewise.
13361         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13362         (inline_call): Likewise.
13363         * ipa-inline.c (caller_growth_limits): Likewise.
13364         (can_inline_edge_p): Likewise.
13365         (can_inline_edge_by_limits_p): Likewise.
13366         (compute_uninlined_call_time): Likewise.
13367         (compute_inlined_call_time): Likewise.
13368         (want_inline_small_function_p): Likewise.
13369         (edge_badness): Likewise.
13370         (update_caller_keys): Likewise.
13371         (update_callee_keys): Likewise.
13372         (inline_small_functions): Likewise.
13373         (inline_to_all_callers_1): Likewise.
13374         (dump_overall_stats): Likewise.
13375         (early_inline_small_functions): Likewise.
13376         (early_inliner): Likewise.
13377         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13378         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13379         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13380         * ipa-split.c (execute_split_functions): Likewise.
13381         * symbol-summary.h: Likewise.
13382         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13384 2018-06-19  Richard Biener  <rguenther@suse.de>
13386         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
13387         (vectorize_loops): ... here.  Fix dbgcnt handling.
13388         (try_vectorize_loop): Wrap try_vectorize_loop_1.
13390 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
13392         PR target/86197
13393         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
13394         ieee128 argument takes up only one (vector) register, not two (floating
13395         point) registers.
13397 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13399         * gimplify.c (gimplify_init_constructor): Really never clear for an
13400         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
13402 2018-06-19  Richard Biener  <rguenther@suse.de>
13404         PR tree-optimization/86179
13405         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
13406         after failed recognition.
13408 2018-06-18  Martin Sebor  <msebor@redhat.com>
13410         PR middle-end/85602
13411         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
13412         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
13413         Handle integer subtraction.
13414         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
13415         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
13417 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13419         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
13420         param from rtx to rtx_insn *.
13421         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
13422         param.
13423         (frv_ifcvt_modify_insn): Likwise.
13424         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
13425         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
13426         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
13427         as_a <rtx_insn *> cast to local "unprotected_region" once
13428         it's been established that it's not NULL or pc_rtx.
13429         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
13430         param "sethi" from rtx to rtx_insn *.
13431         (nds32_group_float_insns): Likewise for param "insn".
13432         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
13433         param.
13434         (vax_output_int_subtract): Likewise.
13435         * config/vax/vax.c (vax_output_int_add): Likewise for param
13436         "insn".
13437         (vax_output_int_subtract): Likewise.
13438         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
13439         (emit_pattern_after): Likewise for param "after".
13440         (emit_insn_after): Likewise.
13441         (emit_jump_insn_after): Likewise.
13442         (emit_call_insn_after): Likewise.
13443         (emit_debug_insn_after): Likewise.
13444         (emit_pattern_before): Likewise for param "before".
13445         (emit_insn_before): Likewise.
13446         (emit_jump_insn_before): Likewise.
13447         * final.c (get_insn_template): Likewise for param "insn", removing
13448         a cast.
13449         * output.h (get_insn_template): Likewise for 2nd param.
13450         * rtl.h (emit_insn_before): Likewise.
13451         (emit_jump_insn_before): Likewise.
13452         (emit_debug_insn_before_noloc): Likewise.
13453         (emit_insn_after): Likewise.
13454         (emit_jump_insn_after): Likewise.
13455         (emit_call_insn_after): Likewise.
13456         (emit_debug_insn_after): Likewise.
13457         (set_insn_deleted): Likewise for param.
13459 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
13461         PR target/85358
13462         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
13463         floating point modes, so that IFmode is numerically greater than
13464         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
13465         to declare the ordering.  This prevents IFmode from being
13466         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
13467         machine.  Include rs6000-modes.h to share the fractional values
13468         between genmodes* and the rest of the compiler.
13469         (IFmode): Likewise.
13470         (KFmode): Likewise.
13471         (TFmode): Likewise.
13472         * config/rs6000/rs6000-modes.h: New file.
13473         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
13474         meaning of rs6000_long_double_size so that 126..128 selects an
13475         appropriate 128-bit floating point type.
13476         (rs6000_option_override_internal): Likewise.
13477         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
13478         (TARGET_LONG_DOUBLE_128): Change the meaning of
13479         rs6000_long_double_size so that 126..128 selects an appropriate
13480         128-bit floating point type.
13481         (LONG_DOUBLE_TYPE_SIZE): Update comment.
13482         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
13483         source and destination to match the standard usage.
13484         (truncifkf2): Likewise.
13485         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
13486         ISA 2.07 to use an explicit clobber, instead of passing in a
13487         temporary.
13488         (copysign<mode>3_soft): Likewise.
13490 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13492         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13493         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
13494         (vect_slp_analyze_instance_dependence): Likewise.
13495         (vect_enhance_data_refs_alignment): Likewise.
13496         (vect_analyze_data_refs_alignment): Likewise.
13497         (vect_slp_analyze_and_verify_instance_alignment
13498         (vect_analyze_data_ref_accesses): Likewise.
13499         (vect_prune_runtime_alias_test_list): Likewise.
13500         (vect_analyze_data_refs): Likewise.
13501         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
13502         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13503         (vect_analyze_scalar_cycles_1): Likewise.
13504         (vect_get_loop_niters): Likewise.
13505         (vect_analyze_loop_form_1): Likewise.
13506         (vect_update_vf_for_slp): Likewise.
13507         (vect_analyze_loop_operations): Likewise.
13508         (vect_analyze_loop): Likewise.
13509         (vectorizable_induction): Likewise.
13510         (vect_transform_loop): Likewise.
13511         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
13512         * tree-vect-slp.c (vect_analyze_slp): Likewise.
13513         (vect_make_slp_decision): Likewise.
13514         (vect_detect_hybrid_slp): Likewise.
13515         (vect_slp_analyze_operations): Likewise.
13516         (vect_slp_bb): Likewise.
13517         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
13518         (vectorizable_bswap): Likewise.
13519         (vectorizable_call): Likewise.
13520         (vectorizable_simd_clone_call): Likewise.
13521         (vectorizable_conversion): Likewise.
13522         (vectorizable_assignment): Likewise.
13523         (vectorizable_shift): Likewise.
13524         (vectorizable_operation): Likewise.
13525         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
13527 2018-06-18  Martin Sebor  <msebor@redhat.com>
13529         PR tree-optimization/81384
13530         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
13531         * builtins.c (expand_builtin_strnlen): New function.
13532         (expand_builtin): Call it.
13533         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
13534         * builtins.def (BUILT_IN_STRNLEN): New.
13535         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
13536         Warn for bounds in excess of maximum object size.
13537         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
13538         single-value ranges.  Handle strnlen.
13539         (handle_builtin_strlen): Handle strnlen.
13540         (strlen_check_and_optimize_stmt): Same.
13541         * doc/extend.texi (Other Builtins): Document strnlen.
13543 2018-06-18  Maya Rashish  <coypu@sdf.org>
13545         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
13546         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
13547         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
13549         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
13550         here to ...
13551         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
13553 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13555         * tree.c (escaped_string::escape): Replace cast to char * by
13556         const_cast<char *> (unescaped).
13558 2018-06-18  Nick Clifton  <nickc@redhat.com>
13560         PR 84195
13561         * tree.c (escaped_string): New class.  Converts an unescaped
13562         string into its escaped equivalent.
13563         (warn_deprecated_use): Use the new class to convert the
13564         deprecation message, if present.
13565         (test_escaped_strings): New self test.
13566         (test_c_tests): Add test_escaped_strings.
13567         * doc/extend.texi (deprecated): Add a note that the
13568         deprecation message is affected by the -fmessage-length
13569         option, and that control characters will be escaped.
13570         (#pragma GCC error): Document this pragma.
13571         (#pragma GCC warning): Likewise.
13572         * doc/invoke.texi (-fmessage-length): Document this option's
13573         effect on the #warning and #error preprocessor directives and
13574         the deprecated attribute.
13576 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
13578         * tree.c (decl_value_expr_lookup): Revert latest change.
13579         (decl_value_expr_insert): Likewise.
13581 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
13583         * gimplify.c (nonlocal_vlas): Delete.
13584         (nonlocal_vla_vars): Likewise.
13585         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
13586         referenced VLAs.
13587         (gimplify_body): Do not create and destroy nonlocal_vlas.
13588         * tree-nested.c: Include diagnostic.h.
13589         (use_pointer_in_frame): Tweak.
13590         (lookup_field_for_decl): Add assertion and declare the transformation.
13591         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
13592         internal error when the reference is in a wrong context.  Do not
13593         create a debug decl by default.
13594         (note_nonlocal_block_vlas): Delete.
13595         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
13596         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
13597         create a debug decl by default.
13598         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
13599         call is in a wrong context.
13600         (fixup_vla_decls): New function.
13601         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
13602         debug variables were created.
13603         * tree.c (decl_value_expr_lookup): Add checking assertion.
13604         (decl_value_expr_insert): Likewise.
13606 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13608         PR middle-end/82479
13609         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
13610         * tree-scalar-evolution.c (interpret_expr): Likewise.
13611         (expression_expensive_p): Likewise.
13612         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
13613         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
13614         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
13615         (ssa_defined_by_minus_one_stmt_p): New.
13617 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13619         PR middle-end/64946
13620         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
13621         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
13622         * dojump.c (do_jump): Likewise.
13623         * expr.c (expand_expr_real_2): Check operand type's sign.
13624         * fold-const.c (const_unop): Handle ABSU_EXPR.
13625         (fold_abs_const): Likewise.
13626         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13627         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13628         (strip_sign_op_1): Likesise.
13629         * match.pd: Add new pattern to generate ABSU_EXPR.
13630         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13631         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13632         * tree-eh.c (operation_could_trap_helper_p): Likewise.
13633         * tree-inline.c (estimate_operator_cost): Likewise.
13634         * tree-pretty-print.c (dump_generic_node): Likewise.
13635         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13636         * tree.def (ABSU_EXPR): New.
13638 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
13640         PR middle-end/86095
13641         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13642         documented as preserved for backward compatibility only.
13643         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13645         PR rtl-optimization/86108
13646         * bb-reorder.c (create_forwarder_block): Renamed to ...
13647         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
13648         jump from new landing pad to the second part.
13649         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13650         Adjust callers.
13652 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
13654         PR middle-end/85878
13655         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13656         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13657         Only call store_expr for halves if the mode is the same.
13659         PR middle-end/86123
13660         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13661         Fix up comment formatting.
13663 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13665         * typed-splay-tree.h (typed_splay_tree::remove): New function.
13666         (typed_splay_tree::closure,
13667         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13668         (typed_splay_tree::typed_splay_tree,
13669         typed_splay_tree::operator =): Declared private.
13670         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13671         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13672         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13673         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13674         typed_splay_tree::splay_tree_splay,
13675         typed_splay_tree::splay_tree_foreach_helper,
13676         typed_splay_tree::splay_tree_insert,
13677         typed_splay_tree::splay_tree_remove,
13678         typed_splay_tree::splay_tree_lookup,
13679         typed_splay_tree::splay_tree_predecessor,
13680         typed_splay_tree::splay_tree_successor,
13681         typed_splay_tree::splay_tree_min,
13682         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13683         (typed_splay_tree::root, typed_splay_tree::comp,
13684         typed_splay_tree::delete_key,
13685         typed_splay_tree::delete_value): New data members.
13686         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13687         typed_splay_tree::remove.
13689 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
13691         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13692         -mginv and -mno-ginv to the assembler.
13693         * config/mips/mips.opt (-mcrc): New option.
13694         (-mginv): Likewise.
13695         * doc/invoke.text (-mcrc): Document.
13696         (-mginv): Likewise.
13698 2018-06-15  Nick Clifton  <nickc@redhat.com>
13700         PR 84195
13701         * tree.c (escaped_string): New class.  Converts an unescaped
13702         string into its escaped equivalent.
13703         (warn_deprecated_use): Use the new class to convert the
13704         deprecation message, if present.
13705         (test_escaped_strings): New self test.
13706         (test_c_tests): Add test_escaped_strings.
13707         * doc/extend.texi (deprecated): Add a note that the
13708         deprecation message is affected by the -fmessage-length
13709         option, and that control characters will be escaped.
13710         (#pragma GCC error): Document this pragma.
13711         (#pragma GCC warning): Likewise.
13712         * doc/invoke.texi (-fmessage-length): Document this option's
13713         effect on the #warning and #error preprocessor directives and
13714         the deprecated attribute.
13716 2018-06-15  Richard Biener  <rguenther@suse.de>
13718         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13719         here, also noting vector size used.
13720         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
13721         size used in MSG_OPTIMIZED_LOCATIONS dump.
13722         (pass_slp_vectorize::execute): Adjust.
13724 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
13726         PR target/85968
13727         * config/arc/arc.c (arc_return_address_register): Fix
13728         if-condition.
13730 2018-06-15  Richard Biener  <rguenther@suse.de>
13732         PR middle-end/86159
13733         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13734         leave useless conversion stripping to force_gimple_operand_gsi.
13735         (gimplify_build2): Likewise.
13736         (gimplify_build1): Likewise.
13738 2018-06-15  Richard Biener  <rguenther@suse.de>
13740         PR middle-end/86076
13741         * tree-cfg.c (move_stmt_op): unshare invariant addresses
13742         before adjusting their block.
13744 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13746         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13747         multilibs for *-*-rtems*.
13748         * config/riscv/t-rtems: New file.
13750 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
13752         PR middle-end/86122
13753         * match.pd ((A +- CST1) +- CST2): Punt if last resort
13754         unsigned_type_for returns NULL.
13756         PR target/85945
13757         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13758         subregs of multi-word pseudos unless the float mode has word size.
13760 2018-06-14  Richard Biener  <rguenther@suse.de>
13762         PR middle-end/86139
13763         * tree-vect-generic.c (build_word_mode_vector_type): Remove
13764         duplicate and harmful type_hash_canon.
13765         * tree.c (type_hash_canon): Assert we didn't find ourselves.
13767 2018-06-14  Richard Biener  <rguenther@suse.de>
13769         PR ipa/86124
13770         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13771         NULL cgraph_node.
13773 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13775         * config/rtems.h (STDINT_LONG32): Define.
13777 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
13778             Prachi Godbole  <prachi.godbole@imgtec.com>
13780         * config/mips/mips-cpus.def: Define P6600.
13781         * config/mips/mips-tables.opt: Regenerate.
13782         * config/mips/mips.c (mips_ucbranch_type): New enum.
13783         (mips_rtx_cost_data): Add support for P6600.
13784         (mips_issue_rate): Likewise.
13785         (mips_multipass_dfa_lookahead): Likewise.
13786         (mips_avoid_hazard): Likewise.
13787         (mips_reorg_process_insns): Likewise.
13788         (mips_classify_branch_p6600): New function.
13789         * config/mips/mips.h (TUNE_P6600): New define.
13790         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13791         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13792         * config/mips/mips.md: Include p6600.md.
13793         (processor): Add p6600.
13794         * config/mips/p6600.md: New file.
13795         * doc/invoke.texi: Add p6600 to supported architectures.
13797 2018-06-13  Martin Sebor  <msebor@redhat.com>
13799         PR tree-optimization/86114
13800         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13801         of integer types.
13802         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13804 2018-06-13  Richard Biener  <rguenther@suse.de>
13806         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13807         Properly set vector type of the intermediate stmt.
13808         * tree-vect-stmts.c (vectorizable_operation): The destination
13809         var always has vectype_out type.
13811 2018-06-13  Jeff Law  <law@redhat.com>
13813         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13814         integer 0 for argument to print_rtl_with_bb.
13815         (rl78_reorg): Likewise.
13817 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
13819         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13820         from rtx to rtx_insn *.
13821         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13822         "label".
13823         (add_sched_insns_for_speculation): Likewise for local "target",
13824         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13825         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13826         from rtx_insn ** to rtx_code_label **.
13827         (reorg_emit_nops): Likewise.
13828         (c6x_reorg): Likewise for local "call_labels".
13829         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13830         rtx to rtx_insn *.
13831         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13832         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13833         the loops over LABEL_REFS.
13834         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13835         braf_label.
13836         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13837         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13838         (split_branches): Strengthen local "olabel" from rtx to
13839         rtx_insn *, adding a safe_as_a cast.
13840         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13841         to "rtx_insn *".
13842         (add_insn_after): Likewise for first two params.
13843         (add_insn_before): Likewise.
13844         (remove_insn): Likewise for param.
13845         (emit_pattern_before_noloc): Likewise for second and third params.
13846         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13847         (emit_call_insn_before_noloc): Likewise.
13848         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13849         to "rtx_insn *".
13850         (emit_barrier_before): Likewise.
13851         (emit_label_before): Strengthen "label" param from "rtx" to
13852         "rtx_code_label *".  Strengthen "before" param from "rtx" to
13853         "rtx_insn *".
13854         (emit_insn_after_1): Strengthen "after" param from "rtx" to
13855         "rtx_insn *".
13856         (emit_pattern_after_noloc): Likewise.
13857         (emit_insn_after_noloc): Likewise.
13858         (emit_jump_insn_after_noloc): Likewise.
13859         (emit_call_insn_after_noloc): Likewise.
13860         (emit_debug_insn_after_noloc): Likewise.
13861         (emit_barrier_after): Likewise.
13862         (emit_label_after): Likewise for both params.
13863         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
13864         "loc" param from "int" to "location_t".
13865         (emit_insn_after_setloc): Likewise.
13866         (emit_jump_insn_after_setloc): Likewise.
13867         (emit_call_insn_after_setloc): Likewise.
13868         (emit_debug_insn_after_setloc): Likewise.
13869         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
13870         "loc" param from "int" to "location_t".
13871         (emit_pattern_before): Convert NULL_RTX to NULL.
13872         (emit_insn_before_setloc): Convert "loc" param from "int" to
13873         "location_t".
13874         (emit_jump_insn_before_setloc): Likewise.
13875         (emit_call_insn_before_setloc): Likewise.
13876         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13877         rtx_insn *.  Convert "loc" param from "int" to "location_t".
13878         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13879         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13880         Convert 3rd param from "int" to "location_t".
13881         (emit_barrier_before, emit_barrier_after, next_real_insn):
13882         Strengthen param from rtx to rtx_insn *.
13883         (emit_label_before): Strengthen 1st param from "rtx" to
13884         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
13885         "rtx_insn *".
13886         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13887         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13888         Strengthen 2nd param from "rtx" to "rtx_insn *".
13889         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13890         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13891         Likewise. Convert 3rd param from "int" to "location_t".
13892         (emit_label_after): Strengthen 1st param from "rtx" to
13893         "rtx_code_label *".
13894         (next_real_insn, remove_insn): Strengthen param from "rtx" to
13895         "rtx_insn *".
13896         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13897         from "rtx" to "rtx_insn *".
13899 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
13901         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13902         bodies streamed in with -Q.
13903         * dumpfile.c (dump_files): Add lto-stream-out dump file.
13904         * dumpfile.h (tree_dump_index): Add lto_stream_out.
13905         * gimple-streamer-out.c: Include gimple-pretty-print.h
13906         (output_bb): Dump stmts streamed.
13907         * lto-section-out.c: Include print-tree.h
13908         (lto_begin_section): Dump sections created.
13909         (lto_output_decl_index): Dump decl encoded.
13910         * lto-streamer-out.c: Include print-tree.h
13911         (create_output_block): Dump output block created.
13912         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13913         (output_function): Dump function output.
13914         (output_constructor): Dump constructor streamed.
13915         (write_global_stream): Output indexes encoded.
13916         (produce_asm_for_decls): Dump streams encoded.
13917         * lto-streamer.c (streamer_dump_file): New global var.
13918         * lto-streamer.h (streamer_dump_file): Declare.
13919         * passes.c (ipa_write_summaries): Initialize streamer dump.
13920         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13921         in.
13923 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
13925         PR target/86048
13926         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13927         offsets for register save directives.  Emit a second batch of save
13928         directives, if need be, when the function accesses prior frames.
13930 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13932         * config/arc/fpu.md (fmasf4): Force operand to register.
13933         (fnmasf4): Likewise.
13935 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13937         * config/arc/arc-protos.h (arc_pad_return): Remove.
13938         * config/arc/arc.c (machine_function): Remove force_short_suffix
13939         and size_reason.
13940         (arc_print_operand): Adjust printing of '&'.
13941         (arc_verify_short): Remove conditional printing of short suffix.
13942         (arc_final_prescan_insn): Remove reference to size_reason.
13943         (pad_return): New function.
13944         (arc_reorg): Call pad_return.
13945         (arc_pad_return): Remove.
13946         (arc_init_machine_status): Remove reference to force_short_suffix.
13947         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13948         (attr length): When attribute iscompact is true force to 2
13949         regardless; in the case of maybe check if we want to force the
13950         instruction to have 4 bytes length.
13951         (nopv): Change it to generate 4 byte long nop as well.
13952         (blockage): New pattern.
13953         (simple_return): Remove call to arc_pad_return.
13954         (p_return_i): Likewise.
13956 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13958         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13960 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13962         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13963         ARC cores.
13965 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13967         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13968         for ARC700 and ARCv2.
13970 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
13972         PR target/86076
13973         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13974         operands[2] instead of operands[1].
13977 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13979         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13980         case, check whether the outer register overlaps an unallocatable
13981         register, not just whether it fits the required class.
13983 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13985         * poly-int.h (can_div_trunc_p): Add new overload in which all values
13986         are poly_ints.
13987         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13988         (memrefs_conflict_p): Likewise.
13989         (init_alias_analysis): Likewise.
13990         * cfgexpand.c (expand_debug_expr): Likewise.
13991         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13992         * cse.c (fold_rtx): Likewise.
13993         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13994         * expr.c (emit_block_move_hints): Likewise.
13995         (clear_storage_hints, push_block, emit_push_insn): Likewise.
13996         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13997         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13998         (emit_group_store): Likewise.
13999         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
14000         to read the PRE/POST_MODIFY increment.
14001         * calls.c (store_one_arg): Use strip_offset.
14002         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
14003         poly_int_rtx_p.
14004         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
14005         by a VEC_SELECT.
14006         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
14007         (simplify_binary_operation_1): Extend CONST_INT handling to
14008         poly_int_rtx_p.
14009         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
14010         than a HOST_WIDE_INT.
14011         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
14012         poly_int64.
14013         (adjust_mems, add_stores): Update accodingly.
14014         (vt_canonicalize_addr): Track polynomial offsets.
14015         (emit_note_insn_var_location): Likewise.
14016         (vt_add_function_parameter): Likewise.
14017         (vt_initialize): Likewise.
14019 2018-06-12  Jeff Law  <law@redhat.com>
14021         * config.gcc (alpha*-*-freebsd*): Remove.
14022         * config/alpha/freebsd.h: Remove.
14024 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
14026         PR other/69968
14027         * spellcheck-tree.c (levenshtein_distance): Rename to...
14028         (get_edit_distance): ...this, and update for underlying renaming.
14029         * spellcheck-tree.h (levenshtein_distance): Rename to...
14030         (get_edit_distance): ...this.
14031         * spellcheck.c (levenshtein_distance): Rename to...
14032         (get_edit_distance): ...this.  Convert from Levenshtein distance
14033         to Damerau-Levenshtein distance by supporting transpositions of
14034         adjacent characters.  Rename "v1" to "v_next" and "v0" to
14035         "v_one_ago".
14036         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
14037         (selftest::test_edit_distance_unit_test_oneway): ...this, and
14038         update for underlying renaming.
14039         (selftest::levenshtein_distance_unit_test): Rename to...
14040         (selftest::test_get_edit_distance_unit): ...this, and update for
14041         underlying renaming.
14042         (selftest::test_find_closest_string): Add example from PR 69968
14043         where transposition helps
14044         (selftest::test_metric_conditions): Update for renaming.
14045         (selftest::test_metric_conditions): Likewise.
14046         (selftest::spellcheck_c_tests): Likewise.
14047         * spellcheck.h (levenshtein_distance): Rename both overloads to...
14048         (get_edit_distance): ...this.
14049         (best_match::consider): Update for renaming.
14051 2018-06-12  Martin Sebor  <msebor@redhat.com>
14053         PR tree-optimization/85259
14054         * builtins.c (compute_objsize): Handle constant offsets.
14055         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
14056         true iff a warning has been issued.
14057         * gimple.h (gimple_nonartificial_location): New function.
14058         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
14059         gimple_nonartificial_location and handle -Wno-system-headers.
14060         (handle_builtin_stxncpy): Same.
14062 2018-06-12  Martin Sebor  <msebor@redhat.com>
14064         PR c/85931
14065         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
14067 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
14069         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
14070         BUILTIN_VEC_XST entries for pointer to double and long long.
14072 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
14074         PR target/85990
14075         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14076         Update comments.
14077         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14078         Likewise.
14080 2018-06-12  Martin Liska  <mliska@suse.cz>
14082         * doc/options.texi: Document IntegerRange.
14084 2018-06-12  Martin Liska  <mliska@suse.cz>
14086         * config/i386/i386.opt: Make MPX-related options as Deprecated.
14087         * opt-functions.awk: Handle Deprecated flag.
14088         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
14089         and report error.
14090         (read_cmdline_option): Report warning for a deprecated option.
14091         * opts.h (struct cl_option): Add new field cl_deprecated.
14092         (CL_ERR_DEPRECATED): New.
14094 2018-06-12  Martin Liska  <mliska@suse.cz>
14096         * doc/options.texi: Document Deprecated option flag.
14098 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14100         * config/arc/arc-arch.h (arc_extras): New enum.
14101         (arc_cpu_t):Add field extra.
14102         (arc_cpu_types): Consider the extras.
14103         * config/arc/arc-cpus.def: Add extras info.
14104         * config/arc/arc-opts.h (processor_type): Consider extra field.
14105         * config/arc/arc.c (arc_override_options): Handle extra field.
14107 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14109         * config/arc/arc-arch.h: Update ARC_OPTX macro.
14110         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
14111         field.
14112         * config/arc/arc.c (arc_init): Update pic warning.
14113         (irq_range): Update irq range parsing warnings.
14114         (arc_override_options): Update various warning messages.
14115         (arc_handle_aux_attribute): Likewise.
14117 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
14119         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
14121 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14123         * doc/sourcebuild.texi: Document usage of line number 0 in verify
14124         compiler messages directives.
14126 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
14128         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
14129         * config/mips/mips-tables.opt: Regenerate.
14130         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
14131         mips64r6.
14132         * doc/invoke.texi: Document -march=i6500.
14134 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
14136         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
14137         (i6400_gpmul): Add cpu_unit.
14138         (i6400_gpdiv): Likewise.
14139         (i6400_msa_add_d): Update reservations.
14140         (i6400_msa_int_add) Likewise.
14141         (i6400_msa_short_logic3) Likewise.
14142         (i6400_msa_short_logic2) Likewise.
14143         (i6400_msa_short_logic) Likewise.
14144         (i6400_msa_move) Likewise.
14145         (i6400_msa_cmp) Likewise.
14146         (i6400_msa_short_float2) Likewise.
14147         (i6400_msa_div_d) Likewise.
14148         (i6400_msa_long_logic1) Likewise.
14149         (i6400_msa_long_logic2) Likewise.
14150         (i6400_msa_mult) Likewise.
14151         (i6400_msa_long_float2) Likewise.
14152         (i6400_msa_long_float4) Likewise.
14153         (i6400_msa_long_float5) Likewise.
14154         (i6400_msa_long_float8) Likewise.
14155         (i6400_fpu_fadd): Include frint type.
14156         (i6400_fpu_store): New define_insn_reservation.
14157         (i6400_fpu_load): Likewise.
14158         (i6400_fpu_move): Likewise.
14159         (i6400_fpu_fcmp): Likewise.
14160         (i6400_fpu_fmadd): Likewise.
14161         (i6400_int_mult): Include imul3nc type and update reservation.
14162         (i6400_int_div): Include idiv3 type and update reservation.
14163         (i6400_int_load): Update to check type not move_type.
14164         (i6400_int_store): Likewise.
14165         (i6400_int_prefetch): Set zero latency.
14167 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
14169         * gcc.c: Document new %@{...} sequence.
14170         (LINK_COMMAND_SPEC): Use it for the -L switches.
14171         (cpp_unique_options): Use it for the -I switches.
14172         (at_file_argbuf): New global variable.
14173         (in_at_file): Likewise.
14174         (alloc_args): Create at_file_argbuf.
14175         (clear_args): Truncate at_file_argbuf.
14176         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
14177         (open_at_file): New function.
14178         (close_at_file): Likewise.
14179         (create_at_file): Delete.
14180         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
14181         <'o'>: Likewise.
14182         <'@'>: New case.
14183         (validate_switches_from_spec): Deal with %@{...} sequence.
14184         (validate_switches): Likewise.
14185         (driver::finalize): Call clear_args.
14187 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
14189         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
14191 2018-06-11  Martin Sebor  <msebor@redhat.com>
14193         * doc/invoke.texi (-Wall): List -Wc++17-compat.
14194         (Wno-class-memaccess): Add @opindex.
14195         (Wno-templates, Wno-multiple-inheritance): Same.
14196         (Wno-virtual-inheritance, Wno-namespaces): Same.
14197         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
14198         (Wno-format-overflow, Wno-format-truncation): Same.
14199         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
14200         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
14201         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
14202         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
14203         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
14204         (Wno-misspelled-isr): Same.
14206 2018-06-11  Martin Sebor  <msebor@redhat.com>
14208         * PR tree-optimization/86083
14209         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
14211 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
14213         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
14215 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
14217         PR target/85755
14218         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
14219         on the correct operand.
14220         (*movdi_internal64): Ditto.
14222 2018-06-11  Martin Liska  <mliska@suse.cz>
14224         PR tree-optimization/86089
14225         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
14227 2018-06-11  Julia Koval  <julia.koval@intel.com>
14229         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
14230         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
14231         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
14233 2018-06-11  Olivier Hainque  <hainque@adacore.com>
14235         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
14236         for Ada with strict dwarf2.
14238 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
14240         PR target/85755
14241         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
14242         addresses.
14244 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
14246         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
14248 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
14250         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
14251         TARGET_ELF.
14253 2018-06-08  Martin Liska  <mliska@suse.cz>
14255         * tree-cfg.h (debug_function): Fix argument type to match
14256         implementation.
14258 2018-06-08  Martin Liska  <mliska@suse.cz>
14260         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
14261         Remove usage of MPX-related (and removed) fields.
14262         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
14264 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
14266         * cfg.c (debug): Use TDF_NONE rather than 0.
14267         * cfghooks.c (debug): Likewise.
14268         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
14269         (struct dump_option_value_info): Convert to...
14270         (struct kv_pair): ...this template type.
14271         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
14272         rather than 0.
14273         (optinfo_verbosity_options): Likewise.
14274         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
14275         OPTGROUP_NONE.
14276         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
14277         than int for "optgroup_flags" param.
14278         (dump_generic_expr_loc): Use dump_flags_t rather than int for
14279         "dump_kind" param.
14280         (dump_dec): Likewise.
14281         (dump_finish): Use TDF_NONE rather than 0.
14282         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
14283         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
14284         than 0.  Update for change to option_ptr.
14285         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
14286         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
14287         0.  Update for changes to optinfo_verbosity_options and
14288         optgroup_options.
14289         (opt_info_switch_p): Convert optgroup_flags from int to
14290         optgroup_flags_t.
14291         (dump_basic_block): Use dump_flags_t rather than int
14292         for "dump_kind" param.
14293         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
14294         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
14295         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
14296         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
14297         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
14298         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
14299         TDF_NONE): Convert from macros to...
14300         (enum dump_flag): ...this new enum.
14301         (dump_flags_t): Update to use enum.
14302         (operator|, operator&, operator~, operator|=, operator&=):
14303         Implement for dump_flags_t.
14304         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
14305         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
14306         Convert from macros to...
14307         (enum optgroup_flag): ...this new enum.
14308         (optgroup_flags_t): New typedef.
14309         (operator|, operator|=): Implement for optgroup_flags_t.
14310         (struct dump_file_info): Convert field "alt_flags" to
14311         dump_flags_t.  Convert field "optgroup_flags" to
14312         optgroup_flags_t.
14313         (dump_basic_block): Use dump_flags_t rather than int for param.
14314         (dump_generic_expr_loc): Likewise.
14315         (dump_dec): Likewise.
14316         (dump_register): Convert param "optgroup_flags" to
14317         optgroup_flags_t.
14318         (opt_info_enable_passes): Likewise.
14319         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
14320         than 0.
14321         * gimple-pretty-print.c (debug): Likewise.
14322         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
14323         (merged_store_group::apply_stores): Likewise.
14324         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
14325         * gimple.c (verify_gimple_pp): Likewise.
14326         * graphite-poly.c (print_pbb_body): Likewise.
14327         * passes.c (pass_manager::register_one_dump_file): Convert
14328         local "optgroup_flags" to optgroup_flags_t.
14329         * print-tree.c (print_node): Use TDF_NONE rather than 0.
14330         (debug): Likewise.
14331         (debug_body): Likewise.
14332         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
14333         to optgroup_flags_t.
14334         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
14335         than 0.
14336         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
14337         (convert_mult_to_fma): Likewise.
14338         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
14339         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
14340         * tree-vect-data-refs.c (dump_lower_bound): Convert param
14341         "dump_kind" to dump_flags_t.
14343 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14345         * config/rs6000/rs6000.c (min, max): Delete.
14347 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14349         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
14350         -mabi=spe and -mabi=no-spe.
14352 2018-06-08  Martin Liska  <mliska@suse.cz>
14354         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
14355         where we expect an existing summary.
14357 2018-06-08  Martin Liska  <mliska@suse.cz>
14359         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
14360         * ipa-inline.h (estimate_edge_growth): Likewise.
14362 2018-06-08  Martin Liska  <mliska@suse.cz>
14364         * cgraph.c (function_version_hasher::hash): Use
14365         cgraph_node::get_uid ().
14366         (function_version_hasher::equal):
14367         * cgraph.h (cgraph_node::get_uid): New method.
14368         * ipa-inline.c (update_caller_keys): Use
14369         cgraph_node::get_uid ().
14370         (update_callee_keys): Likewise.
14371         * ipa-utils.c (searchc): Likewise.
14372         (ipa_reduced_postorder): Likewise.
14373         * lto-cgraph.c (input_node): Likewise.
14374         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
14375         * symbol-summary.h (symtab_insertion): Likewise.
14376         (symtab_removal): Likewise.
14377         (symtab_duplication): Likewise.
14378         * tree-pretty-print.c (dump_function_header): Likewise.
14379         * tree-sra.c (convert_callers_for_node): Likewise.
14381 2018-06-08  Martin Liska  <mliska@suse.cz>
14383         * cgraph.c (symbol_table::create_edge): Always assign a new
14384         unique number.
14385         (symbol_table::free_edge): Do not recycle numbers.
14386         * cgraph.h (cgraph_edge::get): New method.
14387         * symbol-summary.h (symtab_removal): Use it.
14388         (symtab_duplication): Likewise.
14389         (call_summary::hashable_uid): Remove.
14391 2018-06-08  Martin Liska  <mliska@suse.cz>
14393         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
14394         (initialize_growth_caches): Remove.
14395         (free_growth_caches): Likewise.
14396         (do_estimate_edge_time): Use edge_growth_cache.
14397         (do_estimate_edge_size): Likewise.
14398         (do_estimate_edge_hints): Likewise.
14399         * ipa-inline.c (reset_edge_caches): Likewise.
14400         (recursive_inlining): Likewise.
14401         (inline_small_functions): Likewise.
14402         * ipa-inline.h (initialize_growth_caches): Remove.
14403         (estimate_edge_size): Likewise.
14404         (estimate_edge_time): Likewise.
14405         (estimate_edge_hints): Likewise.
14406         (reset_edge_growth_cache): Likewise.
14407         * symbol-summary.h (call_summary::remove): New method.
14409 2018-06-08  Martin Liska  <mliska@suse.cz>
14411         * ipa-cp.c (class edge_clone_summary): New summary.
14412         (grow_edge_clone_vectors): Remove.
14413         (ipcp_edge_duplication_hook): Remove.
14414         (class edge_clone_summary_t): New call_summary class.
14415         (ipcp_edge_removal_hook): Remove.
14416         (edge_clone_summary_t::duplicate): New function.
14417         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
14418         (create_specialized_node): Likewise.
14419         (ipcp_driver): Initialize edge_clone_summaries and do not
14420         register hooks.
14422 2018-06-08  Martin Liska  <mliska@suse.cz>
14424         * symbol-summary.h (get): New function.
14425         (call_summary::m_initialize_when_cloning): New class member.
14427 2018-06-08  Martin Liska  <mliska@suse.cz>
14429         * cgraph.c (cgraph_node::remove): Do not recycle uid.
14430         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
14431         (symbol_table::allocate_cgraph_symbol): Do not set uid.
14432         * passes.c (uid_hash_t): Record removed_nodes by their uids.
14433         (remove_cgraph_node_from_order): Use the removed_nodes set.
14434         (do_per_function_toporder): Likwise.
14435         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
14436         instead of summary_uid.
14437         (symtab_removal): Likewise.
14438         (symtab_duplication): Likewise.
14440 2018-06-08  Martin Liska  <mliska@suse.cz>
14442         * ipa-cp.c (ipcp_store_bits_results): Use
14443         ipcp_transformation_sum.
14444         (ipcp_store_vr_results): Likewise.
14445         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
14446         to ...
14447         (ipcp_transformation_initialize): ... this.
14448         (ipa_set_node_agg_value_chain):
14449         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
14450         (write_ipcp_transformation_info): Likewise.
14451         (read_ipcp_transformation_info): Likewise.
14452         (ipcp_update_bits): Likewise.
14453         (ipcp_update_vr): Likewise.
14454         (ipcp_transform_function): Likewise.
14455         * ipa-prop.h: Rename ipcp_transformation_summary to
14456         ipcp_transformation.
14457         (class ipcp_transformation_t): New function summary.
14458         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
14459         (ipa_get_agg_replacements_for_node): Likewise.
14461 2018-06-08  Martin Liska  <mliska@suse.cz>
14463         * ipa-pure-const.c (struct funct_state_d): Do it class instead
14464         of struct.
14465         (class funct_state_summary_t): New function_summary class.
14466         (has_function_state): Remove.
14467         (get_function_state): Likewise.
14468         (set_function_state): Likewise.
14469         (add_new_function): Likewise.
14470         (funct_state_summary_t::insert): New function.
14471         (duplicate_node_data): Remove.
14472         (remove_node_data): Remove.
14473         (funct_state_summary_t::duplicate): New function.
14474         (register_hooks): Create new funct_state_summaries.
14475         (pure_const_generate_summary): Use it.
14476         (pure_const_write_summary): Likewise.
14477         (pure_const_read_summary): Likewise.
14478         (propagate_pure_const): Likewise.
14479         (propagate_nothrow): Likewise.
14480         (dump_malloc_lattice): Likewise.
14481         (propagate_malloc): Likewise.
14482         (execute): Do not register hooks, just remove summary
14483         instead.
14484         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
14485         constructor.
14487 2018-06-08  Martin Liska  <mliska@suse.cz>
14489         * ipa-reference.c (remove_node_data): Remove.
14490         (duplicate_node_data): Likewise.
14491         (class ipa_ref_var_info_summary_t): New class.
14492         (class ipa_ref_opt_summary_t): Likewise.
14493         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
14494         (get_reference_optimization_summary): Use
14495         ipa_ref_opt_sum_summaries.
14496         (set_reference_vars_info): Remove.
14497         (set_reference_optimization_summary): Likewise.
14498         (ipa_init): Create summaries.
14499         (init_function_info): Use function summary.
14500         (ipa_ref_opt_summary_t::duplicate): New function.
14501         (ipa_ref_opt_summary_t::remove): New function.
14502         (get_read_write_all_from_node): Fix GNU coding style.
14503         (propagate): Use function summary.
14504         (write_node_summary_p): Fix GNU coding style.
14505         (stream_out_bitmap): Likewise.
14506         (ipa_reference_read_optimization_summary): Use function summary.
14507         (ipa_reference_c_finalize): Do not release hooks.
14509 2018-06-08  Martin Liska  <mliska@suse.cz>
14511         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
14512         (analyze_function_body): Extract multiple calls of get_create.
14513         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14514         * ipa-inline.c (recursive_inlining): Use ::get method.
14515         * ipa-inline.h (estimate_edge_growth): Likewise.
14517 2018-06-08  Martin Liska  <mliska@suse.cz>
14519         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
14520         HSA_INVALID.
14521         (hsa_function_summary::hsa_function_summary): Use the new enum
14522         value.
14523         (hsa_gpu_implementation_p): Use hsa_summaries::get.
14524         * hsa-gen.c (hsa_get_host_function): Likewise.
14525         (get_brig_function_name): Likewise.
14526         * ipa-hsa.c (process_hsa_functions): Likewise.
14527         (ipa_hsa_write_summary): Likewise.
14528         * symbol-summary.h (symtab_duplication): Use ::get function/
14529         (get): New function.
14531 2018-06-08  Martin Liska  <mliska@suse.cz>
14533         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
14534         of get.
14535         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
14536         (hsa_register_kernel): Likewise.
14537         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
14538         * hsa-gen.c (hsa_get_host_function): Likewise.
14539         (get_brig_function_name): Likewise.
14540         (generate_hsa): Likewise.
14541         (pass_gen_hsail::execute): Likewise.
14542         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
14543         (devirtualization_time_bonus): Likewise.
14544         (ipcp_propagate_stage): Likewise.
14545         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14546         (edge_set_predicate): Likewise.
14547         (evaluate_conditions_for_known_args): Likewise.
14548         (evaluate_properties_for_edge): Likewise.
14549         (ipa_fn_summary::reset): Likewise.
14550         (ipa_fn_summary_t::duplicate): Likewise.
14551         (dump_ipa_call_summary): Likewise.
14552         (ipa_dump_fn_summary): Likewise.
14553         (analyze_function_body): Likewise.
14554         (compute_fn_summary): Likewise.
14555         (estimate_edge_devirt_benefit): Likewise.
14556         (estimate_edge_size_and_time): Likewise.
14557         (estimate_calls_size_and_time): Likewise.
14558         (estimate_node_size_and_time): Likewise.
14559         (inline_update_callee_summaries): Likewise.
14560         (remap_edge_change_prob): Likewise.
14561         (remap_edge_summaries): Likewise.
14562         (ipa_merge_fn_summary_after_inlining): Likewise.
14563         (ipa_update_overall_fn_summary): Likewise.
14564         (read_ipa_call_summary): Likewise.
14565         (inline_read_section): Likewise.
14566         (write_ipa_call_summary): Likewise.
14567         (ipa_fn_summary_write): Likewise.
14568         (ipa_free_fn_summary): Likewise.
14569         * ipa-hsa.c (process_hsa_functions): Likewise.
14570         (ipa_hsa_write_summary): Likewise.
14571         (ipa_hsa_read_section): Likewise.
14572         * ipa-icf.c (sem_function::merge): Likewise.
14573         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14574         (do_estimate_edge_time): Likewise.
14575         (estimate_size_after_inlining): Likewise.
14576         (estimate_growth): Likewise.
14577         (growth_likely_positive): Likewise.
14578         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14579         (inline_call): Likewise.
14580         * ipa-inline.c (caller_growth_limits): Likewise.
14581         (can_inline_edge_p): Likewise.
14582         (can_inline_edge_by_limits_p): Likewise.
14583         (compute_uninlined_call_time): Likewise.
14584         (compute_inlined_call_time): Likewise.
14585         (want_inline_small_function_p): Likewise.
14586         (edge_badness): Likewise.
14587         (update_caller_keys): Likewise.
14588         (update_callee_keys): Likewise.
14589         (recursive_inlining): Likewise.
14590         (inline_small_functions): Likewise.
14591         (inline_to_all_callers_1): Likewise.
14592         (dump_overall_stats): Likewise.
14593         (early_inline_small_functions): Likewise.
14594         (early_inliner): Likewise.
14595         * ipa-inline.h (estimate_edge_growth): Likewise.
14596         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14597         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14598         * ipa-prop.h (IPA_NODE_REF): Likewise.
14599         (IPA_EDGE_REF): Likewise.
14600         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14601         (propagate_malloc): Likewise.
14602         * ipa-split.c (execute_split_functions): Likewise.
14603         * symbol-summary.h: Rename get to get_create.
14604         (get): Likewise.
14605         (get_create): Likewise.
14606         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14608 2018-06-08  Martin Liska  <mliska@suse.cz>
14610         * symbol-summary.h (release): Move definition out of class
14611         declaration.
14612         (symtab_removal): Likewise.
14613         (symtab_duplication): Likewise.
14615 2018-06-08  Martin Liska  <mliska@suse.cz>
14617         * symbol-summary.h (function_summary): Move constructor
14618         implementation out of class declaration.
14619         (release): Likewise.
14620         (symtab_insertion): Likewise.
14621         (symtab_removal): Likewise.
14622         (symtab_duplication): Likewise.
14623         (get): Likewise.
14625 2018-06-08  Martin Liska  <mliska@suse.cz>
14627         * Makefile.in: Remove support for MPX (macros, related functions,
14628         fields in cgraph_node, ...).
14629         * builtin-types.def (BT_BND): Likewise.
14630         (BT_FN_BND_CONST_PTR): Likewise.
14631         (BT_FN_CONST_PTR_BND): Likewise.
14632         (BT_FN_VOID_PTR_BND): Likewise.
14633         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14634         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14635         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14636         (expand_builtin_mempcpy_with_bounds): Likewise.
14637         (expand_builtin_memset_with_bounds): Likewise.
14638         (expand_builtin_memset_args): Likewise.
14639         (std_expand_builtin_va_start): Likewise.
14640         (expand_builtin): Likewise.
14641         (expand_builtin_with_bounds): Likewise.
14642         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14643         (DEF_LIB_BUILTIN_CHKP): Likewise.
14644         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14645         (DEF_CHKP_BUILTIN): Likewise.
14646         (BUILT_IN_MEMCPY): Likewise.
14647         (BUILT_IN_MEMMOVE): Likewise.
14648         (BUILT_IN_MEMPCPY): Likewise.
14649         (BUILT_IN_MEMSET): Likewise.
14650         (BUILT_IN_STPCPY): Likewise.
14651         (BUILT_IN_STRCAT): Likewise.
14652         (BUILT_IN_STRCHR): Likewise.
14653         (BUILT_IN_STRCPY): Likewise.
14654         (BUILT_IN_STRLEN): Likewise.
14655         (BUILT_IN_MEMCPY_CHK): Likewise.
14656         (BUILT_IN_MEMMOVE_CHK): Likewise.
14657         (BUILT_IN_MEMPCPY_CHK): Likewise.
14658         (BUILT_IN_MEMSET_CHK): Likewise.
14659         (BUILT_IN_STPCPY_CHK): Likewise.
14660         (BUILT_IN_STRCAT_CHK): Likewise.
14661         (BUILT_IN_STRCPY_CHK): Likewise.
14662         * calls.c (store_bounds): Likewise.
14663         (emit_call_1): Likewise.
14664         (special_function_p): Likewise.
14665         (maybe_warn_nonstring_arg): Likewise.
14666         (initialize_argument_information): Likewise.
14667         (finalize_must_preallocate): Likewise.
14668         (compute_argument_addresses): Likewise.
14669         (expand_call): Likewise.
14670         * cfgexpand.c (expand_call_stmt): Likewise.
14671         (expand_return): Likewise.
14672         (expand_gimple_stmt_1): Likewise.
14673         (pass_expand::execute): Likewise.
14674         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14675         (cgraph_node::remove): Likewise.
14676         (cgraph_node::dump): Likewise.
14677         (cgraph_node::verify_node): Likewise.
14678         * cgraph.h (chkp_function_instrumented_p): Likewise.
14679         (symtab_node::get_alias_target): Likewise.
14680         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14681         (cgraph_local_p): Likewise.
14682         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14683         (cgraph_edge::rebuild_references): Likewise.
14684         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14685         (walk_polymorphic_call_targets): Likewise.
14686         (cgraph_node::expand_thunk): Likewise.
14687         (symbol_table::output_weakrefs): Likewise.
14688         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14689         (ix86_handle_option): Likewise.
14690         * config/i386/constraints.md: Likewise.
14691         * config/i386/i386-builtin-types.def (BND): Likewise.
14692         (VOID): Likewise.
14693         (PVOID): Likewise.
14694         (ULONG): Likewise.
14695         * config/i386/i386-builtin.def (BDESC_END): Likewise.
14696         (BDESC_FIRST): Likewise.
14697         (BDESC): Likewise.
14698         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14699         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14700         * config/i386/i386.c (enum reg_class): Likewise.
14701         (ix86_target_string): Likewise.
14702         (ix86_option_override_internal): Likewise.
14703         (ix86_conditional_register_usage): Likewise.
14704         (ix86_valid_target_attribute_inner_p): Likewise.
14705         (ix86_set_indirect_branch_type): Likewise.
14706         (ix86_set_current_function): Likewise.
14707         (ix86_function_arg_regno_p): Likewise.
14708         (init_cumulative_args): Likewise.
14709         (ix86_function_arg_advance): Likewise.
14710         (ix86_function_arg): Likewise.
14711         (ix86_pass_by_reference): Likewise.
14712         (ix86_function_value_regno_p): Likewise.
14713         (ix86_function_value_1): Likewise.
14714         (ix86_function_value_bounds): Likewise.
14715         (ix86_return_in_memory): Likewise.
14716         (ix86_setup_incoming_vararg_bounds): Likewise.
14717         (ix86_va_start): Likewise.
14718         (indirect_thunk_need_prefix): Likewise.
14719         (print_reg): Likewise.
14720         (ix86_print_operand): Likewise.
14721         (ix86_expand_call): Likewise.
14722         (ix86_output_function_return): Likewise.
14723         (reg_encoded_number): Likewise.
14724         (BDESC_VERIFYS): Likewise.
14725         (ix86_init_mpx_builtins): Likewise.
14726         (ix86_init_builtins): Likewise.
14727         (ix86_emit_cmove): Likewise.
14728         (ix86_emit_move_max): Likewise.
14729         (ix86_expand_builtin): Likewise.
14730         (ix86_builtin_mpx_function): Likewise.
14731         (ix86_get_arg_address_for_bt): Likewise.
14732         (ix86_load_bounds): Likewise.
14733         (ix86_store_bounds): Likewise.
14734         (ix86_load_returned_bounds): Likewise.
14735         (ix86_store_returned_bounds): Likewise.
14736         (ix86_class_likely_spilled_p): Likewise.
14737         (ix86_hard_regno_mode_ok): Likewise.
14738         (x86_order_regs_for_local_alloc): Likewise.
14739         (ix86_mitigate_rop): Likewise.
14740         (ix86_bnd_prefixed_insn_p): Likewise.
14741         (ix86_mpx_bound_mode): Likewise.
14742         (ix86_make_bounds_constant): Likewise.
14743         (ix86_initialize_bounds): Likewise.
14744         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14745         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14746         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14747         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14748         (TARGET_CHKP_BOUND_MODE): Likewise.
14749         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14750         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14751         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14752         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14753         * config/i386/i386.h (TARGET_MPX): Likewise.
14754         (TARGET_MPX_P): Likewise.
14755         (VALID_BND_REG_MODE): Likewise.
14756         (FIRST_BND_REG): Likewise.
14757         (LAST_BND_REG): Likewise.
14758         (enum reg_class): Likewise.
14759         (BND_REG_P): Likewise.
14760         (BND_REGNO_P): Likewise.
14761         (BNDmode): Likewise.
14762         (ADJUST_INSN_LENGTH): Likewise.
14763         * config/i386/i386.md: Likewise.
14764         * config/i386/i386.opt: Likewise.
14765         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14766         (defined): Likewise.
14767         (LINK_MPX): Likewise.
14768         (MPX_SPEC): Likewise.
14769         (LIBMPX_SPEC): Likewise.
14770         (LIBMPXWRAPPERS_SPEC): Likewise.
14771         (CHKP_SPEC): Likewise.
14772         * config/i386/predicates.md: Likewise.
14773         * dbxout.c (dbxout_type): Likewise.
14774         * doc/extend.texi: Likewise.
14775         * doc/invoke.texi: Likewise.
14776         * doc/md.texi: Likewise.
14777         * doc/tm.texi: Likewise.
14778         * doc/tm.texi.in: Likewise.
14779         * dwarf2out.c (is_base_type): Likewise.
14780         (gen_formal_types_die): Likewise.
14781         (gen_subprogram_die): Likewise.
14782         (gen_type_die_with_usage): Likewise.
14783         (gen_decl_die): Likewise.
14784         (dwarf2out_late_global_decl): Likewise.
14785         * expr.c (expand_assignment): Likewise.
14786         (emit_storent_insn): Likewise.
14787         (store_expr_with_bounds): Likewise.
14788         (store_expr): Likewise.
14789         (expand_expr_real_1): Likewise.
14790         * expr.h (store_expr_with_bounds): Likewise.
14791         * function.c (use_register_for_decl): Likewise.
14792         (struct bounds_parm_data): Likewise.
14793         (assign_parms_augmented_arg_list): Likewise.
14794         (assign_parm_find_entry_rtl): Likewise.
14795         (assign_parm_is_stack_parm): Likewise.
14796         (assign_parm_load_bounds): Likewise.
14797         (assign_bounds): Likewise.
14798         (assign_parms): Likewise.
14799         (expand_function_start): Likewise.
14800         * gcc.c (CHKP_SPEC): Likewise.
14801         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14802         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14803         (wrestrict_dom_walker::check_call): Likewise.
14804         * gimple.c (gimple_build_call_from_tree): Likewise.
14805         * gimple.h (enum gf_mask): Likewise.
14806         (gimple_call_with_bounds_p): Likewise.
14807         (gimple_call_set_with_bounds): Likewise.
14808         * gimplify.c (gimplify_init_constructor): Likewise.
14809         * ipa-cp.c (initialize_node_lattices): Likewise.
14810         (propagate_constants_across_call): Likewise.
14811         (find_more_scalar_values_for_callers_subset): Likewise.
14812         * ipa-hsa.c (process_hsa_functions): Likewise.
14813         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14814         * ipa-icf.c (sem_function::merge): Likewise.
14815         * ipa-inline.c (early_inliner): Likewise.
14816         * ipa-pure-const.c (warn_function_noreturn): Likewise.
14817         (warn_function_cold): Likewise.
14818         (propagate_pure_const): Likewise.
14819         * ipa-ref.h (enum GTY): Likewise.
14820         * ipa-split.c (find_retbnd): Likewise.
14821         (consider_split): Likewise.
14822         (split_function): Likewise.
14823         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14824         * ipa.c (walk_polymorphic_call_targets): Likewise.
14825         (symbol_table::remove_unreachable_nodes): Likewise.
14826         (process_references): Likewise.
14827         (cgraph_build_static_cdtor_1): Likewise.
14828         * lto-cgraph.c (lto_output_node): Likewise.
14829         (output_refs): Likewise.
14830         (compute_ltrans_boundary): Likewise.
14831         (input_overwrite_node): Likewise.
14832         (input_node): Likewise.
14833         (input_cgraph_1): Likewise.
14834         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14835         * passes.c (pass_manager::execute_early_local_passes): Likewise.
14836         (class pass_chkp_instrumentation_passes): Likewise.
14837         (make_pass_chkp_instrumentation_passes): Likewise.
14838         * passes.def: Likewise.
14839         * rtl.h (struct GTY): Likewise.
14840         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14841         * stor-layout.c (layout_type): Likewise.
14842         * symtab.c: Likewise.
14843         * target.def: Likewise.
14844         * targhooks.c (default_chkp_bound_type): Likewise.
14845         (default_chkp_bound_mode): Likewise.
14846         (default_builtin_chkp_function): Likewise.
14847         (default_chkp_function_value_bounds): Likewise.
14848         (default_chkp_make_bounds_constant): Likewise.
14849         (default_chkp_initialize_bounds): Likewise.
14850         * targhooks.h (default_chkp_bound_type): Likewise.
14851         (default_chkp_bound_mode): Likewise.
14852         (default_builtin_chkp_function): Likewise.
14853         (default_chkp_function_value_bounds): Likewise.
14854         (default_chkp_make_bounds_constant): Likewise.
14855         (default_chkp_initialize_bounds): Likewise.
14856         * toplev.c (compile_file): Likewise.
14857         (process_options): Likewise.
14858         * tree-core.h (DEF_BUILTIN): Likewise.
14859         (DEF_BUILTIN_CHKP): Likewise.
14860         * tree-inline.c (declare_return_variable): Likewise.
14861         (remap_gimple_stmt): Likewise.
14862         (copy_bb): Likewise.
14863         (initialize_inlined_parameters): Likewise.
14864         (expand_call_inline): Likewise.
14865         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14866         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14867         (make_pass_ipa_chkp_produce_thunks): Likewise.
14868         (make_pass_chkp): Likewise.
14869         (make_pass_chkp_opt): Likewise.
14870         (make_pass_chkp_instrumentation_passes): Likewise.
14871         * tree-pretty-print.c (dump_generic_node): Likewise.
14872         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14873         * tree-ssa-dce.c (propagate_necessity): Likewise.
14874         (eliminate_unnecessary_stmts): Likewise.
14875         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14876         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14877         * tree-ssa-sccvn.h: Likewise.
14878         * tree-ssa-strlen.c (get_string_length): Likewise.
14879         (valid_builtin_call): Likewise.
14880         (adjust_last_stmt): Likewise.
14881         (handle_builtin_strchr): Likewise.
14882         (handle_builtin_strcpy): Likewise.
14883         (handle_builtin_stxncpy): Likewise.
14884         (handle_builtin_memcpy): Likewise.
14885         (handle_builtin_strcat): Likewise.
14886         (strlen_check_and_optimize_stmt): Likewise.
14887         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14888         * tree-streamer-in.c: Likewise.
14889         * tree-streamer.c (record_common_node): Likewise.
14890         * tree.c (tree_code_size): Likewise.
14891         (wide_int_to_tree_1): Likewise.
14892         (type_contains_placeholder_1): Likewise.
14893         (build_common_tree_nodes): Likewise.
14894         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14895         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14896         (POINTER_BOUNDS_P): Likewise.
14897         (BOUNDED_TYPE_P): Likewise.
14898         (BOUNDED_P): Likewise.
14899         (CALL_WITH_BOUNDS_P): Likewise.
14900         (pointer_bounds_type_node): Likewise.
14901         * value-prof.c (gimple_ic): Likewise.
14902         * var-tracking.c (vt_add_function_parameters): Likewise.
14903         * varasm.c (make_decl_rtl): Likewise.
14904         (assemble_start_function): Likewise.
14905         (output_constant): Likewise.
14906         (maybe_assemble_visibility): Likewise.
14907         * varpool.c (ctor_for_folding): Likewise.
14908         * chkp-builtins.def: Remove.
14909         * ipa-chkp.c: Remove.
14910         * ipa-chkp.h: Remove.
14911         * rtl-chkp.c: Remove.
14912         * rtl-chkp.h: Remove.
14913         * tree-chkp-opt.c: Remove.
14914         * tree-chkp.c: Remove.
14915         * tree-chkp.h: Remove.
14917 2018-06-07  Carl Love  <cel@us.ibm.com>
14919         * config/rs6000/vsx.md (vextract_fp_from_shorth,
14920         vextract_fp_from_shortl): Add BE support.
14922 2018-06-07  Paul Koning  <ni1d@arrl.net>
14924         * compare-elim.c (try_merge_compare): Don't merge compare if
14925         address contains a side effect.
14926         (try_eliminate_compare): Likewise.
14928 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
14930         * config.gcc: Support "tremont".
14931         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14932         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14933         PROCESSOR_TREMONT.
14934         * config/i386/i386.c (m_TREMONT): Define.
14935         (processor_target_table): Add "tremont".
14936         (PTA_TREMONT): Define.
14937         (ix86_lea_outperforms): Add TARGET_TREMONT.
14938         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14939         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14940         and M_INTEL_GOLDMONT_PLUS.
14941         (fold_builtin_cpu): Add "tremont".
14942         (ix86_add_stmt_cost): Add TARGET_TREMONT.
14943         (ix86_option_override_internal): Add "tremont".
14944         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14945         (processor_type): Add PROCESSOR_TREMONT.
14946         * config/i386/x86-tune.def: Add m_TREMONT.
14947         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14949 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14951         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14952         symbol defined for msp430i* devices to be lower case.
14954 2018-06-07  Richard Biener  <rguenther@suse.de>
14956         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14957         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
14958         Properly wrap signed arithmetic if overflow wraps.
14960 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
14962         PR tree-optimization/69615
14963         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14964         of a cast from a same precision integral SSA_NAME in a bb dominated
14965         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
14966         cast to utype if rhs2 has already a compatible type.
14968 2018-06-07  Richard Biener  <rguenther@suse.de>
14970         PR tree-optimization/85935
14971         * graphite-scop-detection.c (find_params_in_bb): Analyze
14972         condition operands with respect to the correct loop.  Assert
14973         the analysis doesn't fail.
14975 2018-06-04  Carl Love  <cel@us.ibm.com>
14977         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14978         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
14979         as it is slightly cheaper.
14980         (first_match_or_eos_index_<mode>):
14981         Calculate index using natural element order.
14982         (first_match_index_<mode>):
14983         Calculate index using natural element order.
14984         (first_match_or_eos_index_<mode>):
14985         Calculate index using natural order.
14986         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14987         for BE and LE modes.
14988         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14989         P9V_BUILTIN_VCLZLSBB_V16QI.
14990         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14991         specific.
14993 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14995         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14996         indentation and line wrap for many prototypes.  Add missing
14997         @smallexample directives around block of prototypes for vec_xl and
14998         vec_xst.
15000 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
15002         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
15003         track if we pass or return IEEE 128-bit floating point.
15004         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
15005         C++ mangling that is compatible with GCC 8.1.
15006         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
15007         (init_cumulative_args): Note if we pass or return IEEE 128-bit
15008         floating point types.
15009         (rs6000_function_arg_advance_1): Likewise.
15010         (rs6000_mangle_type): Optionally generate mangled names that match
15011         what GCC 8.1 generated for IEEE 128-bit floating point types.
15012         (rs6000_globalize_decl_name): If we have an external function that
15013         passes or returns IEEE 128-bit types, generate a weak reference
15014         from the mangled name used in GCC 8.1 to the current mangled
15015         name.
15016         (rs6000_init_builtins): Make __ibm128 use the long double type if
15017         long double is IBM extended double.  Make __float128 use the long
15018         double type if long double is IEEE 128-bit.
15020         PR target/85657
15021         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
15022         macro for __ibm128 built-in functions.
15023         (PACK_IF): Add __ibm128 pack/unpack functions.
15024         (UNPACK_IF): Likewise.
15025         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
15026         enable long double built-in functions if long double is IEEE
15027         128-bit floating point.
15028         (rs6000_invalid_builtin): Update long double built-in function
15029         error message.
15030         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
15031         functions, adjust the built-in function to use the long double
15032         built-in function if __ibm128 and long double are the same type.
15033         * doc/extend.texi (PowerPC builtins): Update documention for
15034         __builtin_{,un}pack_longdouble.  Add documentation for
15035         __builtin_{,un}pack_ibm128.
15037 2018-06-06  Jim Wilson  <jimw@sifive.com>
15039         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
15040         (struct machine_function): New field interrupt_mode.
15041         (riscv_handle_type_attribute): New function.  Add forward declaration.
15042         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
15043         (riscv_expand_epilogue): Check interrupt_mode field.
15044         (riscv_set_current_function): Check interrupt attribute args and
15045         set interrupt_mode field.
15046         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
15047         (riscv_sret, riscv_uret): New.
15048         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
15049         new arguments to interrupt attribute.
15051 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
15053         PR target/63177
15054         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
15055         Don't handle -mcpu=power8 if -mpower9-vector is also used.
15057 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15059         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
15060         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
15061         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15062         several redundant entries.
15064 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
15066         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
15067         type from "rtx" to "rtx_insn *".
15068         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
15069         for local "call_insn", removing cast.
15070         (ix86_expand_call): Likewise, introducing a "call_insn" local.
15072 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
15074         PR tree-optimization/86066
15075         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
15076         for BIT_INSERT_EXPR stores.
15078 2018-06-06  Richard Biener  <rguenther@suse.de>
15080         PR tree-optimization/86062
15081         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15082         component refs ontop
15083         of to be offsetted base.
15085 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15087         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
15088         to be static and remove check on interrupt attribute name.
15090 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15092         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15093         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
15095 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
15097         PR target/79924
15098         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
15099         second argument.
15100         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
15101         Remove second argument, change how error is called.
15102         (aarch64_layout_arg): Remove second argument from
15103         aarch64_err_no_fpadvsimd call.
15104         (aarch64_init_cumulative_args): Ditto.
15105         (aarch64_gimplify_va_arg_expr): Ditto.
15106         * config/aarch64/aarch64.md (mov<mode>): Ditto.
15108 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
15110         * config/i386/i386.md (simple_return_indirect_internal): New expander.
15111         (*simple_return_indirect_internal<mode>): Rename from
15112         simple_return_indirect_internal.  Use W mode iterator.
15113         (rstorssp): New expander.
15114         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
15115         (clrssbsy): New expander.
15116         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
15118 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15120         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
15121         __typeof__.
15122         (cmse_check_pointed_object): Likewise.
15124 2018-06-05  Martin Liska  <mliska@suse.cz>
15126         PR gcov-profile/47618
15127         * doc/invoke.texi: Document how -fprofile-dir format
15128         is extended.
15130 2018-06-05  Richard Biener  <rguenther@suse.de>
15132         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
15133         removal pretend DOM info isn't available so we do not update
15134         it and only remove edges, not dominated blocks.  Actually free
15135         DOM info in case we removed something.  Remove unreachable blocks.
15136         (mfb_keep_latches): Work with either DOM info or marked backedges.
15137         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
15138         first.  Mark backedges if DOM info isn't available.
15139         (Re-)compute DOM info after cleanup_control_flow_pre.
15141 2018-06-05  Richard Biener  <rguenther@suse.de>
15143         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
15144         (locus_discrim_hasher::hash): Adjust.
15145         (locus_discrim_hasher::equal): Likewise.
15146         (next_discriminator_for_locus): Work on line directly.
15147         (same_line_p): Pass in expanded locus1 as well.
15148         (assign_discriminators): Avoid redundant location expansions.
15150 2018-06-05  Richard Biener  <rguenther@suse.de>
15152         PR tree-optimization/86046
15153         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
15154         if required after clearing TREE_ADDRESSABLE.
15156 2018-06-05  Richard Biener  <rguenther@suse.de>
15158         PR tree-optimization/86047
15159         * tree-ssa-loop.c (for_each_index): Glob handling of all
15160         decls and constants and really handle all of them.
15162 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15164         PR target/81497
15165         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
15166         qualifier_void_pointer and qualifier_const_void_pointer.
15167         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
15168         (arm_init_builtins): Handle the above.
15169         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
15170         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
15171         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
15172         void intrinsics.
15174 2018-06-05  Martin Liska  <mliska@suse.cz>
15176         * auto-profile.c (read_autofdo_file): Do not use
15177         gcov_ctr_summary struct.
15178         (afdo_callsite_hot_enough_for_early_inline): Likewise.
15179         * coverage.c (struct counts_entry): Likewise.
15180         (read_counts_file): Read just single summary entry.
15181         (get_coverage_counts): Use gcov_summary struct.
15182         * coverage.h (get_coverage_counts): Likewise.
15183         * gcov-dump.c (dump_working_sets): Likewise.
15184         (tag_summary): Dump just single summary.
15185         * gcov-io.c (gcov_write_summary): Write just histogram
15186         summary.
15187         (gcov_read_summary): Read just single summary.
15188         (compute_working_sets): Use gcov_summary struct.
15189         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
15190         of GCOV_COUNTERS_SUMMABLE.
15191         (GCOV_COUNTERS_SUMMABLE): Remove.
15192         (GCOV_FIRST_VALUE_COUNTER): Replace with
15193         GCOV_COUNTER_V_INTERVAL.
15194         (struct gcov_ctr_summary): Remove.
15195         (struct gcov_summary): Directly use fields of former
15196         gcov_ctr_summary.
15197         (compute_working_sets): Use gcov_summary struct.
15198         * gcov.c (read_count_file): Do not use ctrs fields.
15199         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
15200         struct.
15201         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
15202         struct.
15203         * profile.c: Likewise.
15204         * profile.h: Likewise.
15206 2018-06-05  Martin Liska  <mliska@suse.cz>
15208         PR gcov-profile/84846
15209         * gcov.c (output_lines): Print working directory only
15210         in intermediate format.
15212 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
15214         * config/s390/s390-builtin-types.def: Add void function type.
15215         * config/s390/s390-builtins.def: Use the function type for the
15216         tbeginc builtin.
15218 2018-06-04  Jim Wilson  <jimw@sifive.com>
15220         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
15221         to int.
15222         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
15223         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
15224         handle EH_RETURN_DATA_REGNO registers properly.
15225         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
15226         (riscv_expand_epilogue): Update comment.  Change argument name and
15227         type.  Update code to use new name and type.  Pass new args to
15228         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
15229         EXCEPTION_RETURN.
15230         * config/riscv/riscv.md (NORMAL_RETURN): New.
15231         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
15232         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
15233         (eh_return): Call gen_eh_return_internal and emit barrier.
15234         (eh_return_internal): Call riscv_expand_epilogue.
15236 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
15238         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
15239         bit_insertion field and declare can_be_merged_into method.
15240         (merged_store_group::can_be_merged_into): New method.
15241         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
15242         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
15243         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
15245 2018-06-04  Richard Biener  <rguenther@suse.de>
15247         PR tree-optimization/85955
15248         * builtins.c (fold_builtin_sincos): Convert pointers to
15249         destination to appropriate type before dereferencing.
15251 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
15253         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
15255 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
15257         * expr.c (expand_expr_real_1): Force the operand into memory if
15258         its TYPE_MODE is BLKmode and if there is no integer mode for
15259         the number of bits being extracted.
15261 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15263         PR target/85832
15264         PR target/86036
15265         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15266         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
15268 2018-06-04  Richard Biener  <rguenther@suse.de>
15270         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
15271         (cleanup_tree_cfg_noloop): ... single caller.  Do
15272         start_recording_case_labels later.
15274 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
15276         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
15277         to _IMMINTRIN_H_INCLUDED.
15278         * config/i386/pconfigintrin.h: Ditto.
15279         * config/i386/waitpkgintrin.h: Ditto.
15280         * config/i386/immintrin.h: Add includes for sgxintrin.h,
15281         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
15282         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
15283         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
15284         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
15285         waitpkgintrin.h and cldemoteintrin.h.
15287 2018-06-04  Richard Biener  <rguenther@suse.de>
15289         PR tree-optimization/86038
15290         * tracer.c (find_best_successor): Check probability for
15291         being initialized, bail out if not.
15293 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
15295         PR target/86003
15296         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
15297         of bits to ignore when comparing architectures.
15299 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15301         PR tree-optimization/69615
15302         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
15303         maximum or minimum of the type, try to merge it also as if
15304         range1 is + [-, x - 1] or + [x + 1, -].
15306         PR c++/86025
15307         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
15309 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
15311         PR tree-optimization/86034
15312         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
15313         the unsigned bitfield type in a bit insertion sequence if it does not
15314         have a larger precision than the bitfield size.
15315         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
15317 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
15319         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
15321 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15323         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
15324         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
15325         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
15326         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
15328 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15330         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
15331         Disable -fdelete-null-pointer-checks for ELF toolchain.
15333 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15334             Kito Cheng  <kito.cheng@gmail.com>
15336         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
15337         (nds32le-*-*, nds32be-*-*): Integrate checking process.
15338         (nds32*-*-*): Add glibc and uclibc conditions.
15339         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
15340         (TARGET_EXCEPT_UNWIND_INFO): Define.
15341         * config/nds32/elf.h: New file.
15342         * config/nds32/linux.h: New file.
15343         * config/nds32/nds32-elf.opt: New file.
15344         * config/nds32/nds32-linux.opt: New file.
15345         * config/nds32/nds32-fp-as-gp.c
15346         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
15347         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
15348         TARGET_LINUX_ABI.
15349         (nds32_asm_file_end): Ditto.
15350         (nds32_print_operand): Ditto.
15351         (nds32_insert_attributes): Ditto.
15352         (nds32_init_libfuncs): New function.
15353         (TARGET_HAVE_TLS): Define.
15354         (TARGET_INIT_LIBFUNCS): Define.
15355         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
15356         spec content.
15357         (TARGET_ELF): Apply different mcmodel setting.
15358         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
15359         been migrated into elf.h and linux.h files.
15360         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
15361         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
15362         (mcmodel): The content has been migrated into nds32-elf.opt and
15363         nds32-linux.opt files.
15364         * config/nds32/t-elf: New file.
15365         * config/nds32/t-linux: New file.
15367 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15368             Shiva Chen  <shiva0217@gmail.com>
15370         * config/nds32/constants.md (unspec_volatile_element): Add
15371         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
15372         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
15373         optimization.
15374         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
15375         (make_pass_nds32_fp_as_gp): Declare.
15376         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
15377         optmization pass.
15378         (nds32_asm_function_end_prologue): Remove unused asm output.
15379         (nds32_asm_function_begin_epilogue): Remove unused asm output.
15380         (nds32_asm_file_start): Output necessary fp_as_gp information.
15381         (nds32_option_override): Adjust register usage.
15382         (nds32_expand_prologue): Consider fp_as_gp situation.
15383         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
15384         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
15385         (epilogue): Ditto.
15386         (return): Ditto.
15387         (simple_return): Ditto.
15388         (omit_fp_begin): Output special directive for fp_as_gp.
15389         (omit_fp_end): Output special directive for fp_as_gp.
15390         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
15391         mforbid-fp-as-gp): New options.
15393 2018-06-01  Mark Wielaard  <mark@klomp.org>
15395         * dwarf2out.c (dwarf2out_finish): Remove generation of
15396         DW_AT_loclists_base.
15398 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
15400         * gimple-ssa-store-merging.c: Include gimple-fold.h.
15401         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
15402         (struct merged_store_group): Add bit_insertion field.
15403         (dump_char_array): Use standard hexadecimal format.
15404         (merged_store_group::merged_store_group): Set bit_insertion to false.
15405         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
15406         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
15407         also print the mask in the dump file.
15408         (pass_store_merging::gate): Minor tweak.
15409         (imm_store_chain_info::coalesce_immediate): Fix wrong association
15410         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
15411         stores with INTEGER_CST stores.
15412         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
15413         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
15414         and use it throughout.  Generate bit insertion sequences if need be.
15415         (pass_store_merging::process_store): Remove redundant condition.
15416         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
15418 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
15420         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
15421         the 128-bit floating point types.  Fix function comment.
15423 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15425         * config/aarch64/aarch64-simd.md
15426         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
15427         mnemonics.
15428         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
15429         mnemonics.
15431 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15433         PR tree-optimization/85989
15434         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
15435         variable.
15436         (backprop::intersect_uses): Check it when deciding whether this
15437         is a backedge reference.
15438         (backprop::process_block): Add each phi to m_visited_phis
15439         after visiting it, then clear it at the end.
15441 2018-06-01  Richard Biener  <rguenther@suse.de>
15443         * tree-vectorizer.h (vect_dr_stmt): New function.
15444         (vect_get_load_cost): Adjust.
15445         (vect_get_store_cost): Likewise.
15446         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
15447         Use vect_dr_stmt instead of DR_SMTT.
15448         (vect_record_base_alignments): Likewise.
15449         (vect_calculate_target_alignment): Likewise.
15450         (vect_compute_data_ref_alignment): Likewise and make static.
15451         (vect_update_misalignment_for_peel): Likewise.
15452         (vect_verify_datarefs_alignment): Likewise.
15453         (vector_alignment_reachable_p): Likewise.
15454         (vect_get_data_access_cost): Likewise.  Pass down
15455         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
15456         (vect_get_peeling_costs_all_drs): Likewise.
15457         (vect_peeling_hash_get_lowest_cost): Likewise.
15458         (vect_enhance_data_refs_alignment): Likewise.
15459         (vect_find_same_alignment_drs): Likewise.
15460         (vect_analyze_data_refs_alignment): Likewise.
15461         (vect_analyze_group_access_1): Likewise.
15462         (vect_analyze_group_access): Likewise.
15463         (vect_analyze_data_ref_access): Likewise.
15464         (vect_analyze_data_ref_accesses): Likewise.
15465         (vect_vfa_segment_size): Likewise.
15466         (vect_small_gap_p): Likewise.
15467         (vectorizable_with_step_bound_p): Likewise.
15468         (vect_prune_runtime_alias_test_list): Likewise.
15469         (vect_analyze_data_refs): Likewise.
15470         (vect_supportable_dr_alignment): Likewise.
15471         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
15472         (vect_gen_prolog_loop_niters): Likewise.
15473         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
15474         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
15475         modify DR_STMT.
15476         (vect_recog_mask_conversion_pattern): Likewise.
15477         (vect_try_gather_scatter_pattern): Likewise.
15478         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
15479         to vect_get_store_cost.
15480         (vect_get_store_cost): Get stmt_info instead of DR.
15481         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
15482         (vect_get_load_cost): Get stmt_info instead of DR.
15484 2018-06-01  Richard Biener  <rguenther@suse.de>
15486         PR middle-end/86017
15487         * gimple-fold.c (var_decl_component_p): Also allow offsetted
15488         vars wrapped in MEM_REFs.
15490 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15492         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15493         Fix subreg tests so that we only return a choice between
15494         GENERAL_REGS and FP_REGS if the original classes included both.
15496 2018-06-01  Richard Biener  <rguenther@suse.de>
15498         PR ipa/85960
15499         * tree-ssa-structalias.c (get_function_part_constraint):
15500         Handle NULL fi->decl.
15501         (find_func_aliases_for_call): Properly handle indirect
15502         fi from direct call.
15503         (find_func_clobbers): Likewise.
15504         (ipa_pta_execute): Likewise.
15505         (create_variable_info_for): For functions that are ifunc_resolver
15506         resolve to a varinfo that contains the result of the resolver call.
15507         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
15508         aliases.
15510 2018-05-31  Michael Collison  <michael.collison@arm.com>
15512         * config/aarch64/aarch64.md:
15513         (*fix_to_zero_extenddfdi2): New pattern.
15514         * gcc.target/aarch64/fix_extend1.c: New testcase.
15516 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
15518         PR middle-end/78809
15519         PR middle-end/83026
15520         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
15521         and BUILT_IN_STRNCMP_EQ.
15522         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
15523         BUILT_IN_STRNCMP_EQ.
15524         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
15525         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15526         (gimple_fold_builtin): Likewise.
15527         * tree-ssa-strlen.c (compute_string_length): New function.
15528         (determine_min_obsize): New function.
15529         (handle_builtin_string_cmp): New function to handle calls to
15530         string compare functions.
15531         (strlen_optimize_stmt): Add handling to builtin string compare
15532         calls.
15533         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
15534         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15535         * tree.c (build_common_builtin_nodes): Add new defines of
15536         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
15538 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
15540         PR target/85984
15541         * bb-reorder.c (pass_partition_blocks::gate): Return false for
15542         functions with naked attribute.
15544 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
15546         * config/i386/sse.md (avx_vec_concat<mode>):
15547         Substitute concat_tg_mode mode attribute with xtg_mode.
15548         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
15549         (concat_tg_mode): Remove mode attribute.
15551 2018-05-31  Martin Sebor  <msebor@redhat.com>
15553         PR c/82063
15554         * calls.c (alloc_max_size): Correct a logic error/typo.
15555         Treat excessive arguments as infinite.  Warn for invalid arguments.
15556         * doc/invoke.texi (-Walloc-size-larger-than): Update.
15558 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
15560         PR target/85829
15561         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
15562         and movx for Haswell.
15564 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
15565             Cesar Philippidis  <cesar@codesourcery.com>
15567         PR middle-end/85879
15568         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
15569         when emitting error on private/firstprivate reductions.
15570         * omp-low.c (lower_omp_target): Avoid reference-type processing
15571         on pointers for firstprivate clause.
15573 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15575         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15576         (st1x2): Likewise.
15577         (st1x3): Likewise.
15578         * config/aarch64/aarch64-simd.md
15579         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
15580         (aarch64_ld1_x3_<mode>): Likewise
15581         (aarch64_st1x2<VALLDIF:mode>): Likewise
15582         (aarch64_st1_x2_<mode>): Likewise
15583         (aarch64_st1x3<VALLDIF:mode>): Likewise
15584         (aarch64_st1_x3_<mode>): Likewise
15585         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
15586         (vld1_s8_x3): Likewise.
15587         (vld1_u16_x3): Likewise.
15588         (vld1_s16_x3): Likewise.
15589         (vld1_u32_x3): Likewise.
15590         (vld1_s32_x3): Likewise.
15591         (vld1_u64_x3): Likewise.
15592         (vld1_s64_x3): Likewise.
15593         (vld1_f16_x3): Likewise.
15594         (vld1_f32_x3): Likewise.
15595         (vld1_f64_x3): Likewise.
15596         (vld1_p8_x3): Likewise.
15597         (vld1_p16_x3): Likewise.
15598         (vld1_p64_x3): Likewise.
15599         (vld1q_u8_x3): Likewise.
15600         (vld1q_s8_x3): Likewise.
15601         (vld1q_u16_x3): Likewise.
15602         (vld1q_s16_x3): Likewise.
15603         (vld1q_u32_x3): Likewise.
15604         (vld1q_s32_x3): Likewise.
15605         (vld1q_u64_x3): Likewise.
15606         (vld1q_s64_x3): Likewise.
15607         (vld1q_f16_x3): Likewise.
15608         (vld1q_f32_x3): Likewise.
15609         (vld1q_f64_x3): Likewise.
15610         (vld1q_p8_x3): Likewise.
15611         (vld1q_p16_x3): Likewise.
15612         (vld1q_p64_x3): Likewise.
15613         (vst1_s64_x2): Likewise.
15614         (vst1_u64_x2): Likewise.
15615         (vst1_f64_x2): Likewise.
15616         (vst1_s8_x2): Likewise.
15617         (vst1_p8_x2): Likewise.
15618         (vst1_s16_x2): Likewise.
15619         (vst1_p16_x2): Likewise.
15620         (vst1_s32_x2): Likewise.
15621         (vst1_u8_x2): Likewise.
15622         (vst1_u16_x2): Likewise.
15623         (vst1_u32_x2): Likewise.
15624         (vst1_f16_x2): Likewise.
15625         (vst1_f32_x2): Likewise.
15626         (vst1_p64_x2): Likewise.
15627         (vst1q_s8_x2): Likewise.
15628         (vst1q_p8_x2): Likewise.
15629         (vst1q_s16_x2): Likewise.
15630         (vst1q_p16_x2): Likewise.
15631         (vst1q_s32_x2): Likewise.
15632         (vst1q_s64_x2): Likewise.
15633         (vst1q_u8_x2): Likewise.
15634         (vst1q_u16_x2): Likewise.
15635         (vst1q_u32_x2): Likewise.
15636         (vst1q_u64_x2): Likewise.
15637         (vst1q_f16_x2): Likewise.
15638         (vst1q_f32_x2): Likewise.
15639         (vst1q_f64_x2): Likewise.
15640         (vst1q_p64_x2): Likewise.
15641         (vst1_s64_x3): Likewise.
15642         (vst1_u64_x3): Likewise.
15643         (vst1_f64_x3): Likewise.
15644         (vst1_s8_x3): Likewise.
15645         (vst1_p8_x3): Likewise.
15646         (vst1_s16_x3): Likewise.
15647         (vst1_p16_x3): Likewise.
15648         (vst1_s32_x3): Likewise.
15649         (vst1_u8_x3): Likewise.
15650         (vst1_u16_x3): Likewise.
15651         (vst1_u32_x3): Likewise.
15652         (vst1_f16_x3): Likewise.
15653         (vst1_f32_x3): Likewise.
15654         (vst1_p64_x3): Likewise.
15655         (vst1q_s8_x3): Likewise.
15656         (vst1q_p8_x3): Likewise.
15657         (vst1q_s16_x3): Likewise.
15658         (vst1q_p16_x3): Likewise.
15659         (vst1q_s32_x3): Likewise.
15660         (vst1q_s64_x3): Likewise.
15661         (vst1q_u8_x3): Likewise.
15662         (vst1q_u16_x3): Likewise.
15663         (vst1q_u32_x3): Likewise.
15664         (vst1q_u64_x3): Likewise.
15665         (vst1q_f16_x3): Likewise.
15666         (vst1q_f32_x3): Likewise.
15667         (vst1q_f64_x3): Likewise.
15668         (vst1q_p64_x3): Likewise.
15670 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15672         * config/msp430/msp430.c (msp430_output_labelref): Prepend
15673         user_label_prefix to name.
15675         * tree-core.h: Update comment about the format of NAME string
15676         passed to handler in attribute_spec.
15678         * config/msp430/msp430.md: Remove erroneous subreg expression from
15679         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15680         zero_extend{q,h}isi2.
15682 2018-05-30  Borislav Petkov  <bp@suse.de>
15684         * doc/extend.texi: Document some architecture specific
15685         constraints and sort entries.
15687 2018-05-30  Martin Sebor  <msebor@redhat.com>
15689         PR middle-end/85369
15690         * builtins.c (expand_builtin_stpcpy_1): New function.
15691         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15692         only if the former succeeds.
15694 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15696         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15697         in saphira.
15699 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15701         * doc/invoke.texi (-flinker-output): Document
15703 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15705         * passes.c (ipa_write_summaries): Only modify statements if body
15706         is in memory.
15707         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15708         incrementally linking.
15709         (ipa_passes): Likewise.
15710         * lto-cgraph.c (lto_output_node): When incrementally linking do not
15711         pass down resolution info.
15712         * common.opt (flag_incremental_link): Update info.
15713         * gcc.c (plugin specs): Turn flinker-output=* to
15714         -plugin-opt=-linker-output-known
15715         * toplev.c (compile_file): Also cut compilation when doing incremental
15716         link.
15717         * flag-types. (enum lto_partition_model): Add
15718         LTO_LINKER_OUTPUT_NOLTOREL.
15719         (invoke.texi): Add -flinker-output docs.
15720         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15721         link same way as WPA; do not stream in dead initializers.
15723         * dwarf2out.c (dwarf2out_die_ref_for_decl,
15724         darf2out_register_external_decl): Support incremental link.
15726 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15728         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15730 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15732         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15733         it down to simple_object_copy_lto_debug_sections.
15734         (run_gcc): Determine incremental LTO link time and configure
15735         lto1 into non-wpa mode, disable renaming of debug sections.
15737 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15739         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15740         descriptions of various incorrectly documented functions.
15742 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15744         Revert:
15745         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15746         address check not strict.
15748 2018-05-30  Richard Biener  <rguenther@suse.de>
15750         PR tree-optimization/85964
15751         * tracer.c (better_p): Drop initialized count check, we only
15752         call the function with initialized counts now.
15753         (find_best_successor): Do find a best edge if one
15754         has uninitialized count.
15755         (find_best_predecessor): Likewise.  Do BB frequency check only
15756         if count is initialized.
15758 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
15760         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15761         (aarch64_ldrstr_offset_compare): New.
15762         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15763         load/store orderings.
15764         (aarch64_gen_adjusted_ldpstp): Likewise.
15766 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
15768         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15769         Check for subset of GENERAL_REGS and FP_REGS.
15770         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15771         r=w alternative.
15773 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
15775         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15776         and wi::to_poly_offset.  Add the current offset and then check
15777         whether the sum fits, rather than using an unchecked addition of
15778         a checked term.  Check for a shwi rather than a uhwi.
15779         * expr.c (get_bit_range): Use tree_to_poly_uint64.
15780         (store_constructor): Use poly_int_tree_p.
15781         (expand_expr_real_1): Likewise.
15782         * function.c (assign_temp): Likewise.
15783         * fold-const.c (const_binop): Use poly_int_tree_p and
15784         wi::to_poly_offset.
15785         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
15786         division.
15787         * ipa-icf-gimple.c (func_checker::compare_operand): Use
15788         to_poly_offset for MEM offsets.
15789         * ipa-icf.c (sem_variable::equals): Likewise.
15790         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15791         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15792         wi::to_poly_offset for BIT_FIELD_REF offsets.
15793         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15794         wi::to_poly_offset.
15795         * var-tracking.c (emit_note_insn_var_location): Use
15796         tree_to_poly_uint64.
15798 2018-05-29  Jim Wilson  <jimw@sifive.com>
15800         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15802 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
15804         PR target/85950
15805         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15806         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15807         sequence.
15808         (sse4_1_round<mode>2): Use nonimmediate_operand
15809         for operand 1 predicate.
15811 2018-05-29  Martin Sebor  <msebor@redhat.com>
15812             Richard Biener  <rguenther@suse.de>
15814         PR testsuite/85888
15815         * calls.c (get_size_range): Call determine_value_range instead
15816         of get_value_range..
15817         * tree-vrp.h (determine_value_range): Declared new function.
15818         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15820 2018-05-29  Richard Biener  <rguenther@suse.de>
15822         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15823         sure to use non-pattern stmts for get_earlier_stmt arguments.
15824         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15825         called on pattern stmts.
15826         (get_later_stmt): Likewise.
15828 2018-05-29  Martin Liska  <mliska@suse.cz>
15830         PR gcov-profile/85759
15831         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15832         env variables.
15834 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
15836         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15837         VEC_UNPACK_*_EXPR.
15838         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15839         VEC_PACK_*_EXPR.
15841         PR target/85918
15842         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15843         VEC_PACK_FLOAT_EXPR): New tree codes.
15844         * tree-pretty-print.c (op_code_prio): Handle
15845         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15846         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15847         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15848         * tree-inline.c (estimate_operator_cost): Likewise.
15849         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15850         * fold-const.c (const_binop): Likewise.
15851         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15852         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15853         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15854         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15855         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15856         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15857         * expr.c (expand_expr_real_2): Likewise.
15858         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15859         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15860         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15861         optabs.
15862         * optabs.c (expand_widen_pattern_expr): For
15863         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15864         sign from result type rather than operand's type.
15865         (expand_binop_directly): For vec_packu_float_optab and
15866         vec_packs_float_optab allow result type to be different from operand's
15867         type.
15868         * optabs-tree.c (optab_for_tree_code): Handle
15869         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15870         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
15871         * tree-vect-generic.c (expand_vector_operations_1):  Handle
15872         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15873         VEC_PACK_FLOAT_EXPR.
15874         * tree-vect-stmts.c (supportable_widening_operation): Handle
15875         FIX_TRUNC_EXPR.
15876         (supportable_narrowing_operation): Handle FLOAT_EXPR.
15877         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15878         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15879         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
15880         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15881         mode attributes.
15882         (vec_pack<floatprefix>_float_<mode>): New expander.
15883         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15884         attributes.
15885         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15886         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15887         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15888         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15889         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15890         Document.
15891         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15892         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15893         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15894         VEC_PACK_FLOAT_EXPR): Document.
15896 2018-05-29  Richard Biener  <rguenther@suse.de>
15898         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15899         member.
15900         (stmt_vec_info_vec): Make pointer.
15901         (init_stmt_vec_info_vec): Remove.
15902         (free_stmt_vec_info_vec): Likewise.
15903         (set_stmt_vec_info_vec): New function.
15904         (free_stmt_vec_infos): Likewise.
15905         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15906         (set_vinfo_for_stmt): Likewise.
15907         (get_earlier_stmt): Likewise.
15908         (get_later_stmt): Likewise.
15909         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15910         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15911         (vec_info::~vec_info): Free stmt_vec_infos.
15912         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15913         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15914         (pass_slp_vectorize::execute): Likewise.
15915         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15916         (free_stmt_vec_info_vec): Likewise.
15917         (set_stmt_vec_info_vec): New function.
15918         (free_stmt_vec_infos): Likewise.
15919         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15920         the global stmt_vec_info_vec.
15921         * tree-parloops.c (gather_scalar_reductions): Use
15922         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15923         vector.
15925 2018-05-29  Richard Biener  <rguenther@suse.de>
15927         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15929 2018-05-29  Martin Liska  <mliska@suse.cz>
15930             David Malcolm  <dmalcolm@redhat.com>
15932         * vec.c (test_reverse): New.
15933         (vec_c_tests): Add new test.
15934         * vec.h (vl_ptr>::reverse): New function.
15936 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
15938         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15940         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15941         and later.
15943 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15945         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15947 2018-05-28  Richard Biener  <rguenther@suse.de>
15949         PR tree-optimization/85933
15950         * tree-vect-data-refs.c (vect_record_base_alignments): Only
15951         look at stmts marked as vectorizable.
15953 2018-05-28  Richard Biener  <rguenther@suse.de>
15955         PR tree-optimization/85934
15956         * tree-vect-generic.c (expand_vector_operations_1): Hoist
15957         vector boolean check before scalar optimization.
15959 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
15961         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15962         for armv5te.
15964 2018-05-28  Mark Wielaard  <mark@klomp.org>
15966         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15967         if it is an expression containing a minus sign.
15969 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
15971         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15973 2018-05-27  Paul Koning  <ni1d@arrl.net>
15975         * config/pdp11/pdp11.md (truncsihi2): Remove.
15977 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
15978             Chung-Ju Wu  <jasonwucj@gmail.com>
15980         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15981         implementation.
15982         (unaligned_store_dw): Ditto.
15983         * config/nds32/nds32-memory-manipulation.c
15984         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15985         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15986         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15987         (emit_setmem_word_loop): Rename to ...
15988         (emit_setmem_doubleword_loop): ... this.
15989         (nds32_gen_dup_4_byte_to_word_value): New function.
15990         (nds32_gen_dup_8_byte_to_double_word_value): New function.
15991         (nds32_expand_setmem_loop): Refine implementation.
15992         (nds32_expand_setmem_loop_v3m): Ditto.
15993         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15994         pattern.
15996 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15998         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
16000 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
16002         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
16003         (nds32_init_machine_status): Initialize machine->attr_naked_p and
16004         machine->attr_no_prologue_p.
16005         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
16006         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
16007         (nds32_expand_epilogue): Consider attr_naked_p.
16008         (nds32_expand_epilogue_v3pop): Likewise.
16009         (nds32_can_use_return_insn): Likewise.
16010         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
16011         attr_no_prologue_p fields.
16012         * config/nds32/nds32.opt (mret-in-naked-func): New option.
16014 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
16016         PR target/85918
16017         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
16018         attributes.
16019         * config/i386/sse.md
16020         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
16021         Rename to ...
16022         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
16023         ... this.
16024         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
16025         Rename to ...
16026         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
16027         ... this.
16028         (*<floatsuffix>floatv2div2sf2): Rename to ...
16029         (*float<floatunssuffix>v2div2sf2): ... this.
16030         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
16031         (float<floatunssuffix>v2div2sf2_mask): ... this.
16032         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
16033         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
16034         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
16035         to ...
16036         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
16037         ... this.
16038         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16039         Rename to ...
16040         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16041         ... this.
16042         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16043         Rename to ...
16044         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16045         ... this.
16046         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
16047         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
16048         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
16049         gen_ufix_truncv8dfv8si2.
16050         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
16051         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
16052         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
16053         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
16054         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
16055         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
16056         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
16057         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
16059 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
16061         PR target/85900
16062         PR target/85345
16063         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
16065 2018-05-25  Jim Wilson  <jimw@sifive.com>
16067         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
16068         * config/riscv/riscv.c (struct machine_function): Add
16069         interrupt_handler_p and attribute_checked_p fields.
16070         (riscv_attribute_table): Add interrupt.
16071         (riscv_interrupt_type_p): New.
16072         (riscv_save_reg_p): Save extra regs for interrupt handler.
16073         (riscv_use_save_libcall): Return false  for interrupt handler.
16074         (riscv_first_stack_step): Add forward declaration.
16075         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
16076         for interrupt handler with large frame.  Use it for saved reg list.
16077         (riscv_expand_prologue): Move flag_stack_usage_info support to
16078         eliminate duplication.
16079         (riscv_expand_epilogue): Generate mret for interrupt handler.
16080         (riscv_epilogue_uses): New.
16081         (riscv_can_use_return_insn): Return false for interrupt handler.
16082         (riscv_function_ok_for_sibcall): Likewise.
16083         (riscv_set_current_function): Add interrupt handler support.
16084         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
16085         * config/riscv/riscv.md (UNSPECV_MRET): New.
16086         (GP_REGNUM): New.
16087         (riscv_frflags, riscv_fsflags): Use tab after opcode.
16088         (riscv_mret): New.
16089         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
16091 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
16093         PR tree-optimization/85712
16094         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
16095         this candidate has already been replaced in-situ by a copy.
16097 2018-05-25  Jason Merrill  <jason@redhat.com>
16099         PR c++/80485 - inline function non-zero address.
16100         * symtab.c (nonzero_address): Check DECL_COMDAT.
16102 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
16104         PR target/83628
16105         * config/alpha/alpha.md (ashlsi3): New insn pattern.
16106         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
16107         extension of SImode operation.  Use const123_operand predicate.
16108         (*saddsi_1): Remove.
16109         (*saddl_se_1): Ditto.
16110         (*ssubsi_1): Ditto.
16111         (*ssubl_se_1): Ditto.
16112         * config/alpha/predicates.md (const123_operand): New predicate.
16113         * config/alpha/constraints.md (P): Use IN_RANGE.
16115 2018-05-25  Richard Biener  <rguenther@suse.de>
16117         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
16118         defaulted to true.
16119         (ref_maybe_used_by_stmt_p): Likewise.
16120         (stmt_may_clobber_ref_p): Likewise.
16121         (stmt_may_clobber_ref_p_1): Likewise.
16122         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
16123         and pass it along.
16124         (ref_maybe_used_by_stmt_p): Likewise.
16125         (stmt_may_clobber_ref_p): Likewise.
16126         (stmt_may_clobber_ref_p_1): Likewise.
16127         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
16128         the alias oracle to disambiguate DRs with stmts DR analysis
16129         couldn't handle.
16130         (vect_analyze_data_refs): Do not give up on not analyzable
16131         DRs for BB vectorization.  Remove code truncating the dataref
16132         vector.
16134 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
16136         PR target/85832
16137         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16138         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
16139         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
16141 2018-05-25  Richard Biener  <rguenther@suse.de>
16143         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
16144         function, combining stmt data ref gathering and fatal analysis
16145         parts.
16146         (vect_analyze_data_refs): Remove now redudnant code and simplify.
16147         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
16148         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
16149         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
16150         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
16152 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16154         PR tree-optimization/85720
16155         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
16156         SCC if all partitions are builtins.
16157         (version_loop_by_alias_check): New parameter.  Generate cancelable
16158         runtime alias check if all partitions are builtins.
16159         (distribute_loop): Update call to above function.
16161 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16163         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
16164         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
16165         (parm_default_def_partition_arg): Ditto.
16166         (set_parm_default_def_partition): Ditto.
16167         (get_parm_default_def_partitions): Ditto and make it static.
16168         (get_undefined_value_partitions): Ditto and make it static.
16169         (remove_ssa_form): Refactor call to init_var_map here.
16170         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
16171         computation for loop region.
16172         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
16173         (register_default_def): Delete.
16174         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
16175         (parm_default_def_partition_arg): Ditto.
16176         (set_parm_default_def_partition): Ditto.
16177         (get_parm_default_def_partitions): Ditto and make it static.
16178         (get_undefined_value_partitions): Ditto and make it static.
16179         (coalesce_with_default, coalesce_with_default): Update comment.
16180         (create_coalesce_list_for_region): New func factored out from
16181         create_outofssa_var_map.
16182         (populate_coalesce_list_for_outofssa): New func factored out from
16183         create_outofssa_var_map and coalesce_ssa_name.
16184         (create_outofssa_var_map): Delete.
16185         (coalesce_ssa_name): Refactor to support live range computation.
16186         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
16187         (get_parm_default_def_partitions): Delete.
16188         (get_undefined_value_partitions): Ditto.
16189         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
16190         computation for loop region.
16191         (new_tree_live_info, loe_visit_block): Ditto.
16192         (live_worklist, set_var_live_on_entry): Ditto.
16193         (calculate_live_on_exit, verify_live_on_entry): Ditto.
16194         * tree-ssa-live.h (struct _var_map): New fields.
16195         (init_var_map): Change decl.
16196         (region_contains_p): New.
16198 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16200         * tree-ssa-live.h (live_merge_and_clear): Delete.
16202 2018-05-25  Richard Biener  <rguenther@suse.de>
16204         PR c++/85912
16205         * tree-dump.c (dequeue_and_dump): Remove access to removed
16206         operand 2 of a SWITCH_EXPR.
16208 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16210         * doc/sourcebuild.texi (vect_double_cond_arith): Include
16211         multiplication and division.
16212         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
16213         (cond_udiv@var{m}, cond_umod@var{m}): Document.
16214         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
16215         (cond_udiv_optab, cond_umod_optab): New optabs.
16216         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
16217         (IFN_COND_RDIV): New internal functions.
16218         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
16219         TRUNC_MOD_EXPR and RDIV_EXPR.
16220         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
16221         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
16222         New unspecs.
16223         (SVE_INT_BINARY): Include mult.
16224         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
16225         (optab, sve_int_op): Handle mult.
16226         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
16227         UNSPEC_COND_DIV.
16228         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
16229         for SVE_INT_BINARY_SD.
16231 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16233         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
16234         (optab, sve_int_op): Handle div and udiv.
16235         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
16236         for SVE_INT_BINARY_SD.
16237         (*<optab><mode>3): New insn for the same.
16239 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16241         * tree-vect-patterns.c: Include predict.h.
16242         (vect_recog_divmod_pattern): Restrict check for division support
16243         to when optimizing for size.
16245 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16247         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
16248         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
16249         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
16250         (gimple_match_op::set_op): Likewise.
16251         (gimple_resimplify4): Declare.
16252         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
16253         (expr::gen_transform): Likewise.
16254         (decision_tree::gen): Generate a simplification routine for 4 operands.
16255         * gimple-match-head.c (gimple_simplify): Add an overload for
16256         4 operands.  In the top-level function, handle up to 4 call
16257         arguments and call gimple_resimplify4.
16258         (gimple_resimplify4): New function.
16259         (build_call_internal): Pass a fourth operand.
16260         (maybe_push_to_seq): Likewise.
16261         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
16262         Fold VEC_COND_EXPRs of an operation and a default value into
16263         an IFN_COND_* function if possible.
16264         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
16265         New unspecs.
16266         (SVE_COND_FP_BINARY): Include them.
16267         (optab, sve_fp_op): Handle them.
16268         (SVE_INT_BINARY_REV): New code iterator.
16269         (SVE_COND_FP_BINARY_REV): New int iterator.
16270         (commutative): New int attribute.
16271         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
16272         Declare.
16273         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
16274         function.
16275         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
16276         (*cond_<optab><mode>): New patterns for reversed operands.
16278 2018-05-25  Richard Biener  <rguenther@suse.de>
16280         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
16281         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
16282         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
16283         (STMT_VINFO_GROUPED_ACCESS): Adjust.
16284         * tree-vect-data-refs.c (everywhere): Adjust users.
16285         * tree-vect-loop.c (everywhere): Likewise.
16286         * tree-vect-slp.c (everywhere): Likewise.
16287         * tree-vect-stmts.c (everywhere): Likewise.
16288         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
16290 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16292         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
16293         Rename to...
16294         (gcc_cv_as_section_exclude): ... this.
16295         Try Solaris as #exclude syntax.
16296         * configure: Regenerate.
16297         * config.in: Regenerate.
16298         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
16299         SECTION_EXCLUDE.
16300         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
16301         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
16303         * varasm.c (default_elf_asm_named_section): Don't check if
16304         HAVE_GAS_SECTION_EXCLUDE is defined.
16306 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16308         * doc/md.texi: Update the documentation of the cond_* optabs
16309         to mention the new final operand.  Fix GET_MODE_NUNITS call.
16310         Describe the scalar case too.
16311         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
16312         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
16313         instead of 2.
16314         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
16315         (get_conditional_internal_fn): Update comment.
16316         * tree-vect-loop.c (vectorizable_reduction): Pass the original
16317         accumulator value as a final argument to conditional functions.
16318         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
16319         a define_expand and add an "else" operand.  Assert for now that
16320         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
16321         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
16322         (*cond_<optab><mode>): New patterns.
16323         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
16324         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16325         (UNSPEC_COND_EOR): Delete.
16326         (optab): Remove associated mappings.
16327         (SVE_INT_BINARY): New code iterator.
16328         (sve_int_op): Remove int attribute and add "minus" to the code
16329         attribute.
16330         (SVE_COND_INT_OP): Delete.
16331         (SVE_COND_FP_OP): Rename to...
16332         (SVE_COND_FP_BINARY): ...this.
16334 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16336         * optabs.c (can_reuse_operands_p): New function.
16337         (maybe_legitimize_operands): Try to reuse the results for
16338         earlier operands.
16340 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16342         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
16343         Add {q} suffix to insn mnemonic.
16345 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16347         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
16348         (msp430_warn_func_return): New.
16350 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
16352         * fold-const.c (tree_nonzero_bits): New function.
16353         * fold-const.h (tree_nonzero_bits): Likewise.
16354         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
16355         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
16357 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
16359         PR target/85900
16360         PR target/85345
16361         * varasm.c (assemble_alias): Check ifunc_resolver only on
16362         FUNCTION_DECL.
16364 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16366         PR target/85903
16367         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
16368         when memory input operand is handled.
16370 2018-05-24  Luis Machado  <luis.machado@linaro.org>
16372         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
16373         global.
16374         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
16376 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16378         * match.pd: Delay FMA folds until after vectorization.
16380 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16382         PR target/83009
16383         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16384         address check not strict.
16386 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16388         * gimple-match.h (gimple_match_op): New class.
16389         (mprts_hook): Replace parameters with a gimple_match_op *.
16390         (maybe_build_generic_op): Likewise.
16391         (gimple_simplified_result_is_gimple_val): Replace parameters with
16392         a const gimple_match_op *.
16393         (gimple_simplify): Replace code_helper * and tree * parameters with
16394         a gimple_match_op * parameter.
16395         (gimple_resimplify1): Replace code_helper *, tree and tree *
16396         parameters with a gimple_match_op * parameter.
16397         (gimple_resimplify2): Likewise.
16398         (gimple_resimplify3): Likewise.
16399         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
16400         parameters with a gimple_match_op * parameter.
16401         * gimple-match-head.c (gimple_simplify): Change prototypes of
16402         auto-generated functions to take a gimple_match_op * instead of
16403         separate code_helper * and tree * parameters.  Make the same
16404         change in the top-level overload and update calls to the
16405         gimple_resimplify routines.  Update calls to the auto-generated
16406         functions and to maybe_push_res_to_seq in the publicly-facing
16407         operation-specific gimple_simplify overloads.
16408         (gimple_match_op::MAX_NUM_OPS): Define.
16409         (gimple_resimplify1): Replace rcode and ops with a single res_op
16410         parameter.  Update call to gimple_simplify.
16411         (gimple_resimplify2): Likewise.
16412         (gimple_resimplify3): Likewise.
16413         (mprts_hook): Replace parameters with a gimple_match_op *.
16414         (maybe_build_generic_op): Likewise.
16415         (build_call_internal): Replace type, nargs and ops with
16416         a gimple_match_op *.
16417         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
16418         with a single gimple_match_op *.  Update calls to mprts_hook,
16419         build_call_internal and gimple_simplified_result_is_gimple_val.
16420         Factor out code that is common to the tree_code and combined_fn cases.
16421         * genmatch.c (expr::gen_transform): Replace tem_code and
16422         tem_ops with a gimple_match_op called tem_op.  Update calls
16423         to the gimple_resimplify functions and maybe_push_res_to_seq.
16424         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
16425         res_ops.  Update call to the gimple_resimplify functions.
16426         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
16427         (decision_tree::gen): Make the functions take a gimple_match_op *
16428         called res_op instead of separate res_code and res_ops parameters.
16429         Update call accordingly.
16430         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
16431         and ops with a single res_op parameter.  Update calls to
16432         maybe_build_generic_op and maybe_push_res_to_seq.
16433         (fold_stmt_1): Update calls to gimple_simplify and
16434         replace_stmt_with_simplification.
16435         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
16436         and gimple_simplified_result_is_gimple_val.
16437         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
16438         gimple_simplify.
16439         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
16440         with a gimple_match_op *.
16441         (vn_nary_build_or_lookup): Likewise.  Update call to
16442         vn_nary_build_or_lookup_1.
16443         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
16444         gimple_match_op *.  Update calls to the gimple_resimplify routines
16445         and to gimple_simplified_result_is_gimple_val.
16446         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
16447         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
16448         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
16449         (visit_nary_op): Likewise.
16450         (visit_reference_op_load): Likewise.
16452 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16454         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
16455         modifier for printing the step amount.
16457 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
16459         PR target/78849
16460         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
16461         types.
16463 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
16465         * doc/sourcebuild.texi (Endianness): New subsubsection.
16467 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16469         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16470         <prefetch_dynamic_strides>: New const bool field.
16471         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16472         prefetch_dynamic_strides.
16473         (exynosm1_prefetch_tune): Likewise.
16474         (thunderxt88_prefetch_tune): Likewise.
16475         (thunderx_prefetch_tune): Likewise.
16476         (thunderx2t99_prefetch_tune): Likewise.
16477         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
16478         false.
16479         (aarch64_override_options_internal): Update to set
16480         PARAM_PREFETCH_DYNAMIC_STRIDES.
16481         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16482         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16483         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16484         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16485         prefetch-dynamic-strides setting.
16487 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16489         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16490         <minimum_stride>: New const int field.
16491         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16492         minimum_stride field defaulting to -1.
16493         (exynosm1_prefetch_tune): Likewise.
16494         (thunderxt88_prefetch_tune): Likewise.
16495         (thunderx_prefetch_tune): Likewise.
16496         (thunderx2t99_prefetch_tune): Likewise.
16497         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
16498         <default_opt_level>: Set to 3.
16499         (aarch64_override_options_internal): Update to set
16500         PARAM_PREFETCH_MINIMUM_STRIDE.
16501         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16502         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16503         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16504         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16505         stride is constant and is below the minimum stride threshold.
16507 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16509         * config/arm/arm-cpus.in (mode26): Delete.
16510         (armv4): Delete mode26 reference.
16511         * config/arm/arm.c (arm_configure_build_target): Delete use of
16512         isa_bit_mode26.
16514 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16516         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
16517         New insn pattern.
16518         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
16519         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
16520         for non-SSE modes.
16521         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
16522         (floatunsdidf2): Ditto.
16524 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16526         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
16527         (fixuns_trunc<mode>si2_avx512f): Ditto.
16528         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
16529         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
16530         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
16532 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16534         PR rtl-optimization/79985
16535         * df-scan.c (df_insn_refs_collect): Remove special case for
16536         global registers and asm statements.
16538 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16540         * extend.texi (Global Register Variables): Rewrite the bullet list.
16541         Note that the register is available for allocation. Note that access
16542         via inline asm must use constraints. Add note about async-signal
16543         handlers. Remove paragraph about automagic register selection.
16545 2018-05-23  Richard Biener  <rguenther@suse.de>
16547         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
16548         of fixed offset from memset VN.
16550 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
16552         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
16553         first_interp field.
16554         (alloc_cand_and_find_basis): Initialize first_interp field.
16555         (slsr_process_mul): Modify first_interp field.
16556         (slsr_process_add): Likewise.
16557         (slsr_process_cast): Modify first_interp field for each new
16558         interpretation.
16559         (slsr_process_copy): Likewise.
16560         (dump_candidate): Dump first_interp field.
16561         (replace_mult_candidate): Process all interpretations, not just
16562         subsequent ones.
16563         (replace_rhs_if_not_dup): Likewise.
16564         (replace_one_candidate): Likewise.
16566 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
16568         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
16569         Add new boolean.
16570         (aarch64_needs_frame_chain): New function.
16571         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
16573 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
16575         PR target/84882
16576         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
16577         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
16578         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
16579         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
16580         as true for strict-align.
16581         (aarch64_can_inline_p): Perform checks even when callee has no
16582         attributes to check for strict alignment.
16583         * doc/extend.texi (AArch64 Function Attributes): Document
16584         no-strict-align.
16585         * doc/invoke.texi: (AArch64 Options): Likewise.
16587 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
16589         PR tree-optimization/85853
16590         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
16591         the handling of the root of the node to...
16592         (vect_slp_analyze_node_operations_1): ...this new function,
16593         and run the whole thing with the child nodes' def types
16594         set according to their SLP node's def type.
16596 2018-05-23  Richard Biener  <rguenther@suse.de>
16598         PR middle-end/85874
16599         * tree-data-ref.c (create_runtime_alias_checks): Defer
16600         and ignore overflow warnings.
16602 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
16604         PR tree-optimization/85822
16605         * tree-vrp.c (is_masked_range_test): Fix handling of negative
16606         constants.
16608 2018-05-23  Richard Biener  <rguenther@suse.de>
16610         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16611         memset constants via native_interpret_expr.
16613 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
16615         PR target/85345
16616         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
16617         attribute.
16618         (cgraph_node::create_alias): Likewise.
16619         (cgraph_node::get_availability): Check ifunc_resolver instead
16620         of looking up ifunc attribute.
16621         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
16622         * varasm.c (do_assemble_alias): Likewise.
16623         (assemble_alias): Likewise.
16624         (default_binds_local_p_3): Likewise.
16625         * cgraph.h (cgraph_node): Add ifunc_resolver.
16626         (cgraph_node::only_called_directly_or_aliased_p): Return false
16627         for IFUNC resolver.
16628         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16629         attribute.
16630         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16631         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16632         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16633         instead of looking up ifunc attribute.
16635 2018-05-22  Luis Machado  <luis.machado@linaro.org>
16637         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16639 2018-05-22  Martin Sebor  <msebor@redhat.com>
16641         PR middle-end/85359
16642         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16643         only when expasion succeeds.
16644         (expand_builtin_strcmp): Same.
16645         (expand_builtin_strncmp): Same.
16647 2018-05-22  Martin Sebor  <msebor@redhat.com>
16649         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16651 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16652             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16654         * config/aarch64/aarch64-ldpstp.md: Replace uses of
16655         aarch64_mem_pair_operand with memory_operand and delete operand swapping
16656         code.
16657         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16658         Add check for legitimate_address.
16659         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16660         (aarch64_swap_ldrstr_operands): New.
16661         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16662         Define prototype.
16664 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16665             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16667         * config/aarch64/aarch64.md: New patterns to generate stp
16668         and ldp.
16669         (store_pair_sw, store_pair_dw): New patterns to generate stp for
16670         single words and double words.
16671         (load_pair_sw, load_pair_dw): Likewise.
16672         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16673         Delete.
16674         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16675         Delete.
16676         * config/aarch64/aarch64-ldpstp.md: Modify peephole
16677         for different mode ldpstp and add peephole for merged zero stores.
16678         Likewise for loads.
16679         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16680         Add size check.
16681         (aarch64_gen_store_pair): Rename calls to match new patterns.
16682         (aarch64_gen_load_pair): Rename calls to match new patterns.
16683         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16684         (load_pair<DREG:mode><DREG2:mode>): ... This.
16685         (store_pair<mode>): Rename to...
16686         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16687         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16688         New mode iterators.
16689         (V_INT_EQUIV): Handle SImode.
16690         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16691         New predicate.
16693 2018-05-22  Martin Sebor  <msebor@redhat.com>
16695         PR c/85623
16696         * calls.c (maybe_warn_nonstring_arg): Use string length to set
16697         or ajust the presumed bound on an operation to avoid unnecessary
16698         warnings.
16700 2018-05-22  Martin Sebor  <msebor@redhat.com>
16702         PR tree-optimization/85826
16703         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16704         assuming that a DECL necesarily has a constant size.
16706 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
16708         PR middle-end/85862
16709         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16711 2018-05-22  Richard Biener  <rguenther@suse.de>
16713         PR tree-optimization/85834
16714         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16715         non-constant and non-zero memset arguments.
16717 2018-05-22  Martin Liska  <mliska@suse.cz>
16719         PR ipa/85607
16720         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16722 2018-05-22  Richard Biener  <rguenther@suse.de>
16724         PR tree-optimization/85863
16725         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16726         comparisons when vectype is specified.
16727         (vectorizable_condition): Do not specify vectype for
16728         vect_is_simple_cond when SLP vectorizing.
16730 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
16732         PR target/85657
16733         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16734         define __ibm128 as long double.
16735         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16736         as a distinct type when IEEE 128-bit support is enabled.
16737         (init_float128_ieee): Fix up conversions between IFmode and IEEE
16738         128-bit types to use the correct functions.
16739         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16740         convert between 128-bit floating point types that have different
16741         modes but the same representation, instead of using gen_lowpart to
16742         makean alias.
16743         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16744         KFmode.
16745         (IFKF_reg): New attributes to give the register constraints for
16746         IFmode and KFmode.
16747         (extend<mode>tf2_internal): New insns to mark an explicit
16748         conversion between 128-bit floating point types that have a
16749         different mode but share the same representation.
16751 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
16753         PR tree-optimization/85814
16754         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16755         a null return from get_strinfo when unsharing the next
16756         strinfo in the chain.
16758 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
16760         PR gcc/84923
16761         * varasm.c (weak_finish): Clean up weak_decls.
16763 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16765         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16766         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16767         UNSPEC_UADALP values.
16768         * config/aarch64/iterators.md (ABAL): New int iterator.
16769         (ABDL2): Likewise.
16770         (ADALP): Likewise.
16771         (sur): Add mappings for the above.
16772         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16773         New define_insn.
16774         (aarch64_<sur>abal<mode>_4): Likewise.
16775         (aarch64_<sur>adalp<mode>_3): Likewise.
16776         (<sur>sadv16qi): New define_expand.
16778 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
16780         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16781         (*movdf_internal): Ditto.
16782         (*rcpsf2_sse): Ditto.
16783         (*rsqrtsf2_sse): Ditto.
16784         (*sqrt<mode>2_sse): Ditto.
16786 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
16788         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16789         eor3q<mode>4.
16790         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16791         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16792         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16793         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16794         vbcaxq_s64): New.
16795         * config/aarch64/arm_neon.h: Likewise.
16796         * config/aarch64/iterators.md (VQ_I): New.
16798 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
16800         * config.gcc: Add arc/t-multilib-linux to tmake_file for
16801         arc*-*-linux*.
16802         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16803         MULTILIB_DIRNAMES
16805 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16807         * config/nds32/constraints.md (S): New constraint.
16808         * config/nds32/nds32.md (call_internal): Use constraint S.
16809         (call_value_internal): Likewise.
16810         (sibcall_internal): Likewise.
16811         (sibcall_value_internal): Likewise.
16813 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16814             Chung-Ju Wu  <jasonwucj@gmail.com>
16816         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16817         into consideration.
16819 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16820             Chung-Ju Wu  <jasonwucj@gmail.com>
16822         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16823         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16824         (nds32_rtx_costs_impl): Simplify.
16825         (nds32_address_cost_impl): Simplify.
16826         (nds32_init_rtx_costs): New function.
16827         (nds32_rtx_costs_speed_prefer): Likewise.
16828         (nds32_rtx_costs_size_prefer): Likewise.
16829         (nds32_address_cost_speed_prefer): Likewise.
16830         (nds32_address_cost_speed_fwprop): Likewise.
16831         (nds32_address_cost_size_prefer): Likewise.
16832         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16833         * config/nds32/nds32.c (nds32_option_override): Use
16834         nds32_init_rtx_costs function.
16836 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16838         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16839         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16840         (TARGET_PIPELINE_N8): Likewise.
16841         (TARGET_PIPELINE_N10): Likewise.
16842         (TARGET_PIPELINE_N13): Likewise.
16843         (TARGET_PIPELINE_GRAYWOLF): Likewise.
16845 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
16847         * config/nds32/nds32-fpu.md: Update copyright year.
16849 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16851         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16853 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16855         * config/nds32/nds32.c
16856         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16857         * config/nds32/nds32.opt (minline-asm-r15): New option.
16859 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16861         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16862         MASK_HW_ABS.
16863         * config/nds32/nds32.md (abssi2): New pattern.
16865 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
16867         * config/i386/i386.md (rex64namesuffix): New mode attribute.
16868         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16869         Merge insn pattern from sse_cvtsi2ss<round_name> and
16870         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16871         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16872         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16873         using SWI48 mode iterator.
16874         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16875         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16876         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16877         pattern from sse_cvttss2si<round_saeonly_name>
16878         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16879         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16880         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16881         using SWI48 mode iterator.
16882         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16883         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16884         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16885         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16886         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16887         using SWI48 mode iterator.
16888         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16889         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16890         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16891         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16892         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16893         SWI48 mode iterator.
16894         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16895         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16896         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16897         pattern from sse_cvttsd2si<round_saeonly_name>
16898         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16900 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16902         * config/nds32/nds32-md-auxiliary.c
16903         (nds32_valid_smw_lwm_base_p): Refine.
16904         (nds32_output_smw_single_word): Refine.
16905         (nds32_output_smw_double_word): New.
16906         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16908 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16910         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16911         (nds32_output_stack_pop): Refine.
16912         (nds32_expand_unaligned_load): Refine.
16913         (nds32_expand_unaligned_store): Refine.
16915 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16916             Chung-Ju Wu  <jasonwucj@gmail.com>
16918         * config/nds32/constants.md: Add TP_REGNUM constant.
16919         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16920         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16921         UNSPEC_ADD32.
16922         * config/nds32/nds32-doubleword.md: Consider flag_pic.
16923         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16924         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16925         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16926         and PIC code generation.
16927         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16928         code generation.
16929         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16930         optimization.
16931         * config/nds32/nds32.md: Support TLS and PIC.
16932         * config/nds32/nds32.c: Support TLS and PIC.
16933         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16934         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16935         predicate.
16937 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16939         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16940         mode with E_ prefix.
16942 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16943             Chung-Ju Wu  <jasonwucj@gmail.com>
16945         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16946         * config/nds32/nds32-md-auxiliary.c
16947         (symbolic_reference_mentioned_p): New.
16948         (nds32_legitimize_ict_address): New.
16949         (nds32_expand_ict_move): New.
16950         (nds32_indirect_call_referenced_p): New.
16951         (nds32_symbol_binds_local_p): Delete.
16952         (nds32_long_call_p): Modify.
16953         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16954         * config/nds32/nds32-protos.h
16955         (symbolic_reference_mentioned_p): Declare.
16956         (nds32_legitimize_ict_address): Declare.
16957         (nds32_expand_ict_move): Declare.
16958         (nds32_indirect_call_referenced_p): Declare.
16959         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16960         (nds32_relax_group): Use nds32_ict_const_p as condition.
16961         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16962         (nds32_asm_file_start): Output ict_model directive in asm code.
16963         (nds32_legitimate_address_p): Consider indirect call.
16964         (nds32_print_operand): Consider indirect call.
16965         (nds32_print_operand_address): Consider indirect call.
16966         (nds32_insert_attributes): Handle "indirect_call" attribute.
16967         (TARGET_LEGITIMATE_ADDRESS_P): Define.
16968         (TARGET_LEGITIMATE_CONSTANT_P): Define.
16969         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16970         (TARGET_DELEGITIMIZE_ADDRESS): Define.
16971         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16972         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16973         (TARGET_ICT_MODEL_SMALL): Define.
16974         (TARGET_ICT_MODEL_LARGE): Define.
16975         * config/nds32/nds32.md (movsi): Consider ict model.
16976         (call, call_value): Consider ict model.
16977         (sibcall, sibcall_value): Consider ict model.
16978         * config/nds32/nds32.opt (mict-model): New option.
16979         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16980         model.
16982 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
16983             Monk Chiang  <sh.chiang04@gmail.com>
16984             Jim Wilson <jimw@sifive.com>
16986         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16987         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
16988         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16989         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16990         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
16991         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16992         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16993         compute save_libcall_adjustment properly.
16994         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16995         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16996         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16997         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16998         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16999         (ABI_SPEC): Handle mabi=ilp32e.
17000         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
17001         (RVE): Add RVE mask.
17002         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
17003         <-march>: Add rv32e as an example.
17005 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
17007         PR c++/82899
17008         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
17009         (intra_create_variable_infos): Handle C++ constructors.
17011 2018-05-18  Martin Liska  <mliska@suse.cz>
17013         * passes.def: Remove a redundant pass.
17015 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
17017         PR bootstrap/85838
17018         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
17020 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17022         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
17023         (ARMv4): Update.
17024         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
17025         (ARMv6m): Update.
17026         (armv2, armv2a, armv3, armv3m): Delete architectures.
17027         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
17028         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
17029         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
17030         Delete cpus.
17031         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
17032         (*mulsidi3adddi): Likewise.
17033         (mulsidi3): Likewise.
17034         (*mulsidi3_nov6): Likewise.
17035         (umulsidi3): Likewise.
17036         (umulsidi3_nov6): Likewise.
17037         (umaddsidi4): Likewise.
17038         (*umulsidi3adddi): Likewise.
17039         (smulsi3_highpart): Likewise.
17040         (*smulsi3_highpart_nov6): Likewise.
17041         (umulsi3_highpart): Likewise.
17042         (*umulsi3_highpart_nov6): Likewise.
17043         * config/arm/arm.h (arm_arch3m): Delete.
17044         * config/arm/arm.c (arm_arch3m): Delete.
17045         (arm_option_override_internal): Update armv3-related comment.
17046         (arm_configure_build_target): Delete use of isa_bit_mode32.
17047         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
17048         (arm_rtx_costs_internal): Delete check of arm_arch3m.
17049         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
17050         (mulsa3): Likewise.
17051         (mulusa3): Likewise.
17052         * config/arm/arm-protos.h (arm_arch3m): Delete.
17053         * config/arm/arm-tables.opt: Regenerate.
17054         * config/arm/arm-tune.md: Likewise.
17055         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
17056         deleted architectures.
17058 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17060         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
17061         (armv5t, armv5te): New features.
17062         (ARMv5, ARMv5e): Delete fgroups.
17063         (ARMv5t, ARMv5te): Adjust for above changes.
17064         (ARMv6m): Likewise.
17065         (armv5, armv5e): Delete arches.
17066         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
17067         arm_arch5.
17068         (*call_reg_arm): Likewise.
17069         (*call_value_reg_armv5): Likewise.
17070         (*call_value_reg_arm): Likewise.
17071         (*call_symbol): Likewise.
17072         (*call_value_symbol): Likewise.
17073         (*sibcall_insn): Likewise.
17074         (*sibcall_value_insn): Likewise.
17075         (clzsi2): Likewise.
17076         (prefetch): Likewise.
17077         (define_split and define_peephole2 dependent on arm_arch5):
17078         Likewise.
17079         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
17080         arm_arch5e.
17081         (TARGET_ARM_QBIT): Likewise.
17082         (TARGET_DSP_MULTIPLY): Likewise.
17083         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
17084         (arm_arch5, arm_arch5e): Delete.
17085         (arm_arch5t, arm_arch5te): Declare.
17086         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
17087         (arm_arch5t): Declare.
17088         (arm_option_reconfigure_globals): Update for the above.
17089         (arm_options_perform_arch_sanity_checks): Update comment, replace
17090         use of arm_arch5 with arm_arch5t.
17091         (use_return_insn): Likewise.
17092         (arm_emit_call_insn): Likewise.
17093         (output_return_instruction): Likewise.
17094         (arm_final_prescan_insn): Likewise.
17095         (arm_coproc_builtin_available): Likewise.
17096         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
17097         arm_arch5e with arm_arch5t and arm_arch5te.
17098         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
17099         (arm_arch5t, arm_arch5te): Declare.
17100         * config/arm/arm-tables.opt: Regenerate.
17101         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
17102         * config/arm/t-multilib: Likewise.
17103         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
17104         instead of arm_arch5.
17105         (*call_reg_thumb1): Likewise.
17106         (*call_value_reg_thumb1_v5): Likewise.
17107         (*call_value_reg_thumb1): Likewise.
17108         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
17109         unreachable path.
17110         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
17112 2018-05-18  Martin Liska  <mliska@suse.cz>
17114         PR gcov-profile/84846
17115         * doc/gcov.texi: Document -t option of gcov tool.
17117 2018-05-18  Martin Liska  <mliska@suse.cz>
17119         PR gcov-profile/84846
17120         * gcov.c (print_usage): Add new -t option.
17121         (process_args): Handle the option.
17122         (generate_results): Use stdout as output when requested by
17123         the option.
17125 2018-05-18  Martin Liska  <mliska@suse.cz>
17127         PR gcov-profile/84846
17128         * coverage.c (coverage_init): Write PWD to .gcno file.
17129         * doc/gcov.texi: Document how working directory is printed.
17130         * gcov-dump.c (dump_gcov_file): Print PWD.
17131         * gcov.c (output_intermediate_file): Likewise.
17132         (read_graph_file): Read PWD string.
17133         (output_lines): Print PWD.
17135 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17137         PR middle-end/85817
17138         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
17139         for retval and return false if all args to phi are zero.
17141 2018-05-18  Richard Biener  <rguenther@suse.de>
17143         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
17144         method.
17145         (evrp_dom_walker::before_dom_children): Call it.
17147 2018-05-18  Richard Biener  <rguenther@suse.de>
17149         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
17150         results when processing array refs with variable index.
17152 2018-05-18  Toon Moene  <toon@moene.org>
17154         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
17155         directly after that of -floop-interchange. Indicate that both
17156         options are enabled by default when specifying -O3.
17158 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17160         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
17161         iterator.  Delete separate integer-mode vec_set<mode> expander.
17162         (aarch64_simd_vec_setv2di): Delete.
17163         (vec_setv2di): Delete.
17164         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
17165         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
17166         the "w, r" alternative.
17168 2018-05-18  Martin Liska  <mliska@suse.cz>
17170         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
17171         * tree-pass.h (make_pass_lower_switch_O0): New function.
17172         * tree-switch-conversion.c (node_has_low_bound): Remove.
17173         (node_has_high_bound): Likewise.
17174         (node_is_bounded): Likewise.
17175         (class pass_lower_switch): Make it a template type and create
17176         two instances.
17177         (pass_lower_switch::execute): Add template argument.
17178         (make_pass_lower_switch): New function.
17179         (make_pass_lower_switch_O0): New function.
17180         (do_jump_if_equal): Remove.
17181         (emit_case_nodes): Simplify to just handle all 3 cases and leave
17182         all the hard work to tree optimization passes.
17184 2018-05-18  Martin Liska  <mliska@suse.cz>
17186         * dbgcnt.c (limit_low): Renamed from limit.
17187         (limit_high): New variable.
17188         (dbg_cnt_is_enabled): Check for upper limit.
17189         (dbg_cnt): Adjust dumping.
17190         (dbg_cnt_set_limit_by_index): Add new argument for high
17191         value.
17192         (dbg_cnt_set_limit_by_name): Likewise.
17193         (dbg_cnt_process_single_pair): Parse new format.
17194         (dbg_cnt_process_opt): Use strtok.
17195         (dbg_cnt_list_all_counters): Remove 'value' and add
17196         'limit_high'.
17197         * doc/invoke.texi: Document changes.
17199 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
17201         * doc/sourcebuild.texi (scalar_all_fma): Document.
17202         * tree.def (FMA_EXPR): Delete.
17203         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
17204         * internal-fn.c (ternary_direct): New macro.
17205         (expand_ternary_optab_fn): Likewise.
17206         (direct_ternary_optab_supported_p): Likewise.
17207         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
17208         * builtins.c (fold_builtin_fma): Delete.
17209         (fold_builtin_3): Don't call it.
17210         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
17211         * expr.c (expand_expr_real_2): Likewise.
17212         * fold-const.c (operand_equal_p): Likewise.
17213         (fold_ternary_loc): Likewise.
17214         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
17215         * gimple.c (DEFTREECODE): Likewise.
17216         * gimplify.c (gimplify_expr): Likewise.
17217         * optabs-tree.c (optab_for_tree_code): Likewise.
17218         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
17219         * tree-eh.c (operation_could_trap_p): Likewise.
17220         (stmt_could_throw_1_p): Likewise.
17221         * tree-inline.c (estimate_operator_cost): Likewise.
17222         * tree-pretty-print.c (dump_generic_node): Likewise.
17223         (op_code_prio): Likewise.
17224         * tree-ssa-loop-im.c (stmt_cost): Likewise.
17225         * tree-ssa-operands.c (get_expr_operands): Likewise.
17226         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
17227         * fold-const-call.h (fold_fma): Delete.
17228         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
17229         CFN_FNMA and CFN_FNMS.
17230         (fold_fma): Delete.
17231         * genmatch.c (combined_fn): New enum.
17232         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
17233         (commutative_op): New function.
17234         (commutate): Use it.  Handle more than 2 operands.
17235         (dt_operand::gen_gimple_expr): Use commutative_op.
17236         (parser::parse_expr): Allow :c to be used with non-binary
17237         operators if the commutative operand is known.
17238         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
17239         CFN_FMS, CFN_FNMA and CFN_FNMS.
17240         (backprop::process_assign_use): Remove FMA_EXPR handling.
17241         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
17242         (gen_hsa_fma): New function.
17243         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
17244         IFN_FNMA and IFN_FNMS.
17245         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
17246         * gimple-fold.h (follow_all_ssa_edges): Declare.
17247         * gimple-fold.c (follow_all_ssa_edges): New function.
17248         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
17249         gimple_build interface and use follow_all_ssa_edges to fold the result.
17250         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
17251         instead of checking for optabs directly.
17252         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
17253         rather than FMA_EXPRs.
17254         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
17255         call to IFN_FMA instead of an FMA_EXPR.
17257 2018-05-17  Jim Wilson  <jimw@sifive.com>
17259         * expr.c (do_tablejump): When converting index to Pmode, if we have a
17260         sign extended promoted subreg, and the range does not have the sign bit
17261         set, then do a sign extend.
17263         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
17264         test, check for sign extended subreg and/or constant operands, and
17265         do a sign extend in that case.
17267 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
17269         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
17270         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
17271         Add untyped.
17272         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
17273         Change logics_shift_reg to logics_shift_imm.
17274         (thunderx2t99_fp_loadpair_basic): Delete.
17275         (thunderx2t99_fp_storepair_basic): Delete.
17276         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
17277         (thunderx2t99_asimd_polynomial): Delete.
17278         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
17279         and neon_fp_mul_d_scalar_q.
17280         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
17281         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
17282         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
17283         (thunderx2t99_asimd_lut): Add missing tbl types.
17284         (thunderx2t99_asimd_ext): Delete.
17285         (thunderx2t99_asimd_load1_1_mult): Delete.
17286         (thunderx2t99_asimd_load1_2_mult): Delete.
17287         (thunderx2t99_asimd_load1_ldp): New.
17288         (thunderx2t99_asimd_load1): New.
17289         (thunderx2t99_asimd_load2): Add missing *load2* types.
17290         (thunderx2t99_asimd_load3): New.
17291         (thunderx2t99_asimd_load4): New.
17292         (thunderx2t99_asimd_store1_1_mult): Delete.
17293         (thunderx2t99_asimd_store1_2_mult): Delete.
17294         (thunderx2t99_asimd_store2_mult): Delete.
17295         (thunderx2t99_asimd_store2_onelane): Delete.
17296         (thunderx2t99_asimd_store_stp): New.
17297         (thunderx2t99_asimd_store1): New.
17298         (thunderx2t99_asimd_store2): New.
17299         (thunderx2t99_asimd_store3): New.
17300         (thunderx2t99_asimd_store4): New.
17302 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
17304         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
17305         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
17307 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
17308             Segher Boessenkool  <segher@kernel.crashing.org>
17310         PR target/85698
17311         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
17312         operand.
17314 2018-05-17  Richard Biener  <rguenther@suse.de>
17316         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
17317         for pruning loop and prune defs feeding only already visited PHIs.
17319 2018-05-17  Richard Biener  <rguenther@suse.de>
17321         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
17323 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
17324             Richard Biener  <rguenther@suse.de>
17326         PR tree-optimization/85793
17327         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
17328         for VMAT_ELEMENTWISE.
17330 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17332         * internal-fn.h (lookup_internal_fn): Declare
17333         * internal-fn.c (lookup_internal_fn): New function.
17334         * gimple.c (gimple_build_call_from_tree): Handle calls to
17335         internal functions.
17336         * gimple-pretty-print.c (dump_gimple_call): Print "." before
17337         internal function names.
17338         * tree-pretty-print.c (dump_generic_node): Likewise.
17339         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
17341 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17343         * gimple-fold.h (gimple_build): Make the function forms take
17344         combined_fn rather than built_in_function.
17345         (gimple_simplify): Likewise.
17346         * gimple-match-head.c (gimple_simplify): Likewise.
17347         * gimple-fold.c (gimple_build): Likewise.
17348         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
17349         rather than gimple_build_call_internal.
17350         (get_initial_defs_for_reduction): Likewise.
17351         (vect_create_epilog_for_reduction): Likewise.
17352         (vectorizable_live_operation): Likewise.
17354 2018-05-17  Martin Liska  <mliska@suse.cz>
17356         * gimple-ssa-sprintf.c (format_directive): Do not use
17357         space in between 'G_' and '('.
17359 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
17361         PR target/85323
17362         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
17363         even if the mask is not all ones.
17365         PR target/85323
17366         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
17367         vector.
17368         (ix86_gimple_fold_builtin): Likewise.
17370         PR target/85323
17371         * config/i386/i386.c: Include tree-vector-builder.h.
17372         (ix86_vector_shift_count): New function.
17373         (ix86_fold_builtin): Fold shift builtins by scalar count.
17374         (ix86_gimple_fold_builtin): Likewise.
17376         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
17377         _mm512_setzero): New intrinsics.
17379 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
17380             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17382         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
17383         code generation for cases where splatting a value is not useful.
17384         * simplify-rtx.c (simplify_ternary_operation): Simplify
17385         vec_merge across a vec_duplicate and a paradoxical subreg forming
17386         a vector mode to a vec_concat.
17388 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
17390         * config.gcc: Support "goldmont-plus".
17391         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17392         "goldmont-plus".
17393         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17394         PROCESSOR_GOLDMONT_PLUS.
17395         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
17396         (processor_target_table): Add "goldmont-plus".
17397         (PTA_GOLDMONT_PLUS): Define.
17398         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
17399         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
17400         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
17401         (fold_builtin_cpu): Add "goldmont-plus".
17402         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
17403         (ix86_option_override_internal): Add "goldmont-plus".
17404         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
17405         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
17406         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
17407         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
17409 2018-05-17  Richard Biener  <rguenther@suse.de>
17411         PR tree-optimization/85757
17412         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
17413         remove defs that only feed that PHI from further processing.
17415 2018-05-16  Jim Wilson  <jimw@sifive.com>
17417         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
17418         asterisk to name.
17419         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
17421 2018-05-16  Mark Wielaard  <mark@klomp.org>
17423         * dwarf2out.c (count_index_strings): New function.
17424         (output_indirect_strings): Call count_index_strings and generate
17425         header for dwarf_version >= 5.
17427 2018-05-16  Mark Wielaard  <mark@klomp.org>
17429         * dwarf2out.c (dwarf_FORM): New function.
17430         (set_indirect_string): Use dwarf_FORM.
17431         (reset_indirect_string): Likewise.
17432         (size_of_die): Likewise.
17433         (value_format): Likewise.
17434         (output_die): Likewise.
17435         (add_skeleton_AT_string): Likewise.
17436         (output_macinfo_op): Likewise.
17437         (index_string): Likewise.
17438         (output_index_string_offset): Likewise.
17439         (output_index_string): Likewise.
17440         (count_index_strings): Likewise.
17442 2018-05-16  Carl Love  <cel@us.ibm.com>
17444         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17445         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
17447 2018-05-16  Martin Jambor  <mjambor@suse.cz>
17449         * ipa-prop.c (ipa_free_all_edge_args): Remove.
17450         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
17452 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
17454         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
17455         (fnma<mode>4): Likewise.
17456         (fms<mode>4): Likewise.
17457         (fnms<mode>4): Likewise.
17458         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
17459         (aarch64_fnma<mode>4): Likewise.
17460         (aarch64_fms<mode>4): Likewise.
17461         (aarch64_fnms<mode>4): Likewise.
17462         (aarch64_fnmadd<mode>4): Likewise.
17464 2018-05-16  Jason Merrill  <jason@redhat.com>
17466         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
17468 2018-05-16  Richard Biener  <rguenther@suse.de>
17470         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
17471         (dump_stmt_cost): Declare.
17472         (add_stmt_cost): Dump cost we add.
17473         (add_stmt_costs): New function.
17474         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
17475         No longer exported.
17476         (vect_analyze_stmt): Adjust prototype.
17477         (vectorizable_condition): Likewise.
17478         (vectorizable_live_operation): Likewise.
17479         (vectorizable_reduction): Likewise.
17480         (vectorizable_induction): Likewise.
17481         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
17482         cost vector to pass to vectorizable_ and record afterwards.
17483         (vect_model_reduction_cost): Take cost vector argument and adjust.
17484         (vect_model_induction_cost): Likewise.
17485         (vectorizable_reduction): Likewise.
17486         (vectorizable_induction): Likewise.
17487         (vectorizable_live_operation): Likewise.
17488         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
17489         SLP_TREE_NUMBER_OF_VEC_STMTS.
17490         (vect_analyze_slp_cost_1): Remove.
17491         (vect_analyze_slp_cost): Likewise.
17492         (vect_slp_analyze_node_operations): Take visited args and
17493         a target cost vector.  Avoid processing already visited stmt sets.
17494         (vect_slp_analyze_operations): Use a local cost vector to gather
17495         costs and register those of non-discarded instances.
17496         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
17497         (vect_schedule_slp_instance): Remove copying of
17498         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
17499         zero.
17500         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
17501         adding cost.  Record cost entry location.
17502         (vect_prologue_cost_for_slp_op): Function to compute cost of
17503         a constant or invariant generated for SLP vect in the prologue,
17504         split out from vect_analyze_slp_cost_1.
17505         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
17506         (vect_model_promotion_demotion_cost): Likewise.
17507         (vect_model_store_cost): Likewise, make static.
17508         (vect_model_load_cost): Likewise.
17509         (vectorizable_bswap): Add cost vector arg and adjust.
17510         (vectorizable_call): Likewise.
17511         (vectorizable_simd_clone_call): Likewise.
17512         (vectorizable_conversion): Likewise.
17513         (vectorizable_assignment): Likewise.
17514         (vectorizable_shift): Likewise.
17515         (vectorizable_operation): Likewise.
17516         (vectorizable_store): Likewise.
17517         (vectorizable_load): Likewise.
17518         (vectorizable_condition): Likewise.
17519         (vectorizable_comparison): Likewise.
17520         (can_vectorize_live_stmts): Likewise.
17521         (vect_analyze_stmt): Likewise.
17522         (vect_transform_stmt): Adjust calls to vectorizable_*.
17523         * tree-vectorizer.c: Include gimple-pretty-print.h.
17524         (dump_stmt_cost): New function.
17526 2018-05-16  Richard Biener  <rguenther@suse.de>
17528         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
17529         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
17530         * tree-ssa-dse.c: Include tree-ssa-loop.h.
17531         (check_name): New callback.
17532         (dse_classify_store): Track cycles via a visited bitmap of PHI
17533         defs and simplify handling of in-loop and across loop dead stores
17534         and properly fail for loop-variant refs.  Handle byte-tracking with
17535         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
17536         limiting the walk.
17538 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
17540         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
17541         (vect_get_mask_type_for_stmt): Likewise.
17542         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
17543         split out from...
17544         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
17545         to determine the statement's vector type and the vector type that
17546         should be used for calculating nunits.  Deal with cases in which
17547         the type has to be deferred.
17548         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
17549         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
17550         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
17551         (vect_determine_vf_for_stmt): New functions, split out from...
17552         (vect_determine_vectorization_factor): ...here.
17553         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
17554         (vect_get_mask_type_for_stmt): New functions, split out from
17555         vect_determine_vectorization_factor.
17557 2018-05-16  Richard Biener  <rguenther@suse.de>
17559         * tree-cfg.c (verify_gimple_assign_ternary): Properly
17560         verify the [VEC_]COND_EXPR embedded comparison.
17562 2018-05-15  Martin Sebor  <msebor@redhat.com>
17564         PR tree-optimization/85753
17565         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
17566         RECORD_TYPE in addition to ARRAY_TYPE.
17568 2018-05-15  Martin Sebor  <msebor@redhat.com>
17570         PR middle-end/85643
17571         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17573 2018-05-15  Richard Biener  <rguenther@suse.de>
17575         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
17576         add by_clobber_p one.  Change algorithm to collect all defs
17577         representing uses we need to walk and try reducing them to
17578         a single one before failing.
17579         (dse_dom_walker::dse_optimize_stmt): Adjust.
17581 2018-05-13  Mark Wielaard  <mark@klomp.org>
17583         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
17584         (size_of_loc_descr): Likewise.
17585         (output_loc_operands): Likewise.
17586         (output_loc_operands_raw): Likewise.
17587         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
17588         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17589         (hash_loc_operands): Likewise.
17590         (compare_loc_operands): Likewise.
17592 2018-05-14  Mark Wielaard  <mark@klomp.org>
17594         * dwarf2out.c (count_index_addrs): New function.
17595         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
17597 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17599         PR tree-optimization/83648
17600         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
17601         return value as malloc candidate.
17603 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17605         PR ipa/85734
17606         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
17607         param as true in call to suggest_attribute.
17609 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
17611         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
17612         -mreadonly-in-sdata.
17614 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17616         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
17617         New pattern.
17618         (aarch64_crypto_aesd_fused): Likewise.
17620 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
17622         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17623         (movsi_aarch64): Likewise.
17624         (load_pairsi): Likewise.
17625         (load_pairdi): Likewise.
17626         (store_pairsi): Likewise.
17627         (store_pairdi): Likewise.
17628         (load_pairsf): Likewise.
17629         (load_pairdf): Likewise.
17630         (store_pairsf): Likewise.
17631         (store_pairdf): Likewise.
17632         (zero_extend): Likewise.
17633         (trunc): Swap alternatives.
17634         (fcvt_target): Add '?' to prefer w over r.
17636 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
17638         PR target/85756
17639         * config/i386/i386.md: Disallow non-commutative arithmetics in
17640         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17641         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
17642         in the peephole2 before it.
17644 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
17646         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17647         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17648         (ix86_handle_option): Handle -mcldemote.
17649         * config.gcc: New header.
17650         * config/i386/cldemoteintrin.h: New file.
17651         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17652         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17653         -mcldemote.
17654         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17655         OPTION_MASK_ISA_CLDEMOTE.
17656         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17657         (ix86_valid_target_attribute_inner_p): Ditto.
17658         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17659         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17660         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17661         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17662         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17663         (cldemote): New.
17664         * config/i386/i386.opt: Add -mcldemote.
17665         * config/i386/x86intrin.h: New header.
17666         * doc/invoke.texi: Add -mcldemote.
17668 2018-05-14  Richard Biener  <rguenther@suse.de>
17670         * doc/match-and-simplify.texi: Adjust :s documentation.
17672 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
17674         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17675         intended memcpy size.
17676         (REORDER_45): Likewise.
17678 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
17680         * sort.cc: New file.
17681         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17682         * vec.c (qsort_chk): Use gcc_qsort.
17683         * Makefile.in (OBJS-libcommon): Add sort.o.
17684         (build/sort.o): New target.  Use it...
17685         (BUILD_RTL): ... here, and...
17686         (build/gencfn-macros): ... here, and...
17687         (build/genmatch): ... here.
17689 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17690             Chung-Ju Wu  <jasonwucj@gmail.com>
17692         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17693         * config/nds32/nds32-graywolf.md: New file.
17694         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17695         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17696         pipeline.
17697         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17698         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17699         * config/nds32/nds32.md (pipeline_model): Add graywolf.
17700         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17701         * config/nds32/pipelines.md: Include n15 settings.
17703 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17704             Chung-Ju Wu  <jasonwucj@gmail.com>
17706         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17707         * config/nds32/nds32-n13.md: New file.
17708         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17709         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17710         pipeline.
17711         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17712         * config/nds32/nds32.md (pipeline_model): Add n13.
17713         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17714         * config/nds32/pipelines.md: Include n13 settings.
17716 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17717             Chung-Ju Wu  <jasonwucj@gmail.com>
17719         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17720         * config/nds32/nds32-n10.md: New file.
17721         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17722         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17723         pipeline.
17724         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17725         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17726         * config/nds32/nds32.md (pipeline_model): Add n10.
17727         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17728         * config/nds32/pipelines.md: Include n10 settings.
17730 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
17731             Kito Cheng  <kito.cheng@gmail.com>
17732             Chung-Ju Wu  <jasonwucj@gmail.com>
17734         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17735         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17736         Add enum values for DSP extension instructions.
17737         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17738         New constraints.
17739         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17740         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17741         New code iterators.
17742         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17743         * config/nds32/nds32-dspext.md: New file for DSP implementation.
17744         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17745         * config/nds32/nds32-intrinsic.md: Likewise.
17746         * config/nds32/nds32_intrinsic.h: Likewise.
17747         * config/nds32/nds32-md-auxiliary.c: Likewise.
17748         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17749         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17750         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17751         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17752         * config/nds32/nds32-protos.h: New declarations for DSP extension.
17753         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17754         TYPE_DMAC in switch statement.
17755         * config/nds32/nds32.c: New checking and implementation for DSP
17756         extension instructions.
17757         * config/nds32/nds32.h: Likewise.
17758         * config/nds32/nds32.md: Likewise.
17759         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17760         * config/nds32/predicates.md: Implement new predicates for DSP
17761         extension.
17763 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
17765         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17766         Reformat alternatives and attributes so it is easier to identify
17767         which constraints/attributes go with which instruction.
17768         (mov<mode>_hardfloat32, FMOVE64): Likewise.
17769         (mov<mode>_softfloat32, FMOVE64): Likewise.
17770         (mov<mode>_hardfloat64, FMOVE64): Likewise.
17771         (mov<mode>_softfloat64, FMOVE64): Likewise.
17773 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17775         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17776         subsection.
17777         (Basic PowerPC Built-in Functions): The new name of the
17778         subsection previously known as "PowerPC Built-in Functions".
17779         (Basic PowerPC Built-in Functions Available on all Configurations):
17780         New subsubsection.
17781         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17782         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17783         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17784         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17786 2018-05-11  Martin Jambor  <mjambor@suse.cz>
17788         PR ipa/85655
17789         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17790         single const.
17792 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17794         PR target/85733
17795         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17797 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
17799         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17800         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17801         (ix86_handle_option): Handle -mwaitpkg.
17802         * config.gcc: New header.
17803         * config/i386/cpuid.h (bit_WAITPKG): New bit.
17804         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17805         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17806         function type.
17807         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17808         OPTION_MASK_ISA_WAITPKG.
17809         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17810         (ix86_option_override_internal): Add PTA_WAITPKG.
17811         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17812         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17813         IX86_BUILTIN_TPAUSE.
17814         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17815         __builtin_ia32_umwait and __builtin_ia32_tpause.
17816         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17817         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17818         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17819         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17820         UNSPECV_TPAUSE): New.
17821         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17822         * config/i386/i386.opt: Add -mwaitpkg.
17823         * config/i386/waitpkgintrin.h: New file.
17824         * config/i386/x86intrin.h: New header.
17825         * doc/invoke.texi: Add -mwaitpkg.
17827 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17829         PR target/85606
17830         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17831         equivalent.
17832         (cortex-m0): Use armv6s-m isa.
17833         (cortex-m0plus): Likewise.
17834         (cortex-m1): Likewise.
17835         (cortex-m0.small-multiply): Likewise.
17836         (cortex-m0plus.small-multiply): Likewise.
17837         (cortex-m1.small-multiply): Likewise.
17839 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
17840             Jakub Jelinek  <jakub@redhat.com>
17842         PR tree-optimization/85692
17843         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17844         source permute as well.
17846 2018-05-11  Martin Liska  <mliska@suse.cz>
17848         PR sanitizer/85556
17849         * doc/extend.texi: Document LLVM style format for no_sanitize
17850         attribute.
17852 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17854         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17855         mode_supports_vsx_dform_quad to mode_supports_dq_form.
17856         (mode_supports_vsx_dform_quad): Likewise.
17857         (mode_supports_vmx_dform): Move these functions to be next to the
17858         other mode_supports functions.
17859         (mode_supports_dq_form): Likewise.
17860         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17861         mode_supports_dq_form.
17862         (reg_offset_addressing_ok_p): Likewise.
17863         (offsettable_ok_by_alignment): Likewise.
17864         (rs6000_legitimate_offset_address_p): Likewise.
17865         (legitimate_lo_sum_address_p): Likewise.
17866         (rs6000_legitimize_address): Likewise.
17867         (rs6000_legitimize_reload_address): Likewise.
17868         (rs6000_secondary_reload_inner): Likewise.
17869         (rs6000_preferred_reload_class): Likewise.
17870         (rs6000_output_move_128bit): Likewise.
17872 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17874         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17875         Generate SImode target register for null target.
17876         <case IX86_BUILTIN_XGETBV>: Ditto.
17877         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17878         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17880 2018-05-10  Carl Love  <cel@us.ibm.com>
17882         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17883         dcbtt and dcbtstt if operands[2] is 0.
17885 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17887         PR target/85693
17888         * config/i386/sse.md (usadv64qi): New expander.
17890 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
17892         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17893         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17894         -maltivec=be support.
17895         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17896         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17897         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17898         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17899         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17900         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17901         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17902         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17903         altivec_vsumsws): Adjust.
17904         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17905         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17906         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17907         support.
17908         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17909         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17910         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17911         (altivec_lve<VI_char>x): Delete expand.
17912         (*altivec_lve<VI_char>x_internal): Rename to...
17913         (altivec_lve<VI_char>x): ... this.
17914         (altivec_lvxl_<mode>): Delete expand.
17915         (*altivec_lvxl_<mode>_internal): Rename to ...
17916         (altivec_lvxl_<mode>): ... this.
17917         (altivec_stvxl_<mode>): Delete expand.
17918         (*altivec_stvxl_<mode>_internal): Rename to ...
17919         (altivec_stvxl_<mode>): ... this.
17920         (altivec_stve<VI_char>x): Delete expand.
17921         (*altivec_stve<VI_char>x_internal): Rename to ...
17922         (altivec_stve<VI_char>x): ... this.
17923         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17924         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17925         reduc_plus_scal_<mode>): Adjust.
17926         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17927         comment.
17928         (rs6000_cpu_cpp_builtins): Adjust.
17929         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17930         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17931         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17932         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17933         -maltivec=be support.
17934         (rs6000_split_vec_extract_var): Adjust.
17935         (rs6000_split_v4si_init): Adjust.
17936         (swap_selector_for_mode): Delete.
17937         (altivec_expand_lvx_be, altivec_expand_stvx_be,
17938         altivec_expand_stvex_be): Delete.
17939         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17940         -maltivec=be support.
17941         (rs6000_gimple_fold_builtin): Ditto.
17942         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17943         Adjust.
17944         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17945         (TARGET_DIRECT_MOVE_64BIT): Adjust.
17946         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17947         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17948         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17949         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17950         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17951         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17952         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17953         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17954         anonymous split): Adjust.
17955         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17956         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17958 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
17960         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17961         when --with-gxx-include-dir is also specified.
17962         * configure: Regenerate.
17964 2018-05-09  Jim Wilson  <jimw@sifive.com>
17966         PR target/84797
17967         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17968         * config/riscv/t-withmultilib: New.
17969         * config/riscv/withmultilib.h: New.
17970         * doc/install.texi: Document RISC-V --with-multilib-list support.
17972 2018-05-09  Richard Biener  <rguenther@suse.de>
17974         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17975         vector.
17976         (vect_bb_vectorization_profitable_p): Adjust.  Compute
17977         actual scalar cost using the cost vector and the add_stmt_cost
17978         machinery.
17980 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17982         PR rtl-optimization/85645
17983         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17984         in the REG_CFA_REGISTER note for LR, don't leave it empty.
17986 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17988         PR rtl-optimization/85645
17989         * shrink-wrap.c (spread_components): Return a boolean saying if
17990         anything was changed.
17991         (try_shrink_wrapping_separate): Iterate spread_components until
17992         nothing changes anymore.
17994 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17996         PR rtl-optimization/85645
17997         * regrename.c (build_def_use): Also kill the chains that include the
17998         destination of a REG_CFA_REGISTER note.
18000 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
18002         PR rtl-optimization/85645
18003         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
18004         insn that has a REG_CFA_REGISTER note.
18006 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
18008         * cfgexpand.c (expand_clobber): New function.
18009         (expand_gimple_stmt_1): Use it.
18010         * tree-vect-stmts.c (vect_clobber_variable): New function,
18011         split out from...
18012         (vectorizable_simd_clone_call): ...here.
18013         (vectorizable_store): Emit a clobber either side of an
18014         IFN_STORE_LANES sequence.
18015         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
18017 2018-05-09  Tom de Vries  <tom@codesourcery.com>
18019         PR target/85626
18020         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
18021         (define_insn "trap_if_false"): Add exit after trap.
18023 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
18025         PR rtl-optimization/85638
18026         * bb-reorder.c: Include common/common-target.h.
18027         (create_forwarder_block): New function extracted from...
18028         (fix_up_crossing_landing_pad): ...here.  Rename into...
18029         (dw2_fix_up_crossing_landing_pad): ...this.
18030         (sjlj_fix_up_crossing_landing_pad): New function.
18031         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
18032         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
18033         from both partitions and exit the loop after one iteration.
18035 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18037         Revert:
18038         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18039         subsection.
18040         (Basic PowerPC Built-in Functions): The new name of the
18041         subsection previously known as "PowerPC Built-in Functions".
18042         (Basic PowerPC Built-in Functions Available on all Configurations):
18043         New subsubsection.
18044         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18045         subsubsection.
18046         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18047         subsubsection.
18048         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18049         subsubsection.
18050         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18051         subsubsection.
18053 2018-05-08  Jim Wilson  <jimw@sifive.com>
18055         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
18056         (LD_EMUL_SUFFIX): New.
18057         (LINK_SPEC): Use it.
18059 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18061         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18062         subsection.
18063         (Basic PowerPC Built-in Functions): The new name of the
18064         subsection previously known as "PowerPC Built-in Functions".
18065         (Basic PowerPC Built-in Functions Available on all Configurations):
18066         New subsubsection.
18067         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18068         subsubsection.
18069         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18070         subsubsection.
18071         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18072         subsubsection.
18073         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18074         subsubsection.
18076 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
18078         PR target/85683
18079         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
18080         after cmpelim optimization.
18082 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
18084         * config.gcc: Support "goldmont".
18085         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
18086         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18087         PROCESSOR_GOLDMONT.
18088         * config/i386/i386.c (m_GOLDMONT): Define.
18089         (processor_target_table): Add "goldmont".
18090         (PTA_GOLDMONT): Define.
18091         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
18092         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
18093         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
18094         (fold_builtin_cpu): Add "goldmont".
18095         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
18096         (ix86_option_override_internal): Add "goldmont".
18097         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
18098         (processor_type): Add PROCESSOR_GOLDMONT.
18099         * config/i386/i386.md: Add CPU "glm".
18100         * config/i386/glm.md: New file.
18101         * config/i386/x86-tune.def: Add m_GOLDMONT.
18102         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
18104 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
18106         PR target/85572
18107         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
18108         E_V4DImode.
18109         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
18110         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
18111         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
18113         PR target/85317
18114         * config/i386/i386.c (ix86_fold_builtin): Handle
18115         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
18117         PR target/85480
18118         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
18119         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
18121 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
18123         PR target/85658
18124         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
18125         (check_arch): Likewise.
18126         (check_fpu): Return the result rather than printing it.
18127         (end arch): Fix operator precedence.
18128         (end cpu): Likewise.
18129         (END): Print the result from check_fpu.
18131 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
18132             Alan Hayward  <alan.hayward@arm.com>
18133             David Sherwood  <david.sherwood@arm.com>
18135         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
18136         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
18137         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
18138         (*fcmuo<mode>_and): New patterns.
18140 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
18142         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
18143         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
18144         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
18145         (cmp_op, sve_imm_con): New code attributes.
18146         (SVE_COND_INT_CMP, imm_con): Delete.
18147         (cmp_op): Remove above unspecs from int attribute.
18148         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
18149         to...
18150         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
18151         comparison-specific unspecs.
18152         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
18153         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
18154         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
18155         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
18156         (*vec_fcm<cmp_op><mode>): Rename to...
18157         (*fcm<cmp_op><mode>): ...this and adjust likewise.
18158         (*vec_fcmuo<mode>): Rename to...
18159         (*fcmuo<mode>): ...this and adjust likewise.
18160         (*pred_fcm<cmp_op><mode>): New pattern.
18161         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
18162         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
18163         functions.
18164         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
18165         and UNORDERED.
18166         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
18167         (aarch64_emit_sve_predicated_cond): New function.
18168         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
18169         (aarch64_emit_unspec_cond_or): Replace with...
18170         (aarch64_emit_sve_or_conds): ...this new function.  Use
18171         aarch64_emit_sve_ptrue_op for the individual comparisons and
18172         aarch64_emit_binop to OR them together.
18173         (aarch64_emit_inverted_unspec_cond): Replace with...
18174         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
18175         aarch64_emit_sve_ptrue_op for the comparison and
18176         aarch64_emit_unop to invert the result.
18177         (aarch64_expand_sve_vec_cmp_float): Update after the above
18178         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
18180 2018-05-07  Nathan Sidwell  <nathan@acm.org>
18182         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
18183         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
18184         (Backwards Compatibility): Likewise.
18186 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18188         PR bootstrap/85681
18189         Revert:
18190         2018-05-07  Luis Machado  <luis.machado@linaro.org>
18192         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18193         <prefetch_dynamic_strides>: New const bool field.
18194         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18195         prefetch_dynamic_strides.
18196         (exynosm1_prefetch_tune): Likewise.
18197         (thunderxt88_prefetch_tune): Likewise.
18198         (thunderx_prefetch_tune): Likewise.
18199         (thunderx2t99_prefetch_tune): Likewise.
18200         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18201         to false.
18202         (aarch64_override_options_internal): Update to set
18203         PARAM_PREFETCH_DYNAMIC_STRIDES.
18204         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18205         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18206         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18207         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18208         prefetch-dynamic-strides setting.
18210         2018-05-07  Luis Machado  <luis.machado@linaro.org>
18212         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18213         <minimum_stride>: New const int field.
18214         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18215         minimum_stride field.
18216         (exynosm1_prefetch_tune): Likewise.
18217         (thunderxt88_prefetch_tune): Likewise.
18218         (thunderx_prefetch_tune): Likewise.
18219         (thunderx2t99_prefetch_tune): Likewise.
18220         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18221         (aarch64_override_options_internal): Update to set
18222         PARAM_PREFETCH_MINIMUM_STRIDE.
18223         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18224         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18225         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18226         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18227         stride is constant and is below the minimum stride threshold.
18229 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18231         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
18232         to 512.
18234 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18236         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18237         <prefetch_dynamic_strides>: New const bool field.
18238         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18239         prefetch_dynamic_strides.
18240         (exynosm1_prefetch_tune): Likewise.
18241         (thunderxt88_prefetch_tune): Likewise.
18242         (thunderx_prefetch_tune): Likewise.
18243         (thunderx2t99_prefetch_tune): Likewise.
18244         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18245         to false.
18246         (aarch64_override_options_internal): Update to set
18247         PARAM_PREFETCH_DYNAMIC_STRIDES.
18248         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18249         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18250         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18251         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18252         prefetch-dynamic-strides setting.
18254 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18256         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18257         <minimum_stride>: New const int field.
18258         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18259         minimum_stride field.
18260         (exynosm1_prefetch_tune): Likewise.
18261         (thunderxt88_prefetch_tune): Likewise.
18262         (thunderx_prefetch_tune): Likewise.
18263         (thunderx2t99_prefetch_tune): Likewise.
18264         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18265         (aarch64_override_options_internal): Update to set
18266         PARAM_PREFETCH_MINIMUM_STRIDE.
18267         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18268         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18269         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18270         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18271         stride is constant and is below the minimum stride threshold.
18273 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
18275         PR c++/85659
18276         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
18277         the type is addressable.  Don't force op into register if it has
18278         BLKmode.
18280 2018-05-05  Roland McGrath  <mcgrathr@google.com>
18282         PR other/77609
18283         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
18284         any section for which we don't know a specific type it should have,
18285         regardless of name.  Previously this was done only for the exact
18286         names ".init_array", ".fini_array", and ".preinit_array".
18287         (default_elf_asm_named_section): Add comment about
18288         relationship with default_section_type_flags and SECTION_NOTYPE.
18289         (get_section): Don't consider it a type conflict if one side has
18290         SECTION_NOTYPE and the other doesn't, as long as neither has the
18291         SECTION_BSS et al used in the default_section_type_flags logic.
18293 2018-05-05  Tom de Vries  <tom@codesourcery.com>
18295         PR target/85653
18296         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
18297         (workaround_barsyncs): New function.
18298         (nvptx_reorg): Use workaround_barsyncs.
18299         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18300         (define_expand "nvptx_membar_cta"): New define_expand.
18301         (define_insn "*nvptx_membar_cta"): New insn.
18303 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
18305         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
18306         To improve optimization opportunities.
18307         * builtin-types.def: The new needed builtin types for the above.
18309 2018-05-04  Richard Biener  <rguenther@suse.de>
18311         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
18312         * gimple-ssa-store-merging.c
18313         (imm_store_chain_info::output_merged_store): Remove redundant create,
18314         release split_store vector contents on failure.
18315         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
18316         scalar stmt vector on cache hit.
18318 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
18320         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
18321         Xilinx FP support.
18322         * config.gcc (powerpc-xilinx-eabi*): Remove.
18323         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
18324         support.
18325         (fusion_addis_mem_combo_load): Ditto.
18326         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
18327         FP support.
18328         (rs6000_cpu_cpp_builtins): Ditto.
18329         * config/rs6000/rs6000-linux.c
18330         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
18331         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
18332         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
18333         support.
18334         (rs6000_setup_reg_addr_masks): Ditto.
18335         (rs6000_init_hard_regno_mode_ok): Ditto.
18336         (rs6000_option_override_internal): Ditto.
18337         (legitimate_lo_sum_address_p): Ditto.
18338         (rs6000_legitimize_address): Ditto.
18339         (rs6000_legitimize_reload_address): Ditto.
18340         (rs6000_legitimate_address_p): Ditto.
18341         (abi_v4_pass_in_fpr): Ditto.
18342         (setup_incoming_varargs): Ditto.
18343         (rs6000_gimplify_va_arg): Ditto.
18344         (rs6000_split_multireg_move): Ditto.
18345         (rs6000_savres_strategy): Ditto.
18346         (rs6000_emit_prologue_components): Ditto.
18347         (rs6000_emit_epilogue_components): Ditto.
18348         (rs6000_emit_prologue): Ditto.
18349         (rs6000_emit_epilogue): Ditto.
18350         (rs6000_elf_file_end): Ditto.
18351         (rs6000_function_value): Ditto.
18352         (rs6000_libcall_value): Ditto.
18353         * config/rs6000/rs6000.h: Ditto.
18354         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
18355         (TARGET_MINMAX): ... this.  New.
18356         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
18357         * config/rs6000/rs6000.md: Remove Xilinx FP support.
18358         (*movsi_internal1_single): Delete.
18359         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
18360         mfpu=, mxilinx-fpu): Delete.
18361         * config/rs6000/singlefp.h: Delete.
18362         * config/rs6000/sysv4.h: Remove Xilinx FP support.
18363         * config/rs6000/t-rs6000: Ditto.
18364         * config/rs6000/t-xilinx: Delete.
18365         * config/rs6000/titan.md: Adjust for fp_type removal.
18366         * config/rs6000/vsx.md: Remove Xilinx FP support.
18367         (VStype_simple): Delete.
18368         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
18369         * config/rs6000/xfpu.h: Delete.
18370         * config/rs6000/xfpu.md: Delete.
18371         * config/rs6000/xilinx.h: Delete.
18372         * config/rs6000/xilinx.opt: Delete.
18373         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
18374         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
18376 2018-05-04  Tom de Vries  <tom@codesourcery.com>
18378         PR libgomp/85639
18379         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
18380         if ignore == 0.
18382 2018-05-04  Richard Biener  <rguenther@suse.de>
18384         PR middle-end/85627
18385         * tree-complex.c (update_complex_assignment): We are always in SSA form.
18386         (expand_complex_div_wide): Likewise.
18387         (expand_complex_operations_1): Likewise.
18388         (expand_complex_libcall): Preserve EH info of the original stmt.
18389         (tree_lower_complex): Handle removed blocks.
18390         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
18391         on complex multiplication and division libcall builtins.
18393 2018-05-04  Richard Biener  <rguenther@suse.de>
18395         PR middle-end/85574
18396         * fold-const.c (negate_expr_p): Restrict negation of operand
18397         zero of a division to when we know that can happen without
18398         overflow.
18399         (fold_negate_expr_1): Likewise.
18401 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
18403         PR libstdc++/85466
18404         * real.h (real_nextafter): Declare.
18405         * real.c (real_nextafter): New function.
18406         * fold-const-call.c (fold_const_nextafter): New function.
18407         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
18408         CASE_CFN_NEXTTOWARD.
18409         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
18410         even when arg1_mode is different from arg0_mode.
18412 2018-05-03  Nathan Sidwell  <nathan@acm.org>
18414         * doc/extend.texi (Deprecated Features): Remove
18415         -ffriend-injection.
18416         (Backwards Compatibility): Likewise.
18417         * doc/invoke.texi (C++ Language Options): Likewise.
18418         (C++ Dialect Options): Likewise.
18420 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
18422         PR target/85530
18423         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
18424         _mm512_mask_mullox_epi64): New intrinsics.
18426 2018-05-03  Tom de Vries  <tom@codesourcery.com>
18428         PR testsuite/85106
18429         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18430         dump files): Add offload-tree.
18432 2018-05-03  Richard Biener  <rguenther@suse.de>
18434         PR tree-optimization/85615
18435         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
18436         to loops not nested in BBs loop father to avoid creating multi-entry
18437         loops.
18439 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18441         PR tree-optimization/70291
18442         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
18443         arguments.  Change return type to tree.  Emit libcall as a new
18444         statement rather than replacing existing one when inplace_p is true.
18445         (expand_complex_multiplication_components): New function.
18446         (expand_complex_multiplication): Expand floating-point complex
18447         multiplication using the above.
18448         (expand_complex_division): Rename inner_type parameter to type.
18449         Update expand_complex_libcall call-site.
18450         (expand_complex_operations_1): Update expand_complex_multiplication
18451         and expand_complex_division call-sites.
18453 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18455         PR target/85582
18456         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18457         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18458         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
18459         the highest significant bit of the shift count mask is clear.  In
18460         check whether and[sq]i3 is needed verify that all significant bits
18461         of the shift count other than the highest are set.
18463 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18465         PR libgomp/82428
18466         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
18467         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
18468         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
18469         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
18470         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
18471         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
18472         __builtin_goacc_parlevel_size.
18474 2018-05-02  Richard Biener  <rguenther@suse.de>
18476         PR tree-optimization/85597
18477         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
18478         do not use split vect_get_vec_defs call but call vect_get_slp_defs
18479         directly.
18481 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18483         PR testsuite/85106
18484         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18485         dump files): Add ltrans-tree.
18487 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18489         PR testsuite/85106
18490         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18491         dump files): Add wpa-ipa.
18493 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
18495         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
18496         powerpc*-*-linux*paired* target.
18497         * config/rs6000/750cl.h: Delete.
18498         * config/rs6000/paired.h: Delete.
18499         * config/rs6000/paired.md: Delete.
18500         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
18501         float support.
18502         * config/rs6000/rs6000-builtin.def: Remove paired float support.
18503         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
18504         comment.  Remove paired float support.
18505         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
18506         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
18507         VECTOR_PAIRED.
18508         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
18509         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
18510         declarations.
18511         * config/rs6000/rs6000.c: Remove paired float support.
18512         (paired_expand_vector_init, paired_expand_vector_move,
18513         paired_emit_vector_compare, paired_emit_vector_cond_expr,
18514         (paired_expand_lv_builtin, paired_expand_stv_builtin,
18515         paired_expand_builtin, paired_expand_predicate_builtin,
18516         paired_init_builtins): Delete.
18517         * config/rs6000/rs6000.h: Remove paired float support.
18518         * config/rs6000/rs6000.md: Remove paired float support.
18519         (move_from_CR_ov_bit): Delete.
18520         * config/rs6000/rs6000.opt (mpaired): Delete.
18521         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
18522         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
18524 2018-05-02  Richard Biener  <rguenther@suse.de>
18526         PR middle-end/85567
18527         * gimplify.c (gimplify_save_expr): When in SSA form allow
18528         SAVE_EXPRs to compute to SSA vars.
18530 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18532         PR target/85582
18533         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18534         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18535         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
18536         clobber operands[2], instead use a new pseudo.  Formatting fixes.
18538 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
18540         PR tree-optimization/85586
18541         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
18542         exit early for statements in the same group if the accesses are
18543         not strided.
18545 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18547         PR lto/85451
18548         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
18549         error message.
18551 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
18553         PR tree-optimization/85143
18554         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
18556 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18558         PR lto/85451
18559         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
18560         not found" error message.
18562 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18564         PR other/83786
18565         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
18566         * vec.c (test_ordered_remove_if): New function.
18567         (vec_c_tests): Call test_ordered_remove_if.
18568         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
18569         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
18570         * tree-vect-patterns.c (vect_pattern_recog_1): Use
18571         VEC_ORDERED_REMOVE_IF.
18573 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18575         PR tree-optimization/82665
18576         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
18577         pointer subtraction where arguments come from a memchr call.
18579 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
18581         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
18582         --push-state --as-needed and --pop-state instead of --as-needed and
18583         --no-as-needed if ld supports it.
18584         * configure: Regenerated.
18586         PR web/85578
18587         * doc/install.texi2html: Replace _002d with - and _002a with * in
18588         generated html files using sed.
18590 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18592         PR c++/85523
18593         * gcc-rich-location.c (blank_line_before_p): New function.
18594         (use_new_line): New function.
18595         (gcc_rich_location::add_fixit_insert_formatted): New function.
18596         * gcc-rich-location.h
18597         (gcc_rich_location::add_fixit_insert_formatted): New function.
18599 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18601         * selftest.c (assert_streq): Rename "expected" and "actual" to
18602         "val1" and "val2".  Extend NULL-handling to cover both inputs
18603         symmetrically, while still requiring both to be non-NULL for a pass.
18604         * selftest.h (assert_streq): Rename "expected" and "actual" to
18605         "val1" and "val2".
18606         (ASSERT_EQ): Likewise.
18607         (ASSERT_EQ_AT): Likewise.
18608         (ASSERT_KNOWN_EQ): Likewise.
18609         (ASSERT_KNOWN_EQ_AT): Likewise.
18610         (ASSERT_NE): Likewise.
18611         (ASSERT_MAYBE_NE): Likewise.
18612         (ASSERT_MAYBE_NE_AT): Likewise.
18613         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
18614         the assertion to pass.
18615         (ASSERT_STREQ_AT): Likewise.
18617 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18619         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
18620         interaction with -pie.
18622 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18624         * selftest.h: Fix alphabetization of per-source-file selftest
18625         declarations.
18627 2018-04-30  Jason Merrill  <jason@redhat.com>
18629         PR c++/61982 - dead stores to destroyed objects.
18630         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18631         of clobber.
18633 2018-04-30  Jason Merrill  <jason@redhat.com>
18635         * tree.c (build_clobber): New.
18636         * tree.h: Declare it.
18637         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18639 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18641         * diagnostic-show-locus.c (layout::layout): Update for
18642         location_get_source_line returning a char_span.
18643         (struct char_span): Move to input.h.
18644         (struct correction): Update for fields in char_span becoming
18645         private.
18646         (struct source_line): Update for location_get_source_line
18647         returning a char_span.
18648         (layout::print_line): Likewise.
18649         * edit-context.c (edited_file::print_content): Likewise.
18650         (edited_file::print_diff_hunk): Likewise.
18651         (edited_file::print_run_of_changed_lines): Likewise.
18652         (edited_file::get_num_lines): Likewise.
18653         (edited_line::edited_line): Likewise.
18654         * final.c (asm_show_source): Likewise.
18655         * input.c (location_get_source_line): Convert return type
18656         from const char * to char_span, losing the final "line_len"
18657         param.
18658         (dump_location_info): Update for the above.
18659         (get_substring_ranges_for_loc): Likewise.  Use a char_span
18660         when handling the literal within the line.
18661         (test_reading_source_line): Update for location_get_source_line
18662         returning a char_span.
18663         * input.h (class char_span): Move here from
18664         diagnostic-show-locus.c, converting from a struct to a class.
18665         Make data members private.
18666         (char_span::operator bool): New.
18667         (char_span::length): New.
18668         (char_span::get_buffer): New.
18669         (char_span::operator[]): New.
18670         (char_span::subspan): Make const.
18671         (char_span::xstrdup): New.
18672         (location_get_source_line): Convert return type from const char *
18673         to char_span, losing the final "line_size" param.
18675 2018-04-30  Jan Hubicka  <jh@suse.cz>
18677         * lto-wrapper.c (ltrans_priorities): New static var.
18678         (cmp_priority): New.
18679         (run_gcc): Read priorities and if doing parallel build order
18680         the Makefile by them.
18682 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18684         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18686 2018-04-30  Richard Biener  <rguenther@suse.de>
18688         * tree-cfg.c (verify_address): Remove base argument, add
18689         flag whether to check TREE_ADDRESSABLE and do that.
18690         (verify_expr): Remove.
18691         (verify_types_in_gimple_reference): Add pieces from verify_expr.
18692         (verify_gimple_assign_single): Likewise.
18693         (verify_gimple_switch): Likewise.
18694         (verify_expr_location_1): Dereference tp once.  Add (disabled)
18695         piece from verify_expr.
18696         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18698 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18700         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18702 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18704         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18705         (small_data_pattern): Likewise.
18706         (arc_rewrite_small_data): Likewise.
18707         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18708         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18709         (get_symbol_alignment): New function.
18710         (legitimate_small_data_address_p): Likewise.
18711         (legitimate_scaled_address): Update, call
18712         legitimate_small_data_address_p.
18713         (output_sdata): New static variable.
18714         (arc_print_operand): Update how we handle small data operands.
18715         (arc_print_operand_address): Likewise.
18716         (arc_legitimate_address_p): Update, use
18717         legitimate_small_data_address_p.
18718         (arc_rewrite_small_data_p): Remove.
18719         (arc_rewrite_small_data_1): Likewise.
18720         (arc_rewrite_small_data): Likewise.
18721         (small_data_pattern): Likewise.
18722         (compact_sda_memory_operand): Update to use
18723         legitimate_small_data_address_p and get_symbol_alignment.
18724         (prepare_move_operands): Don't rewite sdata pattern.
18725         (prepare_extend_operands): Remove.
18726         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18727         pattern.
18728         (zero_extendqisi2): Likewise.
18729         (zero_extendhisi2): Likewise.
18730         (extendqihi2): Likewise.
18731         (extendqisi2): Likewise.
18732         (extendhisi2): Likewise.
18733         (addsi3): Likewise.
18734         (subsi3): Likewise.
18735         (andsi3): Likewise.
18736         * config/arc/constraints.md (Usd): Change it to memory constraint.
18738 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18740         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18741         as source of std instructions.
18742         * config/arc/arc.md (movsi_insn): Update pattern predicate to
18743         allow 6-bit constants as source for store instructions.
18744         (movdi_insn): Update instruction pattern to allow 6-bit constants
18745         as source for store instructions.
18747 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18749         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18751 2018-04-30  Nathan Sidwell  <nathan@acm.org>
18752             Sandra Loosemore <sandra@codesourcery.com>
18754         * dumpfile.c (dump_open): Allow '-' for stdout.
18755         * doc/invoke.texi (Developer Options): Document dump filename
18756         determination early.  Document stdin/stdout selection.
18758 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
18760         Microblaze Target: PIC data text relative
18762         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18763         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18764         Add declaration.
18765         * config/microblaze/microblaze.h (microblaze_constant_address_p):
18766         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18767         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18768         New addressing mode for data-text relative position indepenedent code.
18769         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18770         'ADDRESS_SYMBOLIC_TXT_REL'.
18771         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18772         (microblaze_legitimate_pic_operand): Exclude function calls from
18773         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18774         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18775         addresses cases.
18776         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18777         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18778         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18779         for 'address + offset'.
18780         (microblaze_expand_prologue): Add new function prologue call for
18781         'r20' assignation.
18782         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18783         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18784         table in case of TARGET_PIC_DATA_TEXT_REL.
18785         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18786         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18787         Add new macros 'UNSPEC_TEXT',
18788         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18789         + exclude function calls from 'UNSPEC_PLT' in case of data text
18790         relative mode.
18791         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18792         new target hook for generating address diff vector tables in case of
18793         flag_pic.
18794         * doc/tm.texi : Regenerate.
18795         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18796         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18797         of addr diff vector generation.
18798         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18799         target hook definition.
18800         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18801         Add default function for generate_pic_addr_diff_vec -> flag_pic.
18802         * doc/invoke.texi (Add new pic option): Add new microblaze pic
18803         option for data text relative.
18805 2018-04-30  Richard Biener  <rguenther@suse.de>
18807         * tree-chrec.h (evolution_function_is_constant_p): Remove
18808         redundant check.
18809         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18811 2018-04-30  Richard Biener  <rguenther@suse.de>
18813         PR bootstrap/85571
18814         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18816 2018-04-30  Richard Biener  <rguenther@suse.de>
18818         PR tree-optimization/28364
18819         PR tree-optimization/85275
18820         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18821         copying first exit test.
18823 2018-04-28  Mark Wielaard  <mark@klomp.org>
18825         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18826         dwarf_version >= 5.
18827         (dwarf_AT): Handle DW_AT_addr_base.
18828         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18830 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
18832         PR target/84431
18833         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18834         (*ashl<dwi>3_doubleword_mask_1): Ditto.
18835         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18836         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18838 2018-04-28  Richard Biener  <rguenther@suse.de>
18840         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18841         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18842         to reflect use.  Only add interesting stmts.
18844 2018-04-27  Martin Jambor  <mjambor@suse.cz>
18846         PR ipa/85549
18847         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18848         the jump function allows for passing through aggregate values.
18850 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
18852         * input.h (in_system_header_at): Convert from macro to inline
18853         function.
18854         (from_macro_expansion_at): Likewise.
18855         (from_macro_definition_at): Likewise.
18857 2018-04-27  Jeff Law  <law@redhat.com>
18859         * config.gcc: Mark tile* targets as deprecated/obsolete.
18861 2018-04-27  Richard Biener  <rguenther@suse.de>
18863         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18864         fix for ILP32.
18866 2018-04-27  Richard Biener  <rguenther@suse.de>
18868         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18870 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
18872         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18873         with Yd constraint. Set "preferred_for_speed" attribute from
18874         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18875         with Yd constraint.
18876         (*movdi_internal): Ditto.
18877         (movti_interunit splitters): Remove
18878         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18879         (movdi_interunit splitters): Ditto.
18880         * config/i386/constraints.md (Ye): Remove.
18881         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18883 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18885         PR target/85512
18886         * config/aarch64/constraints.md (Usg): Limit to 31.
18887         (Usj): Limit to 63.
18889 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
18891         PR tree-optimization/85529
18892         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18893         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18894         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18895         zero extension or masking of the MSB bit.
18896         (optimize_range_tests): Add FIRST_BB argument, pass it through
18897         to optimize_range_tests_var_bound.
18898         (maybe_optimize_range_tests, reassociate_bb): Adjust
18899         optimize_range_tests callers.
18901 2018-04-26  Richard Biener  <rguenther@suse.de>
18902             Jakub Jelinek  <jakub@redhat.com>
18904         * cgraph.h (symbol_table): Just declare debug method here.
18905         * symtab.c (symbol_table::debug): Define.
18907 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
18909         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18911 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
18913         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18914         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18915         (*movdi_internal): Substitute Yi and Yj constraint with x
18916         and Ym and Yn constraint with y constraint.  Update "isa"
18917         attribute and set "preferred_for_speed" attribute from
18918         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18919         (*movsi_internal): Ditto.
18920         (*movdf_internal): Ditto.
18921         (*movsf_internal): Ditto.
18922         (*zero_extendsidi2): Ditto.
18923         * config/i386/sse.md (vec_set<mode>_0): Ditto.
18924         (sse2_loadld): Ditto.
18925         (*vec_extract<ssevecmodelower>_0): Ditto.
18926         (*vec_extractv4si_0_zext_sse4): Ditto.
18927         (vec_concatv2di): Ditto.
18928         (*vec_dup<mode>): Ditto.
18929         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18930         * config/i386/constraints.md (Yi): Remove.
18931         (Yj): Remove.
18932         (Ym): Remove.
18933         (Yn): Remove.
18935 2018-04-26  Nathan Sidwell  <nathan@acm.org>
18937         * dumpfile.c (dump_open): New.
18938         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18939         (dump_finish): Detect stdio/stderr by value not name.
18941 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
18943         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18945 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18947         PR target/84952
18948         * config/nvptx/nvptx.c (verify_neutering_jumps)
18949         (verify_neutering_labels): New function
18950         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18952 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18954         PR target/84025
18955         * config/nvptx/nvptx.c (needs_neutering_p): New function.
18956         (nvptx_single): Use needs_neutering_p to skip over insns that do not
18957         need neutering.
18959 2018-04-26  Richard Biener <rguenther@suse.de>
18960             Tom de Vries  <tom@codesourcery.com>
18962         PR lto/85422
18963         * lto-streamer-out.c (output_function): Fixup loops if required to match
18964         discovery done in the reader.
18966 2018-04-26  Richard Biener  <rguenther@suse.de>
18968         PR tree-optimization/85116
18969         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18970         have a loop exit from the single latch predecessor.  Remove
18971         case of header with just condition.
18972         (ch_base::copy_headers): Exclude infinite loops from any
18973         processing.
18974         (pass_ch::execute): Record exits.
18976 2018-04-26  Richard Biener  <rguenther@suse.de>
18978         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18979         prologue cost vector and pass it to vect_get_load_cost.
18980         (vect_get_peeling_costs_all_drs): Likewise.
18981         (vect_peeling_hash_get_lowest_cost): Likewise.
18982         (vect_enhance_data_refs_alignment): Likewise.
18984 2018-04-26  Richard Biener  <rguenther@suse.de>
18986         PR middle-end/85450
18987         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18988         checking of integer<->pointer conversions.
18989         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18990         sign-/zero-extending pointer types.
18991         (expand_omp_for_static_chunk): Likewise.
18993 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
18994             Jean Lee  <xiaoyur347@gmail.com>
18996         * config/mips/mips.c (mips_asan_shadow_offset): New function.
18997         (TARGET_ASAN_SHADOW_OFFSET): Define.
18998         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18999         true for -fsanitize=address.
19001 2018-04-25  Mark Wielaard  <mark@klomp.org>
19003         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
19004         shorter ones.
19006 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
19008         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
19009         than "alu", remove explicit "memory" and "imm_disp" attributes.
19010         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
19012         PR middle-end/85414
19013         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
19014         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
19015         gen_lowpart_no_emit.
19017 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
19019         PR target/85473
19020         * config/i386/i386.c (ix86_expand_builtin): Change memory
19021         operand to XI, extend p0 to Pmode.
19022         * config/i386/i386.md: Change unspec volatile and operand
19023         1 mode to XI, change operand 0 mode to P.
19025 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19027         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
19028         GET_MODE_MASK before any checking.
19029         (nds32_can_use_bset_p): Likewise.
19030         (nds32_can_use_btgl_p): Likewise.
19032 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19034         * config/nds32/nds32-doubleword.md: New define_split pattern for
19035         illegal register number.
19037 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19039         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
19041 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19043         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
19045 2018-04-25  Richard Biener  <rguenther@suse.de>
19047         * lto-streamer.h (LTO_major_version): Bump to 8.
19049 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
19051         * BASE-VER: Set to 9.0.0.
19053 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
19055         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
19056         in __abskf2 and __powikf2.
19058 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19060         PR target/85512
19061         * config/aarch64/constraints.md (Usg, Usj): New constraints.
19062         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
19063         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
19064         Use the above on operand 2.  Reindent.
19065         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
19067 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
19069         PR target/85485
19070         * common/config/i386/i386-common.c (ix86_handle_option): Don't
19071         handle OPT_mcet.
19072         * config/i386/i386.opt (mcet): Removed.
19073         * doc/install.texi: Remove -mcet documentation.
19074         * doc/invoke.texi: Likewise.
19076 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
19078         PR target/85485
19079         * doc/install.texi: Remove -mcet from bootstrap-cet.
19081 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
19083         PR target/85511
19084         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
19085         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
19086         if TARGET_64BIT.
19088         PR target/85503
19089         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
19090         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
19091         containing a CONST_VECTOR.
19093 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
19095         * doc/install.texi: Update newlib dependency for nvptx.
19097 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
19099         PR target/85508
19100         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
19101         instead of INTVAL when shifting x left.
19103 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
19105         PR tree-optimization/85478
19106         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
19107         vect_grouped_store_supported for single element vectors.
19109 2018-04-24  Richard Biener  <rguenther@suse.de>
19111         PR target/85491
19112         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
19113         load cost increase to the case of non-constant step.
19115 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
19117         PR target/84828
19118         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
19119         destination if any_malformed_asm.
19121 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
19123         PR middle-end/85496
19124         * expr.c (store_field): In the bitfield case, if the value comes from
19125         a function call and is returned in registers by means of a PARALLEL,
19126         do not change the mode of the temporary unless BLKmode and VOIDmode.
19128 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
19130         PR rtl-optimization/85423
19131         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
19132         dependencies to debug insns when the previous insn is non-debug.
19134 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
19136         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
19137         enums into a single definition.
19138         (fls): Fix predicates and printing.
19139         (seti): Likewise.
19141 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
19143         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
19144         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
19145         and short u6 immediate.
19146         (check_if_valid_sleep_operand): Remove.
19147         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
19149 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19151         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
19152         flag_always_save_lp condition.
19153         * config/nds32/nds32.opt (malways-save-lp): New option.
19155 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19157         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
19158         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
19159         * config/nds32/nds32.h
19160         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
19161         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
19163 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19165         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
19166         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
19168 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19169             Chung-Ju Wu  <jasonwucj@gmail.com>
19171         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
19172         Declare.
19173         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
19174         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
19176 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19178         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
19180 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19182         * config/nds32/nds32-protos.h (nds32_data_alignment,
19183         nds32_local_alignment): Declare.
19184         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
19185         nds32_local_alignment): New functions.
19186         (TARGET_CONSTANT_ALIGNMENT): Define.
19187         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
19189 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19191         * config/nds32/nds32.c
19192         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
19193         (TARGET_MODES_TIEABLE_P): Likewise.
19195 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19197         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
19198         level Ofast and Og.
19200 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
19201             Chung-Ju Wu  <jasonwucj@gmail.com>
19203         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19204         for unaligned access.
19205         * config/nds32/nds32-intrinsic.c: Implementation of expanding
19206         unaligned access.
19207         * config/nds32/nds32-intrinsic.md: Likewise.
19208         * config/nds32/nds32_intrinsic.h: Likewise.
19209         * config/nds32/nds32.h (nds32_builtins): Likewise.
19210         * config/nds32/nds32.opt (munaligned-access): New option.
19211         * config/nds32/nds32.c (nds32_asm_file_start): Display
19212         flag_unaligned_access status.
19214 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
19216         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
19217         -mno-relax is present.
19218         * config/riscv/linux.h (LINK_SPEC): Ditto.
19220 2018-04-20  Martin Sebor  <msebor@redhat.com>
19222         PR c/85365
19223         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
19224         for null pointers.
19225         (gimple_fold_builtin_stxcpy_chk): Same.
19226         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
19228 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
19230         PR target/85456
19231         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
19232         __powikf2 when long double is IEEE 128-bit.
19234 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
19236         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
19237         step to make sure stack always aligned.
19239 2018-04-20  Carl Love  <cel@us.ibm.com>
19241         PR target/83402
19242         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
19243         size check for arg0.
19245 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
19246             Tom de Vries  <tom@codesourcery.com>
19248         PR target/85445
19249         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
19250         Emit insns for calls too.
19251         (nvptx_find_par): Always look for worker-level predecessor insn.
19252         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
19253         calls.
19254         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
19255         (nvptx_process_pars): Propagate frames for calls.
19257 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
19259         PR target/85469
19260         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
19261         Removed.
19262         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
19263         (ix86_handle_option): Don't handle OPT_mibt.
19264         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
19265         __SHSTK__.
19266         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
19267         has_ibt and ibt.
19268         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
19269         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
19270         (ix86_target_macros): Define __CET__ with flag_cf_protection
19271         for -fcf-protection.
19272         * config/i386/i386.c (isa2_opts): Remove -mibt.
19273         * config/i386/i386.h (TARGET_IBT): Removed.
19274         (TARGET_IBT_P): Likewise.
19275         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
19276         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
19277         * config/i386/i386.opt (mcet): Update help message.
19278         (mshstk): Likewise.
19279         (mibt): Removed.
19280         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
19281         -mcet as an alias for -mshstk.
19283 2018-04-20  Richard Biener <rguenther@suse.de>
19285         PR middle-end/85475
19286         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
19287         complexity by forcing a single use of the multiply operand.
19289 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19291         ipa/85449
19292         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
19293         recursion dependency to only apply to non-clones.
19295 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19297         ipa/85447
19298         * ipa-cp.c (create_specialized_node): Check that clones of
19299         self-recursive edges exist during IPA-CP.
19301 2018-04-19  Toon Moene  <toon@moene.org>
19303         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
19304         by -O3.
19306 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19308         PR tree-optimization/85467
19309         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
19310         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
19311         VECTOR_CST element to type.
19313 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19315         PR target/85397
19316         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
19317         * config/i386/i386.md (builtin_setjmp_setup): Removed.
19318         (builtin_longjmp): Likewise.
19319         (save_stack_nonlocal): New pattern.
19320         (restore_stack_nonlocal): Likewise.
19322 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19324         PR target/85404
19325         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19326         Replace ASM_OUTPUT_LABEL with fprintf.
19328 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19330         PR target/85417
19331         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19332         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
19333         * config/i386/i386-c.c (ix86_target_macros_internal): Also
19334         define __IBT__ and __SHSTK__ for -fcf-protection.
19335         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
19336         TARGET_IBT.
19337         (ix86_trampoline_init): Likewise.
19338         (x86_output_mi_thunk): Likewise.
19339         (ix86_notrack_prefixed_insn_p): Likewise.
19340         (ix86_option_override_internal): Don't disallow -fcf-protection.
19341         * config/i386/i386.md (rdssp<mode>): Also enable for
19342         -fcf-protection.
19343         (incssp<mode>): Likewise.
19344         (nop_endbr): Likewise.
19345         * config/i386/i386.opt (mcet): Change help message to built-in
19346         functions only.
19347         (mibt): Likewise.
19348         (mshstk): Likewise.
19349         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
19350         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
19351         enable CET built-in functions.
19353 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
19355         * common/config/i386/i386-common.c
19356         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
19357         OPTION_MASK_ISA_MOVDIRI_UNSET,
19358         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
19359         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
19360         * config.gcc (movdirintrin.h): New header.
19361         * config/i386/cpuid.h (bit_MOVDIRI,
19362         bit_MOVDIR64B): New bits.
19363         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
19364         and -mmvodir64b.
19365         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
19366         (VOID, PVOID, PCVOID)): New function types.
19367         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
19368         __builtin_ia32_directstoreu_u64,
19369         __builtin_ia32_movdir64b): New builtins.
19370         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
19371         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
19372         and -mmovdiri.
19373         (ix86_valid_target_attribute_inner_p): Ditto.
19374         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
19375         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
19376         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
19377         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
19378         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
19379         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
19380         (movdiri<mode>, movdir64b_<mode>): New.
19381         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
19382         * config/i386/immintrin.h: Include movdirintrin.h.
19383         * config/i386/movdirintrin.h: New file.
19384         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
19386 2018-04-19  Richard Biener  <rguenther@suse.de>
19388         PR middle-end/85455
19389         * cfg.c (clear_bb_flags): When loop state says we have
19390         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
19392 2018-04-19  Richard Biener  <rguenther@suse.de>
19394         PR tree-optimization/84737
19395         * tree-vect-data-refs.c (vect_copy_ref_info): New function
19396         copying restrict info.
19397         (vect_setup_realignment): Use it.
19398         * tree-vectorizer.h (vect_copy_ref_info): Declare.
19399         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
19400         the first DR to all generated stores.
19401         (vectorizable_load): Likewise for loads.
19403 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19405         PR tree-optimization/85446
19406         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
19407         the integral and pointer types to have the same precision.
19409         * doc/install.texi: Document --disable-cet being the default and
19410         --enable-cet=auto.
19412 2018-04-18  Martin Liska  <mliska@suse.cz>
19414         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
19415         style.
19417 2018-04-18  Martin Liska  <mliska@suse.cz>
19419         Revert
19420         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
19422         PR ipa/83983
19423         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19424         arguments if they are comparable.
19426 2018-04-18  Martin Liska  <mliska@suse.cz>
19428         Revert
19429         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19431         PR lto/84805
19432         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19433         incomplete types.
19435 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
19437         PR target/85388
19438         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
19439         ENDBR after calling __morestack.
19441 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
19443         PR jit/85384
19444         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
19445         by using gcc_base_ver to generate a gcc_driver_version, and use
19446         it when generating GCC_DRIVER_NAME.
19447         * configure: Regenerate.
19449 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
19451         PR target/81084
19452         * config.gcc: Obsolete powerpc*-*-*spe*.
19454 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19456         PR debug/84637
19457         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
19458         (stabstr_D): Change type of unum from unsigned int to
19459         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
19460         type.
19462 2018-04-17  Jim Wilson  <jimw@sifive.com>
19464         PR 84856
19465         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
19466         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
19467         Set arg_pointer_offset after using pretend_args_size.
19469 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19471         PR rtl-optimization/85431
19472         * dse.c (record_store): Ignore zero width stores.
19474         PR sanitizer/85230
19475         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
19476         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
19477         __builtin_stack_restore rather than after it.
19478         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
19479         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
19480         argument instead of virtual_dynamic_stack_rtx.
19482 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19484         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
19485         New prototype.
19486         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19487         Add note to error message to explain internal mapping of overloaded
19488         built-in function name to non-overloaded built-in function name.
19489         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
19490         function.
19492 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
19494         PR target/85424
19495         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
19496         where the inputs overlap with the output.
19498 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19500         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
19501         (=v, v) alternative and explicit "memory" attribute.
19502         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
19503         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19504         attributes.
19505         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
19506         "sselog1" type instead of "sselog".
19507         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
19508         "sselog".  Remove explicit "memory" attribute.
19509         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
19510         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19511         attributes.
19512         (vec_extract_hi_v32hi): Merge all alternatives into one, use
19513         "sselog1" type instead of "sselog".  Remove explicit "memory"
19514         attribute.
19515         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
19516         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19517         attribute.
19518         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
19519         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19520         attributes.
19521         (vec_extract_hi_v64qi): Merge all alternatives into one, use
19522         "sselog1" type instead of "sselog".  Remove explicit "memory"
19523         attribute.
19524         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
19525         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19526         attribute.
19528         PR target/85430
19529         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
19531         PR middle-end/85414
19532         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
19533         on a SUBREG.
19535 2018-04-17  Martin Jambor  <mjambor@suse.cz>
19537         PR ipa/85421
19538         * ipa-cp.c (create_specialized_node): Call
19539         expand_all_artificial_thunks if necessary.
19541 2018-04-17  Martin Liska  <mliska@suse.cz>
19543         PR lto/85405
19544         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
19545         in message, remote space in between '_G' and '('.
19547 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19549         PR target/85281
19550         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
19551         avx512f_vmcmp<mode>3<round_saeonly_name>,
19552         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
19553         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
19554         avx512f_rndscale<mode><round_saeonly_name>,
19555         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
19556         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
19557         Use %<iptr>2 instead of %2 for -masm=intel.
19558         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
19559         avx512f_vcvttss2usi<round_saeonly_name>,
19560         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
19561         -masm=intel.
19562         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
19563         avx512f_vcvttsd2usi<round_saeonly_name>,
19564         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
19565         Use %q1 instead of %1 for -masm=intel.
19566         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
19567         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
19568         of %3 for -masm=intel.
19569         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
19570         -masm=intel.
19571         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
19572         -masm=intel.
19573         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
19574         -masm=intel.
19575         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
19576         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
19577         %g1.
19578         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
19579         -masm=intel.
19580         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
19581         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
19582         %g1 and one with %0 and %1.
19583         (avx512er_vmrcp28<mode><round_saeonly_name>,
19584         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
19585         %1 for -masm=intel.
19586         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
19587         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
19588         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
19589         of %0 and %{%4%} for -masm=intel.
19590         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
19591         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
19592         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
19593         order of %0 and %{%5%}%{z%} for -masm=intel.
19595 2018-04-17  Jan Hubicka  <jh@suse.cz>
19597         PR lto/85405
19598         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
19600 2018-04-17  Martin Liska  <mliska@suse.cz>
19602         PR ipa/85329
19603         * multiple_target.c (create_dispatcher_calls): Set apostrophes
19604         for target_clone error message.  Make default implementation
19605         clone to be a local declaration.
19606         (separate_attrs): Add new argument and check for an empty
19607         string.
19608         (expand_target_clones): Handle it.
19609         (ipa_target_clone): Make redirection just for target_clones
19610         functions.
19612 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
19613             Tom de Vries  <tom@codesourcery.com>
19615         PR middle-end/84955
19616         * omp-expand.c (expand_oacc_for): Add dummy false branch for
19617         tiled basic blocks without omp continue statements.
19619 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
19621         PR target/83660
19622         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19623         vec_extract expression as having side effects to make sure it gets
19624         a cleanup point.
19626 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
19628         PR target/85403
19629         * config/i386/i386.c (get_builtin_code_for_version): Check
19630         error_mark_node.
19632 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
19634         PR target/84331
19635         * config.gcc: Support "skylake".
19636         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19637         PROCESSOR_SKYLAKE.
19638         * config/i386/i386.c (m_SKYLAKE): Define.
19639         (processor_target_table): Add "skylake".
19640         (ix86_option_override_internal): Add "skylake".
19641         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19642         PROCESSOR_CANNONLAKE.
19643         (get_builtin_code_for_version): Fix priority for
19644         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19645         PROCESSOR_SKYLAKE-AVX512.
19646         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19647         (processor_type): Add PROCESSOR_SKYLAKE.
19649 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
19650             Jason Merrill  <jason@redhat.com>
19652         PR c++/85112
19653         * convert.c (convert_to_integer_1): Use direct recursion for
19654         enumeral types and types with a precision less than the number
19655         of bits in their mode.
19657 2018-04-16  Julia Koval  <julia.koval@intel.com>
19659         PR target/84413
19660         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19661         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19663 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
19665         PR target/85293
19666         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19667         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19668         and -mno-direct-move.
19670 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
19672         PR target/83402
19673         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19674         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19675         Ensure negative shifts result in {0}.
19677 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
19679         PR rtl-optimization/79916
19680         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19681         regs (if any) to define how to gnerate SD moves when LRA is in
19682         progress.
19684 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
19686         PR rtl-optimization/85393
19687         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19688         * except.c (expand_dw2_landing_pad_for_region): Make static.
19689         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19690         a label and unconditional jump to old_bb, rather than
19691         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19692         basic block.
19694         PR rtl-optimization/85376
19695         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19696         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19697         instead of a specific value.
19699 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19700             Bin Cheng  <bin.cheng@arm.com>
19702         PR tree-optimization/82965
19703         PR tree-optimization/83991
19704         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19705         by_profile_only parameter.
19706         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19707         information if the loop was predicted to iterate too many times.
19708         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19710 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19712         PR lto/71991
19713         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19714         always inline.
19716 2018-04-13  Martin Liska  <mliska@suse.cz>
19717             Jakub Jelinek  <jakub@redhat.com>
19719         PR middle-end/81657
19720         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19721         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19722         * builtins.c (expand_builtin_memory_copy_args): Use
19723         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19724         handle dest_addr == pc_rtx.
19726 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
19728         PR target/85291
19729         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19730         asked to not generate direct moves.
19731         (fix_trunc<mode>si2_stfiwx): Similar.
19732         (fix_trunc<mode>si2_internal): Similar.
19734 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19736         PR debug/83157
19737         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19738         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19739         lookup if dest in some wider mode is known to be const0_rtx and
19740         if so, record permanent equivalence for it to be ZERO_EXTEND of
19741         the narrower mode destination.
19743 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19745         * lto-streamer-out.c (output_function): Revert 259346.
19746         * omp-expand.c (expand_oacc_for): Likewise.
19748 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
19750         PR rtl-optimization/85354
19751         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19752         * sel-sched.c (sel_global_init): ... here.
19754 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
19756         PR target/85238
19757         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19758         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19759         mode for PE-COFF targets.
19760         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19761         (i386_pe_asm_lto_end): Likewise.
19762         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19763         (TARGET_ASM_LTO_END): Likewise.
19764         * config/i386/winnt.c (saved_debug_info_level): New static variable.
19765         (i386_pe_asm_lto_start): New function.
19766         (i386_pe_asm_lto_end): Likewise.
19768 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19769             Richard Biener  <rguenther@suse.de>
19771         PR middle-end/84955
19772         * lto-streamer-out.c (output_function): Fix CFG loop state before
19773         streaming out.
19774         * omp-expand.c (expand_oacc_for): Handle calls to internal
19775         functions like regular functions.
19777 2018-04-12  Richard Biener  <rguenther@suse.de>
19779         PR lto/85371
19780         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19781         for the early LTO debug to properly generate references to it
19782         during DIE emission.  Do not re-use that for the skeleton for
19783         split-dwarf.
19784         (dwarf2out_early_finish): Likewise.
19786 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19788         PR target/85328
19789         * config/i386/sse.md
19790         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19791         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19792         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19793         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19794         and output is a reg, avoid creating invalid lowpart subreg, but
19795         instead split into a 512-bit move.  Don't split if not AVX512VL,
19796         input is xmm16+ reg and output is a mem.
19797         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19798         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19799         xmm16+ reg and output is a mem.
19801 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19803         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19804         also for flag_dwarf2_cfi_asm.
19806 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19808         PR rtl-optimization/85342
19809         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19810         a bool scalar var inside of the loop instead.  Don't try to update
19811         recog_data.operand after failed apply_change_group.
19813 2018-04-12  Tom de Vries  <tom@codesourcery.com>
19815         PR target/85296
19816         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19817         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
19818         array with flexible array member as array without given dimension.
19819         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19820         argument for undefined param to true.
19822 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
19824         PR target/85321
19825         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19826         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19827         from PowerPC section.
19828         * config/rs6000/sysv4.opt (mcall-): Improve help text.
19829         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19830         help text that is too long.
19831         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19832         help text that is too long.
19833         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19834         help text that is too long.
19836 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
19838         * config/alpha/alpha.md (stack_probe_internal): Rename
19839         from "probe_stack".  Update all callers.
19841 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19843         PR rtl-optimization/84566
19844         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19845         sched_macro_fuse_insns.
19847 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19849         PR target/84301
19850         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19851         (compute_block_dependences): ... from here.
19853 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19855         PR tree-optimization/85331
19856         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19857         from int to HOST_WIDE_INT.
19859 2018-04-11  Martin Jambor  <mjambor@suse.cz>
19861         PR ipa/84149
19862         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19863         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19864         not the same as the source val.
19865         (cgraph_edge_brings_value_p): New parameter.
19866         (gather_edges_for_value): Pass destination value to
19867         cgraph_edge_brings_value_p.
19868         (perhaps_add_new_callers): Likewise.
19869         (get_info_about_necessary_edges): Likewise and exclude values brought
19870         only by self-recursive edges.
19871         (create_specialized_node): Redirect only clones of self-calling edges.
19872         (+self_recursive_pass_through_p): New function.
19873         (find_more_scalar_values_for_callers_subset): Use it.
19874         (find_aggregate_values_for_callers_subset): Likewise.
19875         (known_aggs_to_agg_replacement_list): Removed.
19876         (decide_whether_version_node): Re-calculate known constants for all
19877         remaining context clones.
19879 2018-04-11  Richard Biener  <rguenther@suse.de>
19881         PR lto/85339
19882         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19883         from early DWARF output.
19884         (dwarf2out_early_finish): Output line info unconditionally into
19885         early DWARF and add reference to it.
19887 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19889         PR target/85281
19890         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19891         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19892         other than V2DFmode using iptr mode attribute.
19893         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19895 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19897         PR rtl-optimization/84659
19898         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19900 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19902         PR debug/85302
19903         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19904         SIZEP is NULL.
19905         (output_loc_list): Pass address of a dummy size variable even in the
19906         locview handling loop.
19907         (index_location_lists): Add comment on why skip_loc_list_entry can't
19908         call size_of_locs.
19910 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19912         PR target/85261
19913         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19914         into register.
19916 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
19918         PR target/85321
19919         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19920         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19921         and -mstring-compare-inline-limit.
19923 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19925         PR target/85287
19926         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19927         for stack clash protection in a register whenever we need it to be in
19928         a register.
19930 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19932         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19933         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19935 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19937         PR target/85321
19938         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19939         the help text.
19940         (mlong-double-): Ditto.
19941         * config/rs6000/sysv4.opt (msdata=): Ditto.
19942         (mtls-size=): Ditto.
19944 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19946         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19947         erroneous entries for
19948         "vector int vec_ldl (int, long int *)", and
19949         "vector unsigned int vec_ldl (int, unsigned long int *)".
19950         Add comments and entries for
19951         "vector bool char vec_ldl (int, bool char *)",
19952         "vector bool short vec_ldl (int, bool short *)",
19953         "vector bool int vec_ldl (int, bool int *)",
19954         "vector bool long long vec_ldl (int, bool long long *)",
19955         "vector pixel vec_ldl (int, pixel *)",
19956         "vector long long vec_ldl (int, long long *)",
19957         "vector unsigned long long vec_ldl (int, unsigned long long *)".
19958         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19959         type tree bool_long_long_type_node and correct definition of
19960         bool_V2DI_type_node to make reference to this new type tree.
19961         (rs6000_mangle_type): Replace erroneous reference to
19962         bool_long_type_node with bool_long_long_type_node.
19963         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19964         comments to emphasize sign distinctions for char and int types and
19965         replace RS6000_BTI_bool_long constant with
19966         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
19967         use of RS6000_BTI_pixel.
19968         (bool_long_type_node): Remove this macro definition.
19969         (bool_long_long_type_node): New macro definition
19971 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19973         PR rtl-optimization/85300
19974         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19975         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19976         simplify_unary_operation fails.
19978 2018-04-10  Martin Liska  <mliska@suse.cz>
19980         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19981         cgraph_edge and ipa_ref.
19983 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19985         PR target/85177
19986         PR target/85255
19987         * config/i386/sse.md
19988         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19989         computation of the VEC_MERGE selector from mask.
19990         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19991         Fix decoding of the VEC_MERGE selector into mask.
19993 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19995         PR tree-optimization/85286
19996         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19998 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
20000         * final.c (final_1): Set insn_last_address as well as
20001         insn_current_address.
20003 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20005         PR target/85173
20006         * explow.c (emit_stack_probe): Call validize_mem on memory location
20007         before passing it to gen_probe_stack.  Create address operand and
20008         legitimize it for the probe_stack_address case.
20010 2018-04-09  Jan Hubicka  <jh@suse.cz>
20012         PR lto/85078
20013         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
20014         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
20015         * tree.c (free_lang_data_in_type): Fix handling of binfos;
20016         walk basetypes.
20017         (free_lang_data): Rebuild type inheritance graph.
20019 2018-04-09  Martin Sebor  <msebor@redhat.com>
20021         * invoke.texi (-finline-small-functions): Mention other optimization
20022         options.
20023         (-findirect-inlining, -fpartial-inlining): Same.
20024         (-finline-functions-called-once): Same.
20025         (-freorder-blocks-and-partition): Same.
20027 2018-04-09  Jan Hubicka  <jh@suse.cz>
20029         PR rtl/84058
20030         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
20031         jumps; choose last target that matches the criteria (i.e.
20032         no partition changes for non-crossing jumps).
20033         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
20034         support for redirecting crossing jumps to non-crossing.
20036 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
20038         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
20039         also for naked functions.
20041 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
20043         * config/arc/arc.md (add_shift): New pattern.
20044         (add_shift2): Likewise.
20045         (sub_shift): Likewise.
20046         (sub_shift_cmp0_noout): Likewise.
20047         (compare_si_ashiftsi): Likewise.
20048         (xbfu_cmp0_noout): New combine pattern.
20049         (xbfu_cmp0"): Likewise.
20050         (movsi_set_cc_insn): Place the predicable variant first.
20051         (commutative_binary_cmp0_noout): Remove clobber.
20052         (commutative_binary_cmp0): New pattern.
20053         (noncommutative_binary_cmp0): Likewise.
20054         (noncommutative_binary_cmp0_noout): Likewise.
20055         (noncommutative_binary_comparison_result_used): Removed.
20056         (rsub_cmp0): New pattern.
20057         (rsub_cmp0_noout): Likewise.
20058         (extzvsi): Changed, keep only meaningful variants.
20059         (SQH, SEZ): New iterators.
20060         (SQH_postfix): New mode attribute.
20061         (SEZ_prefix): New code attribute.
20062         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
20063         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
20064         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
20065         of numerical value.
20066         (noncommutative_operator): Check the availability of barrel
20067         shifter option.
20069 2018-04-09  Richard Biener  <rguenther@suse.de>
20071         PR tree-optimization/85284
20072         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
20073         Only use the niter constraining form of simple_iv when the exit
20074         is always executed.
20076 2018-04-09  Tom de Vries  <tom@codesourcery.com>
20078         PR target/84041
20079         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20080         (define_expand "*memory_barrier"): New define_expand.
20081         (define_insn "memory_barrier"): New insn.
20083 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20085         PR rtl-optimization/80463
20086         PR rtl-optimization/83972
20087         PR rtl-optimization/83480
20089         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
20090         correct producer for the insn.
20091         (tidy_control_flow): Fixup seqnos in case of debug insns.
20093 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20095         PR rtl-optimization/83913
20097         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
20098         different sched-times when merging exprs.
20100 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20102         PR rtl-optimization/83962
20104         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
20105         tidy_fallthru_edge and tidy_control_flow.
20107 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20109         PR rtl-optimization/83530
20111         * sel-sched.c (force_next_insn): New global variable.
20112         (remove_insn_for_debug): When force_next_insn is true, also leave only
20113         next insn in the ready list.
20114         (sel_sched_region): When the region wasn't scheduled, make another pass
20115         over it with force_next_insn set to 1.
20117 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
20119         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
20120         into tm_file.
20121         * config/nds32/constants.md (unspec_volatile_element): Add enum values
20122         for interrupt control.
20123         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
20124         functions for interrupt control.
20125         * config/nds32/nds32-intrinsic.md: Likewise.
20126         * config/nds32/nds32_intrinsic.h: Likewise.
20127         * config/nds32/nds32.h (nds32_builtins): Likewise.
20129 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
20131         * config/nds32/nds32.c (nds32_init_machine_status,
20132         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
20133         strict_aligned_p field.
20134         (nds32_expand_to_rtl_hook): New function.
20135         (TARGET_EXPAND_TO_RTL_HOOK): Define.
20136         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
20138 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20139             Chung-Ju Wu  <jasonwucj@gmail.com>
20141         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
20142         * config/nds32/nds32-n7.md: New file.
20143         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
20144         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
20145         pipeline.
20146         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
20147         * config/nds32/nds32.md (pipeline_model): Add n7.
20148         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
20149         * config/nds32/pipelines.md: Include n7 settings.
20151 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20152             Chung-Ju Wu  <jasonwucj@gmail.com>
20154         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
20155         * config/nds32/nds32-e8.md: New file.
20156         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
20157         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
20158         pipeline.
20159         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
20160         * config/nds32/nds32.md (pipeline_model): Add e8.
20161         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
20162         * config/nds32/pipelines.md: Include e8 settings.
20164 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20165             Chung-Ju Wu  <jasonwucj@gmail.com>
20167         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
20168         * config/nds32/nds32-n8.md: New file.
20169         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
20170         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
20171         pipeline.
20172         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
20173         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
20174         * config/nds32/nds32.md (pipeline_model): Add n8.
20175         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
20176         * config/nds32/pipelines.md: Include n8 settings.
20178 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20179             Chung-Ju Wu  <jasonwucj@gmail.com>
20181         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
20182         * config/nds32/nds32-n9-2r1w.md: New file.
20183         * config/nds32/nds32-n9-3r2w.md: New file.
20184         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
20185         nds32_register_ports): New or modify for cpu n9.
20186         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
20187         pipeline.
20188         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
20189         * config/nds32/nds32-utils.c: New file.
20190         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
20191         TARGET_MUL_SLOW): Define.
20192         * config/nds32/nds32.md (pipeline_model): New attribute.
20193         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
20194         New options that support cpu n9.
20195         * config/nds32/pipelines.md: Include n9 settings.
20196         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
20198 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
20200         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
20201         information if necessary.
20202         (output_cond_branch_compare_zero): Likewise.
20203         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
20204         (nds32_target_alignment): Refine for alignment.
20205         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
20206         (FUNCTION_BOUNDARY): Modify.
20207         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
20208         align case.
20209         * config/nds32/nds32.opt (malways-align, malign-functions): New.
20211 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
20213         * config/nds32/constants.md (unspec_volatile_element): Add values for
20214         TLB operation and data prefetch.
20215         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
20216         functions for TLB operation and data prefetch.
20217         * config/nds32/nds32-intrinsic.md: Likewise.
20218         * config/nds32/nds32_intrinsic.h: Likewise.
20219         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
20220         (nds32_print_operand): Likewise.
20221         * config/nds32/nds32.h (nds32_builtins): Likewise.
20223 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
20224         Andrew Pinski <pinsika@gcc.gnu.org>
20226         PR middle-end/82976
20227         * match.pd: Use constant_boolean_node of correct type instead of
20228         boolean_true_node or boolean_false_node for simplifying
20229         pointer comparisons to zero.
20231 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20233         PR tree-optimization/80021
20234         * tree.c (verify_type_variant): Make error call in verify_variant_match
20235         translatable and remove final full stop.
20237 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20239         * config/nds32/constants.md (unspec_volatile_element): Add
20240         UNSPEC_VOLATILE_EH_RETURN.
20241         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
20242         nds32_output_stack_pop): Support dwarf exception handling process.
20243         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
20244         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
20245         exception handling process.
20246         (nds32_compute_stack_frame): Likewise.
20247         (nds32_return_addr_rtx): Likewise.
20248         (nds32_initial_elimination_offset): Likewise.
20249         (nds32_expand_prologue): Likewise.
20250         (nds32_expand_epilogue): Likewise.
20251         (nds32_dynamic_chain_address): New function.
20252         * config/nds32/nds32.h (machine_function): Add fields for dwarf
20253         exception handling.
20254         (DYNAMIC_CHAIN_ADDRESS): Define.
20255         (EH_RETURN_DATA_REGNO): Define.
20256         (EH_RETURN_STACKADJ_RTX): Define.
20257         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
20258         patterns for dwarf exception handling.
20260 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20262         * config/nds32/nds32.h: Clean up obsolete macros.
20264 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20266         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20267         Add enum values for particular instructions.
20268         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20269         particular intrinsic functions.
20270         * config/nds32/nds32-intrinsic.md: Likewise.
20271         * config/nds32/nds32_intrinsic.h: Likewise.
20272         * config/nds32/nds32.h (nds32_builtins): Likewise.
20273         * config/nds32/nds32.md (type): Add pbsad and pbsada.
20274         (btst, ave): New patterns for particular instructions.
20276 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20278         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20279         Add enum values for atomic load/store and memory sync.
20280         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
20281         and memory sync.
20282         * config/nds32/nds32-intrinsic.md: Likewise.
20283         * config/nds32/nds32_intrinsic.h: Likewise.
20284         * config/nds32/nds32.h (nds32_builtins): Likewise.
20286 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20288         PR tree-optimization/85257
20289         * fold-const.c (native_encode_vector): If not all elts could fit
20290         and off is -1, return 0 rather than offset.
20291         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
20292         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
20293         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
20294         adjust buffer in native_interpret_expr call.
20296 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20298         * config/nds32/constants.md (unspec_volatile_element): Add cache
20299         control enum values.
20300         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
20301         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
20302         * config/nds32/nds32.c (nds32_cctl_names): New.
20303         (nds32_print_operand): Handle cache control register names.
20304         * config/nds32/nds32.h (nds32_builtins): New enum values.
20305         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
20306         macros.
20307         * config/nds32/nds32.md (type): Add mmu.
20308         * config/nds32/pipelines.md (simple_insn): Add mmu.
20310 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20312         * config/nds32/nds32.md (type): Remove call.
20313         * config/nds32/pipelines.md (simple_insn): Likewise.
20315 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20317         * config/nds32/constants.md (unspec_volatile_element): Add
20318         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
20319         UNSPEC_VOLATILE_FMFCFG.
20320         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
20321         description for fmfcfg and fmfcsr.
20322         (bdesc_1arg): Add fmtcsr.
20323         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
20324         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
20325         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
20326         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
20327         unspec_fmfcfg): New patterns.
20328         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
20329         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
20330         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
20331         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
20332         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
20333         __nds32__fmfcfg): Define.
20335 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20337         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
20338         intrinsic register names.
20339         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
20340         intrinsic register enum values and macros.
20342 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20344         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
20345         for load/store addressing form.
20346         (nds32_print_operand_address): Likewise.
20348 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
20350         PR target/85196
20351         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
20352         based on LABEL_REF.  Remove useless assertion.
20353         (pic_address_needs_scratch): Fix formatting.
20354         (sparc_legitimize_pic_address): Minor tweaks.
20355         (sparc_delegitimize_address): Adjust assertion accordingly.
20356         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
20357         into symbolic_operand.
20358         (movsi_high_pic_label_ref): Likewise.
20359         (movsi_lo_sum_pic_label_ref): Likewise.
20360         (movdi_pic_label_ref): Likewise.
20361         (movdi_high_pic_label_ref): Likewise.
20362         (movdi_lo_sum_pic_label_ref): Likewise.
20364 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
20366         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
20367         custom LIB_SPEC setup.
20369 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
20370             Kito Cheng  <kito.cheng@gmail.com>
20372         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
20373         * config/riscv/freebsd.h: New.
20375 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20377         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
20378         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
20379         file.
20381 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20382             Kito Cheng  <kito.cheng@gmail.com>
20384         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
20385         nds32_output_call, nds32_symbol_binds_local_p): New functions.
20386         * config/nds32/nds32-protos.h (nds32_output_call,
20387         nds32_output_return): Declare.
20388         * config/nds32/nds32.md: Refine all the call and return patterns.
20390 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20392         PR debug/85252
20393         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
20394         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
20396         PR rtl-optimization/84872
20397         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
20398         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
20399         EDGE_CROSSING edge.
20401 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
20403         * expr.c (copy_blkmode_to_reg): Revert 254862.
20404         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
20406 2018-04-06  Richard Biener  <rguenther@suse.de>
20408         PR middle-end/85244
20409         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
20410         after seeing a component reference with an adjacent field.  Treat
20411         refs to arrays at struct end of external decls similar to
20412         refs to unconstrained commons.
20414 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20416         PR sanitizer/85213
20417         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
20418         look through SAVE_EXPRs with non-side-effects argument.  Adjust
20419         recursive calls.
20420         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
20421         save_p here.
20423 2018-04-06  Richard Biener  <rguenther@suse.de>
20425         PR middle-end/85180
20426         * alias.c (find_base_term): New wrapper around find_base_term
20427         unwinding CSELIB_VAL_PTR changes.
20428         (find_base_term): Do not restore CSELIB_VAL_PTR during the
20429         recursion.
20431 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20433         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
20434         instructions.
20435         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
20436         constant definitions.
20437         ("nop"): lr 0,0 -> nopr r0
20438         ("nop_lr0", "nop_lr1"): New insn definitions.
20440 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20442         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
20443         NDS32_V3PUSH_AVAILABLE_P macro.
20445 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
20446             Chung-Ju Wu  <jasonwucj@gmail.com>
20448         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
20449         (nds32*-*-*): Add float and fpu_config into supported_defaults.
20450         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
20451         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
20452         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
20453         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
20454         * config/nds32/constraints.md: New constraints and checking for hard
20455         float configuration.
20456         * config/nds32/iterators.md: New mode iterator and attribute for hard
20457         float configuration.
20458         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
20459         patterns.
20460         * config/nds32/nds32-fpu.md: New file.
20461         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
20462         deal with hard float code generation.
20463         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
20464         ARCH_V3S.
20465         (abi_type, float_reg_number): New enum type.
20466         * config/nds32/nds32-predicates.c: New predicates for hard float.
20467         * config/nds32/nds32-protos.h: Declare functions for hard float.
20468         * config/nds32/nds32.c: Implementation for hard float configuration.
20469         * config/nds32/nds32.h: Definitions for hard float configuration.
20470         * config/nds32/nds32.md: Include hard float machine description and
20471         modify patterns for hard float configuration.
20472         * config/nds32/nds32.opt: New options for hard float configuration.
20473         * config/nds32/predicates.md: New predicates for hard float
20474         configuration.
20476 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20478         * common/config/nds32/nds32-common.c
20479         (nds32_option_optimization_table): Enable -mreleax-hint by default.
20481 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
20483         PR middle-end/85195
20484         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
20485         CONSTRUCTOR_ELT (ctor, ...)->value.
20487 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
20489         PR target/85193
20490         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
20492 2018-04-05  Tom de Vries  <tom@codesourcery.com>
20494         PR target/85204
20495         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
20496         cond jump.
20498 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20499             Kito Cheng  <kito.cheng@gmail.com>
20501         * config/nds32/constraints.md (U33): Fine-tune checking condition.
20502         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
20503         * config/nds32/nds32.h (nds32_16bit_address_type): Add
20504         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
20506 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20507             Kito Cheng  <kito.cheng@gmail.com>
20509         * config/nds32/constraints.md (Ufe): New memory constraint.
20510         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
20511         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
20512         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
20513         operands.
20514         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
20515         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
20517 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20519         * config/nds32/nds32.md: Use optimize_size in the condition for
20520         alu-shift instructions.
20522 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20524         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
20526 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20528         * config/nds32/nds32.md (negsi2): Refine pattern.
20530 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20531             Chung-Ju Wu  <jasonwucj@gmail.com>
20533         * config/nds32/iterators.md (shift_rotate): New code iterator.
20534         (shift): New code attribute.
20535         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
20536         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
20537         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
20538         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
20539         bit-wise operations.
20540         (andsi3, *andsi3): Ditto.
20541         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
20542         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
20543         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
20544         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
20545         nds32_ior_operand, nds32_xor_operand): New predicates.
20547 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20549         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
20550         (addsi3, subsi3): ... this.
20552 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20554         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
20556 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20558         * config/nds32/nds32.md: Adjust indention.
20560 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20562         * config/nds32/nds32.md (feature): New attribute.
20564 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20566         * config/nds32/nds32.md (subtype): New attribute.
20568 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20570         PR target/85203
20571         * config/arm/arm-builtins.c (arm_expand_builtin): Change
20572         expansion to perform a bitwise AND of the argument followed by a
20573         boolean negation of the result.
20575 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
20577         PR rtl-optimization/84878
20578         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
20579         the basic block.  Assert the use reference is not artificial and that
20580         it has an associated insn.
20582 2018-04-04  Michael Matz  <matz@suse.de>
20584         * builtins.c (compute_objsize): Pass correct operand
20585         to array_at_struct_end_p.
20587 2018-04-04  Richard Biener  <rguenther@suse.de>
20589         PR lto/85176
20590         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
20591         from contexts for DINFO_LEVEL_TERSE and below.
20593 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20595         * config/nds32/nds32-doubleword.md (move_<mode>): Require
20596         resiter_operand condition.
20597         * config/nds32/nds32.md (*move<mode>): Ditto.
20599 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20600             Monk Chiang  <sh.chiang04@gmail.com>
20602         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
20604 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20606         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
20608 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20609             Kito Cheng  <kito.cheng@gmail.com>
20611         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
20612         nds32_cond_code_str, output_cond_branch,
20613         output_cond_branch_compare_zero, nds32_expand_cbranch,
20614         nds32_expand_cstore, nds32_expand_movcc,
20615         nds32_output_cbranchsi4_equality_zero,
20616         nds32_output_cbranchsi4_equality_reg,
20617         nds32_output_cbranchsi4_equality_reg_or_const_int,
20618         nds32_output_cbranchsi4_greater_less_zero: New functions.
20619         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
20620         nds32_expand_cstore, nds32_expand_movcc,
20621         nds32_output_cbranchsi4_equality_zero,
20622         nds32_output_cbranchsi4_equality_reg,
20623         nds32_output_cbranchsi4_equality_reg_or_const_int,
20624         nds32_output_cbranchsi4_greater_less_zero): Declare.
20625         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20626         nds32_rimm11s_operand): New predicates.
20627         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20628         * config/nds32/nds32.md: Rewrite all the branch and conditional move
20629         patterns.
20631 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20633         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20634         * config/nds32/nds32.md: Ditto.
20635         * config/nds32/pipelines.md: Ditto.
20637 2018-04-04  Richard Biener  <rguenther@suse.de>
20639         PR tree-optimization/85168
20640         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20641         propagating abnormals.
20643 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20645         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20647 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20648             Kito Cheng  <kito.cheng@gmail.com>
20650         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20651         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20652         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20653         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20654         * config/nds32/nds32.md (sibcall_internal): New.
20655         (sibcall_register): Remove.
20656         (sibcall_immediate): Remove.
20657         (sibcall_value_internal): New.
20658         (sibcall_value_register): Remove.
20659         (sibcall_value_immediate): Remove.
20660         * config/nds32/predicates.md (nds32_general_register_operand): New.
20661         (nds32_call_address_operand): New.
20663 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20665         PR rtl-optimization/85167
20666         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20667         bb_defs if *split_p, instead preinitialize it to NULL.
20669         PR tree-optimization/85156
20670         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20671         evaluating the argument multiple times.
20673 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
20675         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20676         than vector.
20677         (_mm_cvtpd_ps): Likewise.
20678         (_mm_cvttpd_epi32): Likewise.
20679         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20680         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20681         vector, pixel, and bool following altivec.h include.
20683 2018-04-03  Martin Sebor  <msebor@redhat.com>
20685         * doc/extend.texi (Common Function Attributes): Clarify.
20686         (const attribute): Likewise.
20687         (pure attribute): Likewise.
20689 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20691         PR target/85169
20692         * config/i386/i386.c (ix86_expand_vector_set): Use
20693         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
20695 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
20697         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20698         instructions when changing rounding bits to preserve precision bits
20699         in the x87 control word.
20701 2018-04-03  Martin Liska  <mliska@suse.cz>
20703         PR tree-optimization/82491
20704         * rtl.h (strip_offset_and_add): Replace += suboffset with
20705         poly_uint64 () + suboffset.
20707 2018-03-29  Martin Liska  <mliska@suse.cz>
20708             Martin Jambor  <mjambor@suse.cz>
20710         PR ipa/84947
20711         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20712         param_type is not an integral or pointer type.
20714 2018-04-03  Richard Biener  <rguenther@suse.de>
20716         * sese.h (recompute_all_dominators): Remove.
20718 2018-04-02  Martin Sebor  <msebor@redhat.com>
20720         * doc/invoke.texi (-Wrestrict): Fix typos.
20722 2018-04-02  Jim Wilson  <jimw@sifive.com>
20724         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20725         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20726         (<optab>di3, <optab>si3_extend): Likewise.
20727         (<optab>si3_mask, <optab>si3_mask_1): New.
20728         (<optab>di3_mask, <optab>di3_mask_1): New.
20729         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20730         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20731         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20733 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
20735         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20736         example.
20738 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20740         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20741         (nds32_canonicalize_comparison): New function.
20743 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20744             Kito Cheng  <kito.cheng@gmail.com>
20745             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20747         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20748         * config/nds32/constants.md (unspec_volatile_element): Add
20749         UNSPEC_VOLATILE_RELAX_GROUP.
20750         * config/nds32/nds32-relax-opt.c: New file.
20751         * config/nds32/nds32-predicates.c
20752         (nds32_symbol_load_store_p): New function.
20753         * config/nds32/nds32-protos.h
20754         (nds32_symbol_load_store_p): Declare function.
20755         (make_pass_nds32_relax_opt): Declare new rtl pass function.
20756         * config/nds32/nds32.c
20757         (nds32_register_pass): New function to register pass.
20758         (nds32_register_passes): New function to register passes.
20759         * config/nds32/nds32.md (relax_group): New pattern.
20760         * config/nds32/nds32.opt (mrelax-hint): New option.
20761         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20763 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
20765         * config/nds32/t-nds32: Modify files dependency.
20767 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20769         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20770         (PROFILE_HOOK): Define its implementation.
20772 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20774         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20775         type and 32-bit size.
20777 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
20779         PR middle-end/85090
20780         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20781         (V_128_256): New mode iterator.
20782         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20783         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20784         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20785         of V.
20786         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20787         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20789 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
20791         PR target/83315
20792         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20793         NaN inputs correctly.
20795 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
20797         PR target/80546
20798         * config/rs6000/vsx.md (??r): New mode attribute.
20799         (*vsx_mov<mode>_64bit): Use it.
20800         (*vsx_mov<mode>_32bit): Likewise.
20802 2018-03-30  Martin Sebor  <msebor@redhat.com>
20804         PR tree-optimization/84818
20805         * builtins.c (check_access): Use warning_n.
20807 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20809         PR target/83822
20810         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20811         condition.
20812         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20813         condition.
20815 2018-03-30  Julia Koval  <julia.koval@intel.com>
20817         PR target/84413
20818         * x86-tune.def (movx, partial_reg_dependency): Enable for
20819         m_SKYLAKE_AVX512.
20821 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
20823         PR inline-asm/84985
20824         * lra-constraints.c (process_alt_operands): Move setting
20825         this_alternative_matches below.
20827 2018-03-29  Martin Liska  <mliska@suse.cz>
20829         PR lto/84995.
20830         * doc/invoke.texi: Document how LTO works with debug info.
20831         Describe auto-load support of binutils.  Mention 'x86-64'
20832         as valid option value of -march option.
20834 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
20836         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20838         PR c/85094
20839         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20840         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20841         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20842         checking.
20844 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20846         PR target/84912
20847         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20848         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20849         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20850         for RS6000_BTM_POWERPC64.
20851         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20852         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20853         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20854         definition.
20855         (DIVDE): Use it.
20856         (DIVDEU): Likewise.
20858 2018-03-28  Carl Love  <cel@us.ibm.com>
20860         Revert
20861         2017-09-27  Carl Love  <cel@us.ibm.com>
20863         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20864         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20865         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20866         fctiw instruction.
20868 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20870         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20871         instead of __vector bool.
20872         (_mm_max_pu8): Likewise.
20873         (_mm_min_pi16): Likewise.
20875 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20877         PR target/84912
20878         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20879         (DIVWEUO): Likewise.
20880         (DIVDEO): Likewise.
20881         (DIVDEUO): Likewise.
20882         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20883         DIVWEUO and DIVDEUO.
20884         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20885         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20886         (div_extend): Likewise.
20887         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20888         builtin function.
20889         (__builtin_divweuo): Likewise.
20890         (__builtin_divdeo): Likewise.
20891         (__builtin_divdeuo): Likewise.
20893 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20895         PR target/85095
20896         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20897         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20899         PR tree-optimization/82004
20900         * gimple-match-head.c (optimize_pow_to_exp): New function.
20901         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20902         Don't fold to exp if optimize_pow_to_exp is false.
20904 2018-03-28  Martin Liska  <mliska@suse.cz>
20906         PR other/84819
20907         * calls.c (initialize_argument_information): Fix trailing space.
20908         * common.opt: Fix typo and provide better explanation for
20909         -fsanitize-coverage option.
20910         * config/i386/i386.opt: Fix typo.
20912 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20913             Martin Liska  <mliska@suse.cz>
20915         PR sanitizer/85081
20916         * gimplify.c (asan_poison_variable): Don't do the check for
20917         gimplify_omp_ctxp here.
20918         (gimplify_decl_expr): Do it here.
20919         (gimplify_target_expr): Likewise.
20921 2018-03-28  Martin Liska  <mliska@suse.cz>
20923         PR target/84988
20924         * config/i386/i386.c (ix86_function_arg_advance): Do not call
20925         chkp_type_bounds_count if MPX is not enabled.
20927 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20929         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20931 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
20933         PR target/84914
20934         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20935         function to create the function decl for complex long double
20936         multiply and divide for -mabi=ieeelongdouble.
20937         (init_float128_ieee): Call it.
20939 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
20941         PR target/85044
20942         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20943         -fcf-protection=branch -mibt.
20944         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20946 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20948         PR target/81863
20949         * config/arm/arm.c (arm_valid_symbolic_address): Handle
20950         arm_word_relocations.
20952 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
20954         PR target/85056
20955         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20956         extern array declarations.
20958 2018-03-27  Richard Biener  <rguenther@suse.de>
20960         PR middle-end/84067
20961         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20962         explicit single_use checks.
20964 2018-03-27  Richard Biener  <rguenther@suse.de>
20966         PR tree-optimization/85082
20967         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20968         Valueize the VUSE.
20970 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20972         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20973         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20974         Turn on fasynchronous-unwind-tables and funwind-tables.
20976 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
20978         PR target/85073
20979         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20980         (*bmi_blsr_<mode>_ccz): Ditto.
20982 2018-03-26  Tom de Vries  <tom@codesourcery.com>
20984         PR tree-optimization/85063
20985         * omp-general.c (offloading_function_p): New function.  Factor out
20986         of ...
20987         * omp-offload.c (pass_omp_target_link::gate): ... here.
20988         * omp-general.h (offloading_function_p): Declare.
20989         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20990         with attribute omp declare target for offloading functions.
20992 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
20994         PR tree-optimization/84005
20995         * tree-data-ref.h (get_base_for_alignment): Declare.
20996         * tree-data-ref.c (get_base_for_alignment_1): New function.
20997         (get_base_for_alignment): Likewise.
20998         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20999         get_base_for_alignment to find a suitable base object, instead
21000         of always using drb->base_address.
21002 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
21004         PR inline-asm/85022
21005         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
21006         known size by default.
21008 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
21010         PR inline-asm/85030
21011         * lra-constraints.c (process_alt_operands): Don't match BLKmode
21012         and non BLKmode operands.
21014 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21016         PR target/85026
21017         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
21018         Clean up attributes.
21020 2018-03-23  Richard Biener  <rguenther@suse.de>
21022         PR debug/85020
21023         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
21024         we are going to emit early debug for LTO.
21026 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
21028         PR inline-asm/85034
21029         * function.c (match_asm_constraints_1): Don't optimize if input
21030         doesn't satisfy general_operand predicate for output's mode.
21032         PR inline-asm/85022
21033         * alias.c (write_dependence_p): Don't require for x_canonicalized
21034         non-VOIDmode if x has VOIDmode.
21036         PR sanitizer/85029
21037         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
21038         just don't try to optimize it rather than assert it never happens.
21040 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21042         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
21043         macro expansions for definition of ST_INTERNAL_<mode> and
21044         LD_INTERNAL_<mode> builtins.
21045         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
21046         Remove prototype.
21047         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
21048         function.
21049         (altivec_expand_st_builtin): Likewise.
21050         (altivec_expand_builtin): Remove calls to deleted functions.
21051         (rs6000_address_for_altivec): Delete this function.
21052         * config/rs6000/vector.md: Remove expands for
21053         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
21055 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
21057         PR target/84826
21058         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
21059         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
21060         re-computing once computed.
21061         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
21062         (arm_init_machine_status): Initialize
21063         machine->static_chain_stack_bytes.
21065 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21067         PR target/84760
21068         * doc/extend.texi: Add four new prototypes for vec_ld.
21069         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
21070         definitions for more logical presentation.
21071         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
21072         entries for V1TI variants of __builtin_altivec_ld builtin.
21073         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
21074         handling of V1TI variant of LVX icode pattern.
21075         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
21076         (rs6000_gimple_fold_builtin): Likewise.
21077         (altivec_init_builtins): Add code to define
21078         __builtin_altivec_lvx_v1ti function.
21080 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
21082         PR inline-asm/84941
21083         * function.c (match_asm_constraints_1): Don't do the optimization
21084         if input isn't a REG, SUBREG, MEM or constant.
21086 2018-03-22  Tom de Vries  <tom@codesourcery.com>
21088         PR tree-optimization/84956
21089         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
21090         bb_has_abnormal_pred.
21092 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
21094         PR sanitizer/85018
21095         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
21096         DECL_INITIAL (decl) to decl at the end.
21097         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
21098         adjust the comment.
21100 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
21102         * doc/extend.texi (__builtin_tgmath): Document when complex
21103         integer types are treated as _Complex _Float64.
21105 2018-03-21  Tom de Vries  <tom@codesourcery.com>
21107         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
21109 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
21111         PR tree-optimization/84960
21112         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
21113         if it is ENTRY block, move them into single succ of ENTRY in that case.
21115 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
21117         PR tree-optimization/84811
21118         * poly-int.h (poly_span_traits): Remove the T3 parameter and
21119         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
21120         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
21121         (known_subrange_p): Update accordingly.  Cast each value involved
21122         in the size comparison, rather than casting the result of the
21123         subtraction.
21125 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
21127         PR tree-optimization/84982
21128         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
21129         by flipping the least significant bit rather than all bits from
21130         bitpos to bitpos + bitsize - 1.
21132 2018-03-21  Nathan Sidwell  <nathan@acm.org>
21134         * doc/extend.texi (Deprecated Features): Remove mention of
21135         long-deleted deprecations.
21137 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21139         PR jit/84288
21140         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
21141         * configure: Regenerate.
21143 2018-03-21  Tom de Vries  <tom@codesourcery.com>
21145         PR tree-optimization/83126
21146         * tree-parloops.c (num_phis): New function.
21147         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
21149 2018-03-21  Nathan Sidwell  <nathan@acm.org>
21151         * doc/extend.texi (Deprecated Features): Update deprecated flags,
21152         mention anon-struct/union members and trailing attributes.
21154 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
21156         PR tree-optimization/84969
21157         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
21158         builtin memset partitions if they set different rhs values.
21160 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
21162         PR rtl-optimization/84989
21163         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
21164         VEC_DUPLICATE with scalar result mode.
21166 2018-03-21  Martin Liska  <mliska@suse.cz>
21168         PR ipa/84963
21169         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
21170         not intended return statement.
21172 2018-03-21  Martin Liska  <mliska@suse.cz>
21174         PR target/84988
21175         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
21176         (chkp_find_bound_slots_1): Limit number of iterations.
21178 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
21180         PR target/84838
21181         * Minor grammar fixes for x86 options.
21183 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21185         PR debug/84875
21186         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
21187         holding REG_CFA_RESTORE notes, instead turn them into a USE.
21189 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
21191         PR target/83789
21192         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
21193         (altivec_lvx_<mode>_1op): Likewise.
21194         (altivec_stvx_<mode>_2op): Likewise.
21195         (altivec_stvx_<mode>_1op): Likewise.
21196         (altivec_lvx_<VM2:mode>): New define_expand.
21197         (altivec_stvx_<VM2:mode>): Likewise.
21198         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
21199         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21200         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
21201         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21202         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
21203         (rs6000_gen_lvx): Likewise.
21204         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
21205         (altivec_expand_stv_builtin): Likewise.
21206         (altivec_expand_builtin): Likewise.
21207         * config/rs6000/vector.md: Likewise.
21209 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21211         PR target/82518
21212         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
21213         BYTES_BIG_ENDIAN.
21215 2018-03-20  Richard Biener  <rguenther@suse.de>
21217         PR target/84986
21218         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
21219         sign-conversions as zero, fall back to standard scalar_stmt
21220         cost for the rest.
21222 2018-03-20  Martin Liska  <mliska@suse.cz>
21224         PR ipa/84825
21225         * predict.c (rebuild_frequencies): Handle case when we have
21226         PROFILE_ABSENT, but flag_guess_branch_prob is false.
21228 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21230         PR target/84990
21231         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
21232         flag_section_anchors.
21233         * varasm.c (use_blocks_for_decl_p): Remove hack for
21234         dw2_force_const_mem.
21236         PR target/84845
21237         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
21238         to ...
21239         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
21240         be created, use lowpart_subreg of operands[0] rather than operands[0]
21241         itself.
21242         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
21243         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
21244         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
21245         and n constraint instead of aarch64_shift_imm_di and Usd.
21246         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
21247         (*aarch64_<optab>_reg_minus<mode>3): ... this.
21249 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
21251         PR target/82989
21252         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
21253         to favor GPR over NEON registers.
21254         (<shift>di3_neon): Likewise.
21256 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21258         PR target/84952
21259         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
21260         (nvptx_process_pars): Emit bar.sync asap and alap.
21262 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21264         PR target/84954
21265         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
21266         seen_label if seen_label is already set.
21268 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21270         PR target/84945
21271         * config/i386/i386.c (fold_builtin_cpu): For features above 31
21272         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
21273         Use 1U instead of 1.  Formatting fixes.
21275         PR c/84953
21276         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
21277         instead of TREE_TYPE (s1) for the return value.
21279 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
21281         PR tree-optimization/84946
21282         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
21283         bitsize + bitsize in poly_uint64 rather than poly_int64.
21285         PR sanitizer/78651
21286         * dwarf2asm.c: Include fold-const.c.
21287         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
21288         of decl rather than decl itself.
21290         PR rtl-optimization/84643
21291         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
21293 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
21295         PR sanitizer/78651
21296         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
21297         calling assemble_variable.
21299 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
21301         PR target/81647
21302         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
21303         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
21305 2018-03-19  Jim Wilson  <jimw@sifive.com>
21307         PR bootstrap/84856
21308         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
21309         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
21310         (riscv_first_stack_step): Likewise.
21311         (riscv_option_override): Use STACK_BOUNDARY instead of
21312         MIN_STACK_BOUNDARY.
21313         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
21314         MIN_STACK_BOUNDARY.
21315         (BIGGEST_ALIGNMENT): Set to 128.
21316         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
21317         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
21318         STACK_BOUNDARY.
21320 2018-03-19  Richard Biener  <rguenther@suse.de>
21322         PR tree-optimization/84933
21323         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
21324         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
21326 2018-03-19  Richard Biener  <rguenther@suse.de>
21328         PR tree-optimization/84859
21329         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
21330         (cond_if_else_store_replacement): Perform sinking operation on
21331         single-store BBs regardless of MAX_STORES_TO_SINK setting.
21332         Generalize what a BB with a single eligible store is.
21334 2018-03-19  Richard Biener  <rguenther@suse.de>
21336         PR tree-optimization/84929
21337         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
21338         chrec_is_positive against non-chrec arg.
21340 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
21342         PR target/84711
21343         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
21345 2018-03-18  Martin Liska  <mliska@suse.cz>
21347         PR rtl-optimization/84635
21348         * regrename.c (build_def_use): Use matches_mode only when
21349         matches >= 0.
21351 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
21353         PR tree-optimization/84913
21354         * tree-vect-loop.c (vectorizable_reduction): Don't try to
21355         vectorize chains of COND_EXPRs.
21357 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21359         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
21361 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21363         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
21365 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21367         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21369 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21370             Kito Cheng  <kito.cheng@gmail.com>
21372         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
21373         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
21374         (nds32_adjust_reg_alloc_order): New function.
21375         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
21377 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
21379         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
21380         nds32_print_operand, nds32_print_operand_address): Use
21381         HOST_WIDE_INT_PRINT_DEC instead.
21383 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21385         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
21387 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
21389         PR target/84902
21390         * config/i386/i386.c (initial_ix86_tune_features,
21391         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
21392         unsigned long long.
21393         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
21394         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
21395         rather than 1u << ix86_tune.  Formatting fix.
21396         (ix86_option_override_internal): Change ix86_arch_mask from
21397         unsigned int to unsigned HOST_WIDE_INT, initialize to
21398         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
21399         (ix86_function_specific_restore): Likewise.
21401 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21403         PR target/84899
21404         * postreload.c (reload_combine_recognize_pattern): Perform
21405         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
21406         truncate_int_for_mode the result for the destination's mode.
21408         PR c/84909
21409         * hsa-gen.c (mem_type_for_type): Fix comment typo.
21410         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
21411         Likewise.
21412         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21413         Likewise.
21415 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
21417         PR target/84876
21418         * lra-assigns.c (lra_split_hard_reg_for): Don't use
21419         regno_allocno_class_array and sorted_pseudos.
21420         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
21421         insns where regno is used.
21423 2018-03-16  Martin Liska  <mliska@suse.cz>
21425         PR ipa/84833
21426         * multiple_target.c (create_dispatcher_calls): Redirect
21427         reference in the symbol table.
21429 2018-03-16  Martin Liska  <mliska@suse.cz>
21431         PR ipa/84722
21432         * multiple_target.c (create_dispatcher_calls): Redirect also
21433         an alias.
21435 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21437         PR c++/79937
21438         PR c++/82410
21439         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
21440         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
21441         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
21443 2018-03-16  Julia Koval  <julia.koval@intel.com>
21445         * doc/invoke.texi (Skylake Server): Add CLWB.
21446         Cannonlake): Remove CLWB.
21448 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21450         PR tree-optimization/84841
21451         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
21452         1 << 3.
21453         (FLOAT_ONE_CONST_TYPE): Define.
21454         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
21455         (sort_by_operand_rank): Put entries with higher constant_type last
21456         rather than first to match comments.
21458 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
21460         * config/nios2/nios2.md (movsi_internal): Fix thinko in
21461         split predicate.
21463 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21465         PR c++/79085
21466         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
21467         check and use address of target always.
21469 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
21471         PR target/84574
21472         * config/i386/i386.c (indirect_thunk_needed): Update comments.
21473         (indirect_thunk_bnd_needed): Likewise.
21474         (indirect_thunks_used): Likewise.
21475         (indirect_thunks_bnd_used): Likewise.
21476         (indirect_return_needed): New.
21477         (indirect_return_bnd_needed): Likewise.
21478         (output_indirect_thunk_function): Add a bool argument for
21479         function return.
21480         (output_indirect_thunk_function): Don't generate alias for
21481         function return thunk.
21482         (ix86_code_end): Call output_indirect_thunk_function to generate
21483         function return thunks.
21484         (ix86_output_function_return): Set indirect_return_bnd_needed
21485         and indirect_return_needed instead of indirect_thunk_bnd_needed
21486         and indirect_thunk_needed.
21488 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
21490         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
21491         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
21492         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
21494 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
21495             Paul Hua <paul.hua.gm@gmail.com>
21497         PR c/84852
21498         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
21500 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
21502         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
21503         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
21504         resp. SFmode cases.
21506 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
21508         PR target/84711
21509         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
21510         instead of GET_MODE_SIZE when comparing Units.
21512 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
21514         PR target/68256
21515         * varasm.c (hash_section): Return an unchangeble hash value
21516         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
21517         Return !aarch64_can_use_per_function_literal_pools_p ().
21519 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21521         PR target/84860
21522         * optabs.c (emit_conditional_move): Pass address of cmode's copy
21523         rather than address of cmode as last argument to prepare_cmp_insn.
21525 2018-03-15  Julia Koval  <julia.koval@intel.com>
21527         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
21528         F_AVX512VNNI, F_AVX512BITALG): New.
21530 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
21532         PR target/83451
21533         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
21534         insn for floating-point loads and stores.
21536 2018-03-14  Carl Love  <cel@us.ibm.com>
21538         * config/rs6000/rs6000-c.c: Add macro definitions for
21539         ALTIVEC_BUILTIN_VEC_PERMXOR.
21540         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
21541         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
21542         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
21543         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
21544         UNSPEC_VPERMXOR.
21545         * config/doc/extend.texi: Add prototypes for vec_permxor.
21547 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
21549         PR c/84852
21550         * diagnostic-show-locus.c (class layout_point): Convert m_line
21551         from int to linenum_type.
21552         (line_span::comparator): Use linenum "compare" function when
21553         comparing line numbers.
21554         (test_line_span): New function.
21555         (layout_range::contains_point): Convert param "row" from int to
21556         linenum_type.
21557         (layout_range::intersects_line_p): Likewise.
21558         (layout::will_show_line_p): Likewise.
21559         (layout::print_source_line): Likewise.
21560         (layout::should_print_annotation_line_p): Likewise.
21561         (layout::print_annotation_line): Likewise.
21562         (layout::print_leading_fixits): Likewise.
21563         (layout::annotation_line_showed_range_p): Likewise.
21564         (struct line_corrections): Likewise for field m_row.
21565         (line_corrections::line_corrections): Likewise for param "row".
21566         (layout::print_trailing_fixits): Likewise.
21567         (layout::get_state_at_point): Likewise.
21568         (layout::get_x_bound_for_row): Likewise.
21569         (layout::print_line): Likewise.
21570         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
21571         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
21572         * input.c (selftest::test_linenum_comparisons): New function.
21573         (selftest::input_c_tests): Call it.
21574         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
21575         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
21576         * selftest.h (ASSERT_GT): New macro.
21577         (ASSERT_GT_AT): New macro.
21578         (ASSERT_LT): New macro.
21579         (ASSERT_LT_AT): New macro.
21581 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
21583         PR rtl-optimization/84780
21584         * combine.c (distribute_links): Don't make a link based on pc_rtx.
21586 2018-03-14  Martin Liska  <mliska@suse.cz>
21588         * tree.c (record_node_allocation_statistics): Use
21589         get_stats_node_kind.
21590         (get_stats_node_kind): New function extracted from
21591         record_node_allocation_statistics.
21592         (free_node): Use get_stats_node_kind.
21594 2018-03-14  Richard Biener  <rguenther@suse.de>
21596         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
21597         that the value-set of ANTIC_IN doesn't grow.
21599         Revert
21600         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21601         member.
21602         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21603         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21605 2018-03-14  Julia Koval  <julia.koval@intel.com>
21607         * config.gcc (icelake-client, icelake-server): New.
21608         (icelake): Remove.
21609         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
21610         (initial_ix86_arch_features): Ditto.
21611         (PTA_SKYLAKE): Add SGX.
21612         (PTA_ICELAKE): Remove.
21613         (PTA_ICELAKE_CLIENT): New.
21614         (PTA_ICELAKE_SERVER): New.
21615         (ix86_option_override_internal): Split up icelake on icelake client and
21616         icelake server.
21617         (get_builtin_code_for_version): Ditto.
21618         (fold_builtin_cpu): Ditto.
21619         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
21620         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
21621         * config/i386/i386.h (processor_type): Ditto.
21622         * doc/invoke.texi: Ditto.
21624 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
21626         PR sanitizer/83392
21627         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21628         INTEGER_CST offset, add it together with bitpos / 8 and
21629         sign extend based on POINTER_SIZE.
21631         PR target/84844
21632         Revert
21633         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
21635         PR target/78090
21636         * config/i386/constraints.md (Yc): New register constraint.
21637         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21638         Use Yc constraint for alternative 2 of operand 0.  Remove
21639         preferred_for_speed attribute.
21641 2018-03-14  Richard Biener  <rguenther@suse.de>
21643         PR tree-optimization/84830
21644         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21645         with the old one to avoid oscillations.
21647 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
21649         PR target/83712
21650         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21651         pseudos.
21652         (assign_by_spills): Return a flag of reload assignment failure.
21653         Do not process the reload assignment failures.  Do not spill other
21654         reload pseudos if they has the same reg class.  Update n if
21655         necessary.
21656         (lra_assign): Add a return arg.  Set up from the result of
21657         assign_by_spills call.
21658         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21659         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21660         usage_insns if it is not NULL.
21661         (spill_hard_reg_in_range): New function.
21662         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21663         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21664         function prototypes.
21665         (lra_assign): Change prototype.
21666         * lra.c (lra): Add code to deal with fails by splitting hard reg
21667         live ranges.
21669 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
21671         * config/riscv/riscv.opt (mrelax): New option.
21672         * config/riscv/riscv.c (riscv_file_start): Emit ".option
21673         "norelax" when riscv_mrelax is disabled.
21674         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21676 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21678         PR target/84743
21679         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21680         reassociation for int modes.
21682 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21684         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21685         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21686         for big-endian.
21687         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21688         * config/aarch64/aarch64-sve.md
21689         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21690         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21691         (*extend<mode><Vwide>2): Rename to...
21692         (aarch64_sve_extend<mode><Vwide>2): ...this.
21693         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21694         renaming the old pattern to...
21695         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
21696         unsigned packs.
21697         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21698         define_expand, renaming the old pattern to...
21699         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21700         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21701         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21702         account when deciding which SVE instruction the optab should use.
21703         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21705 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21707         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21708         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21709         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21710         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21711         (tlsdesc_small_<mode>): Turn a define_expand and use
21712         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
21713         (tlsdesc_small_advsimd_<mode>): ...this.
21714         (tlsdesc_small_sve_<mode>): New pattern.
21716 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21718         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21719         (UNSPEC_UMUL_HIGHPART): New constants.
21720         (MUL_HIGHPART): New int iteraor.
21721         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21722         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21723         define_expand.
21724         (*<su>mul<mode>3_highpart): New define_insn.
21726 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21728         PR lto/84805
21729         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21730         incomplete types.
21732 2018-03-13  Martin Liska  <mliska@suse.cz>
21734         PR ipa/84658.
21735         * (sem_item_optimizer::sem_item_optimizer): Initialize new
21736         vector.
21737         (sem_item_optimizer::~sem_item_optimizer): Release it.
21738         (sem_item_optimizer::merge_classes): Register variable aliases.
21739         (sem_item_optimizer::fixup_pt_set): New function.
21740         (sem_item_optimizer::fixup_points_to_sets): Likewise.
21741         * ipa-icf.h: Declare new variables and functions.
21743 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
21745         PR middle-end/84834
21746         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21747         integer_pow2p@2 and test integer_pow2p in condition.
21748         (A < 0 ? C : 0): Similarly for @1.
21750         PR middle-end/84831
21751         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21752         characters starting at p contain '\0' character, don't look beyond
21753         that.
21755         PR target/84827
21756         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21757         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21759         PR target/84828
21760         * reg-stack.c (change_stack): Change update_end var from int to
21761         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21762         also call set_block_for_insn on the newly added insns and rescan.
21764         PR target/84786
21765         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21766         on the last operand.
21768         PR c++/84704
21769         * tree.c (stabilize_reference_1): Return save_expr (e) for
21770         STATEMENT_LIST even if it doesn't have side-effects.
21772 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
21774         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21776 2018-03-12  Renlin Li  <renlin.li@arm.com>
21778         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21779         aarch64_output_scalar_simd_mov_immediate.
21781 2018-03-12  Martin Sebor  <msebor@redhat.com>
21783         PR tree-optimization/83456
21784         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21785         for perfectly overlapping calls to memcpy.
21786         (gimple_fold_builtin_memory_chk): Same.
21787         (gimple_fold_builtin_strcpy): Handle no-warning.
21788         (gimple_fold_builtin_stxcpy_chk): Same.
21789         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21791 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21793         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21794         parameter.  Use it for SFmode.
21795         (rs6000_function_arg_advance_1): Adjust.
21796         (rs6000_function_arg): Adjust.
21797         (rs6000_gimplify_va_arg): Pass false for that new parameter.
21799 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21801         PR rtl-optimization/84169
21802         PR rtl-optimization/84780
21803         * combine.c (can_combine_p): Check for a 2-insn combination whether
21804         the destination register is used between the two insns, too.
21806 2018-03-12  Richard Biener  <rguenther@suse.de>
21808         PR tree-optimization/84803
21809         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21810         for refs DR analysis didn't process.
21812 2018-03-12  Richard Biener  <rguenther@suse.de>
21814         PR tree-optimization/84777
21815         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21816         force-vectorize loops ignore whether we are optimizing for size.
21818 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
21820         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21821         (TARGET_MD_ASM_ADJUST): Define.
21823 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
21824             Kito Cheng  <kito.cheng@gmail.com>
21825             Chung-Ju Wu  <jasonwucj@gmail.com>
21827         * config/nds32/nds32.c (nds32_compute_stack_frame,
21828         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21829         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21830         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21831         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21832         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21833         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21834         * config/nds32/nds32.md (prologue, epilogue): Use macro
21835         NDS32_V3PUSH_AVAILABLE_P to do checking.
21837 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
21839         PR debug/58150
21840         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21841         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21842         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
21843         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
21844         addition of most attributes on !orig_type_die or the attribute not
21845         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21847 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21848             Chung-Ju Wu  <jasonwucj@gmail.com>
21850         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21851         __NDS32_VH__ macro.
21852         * config/nds32/nds32.opt (mvh): New option.
21854 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21855             Chung-Ju Wu  <jasonwucj@gmail.com>
21857         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21858         function.
21859         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21860         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21861         definition.
21863 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21864             Chung-Ju Wu  <jasonwucj@gmail.com>
21866         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21867         function.
21868         * config/nds32/nds32-multiple.md (strlensi): New pattern.
21869         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21871 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
21872             Kito Cheng  <kito.cheng@gmail.com>
21873             Chung-Ju Wu  <jasonwucj@gmail.com>
21875         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21876         UNSPEC_FFMISM and UNSPEC_FLMISM.
21877         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21878         for ffb, ffmism and flmism.
21879         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21880         (unspec_ffmism): Ditto.
21881         (unspec_flmism): Ditto.
21882         (nds32_expand_builtin_impl): Check if string extension is available.
21883         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21884         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21886 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
21888         Reverting patch:
21889         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21891         PR target/83712
21892         * lra-assigns.c (assign_by_spills): Return a flag of reload
21893         assignment failure.  Do not process the reload assignment
21894         failures.  Do not spill other reload pseudos if they has the same
21895         reg class.
21896         (lra_assign): Add a return arg.  Set up from the result of
21897         assign_by_spills call.
21898         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21899         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21900         usage_insns if it is not NULL.
21901         (spill_hard_reg_in_range): New function.
21902         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21903         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21904         function prototypes.
21905         (lra_assign): Change prototype.
21906         * lra.c (lra): Add code to deal with fails by splitting hard reg
21907         live ranges.
21909 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
21911         PR target/84807
21912         * config/i386/i386.opt: Replace Enforcment with Enforcement.
21914 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
21916         PR debug/84620
21917         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21918         (dw_val_node): Add val_symbolic_view.
21919         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21920         (symview_upper_bound): New.
21921         (new_line_info_table): Initialize symviews_since_reset.
21922         (dwarf2out_source_line): Count symviews_since_reset and set
21923         symview_upper_bound.
21924         (dw_val_equal_p): Handle symview.
21925         (add_AT_symview): New.
21926         (print_dw_val): Handle symview.
21927         (attr_checksum, attr_checksum_ordered): Likewise.
21928         (same_dw_val_p, size_of_die): Likewise.
21929         (value_format, output_die): Likewise.
21930         (add_high_low_attributes): Use add_AT_symview for entry_view.
21931         (dwarf2out_finish): Reset symview_upper_bound, clear
21932         zero_view_p.
21934 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
21936         PR target/83969
21937         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21938         Add strict argument and use it.
21939         (rs6000_split_multireg_move): Update for new strict argument.
21940         (mem_operand_gpr): Disallow all non-offsettable addresses.
21941         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21943 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
21945         PR target/84772
21946         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21947         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21948         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21950         PR c++/84767
21951         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21952         decl, use remap_type if we want to use the type.
21954 2018-03-09  Martin Sebor  <msebor@redhat.com>
21956         PR tree-optimization/84526
21957         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21958         Remove dead code.
21959         (builtin_access::generic_overlap): Be prepared to handle non-array
21960         base objects.
21962 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
21964         PR rtl-optimization/84682
21965         * lra-constraints.c (process_address_1): Check is_address flag
21966         for address constraints.
21967         (process_alt_operands): Likewise.
21968         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21969         preprocess_constraints.
21970         * recog.h (preprocess_constraints): Add oploc parameter.
21971         Adjust callers.
21972         * recog.c (preprocess_constraints): Test address_operand for
21973         CT_ADDRESS constraints.
21975 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21977         PR target/83712
21978         * lra-assigns.c (assign_by_spills): Return a flag of reload
21979         assignment failure.  Do not process the reload assignment
21980         failures.  Do not spill other reload pseudos if they has the same
21981         reg class.
21982         (lra_assign): Add a return arg.  Set up from the result of
21983         assign_by_spills call.
21984         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21985         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21986         usage_insns if it is not NULL.
21987         (spill_hard_reg_in_range): New function.
21988         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21989         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21990         function prototypes.
21991         (lra_assign): Change prototype.
21992         * lra.c (lra): Add code to deal with fails by splitting hard reg
21993         live ranges.
21995 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21997         PR target/83193
21998         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21999         Accept complain bool parameter.  Only emit errors if it is true.
22000         (arm_parse_cpu_option_name): Likewise.
22001         (arm_target_thumb_only): Adjust callers of the above.
22002         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
22003         prototype to take a default true bool parameter.
22004         (arm_parse_arch_option_name): Likewise.
22006 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
22007             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
22009         PR jit/64089
22010         PR jit/84288
22011         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
22012         * configure: Regenerate.
22013         * configure.ac ("linker --version-script option"): New.
22014         ("linker soname option"): New.
22016 2018-03-09  Richard Biener  <rguenther@suse.de>
22018         PR tree-optimization/84775
22019         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
22020         immediate uses of predicate stmts and mark them modified.
22022         Revert
22023         PR tree-optimization/84178
22024         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22025         to caller.
22026         (version_loop_for_if_conversion): Delay update_ssa call.
22027         (tree_if_conversion): Delay update_ssa until after predicate
22028         insertion.
22030 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
22032         PR target/84763
22033         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
22034         when the function accesses prior frames.
22036 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
22038         PR debug/84456
22039         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
22040         gen_llsym, otherwise call maybe_gen_llsym.
22042         PR inline-asm/84742
22043         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
22044         has ',' character inside of it.
22046 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22048         PR target/84748
22049         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
22050         as clobbering CC_REGNUM.
22052 2018-03-08  Richard Biener  <rguenther@suse.de>
22054         PR middle-end/84552
22055         * tree-scalar-evolution.c: Include tree-into-ssa.h.
22056         (follow_copies_to_constant): Do not follow SSA names registered
22057         for update.
22059 2018-03-08  Richard Biener  <rguenther@suse.de>
22061         PR tree-optimization/84178
22062         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22063         to caller.
22064         (version_loop_for_if_conversion): Delay update_ssa call.
22065         (tree_if_conversion): Delay update_ssa until after predicate
22066         insertion.
22068 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
22070         PR tree-optimization/84178
22071         * tree-if-conv.c (release_bb_predicate): Remove the
22072         the assertion that the stmts have NULL use_ops.
22073         Discard the statements, asserting that they haven't
22074         yet been added to a BB.
22076 2018-03-08  Richard Biener  <rguenther@suse.de>
22078         PR tree-optimization/84746
22079         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
22080         (phi_translate): Pass in destination ANTIC_OUT set.
22081         (phi_translate_1): Likewise.  For a simplified result lookup
22082         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
22083         (phi_translate_set): Adjust.
22084         (do_pre_regular_insertion): Likewise.
22085         (do_pre_partial_partial_insertion): Likewise.
22087 2018-03-08  Martin Liska  <mliska@suse.cz>
22089         PR gcov-profile/84735
22090         * doc/gcov.texi: Document usage of profile files.
22091         * gcov-io.h: Document changes in the format.
22093 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
22095         PR debug/84404
22096         PR debug/84408
22097         * dwarf2out.c (struct dw_line_info_table): Update comments for
22098         view == -1.
22099         (FORCE_RESET_NEXT_VIEW): New.
22100         (FORCE_RESETTING_VIEW_P): New.
22101         (RESETTING_VIEW_P): Check for -1 too.
22102         (ZERO_VIEW_P): Likewise.
22103         (new_line_info_table): Force-reset next view.
22104         (dwarf2out_begin_function): Likewise.
22105         (dwarf2out_source_line): Simplify zero_view_p initialization.
22106         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
22107         view directly.  Omit view when omitting .loc at line 0.
22109 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
22111         PR tree-optimization/84740
22112         * tree-switch-conversion.c (process_switch): Call build_constructors
22113         only if info.phi_count is non-zero.
22115         PR tree-optimization/84739
22116         * tree-tailcall.c (find_tail_calls): Check call arguments against
22117         DECL_ARGUMENTS (current_function_decl) rather than
22118         DECL_ARGUMENTS (func) when checking for tail recursion.
22120 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
22122         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
22123         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
22124         Volker Reichelt's entry and add entries for people that perform
22125         GCC fuzzy testing and report numerous bugs.
22127 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
22129         PR target/82411
22130         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
22131         readonly data in sdata, if that is disabled.
22132         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
22133         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
22134         -mreadonly-in-sdata option.
22136 2018-03-07  Martin Sebor  <msebor@redhat.com>
22138         PR tree-optimization/84468
22139         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
22140         basic block when looking for nul assignment.
22142 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
22144         PR target/84277
22145         * except.h (output_function_exception_table): Adjust prototype.
22146         * except.c (output_function_exception_table): Remove FNNAME parameter
22147         and add SECTION parameter.  Ouput one part of the table at a time.
22148         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
22149         the first part of the exception table and emit unwind directives.
22150         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
22151         (i386_pe_seh_cold_init): Likewise.
22152         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
22153         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
22154         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
22155         (ix86_output_call_insn): Emit a nop in one more case for SEH.
22156         * config/i386/winnt.c: Include except.h.
22157         (struct seh_frame_state): Add reg_offset, after_prologue and
22158         in_cold_section fields.
22159         (i386_pe_seh_end_prologue): Set seh->after_prologue.
22160         (i386_pe_seh_cold_init): New function.
22161         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
22162         to seh->in_cold_section.
22163         (seh_emit_push): Record the offset of the push.
22164         (seh_emit_save): Record the offet of the save.
22165         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
22166         Test seh->after_prologue to disregard the epilogue.
22167         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
22168         (i386_pe_end_cold_function): New function.
22170 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
22172         PR fortran/84565
22173         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
22174         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
22176         PR c++/84704
22177         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
22178         on tmp_var.
22179         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
22180         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
22182         PR middle-end/84723
22183         * multiple_target.c: Include tree-inline.h and intl.h.
22184         (expand_target_clones): Diagnose and fail if node->definition and
22185         !tree_versionable_function_p (node->decl).
22187 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
22189         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
22190         sprint_ul.
22191         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
22192         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
22193         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
22195 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
22197         PR target/84710
22198         * combine.c (try_combine): Use reg_or_subregno instead of handling
22199         just paradoxical SUBREGs and REGs.
22201 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
22203         * config/arc/arc.c (arc_finalize_pic): Remove function.
22204         (arc_must_save_register): We use single base PIC register, remove
22205         checks to save/restore the PIC register.
22206         (arc_expand_prologue): Likewise.
22207         * config/arc/arc-protos.h (arc_set_default_type_attributes):
22208         Remove.
22209         (arc_verify_short): Likewise.
22210         (arc_attr_type): Likewise.
22211         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
22212         (walk_stores): Likewise.
22213         (arc_address_cost): Make it static.
22214         (arc_verify_short): Likewise.
22215         (branch_dest): Likewise.
22216         (arc_attr_type): Likewise.
22217         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
22218         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
22219         (arc_final_prescan_insn): Remove inserting the nops due to
22220         hardware hazards.  It is done in reorg step.
22221         (insn_length_variant_t): Remove.
22222         (insn_length_parameters_t): Likewise.
22223         (arc_insn_length_parameters): Likewise.
22224         (arc_get_insn_variants): Likewise.
22225         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
22227 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
22229         PR inline-asm/84683
22230         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
22231         assertion failure.
22233         PR tree-optimization/84687
22234         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
22235         on new_node->decl.
22236         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
22238 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22240         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
22241         Rename to ppc_speculation_barrier.
22242         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
22243         __builtin_ppc_speculation_barrier.
22245 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22247         PR target/84700
22248         * combine.c (combine_simplify_rtx): Don't try to simplify if
22249         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
22250         are equal to x.
22252 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
22254         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
22255         to 32 bytes when compiling for POWER9.
22257 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22259         PR target/84564
22260         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
22261         regparm >= 3 with no arg reg available also for calls with
22262         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
22264         PR target/84524
22265         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
22266         orig,vex.
22267         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
22269 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
22271         PR target/84264
22272         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
22274 2018-03-05  Richard Biener  <rguenther@suse.de>
22276         PR tree-optimization/84486
22277         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
22278         When inserting a __builtin_assume_aligned call set the LHS
22279         SSA name alignment info accordingly.
22281 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
22283         PR tree-optimization/84114
22284         * config/aarch64/aarch64.c (aarch64_reassociation_width)
22285         Avoid reassociation of FLOAT_MODE addition.
22287 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
22289         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
22290         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
22291         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
22292         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
22293         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
22294         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
22295         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
22296         and -mwbnoinvd.
22297         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
22298         __builtin_ia32_wbinvd): New builtins.
22299         (SPECIAL_ARGS2): New.
22300         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
22301         (SPECIAL_ARGS2): New.
22302         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
22303         (ix86_valid_target_attribute_inner_p): Ditto.
22304         (ix86_init_mmx_sse_builtins): Add special_args2.
22305         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
22306         TARGET_WBNOINVD_P): New.
22307         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
22308         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
22309         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
22310         * config/i386/immintrin.h (_wbinvd): New intrinsic.
22311         * config/i386/pconfigintrin.h: New file.
22312         * config/i386/wbnoinvdintrin.h: Ditto.
22313         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
22314         wbnoinvdintrin.h.
22315         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
22317 2018-03-05  Richard Biener  <rguenther@suse.de>
22319         PR tree-optimization/84670
22320         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22321         member.
22322         (BB_VISITED_WITH_VISITED_SUCCS): New define.
22323         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22324         (compute_antic_aux): Only assert the number of values in ANTIC_IN
22325         doesn't grow if all successors (recursively) were visited at least
22326         once.
22328 2018-03-05  Richard Biener  <rguenther@suse.de>
22330         PR tree-optimization/84650
22331         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
22332         if executed in the loop pipeline.
22334 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
22336         * doc/configfiles.texi (Configuration Files): Move info about
22337         conditionalizing $target-protos.h to...
22338         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
22339         differs from $target-protos.h.
22341 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
22342             Chung-Ju Wu  <jasonwucj@gmail.com>
22344         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
22345         * config/nds32/nds32-multiple.md (setmemsi): Define.
22346         * config/nds32/nds32-memory-manipulation.c
22347         (nds32_gen_dup_4_byte_to_word_value): New.
22348         (emit_setmem_word_loop): New.
22349         (emit_setmem_byte_loop): New.
22350         (nds32_expand_setmem_loop): New.
22351         (nds32_expand_setmem_loop_v3m): New.
22352         (nds32_expand_setmem_unroll): New.
22353         (nds32_expand_setmem): New.
22355 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22356             Chung-Ju Wu  <jasonwucj@gmail.com>
22358         * config/nds32/nds32-memory-manipulation.c
22359         (nds32_emit_load_store): New.
22360         (nds32_emit_post_inc_load_store): New.
22361         (nds32_emit_mem_move): New.
22362         (nds32_emit_mem_move_block): New.
22363         (nds32_expand_movmemsi_loop_unknown_size): New.
22364         (nds32_expand_movmemsi_loop_known_size): New.
22365         (nds32_expand_movmemsi_loop): New.
22366         (nds32_expand_movmemsi_unroll): New.
22367         (nds32_expand_movmemqi): Rename ...
22368         (nds32_expand_movmemsi): ... to this.
22369         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
22370         (movmemsi): ... to this.
22371         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
22372         (nds32_expand_movmemsi): ... to this.
22374 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22375             Monk Chiang  <sh.chiang04@gmail.com>
22376             Chung-Ju Wu  <jasonwucj@gmail.com>
22378         * config/nds32/nds32-protos.h
22379         (nds32_expand_load_multiple): New arguments.
22380         (nds32_expand_store_multiple): Ditto.
22381         (nds32_valid_multiple_load_store): Rename ...
22382         (nds32_valid_multiple_load_store_p): ... to this.
22383         * config/nds32/nds32-memory-manipulation.c
22384         (nds32_expand_load_multiple): Refine implementation.
22385         (nds32_expand_store_multiple): Ditto.
22386         * config/nds32/nds32-multiple.md
22387         (load_multiple): Update nds32_expand_load_multiple interface.
22388         (store_multiple): Update nds32_expand_store_multiple interface.
22389         * config/nds32/nds32-predicates.c
22390         (nds32_valid_multiple_load_store): Rename ...
22391         (nds32_valid_multiple_load_store_p): ... to this and refine
22392         implementation.
22393         * config/nds32/predicates.md
22394         (nds32_load_multiple_and_update_address_operation): New predicate.
22395         (nds32_store_multiple_and_update_address_operation): New predicate.
22397 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22398             Chung-Ju Wu  <jasonwucj@gmail.com>
22400         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
22401         (combo): New attribute.
22402         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
22404 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
22406         * config/nds32/nds32.opt: Change -mcmodel= default value.
22408 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
22409             Monk Chiang  <sh.chiang04@gmail.com>
22410             Chung-Ju Wu  <jasonwucj@gmail.com>
22412         * config/nds32/constants.md (unspec_element): New enum.
22413         * config/nds32/constraints.md (Umw): New constraint.
22414         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
22415         * config/nds32/nds32-intrinsic.md: Likewise.
22416         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
22417         (nds32_valid_smw_lwm_base_p): New.
22418         (nds32_output_smw_single_word): New.
22419         (nds32_output_lmw_single_word): New.
22420         (nds32_expand_unaligned_load): New.
22421         (nds32_expand_unaligned_store): New.
22422         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
22423         (nds32_output_smw_single_word): Declare.
22424         (nds32_output_lmw_single_word): Declare.
22425         (nds32_expand_unaligned_load): Declare.
22426         (nds32_expand_unaligned_store): Declare.
22427         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
22428         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
22429         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
22430         NDS32_BUILTIN_UASTORE_DW.
22431         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
22432         predicate.
22434 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
22435             Kito Cheng  <kito.cheng@gmail.com>
22436             Chung-Ju Wu  <jasonwucj@gmail.com>
22438         * config/nds32/nds32-intrinsic.c
22439         (nds32_expand_builtin_null_ftype_reg): Delete.
22440         (nds32_expand_builtin_reg_ftype_imm): Ditto.
22441         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
22442         (nds32_read_argument): New.
22443         (nds32_legitimize_target): Ditto.
22444         (nds32_legitimize_argument): Ditto.
22445         (nds32_check_constant_argument): Ditto.
22446         (nds32_expand_unop_builtin): Ditto.
22447         (nds32_expand_unopimm_builtin): Ditto.
22448         (nds32_expand_binop_builtin): Ditto.
22449         (nds32_builtin_decl_impl): Ditto.
22450         (builtin_description): Ditto.
22451         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
22452         (nds32_init_builtins_impl): Ditto.
22453         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
22454         (nds32_builtin_decl): New.
22455         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
22456         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
22458 2018-03-02  Jeff Law  <law@redhat.com>
22460         * reorg.c (stop_search_p): Handle DEBUG_INSN.
22461         (redundant_insn, fill_simple_delay_slots): Likewise.
22462         (fill_slots_from_thread): Likewise.
22463         * resource.c (mark_referenced_resources): Likewise.
22464         (mark_set_resources, find_dead_or_set_registers): Likewise.
22466 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22468         * substring-locations.h (format_warning_va): Formatting fix for
22469         ATTRIBUTE_GCC_DIAG.
22470         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
22471         argument.
22472         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
22473         * substring-locations.c: Include intl.h.
22474         (format_warning_va): Turned into small wrapper around
22475         format_warning_n_va, renamed to ...
22476         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
22477         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
22478         use ngettext.
22479         (format_warning_at_substring_n): New function.
22480         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
22481         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
22482         format_warning_at_substring with just a shorter name instead of
22483         const function pointer.
22484         (fmtwarn_n): New function.
22485         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
22486         appropriate, get rid of all the fmtstr temporaries, move conditionals
22487         with G_() wrapped string literals directly into fmtwarn arguments,
22488         cast dir.len to (int), formatting fixes.
22490 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
22492         * doc/invoke.texi: Remove "Cilk Plus" references.
22494 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22495             Richard Biener  <rguenther@suse.de>
22497         PR ipa/84628
22498         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
22499         for error or warning attributes if CALL_FROM_THUNK_P is set.
22500         Formatting fixes.
22502 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22504         PR target/56540
22505         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
22506         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
22508         PR target/56540
22509         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
22510         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
22512         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
22513         instead of -1U in last predictors element's probability member.
22515 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
22517         PR ipa/83983
22518         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22519         arguments if they are comparable.
22521 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22523         PR tree-optimization/84634
22524         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
22525         masks and masked_loop_p with a single loop_masks, making sure it's
22526         null for bb vectorization.
22528 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22530         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
22531         (vect_analyze_data_ref_access): Use loop->safe_len rather than
22532         loop->force_vectorize to check whether there is no alias.
22534 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22536         PR target/84614
22537         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
22538         prototypes.
22539         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
22540         comments.
22541         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
22542         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
22543         instead of a loop around prev_real_insn.
22544         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
22545         prev_real_insn.
22547         PR inline-asm/84625
22548         * config/i386/i386.c (ix86_print_operand): Use conditional
22549         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
22550         zero vector.
22552 2018-03-02  Richard Biener  <rguenther@suse.de>
22554         PR tree-optimization/84427
22555         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
22556         (bitmap_set_subtract_values): Rewrite to handle multiple
22557         exprs per value.
22558         (clean): Likewise.
22559         (prune_clobbered_mems): Likewise.
22560         (phi_translate): Take edge instead of pred/phiblock.
22561         (phi_translate_1): Likewise.
22562         (phi_translate_set): Likewise.  Insert all translated
22563         exprs for a value into the set, keeping possibly multiple
22564         expressions per value.
22565         (compute_antic_aux): Adjust for phi_translate changes.
22566         When intersecting union the expressions and prune those
22567         not in the final value set, keeping possibly multiple
22568         expressions per value.  Do not use value-insertion
22569         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
22570         all expressions.  Add verification that the value-sets
22571         only shrink during iteration.
22572         (compute_partial_antic_aux): Adjust for the phi_translate changes.
22573         (do_pre_regular_insertion): Likewise.
22574         (do_pre_partial_partial_insertion): Likewise.
22576 2018-03-02  Richard Biener  <rguenther@suse.de>
22578         PR target/82005
22579         * config/darwin.c (saved_debug_info_level): New static global.
22580         (darwin_asm_lto_start): Disable debug info generation for LTO out.
22581         (darwin_asm_lto_end): Restore debug info generation settings.
22583 2018-03-01  Martin Liska  <mliska@suse.cz>
22585         PR sanitizer/82484
22586         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
22587         volatile arguments.
22589 2018-03-01  Richard Biener  <rguenther@suse.de>
22591         PR debug/84645
22592         * dwarf2out.c (gen_variable_die): Properly handle late VLA
22593         type annotation with LTO when debug was disabled at compile-time.
22595 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
22597         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
22598         XINT with INTVAL.
22599         (mips_final_postscan_insn): Likewise.
22601 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
22603         PR rtl-optimization/84528
22604         * alias.c (init_alias_target): Add commentary.
22605         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
22606         a unique base value if the frame pointer is not eliminated
22607         to the stack pointer.
22609 2018-03-01  Tom de Vries  <tom@codesourcery.com>
22611         PR rtl-optimization/83327
22612         * lra-int.h (hard_regs_spilled_into): Declare.
22613         * lra.c (hard_regs_spilled_into): Define.
22614         (init_reg_info): Init hard_regs_spilled_into.
22615         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22616         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
22617         (process_bb_lives): Handle hard_regs_spilled_into.
22618         (lra_create_live_ranges_1): Before doing liveness propagation, clear
22619         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
22621 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
22623         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22624         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22625         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22626         * config/rs6000/aix72.h: New file.
22628 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
22630         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22631         instead of warning_at with conditional singular and plural messages
22632         where possible.
22634         PR target/52991
22635         * stor-layout.c (update_alignment_for_field): For
22636         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22637         && !DECL_PACKED (field), do the alignment update, just use
22638         only desired_align instead of MAX (type_align, desired_align)
22639         as the alignment.
22640         (place_field): Don't do known_align < desired_align handling
22641         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22642         is non-NULL, instead do it after rli->prev_field handling and
22643         only if not within a bitfield word.  For DECL_PACKED (field)
22644         use type_align of BITS_PER_UNIT.
22646 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
22648         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22649         superfluous parentheses and trailing spaces.
22651 2018-02-28  Richard Biener  <rguenther@suse.de>
22653         PR tree-optimization/84584
22654         * graphite-scop-detection.c (scop_detection::add_scop): Discard
22655         SCoPs with fake exit edge.
22657 2018-02-28  Martin Liska  <mliska@suse.cz>
22659         PR testsuite/84597
22660         * timevar.c (timer::print): Fix format to properly print 100%
22661         values.
22663 2018-02-28  Richard Biener  <rguenther@suse.de>
22665         PR middle-end/84607
22666         * genmatch.c (capture_info::walk_match): Do not mark
22667         captured expressions without operands as expr_p given
22668         they act more like predicates and should be subject to
22669         "lost tail" side-effect preserving.
22671 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
22673         PR rtl-optimization/81611
22674         * auto-inc-dec.c (attempt_change): Move dead note from
22675         mem_insn if it's the next use of regno
22676         (find_address): Take address use of reg holding
22677         non-incremented value.  Add parm to limit search to the named
22678         reg only.
22679         (merge_in_block): Attempt to use a mem insn that is the next
22680         use of the original regno.
22682 2018-02-27  Martin Sebor  <msebor@redhat.com>
22684         PR c++/83871
22685         * doc/invoke.texi (-Wmissing-attributes): New option.
22686         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22688 2018-02-27  Martin Sebor  <msebor@redhat.com>
22690         PR translation/84207
22691         * diagnostic-core.h (warning_n, error_n, inform_n): Change
22692         n argument to unsigned HOST_WIDE_INT.
22693         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22694         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
22695         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22696         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22698 2018-02-27  Richard Biener  <rguenther@suse.de>
22700         PR tree-optimization/84512
22701         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22702         Do not use the estimate returned from record_stmt_cost for
22703         the scalar iteration cost but sum properly using add_stmt_cost.
22705 2018-02-27  Richard Biener  <rguenther@suse.de>
22707         PR tree-optimization/84466
22708         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22709         Adjust last change to less strictly validate use operands.
22711 2018-02-27  Martin Liska  <mliska@suse.cz>
22713         PR gcov-profile/84548
22714         * gcov.c (process_file): Allow partial overlap and consider it
22715         also as group functions.
22716         (output_lines): Properly calculate range of lines for a group.
22718 2018-02-27  Martin Liska  <mliska@suse.cz>
22720         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22721         'ggc' suffixes.  Change first column width.
22722         (timer::print): Fix formatting of the column.
22724 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
22726         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22727         preserve inline entry blocks for the sake of debug inline
22728         entry point markers alone.
22729         (remove_unused_locals): Suggest in comments a better place to
22730         force the preservation of inline entry blocks that are
22731         otherwise unused, but do not preserve them.
22733 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22735         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22737 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22739         PR target/84039
22740         * config/i386/constraints.md (Bs): Replace
22741         ix86_indirect_branch_register with
22742         TARGET_INDIRECT_BRANCH_REGISTER.
22743         (Bw): Likewise.
22744         * config/i386/i386.md (indirect_jump): Likewise.
22745         (tablejump): Likewise.
22746         (*sibcall_memory): Likewise.
22747         (*sibcall_value_memory): Likewise.
22748         Peepholes of indirect call and jump via memory: Likewise.
22749         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22750         (*sibcall_value_GOT_32): Likewise.
22751         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22752         (GOT_memory_operand): Likewise.
22753         (call_insn_operand): Likewise.
22754         (sibcall_insn_operand): Likewise.
22755         (GOT32_symbol_operand): Likewise.
22756         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22758 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22760         PR rtl-optimization/83496
22761         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22762         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
22763         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22764         redundant insn, if any.
22765         (relax_delay_slots): Likewise.
22766         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22768 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
22770         PR tree-optimization/83965
22771         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22772         that grouped statements are part of a reduction chain.  Return
22773         true if the statement is not marked as a reduction itself but
22774         is part of a group.
22775         (vect_recog_dot_prod_pattern): Don't check whether the statement
22776         is part of a group here.
22777         (vect_recog_sad_pattern): Likewise.
22778         (vect_recog_widen_sum_pattern): Likewise.
22780 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22782         PR debug/84545
22783         * final.c (rest_of_clean_state): Also look for calls inside sequences.
22785 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22787         PR target/84530
22788         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22789         the bool argument.
22790         (ix86_output_indirect_function_return): New prototype.
22791         (ix86_split_simple_return_pop_internal): Likewise.
22792         * config/i386/i386.c (indirect_return_via_cx): New.
22793         (indirect_return_via_cx_bnd): Likewise.
22794         (indirect_thunk_name): Handle return va CX_REG.
22795         (output_indirect_thunk_function): Create alias for
22796         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22797         (ix86_output_indirect_jmp): Remove the bool argument.
22798         (ix86_output_indirect_function_return): New function.
22799         (ix86_split_simple_return_pop_internal): Likewise.
22800         * config/i386/i386.md (*indirect_jump): Don't pass false
22801         to ix86_output_indirect_jmp.
22802         (*tablejump_1): Likewise.
22803         (simple_return_pop_internal): Change it to define_insn_and_split.
22804         Call ix86_split_simple_return_pop_internal to split it for
22805         -mfunction-return=.
22806         (simple_return_indirect_internal): Call
22807         ix86_output_indirect_function_return instead of
22808         ix86_output_indirect_jmp.
22810 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
22812         PR bootstrap/84405
22813         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22814         memset and value initialization afterwards.
22816 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
22818         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22820 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22822         PR target/84521
22823         * common/config/aarch64/aarch64-common.c
22824         (aarch_option_optimization_table[]): Switch
22825         off fomit-frame-pointer
22827 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22828             Chung-Ju Wu  <jasonwucj@gmail.com>
22830         * config/nds32/nds32-multiple.md (load_multiple): Disallow
22831         volatile memory.
22832         (store_multiple): Ditto.
22834 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22836         * config.gcc: Add --with-cpu support for nds32 target.
22837         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22838         * config/nds32/nds32.opt: Add -mcpu= option.
22840 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
22842         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22843         isel=yes): Warn for these deprecated options.
22845 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
22847         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22848         ISA_2_5_MASKS_EMBEDDED.
22850 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
22852         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22853         p->max as pointers rather than using iterative_hash_expr.
22855 2018-02-23  Carl Love  <cel@us.ibm.com>
22857         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22858         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22859         BU_P8V_OVERLOAD_2.
22860         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22861         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22862         P8V_BUILTIN_VEC_VUNSIGNED2.
22864 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
22866         PR target/81572
22867         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22868         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22869         LRA_UNKNOWN_ALT.
22870         * lra-constraints.c (curr_insn_transform): Set up
22871         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
22872         LRA_UNKNOWN_ALT.
22873         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22874         * lra-eliminations.c (spill_pseudos): Ditto.
22875         (process_insn_for_elimination): Ditto.
22876         * lra-lives.c (reg_early_clobber_p): Use the new macros.
22877         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22878         LRA_NON_CLOBBERED_ALT.
22880 2018-02-22  Martin Sebor  <msebor@redhat.com>
22882         PR tree-optimization/84480
22883         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22884         to maybe_diag_stxncpy_trunc.  Call it.
22885         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22886         from gimple_fold_builtin_strcpy.  Print inlining stack.
22887         (handle_builtin_stxncpy): Print inlining stack.
22888         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22890 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
22892         PR target/84176
22893         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22894         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22895         and -fcheck-pointer-bounds are used together.
22896         (indirect_thunk_prefix): New enum.
22897         (indirect_thunk_need_prefix): New function.
22898         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
22899         "_nt" instead of "_bnd" for NOTRACK prefix.
22900         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22901         (output_indirect_thunk_function): Likewise.
22902         (): Likewise.
22903         (ix86_code_end): Update output_indirect_thunk_function calls.
22904         (ix86_output_indirect_branch_via_reg): Replace
22905         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22906         (ix86_output_indirect_branch_via_push): Likewise.
22907         (ix86_output_function_return): Likewise.
22908         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22909         incompatible with -fcf-protection=branch and
22910         -fcheck-pointer-bounds.
22912 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22914         PR target/83335
22915         * config/aarch64/aarch64.c (aarch64_print_address_internal):
22916         Change gcc_assert call to output_operand_lossage.
22918 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22920         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22922 2018-02-22  DJ Delorie  <dj@redhat.com>
22923             Sebastian Perta  <sebastian.perta@renesas.com>
22924             Oleg Endo  <olegendo@gcc.gnu.org>
22926         * config/rx/rx.c (rx_rtx_costs): New function.
22927         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22929 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22931         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22933 2018-02-22  Martin Liska  <mliska@suse.cz>
22935         PR driver/83193
22936         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22937         Add "native" as a possible value.
22939 2018-02-22  Martin Liska  <mliska@suse.cz>
22941         PR driver/83193
22942         * config/i386/i386.c (ix86_option_override_internal):
22943         Add "native" as a possible value for -march and -mtune.
22945 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
22947         PR target/84502
22948         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22949         to all type variants.
22951         PR tree-optimization/84503
22952         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22953         width as info->bitpos + info->bitsize - start.
22954         (merged_store_group::merge_overlapping): Simplify width computation.
22955         (check_no_overlap): New function.
22956         (imm_store_chain_info::try_coalesce_bswap): Compute expected
22957         start + width and last_order of the group, fail if check_no_overlap
22958         fails.
22959         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22960         to group if check_no_overlap fails.
22962 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22964         * config/rs6000/altivec.md: Delete contraint arguments to
22965         define_expand, define_split, and define_peephole2, and in
22966         define_insn_and_split if always unused.
22967         * config/rs6000/darwin.md: Ditto.
22968         * config/rs6000/dfp.md: Ditto.
22969         * config/rs6000/rs6000.md: Ditto.
22970         * config/rs6000/sync.md: Ditto.
22971         * config/rs6000/vector.md: Ditto.
22972         * config/rs6000/vsx.md: Ditto.
22974 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22976         * config/rs6000/altivec.md: Write output control strings as braced
22977         blocks instead of double-quoted strings.
22978         * config/rs6000/darwin.md: Ditto.
22979         * config/rs6000/rs6000.md: Ditto.
22980         * config/rs6000/vector.md: Ditto.
22981         * config/rs6000/vsx.md: Ditto.
22983 2018-02-21  Jason Merrill  <jason@redhat.com>
22985         PR c++/84314 - ICE with templates and fastcall attribute.
22986         * attribs.c (build_type_attribute_qual_variant): Remove assert.
22988 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22990         * ipa-cp.c (determine_versionability): Fix comment typos.
22992 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22994         PR c/84229
22995         * ipa-cp.c (determine_versionability): Do not version functions caling
22996         va_arg_pack.
22998 2018-02-21  Martin Liska  <mliska@suse.cz>
23000         PR driver/83193
23001         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
23002         Add "native" as a possible value.
23003         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
23004         the macro when native cpu detection is available.
23006 2018-02-21  Martin Liska  <mliska@suse.cz>
23008         PR driver/83193
23009         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
23010         Add "native" as a possible value.
23011         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
23012         when native cpu detection is available.
23014 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
23015             Martin Sebor  <msebor@redhat.com>
23017         PR tree-optimization/84478
23018         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
23019         false.
23020         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
23021         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
23022         support which is conservatively correct, for 2 only stay conservative
23023         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
23024         argument to the 2 argument get_range_strlen, adjust 6 arg
23025         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
23026         false.
23027         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
23028         (gimple_fold_builtin_strlen): Pass true as last argument to
23029         get_range_strlen.
23031 2018-02-20  Martin Sebor  <msebor@redhat.com>
23033         PR middle-end/84095
23034         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
23035         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
23036         (builtin_memref::builtin_memref): Factor out parts into
23037         set_base_and_offset and call it.
23039 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
23041         PR middle-end/84406
23042         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
23043         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
23044         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
23045         search at the associated MODE_INT.
23047 2018-02-20  Jeff Law  <law@redhat.com>
23049         PR middle-end/82123
23050         PR tree-optimization/81592
23051         PR middle-end/79257
23052         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
23053         for range data rather than using global data.
23054         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
23055         range data rather than using global data.
23056         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
23057         pass it to children as needed.
23058         (struct directive::fmtresult): Similarly.
23059         (struct directive::set_width): Similarly.
23060         (struct directive::set_precision): Similarly.
23061         (format_integer, format_directive, parse_directive): Similarly.
23062         (format_none): Accept unnamed vr_values parameter.
23063         (format_percent, format_floating, format_character): Similarly.
23064         (format_string, format_plain): Similarly.
23065         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
23066         the EVRP range analyzer for range data rather than using global data.
23067         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
23068         gimple-ssa-evrp-analyze.h
23069         (class sprintf_dom_walker): Add after_dom_children member function.
23070         Add evrp_range_analyzer member.
23071         (sprintf_dom_walker::before_dom_children): Call into the EVRP
23072         range analyzer as needed.
23073         (sprintf_dom_walker::after_dom_children): New member function.
23074         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
23075         if not optimizing.
23076         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
23077         (evrp_range_analyzer::pop_to_marker): Likewise.
23079 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
23081         PR tree-optimization/84419
23082         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
23083         with the required type if its current type is compatible but
23084         different.
23086 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
23088         PR middle-end/82004
23089         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
23090         after vectorization.
23092 2018-02-20  Martin Liska  <mliska@suse.cz>
23094         PR driver/83193
23095         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
23096         possible values if we don't have a hint.
23098 2018-02-20  Martin Liska  <mliska@suse.cz>
23100         PR c/84310
23101         PR target/79747
23102         * final.c (shorten_branches): Build align_tab array with one
23103         more element.
23104         * opts.c (finish_options): Add alignment option limit check.
23105         (MAX_CODE_ALIGN): Likewise.
23106         (MAX_CODE_ALIGN_VALUE): Likewise.
23107         * doc/invoke.texi: Document maximum allowed option value for
23108         all -falign-* options.
23110 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
23112         PR target/84146
23113         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
23114         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
23115         * var-tracking.c (emit_note_insn_var_location): Remove all references
23116         to NOTE_INSN_CALL_ARG_LOCATION.
23117         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
23118         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
23119         Use copy_rtx_if_shared.
23120         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
23121         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
23122         (dwarf2out_var_location): Remove handling of
23123         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
23124         on call_insn.
23125         * final.c (final_scan_insn): Remove all references to
23126         NOTE_INSN_CALL_ARG_LOCATION.
23127         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
23128         before dumping final insns.
23129         * except.c (emit_note_eh_region_end): Remove all references to
23130         NOTE_INSN_CALL_ARG_LOCATION.
23131         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
23132         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
23133         * config/arc/arc.c (hwloop_optimize): Likewise.
23134         * config/arm/arm.c (create_fix_barrier): Likewise.
23135         * config/s390/s390.c (s390_chunkify_start): Likewise.
23136         * config/sh/sh.c (find_barrier): Likewise.
23137         * config/i386/i386.c (rest_of_insert_endbranch,
23138         ix86_seh_fixup_eh_fallthru): Likewise.
23139         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
23140         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
23141         * config/frv/frv.c (frv_function_prologue): Likewise.
23142         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
23143         reg note.
23144         (note_outside_basic_block_p): Remove all references to
23145         NOTE_INSN_CALL_ARG_LOCATION.
23146         * gengtype.c (adjust_field_rtx_def): Likewise.
23147         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
23148         Likewise.
23149         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
23150         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
23152         PR c++/84444
23153         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
23154         is ADDR_EXPR.
23156         PR tree-optimization/84452
23157         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
23158         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
23159         is NULL.
23161 2018-02-19  Martin Liska  <mliska@suse.cz>
23163         PR sanitizer/82183
23164         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
23166 2018-02-19  Martin Liska  <mliska@suse.cz>
23167             Richard Sandiford  <richard.sandiford@linaro.org>
23169         PR tree-optimization/82491
23170         * gimple-fold.c (get_base_constructor): Make earlier bail out
23171         to prevent ubsan.
23173 2018-02-19  Carl Love  <cel@us.ibm.com>
23175         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
23176         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
23177         BU_P8V_OVERLOAD_1.
23178         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
23179         P8V_BUILTIN_VEC_NEG.
23181 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
23183         * config/rl78/rl78.md (movdf): New define expand.
23185 2018-02-19  Martin Liska  <mliska@suse.cz>
23187         PR other/80589
23188         * doc/invoke.texi: Fix typo.
23189         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
23191 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
23193         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
23194         handle rs6000_single_float and rs6000_double_float specially for
23195         e500 family CPUs.
23197 2018-02-16  Jeff Law  <law@redhat.com>
23199         * config/rx/rx.c (add_pop_cfi_notes): New function.;
23200         (pop_regs): Use it.
23202 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
23204         PR ipa/84425
23205         * ipa-inline.c (inline_small_functions): Fix a typo.
23207 2018-02-16  Nathan Sidwell  <nathan@acm.org>
23209         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
23211 2018-02-16  Carl Love  <cel@us.ibm.com>
23213         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
23214         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
23215         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
23216         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
23217         expansion to P8V_BUILTIN_VEC_FLOAT2.
23219 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
23221         PR rtl-optimization/70023
23222         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
23223         src_regno into account.
23225 2018-02-16  Carl Love  <cel@us.ibm.com>
23227         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
23228         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
23229         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
23230         * config/rs6000/rs6000.c: Remove case statements for
23231         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
23232         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
23233         and P9V_BUILTIN_VEC_VINSERT4B.
23234         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
23235         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
23236         * config/rs6000/vsx.md:
23237         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
23238         vec_insert4b.
23240 2018-02-16  Carl Love  <cel@us.ibm.com>
23242         * config/rs6000/altivec.h: Add builtin names vec_extract4b
23243         vec_insert4b.
23244         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
23245         definitions.
23246         * config/rs6000/rs6000-c.c: Add the definitions for
23247         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
23248         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
23249         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
23250         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
23251         definition for insert4b and define insn *insert3b_internal.
23252         * doc/extend.texi: Add documentation for vec_extract4b.
23254 2018-02-16  Nathan Sidwell  <nathan@acm.org>
23256         * doc/extend.texi (Backwards Compatibility): Mention friend
23257         injection.  Note for-scope is deprecated.
23258         * doc/invoke.texi (-ffriend-injection): Deprecate.
23260 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
23262         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
23263         that moved to I2, also allow destinations that are a paradoxical
23264         subreg (instead of a normal reg).
23266 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
23268         PR target/83831
23269         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
23270         to QImode.
23272 2018-02-16  Richard Biener  <rguenther@suse.de>
23274         PR tree-optimization/84037
23275         PR tree-optimization/84016
23276         PR target/82862
23277         * config/i386/i386.c (ix86_builtin_vectorization_cost):
23278         Adjust vec_construct for the fact we need additional higher latency
23279         128bit inserts for AVX256 and AVX512 vector builds.
23280         (ix86_add_stmt_cost): Scale vector construction cost for
23281         elementwise loads.
23283 2018-02-16  Richard Biener  <rguenther@suse.de>
23285         PR tree-optimization/84417
23286         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
23287         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
23288         (non_rewritable_lvalue_p): Likewise, use poly-ints.
23290 2018-02-16  Martin Liska  <mliska@suse.cz>
23292         PR sanitizer/84307
23293         * internal-fn.def (ASAN_CHECK): Set proper flags.
23294         (ASAN_MARK): Likewise.
23296 2018-02-16  Julia Koval  <julia.koval@intel.com>
23298         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
23299         from PTA_CANNONLAKE.
23301 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
23303         PR target/84272
23304         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
23305         Use ++iter rather than iter++ for std::list iterators.
23306         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
23307         defer deleting them until all nodes in the forest are processed.  Do
23308         free even leaf nodes.  Change to_process into auto_vec.
23310         PR bootstrap/84405
23311         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
23312         * vec.h (vec_default_construct): Use memset instead of placement new
23313         if BROKEN_VALUE_INITIALIZATION is defined.
23314         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
23315         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
23316         is defined.
23318         PR rtl-optimization/83723
23319         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
23320         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
23321         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
23322         recursive calls.
23323         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
23324         callers.
23325         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
23327 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
23329         PR rtl-optimization/81443
23330         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
23331         from inner REGs to paradoxical SUBREGs.
23333 2018-02-16  Richard Biener  <rguenther@suse.de>
23335         PR tree-optimization/84399
23336         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23337         For operands we can analyze at their definition make sure we can
23338         analyze them at each use as well.
23340 2018-02-16  Richard Biener  <rguenther@suse.de>
23342         PR tree-optimization/84190
23343         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
23344         volatile accesses if the decl isn't volatile.
23346 2018-02-15  Jason Merrill  <jason@redhat.com>
23348         PR c++/84314 - ICE with templates and fastcall attribute.
23349         * attribs.c (build_type_attribute_qual_variant): Don't clobber
23350         TYPE_CANONICAL on an existing type.
23352 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
23354         PR tree-optimization/84383
23355         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
23356         dstoff nor call operand_equal_p if dstbase is NULL.
23358         PR tree-optimization/84334
23359         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
23360         also a CONSTANT_CLASS_P, punt.
23362 2018-02-14  Jim Wilson  <jimw@sifive.com>
23364         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
23365         first SMALL_OPERAND check.  New local min_second_step.  Move assert
23366         to where locals are set.  Add TARGET_RVC support.
23367         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
23369 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
23371         * doc/invoke.texi: Correct -Wformat-overflow code sample.
23373 2018-02-14  Martin Sebor  <msebor@redhat.com>
23375         PR tree-optimization/83698
23376         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
23377         arrays constrain the offset range to their bounds.
23378         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
23379         (builtin_access::overlap): Avoid setting the size of overlap if it's
23380         already been set.
23381         (maybe_diag_overlap): Also consider arrays when deciding what values
23382         of offsets to include in diagnostics.
23384 2018-02-14  Martin Sebor  <msebor@redhat.com>
23386         PR c/84108
23387         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
23388         that correspond to the kind of a declaration.
23390 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
23392         PR target/83984
23393         * config/pa/pa.md: Load address of PIC label using the linkage table
23394         if the label is nonlocal.
23396 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23398         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
23399         warning message if user requests -maltivec=be.
23400         * doc/invoke.texi: Document deprecation of -maltivec=be.
23402 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
23404         PR target/84220
23405         * config/rs6000/rs6000-c.c: Update definitions for
23406         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
23407         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
23409 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23411         PR target/84239
23412         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
23413         add _get_ssp intrinsics. Remove argument from
23414         __builtin_ia32_rdssp[d|q].
23415         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
23416         * config/i386/i386-builtin.def: Remove argument from
23417         __builtin_ia32_rdssp[d|q].
23418         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
23419         ix86_expand_special_args_builtin for _rdssp[d|q].
23420         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
23421         Clear register before usage.
23422         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
23423         Add documentation for new _get_ssp and _inc_ssp intrinsics.
23425 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
23427         PR tree-optimization/84357
23428         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
23429         operand 1 of an ARRAY_REF too.
23431 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
23433         PR target/83831
23434         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
23435         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23436         declarations.
23437         (set_of_reg): New struct.
23438         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
23439         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
23440         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23441         functions.
23442         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
23443         Split into bitclr, bitset, bitinvert patterns if appropriate.
23444         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
23445         use rx_fuse_in_memory_bitop.
23446         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
23447         to named insn, correct maximum insn length.
23449 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
23451         PR target/79242
23452         * machmode.def: Define a complex mode for PARTIAL_INT.
23453         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
23454         MODE_PARTIAL_INT.
23455         * doc/rtl.texi: Document CSPImode.
23456         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
23457         handling.
23458         (msp430_hard_regno_nregs_with_padding): Likewise.
23460 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
23462         PR target/84279
23463         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
23465 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
23467         PR rtl-optimization/84169
23468         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
23469         we generated a parallel as new i3 and we split that to new i2 and i3
23470         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
23471         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
23472         those to i2, not i1.  Partially rewrite this scan code.
23474 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23476         PR c/82210
23477         * stor-layout.c (place_field): For variable length fields, adjust
23478         offset_align afterwards not just based on the field's alignment,
23479         but also on the size.
23481         PR middle-end/84309
23482         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
23483         of exps and logs in the use_exp2 case.
23485 2018-02-13  Jeff Law  <law@redhat.com>
23487         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
23488         entry for "vector".
23490         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
23491         ARGS as unused.
23493 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
23495         PR debug/84342
23496         PR debug/84319
23497         * common.opt (gas-loc-support, gas-locview-support): New.
23498         (ginline-points, ginternal-reset-location-views): New.
23499         * doc/invoke.texi: Document them.  Use @itemx where intended.
23500         (gvariable-location-views): Adjust.
23501         * target.def (reset_location_view): New.
23502         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
23503         (TARGET_RESET_LOCATION_VIEW): New.
23504         * doc/tm.texi: Rebuilt.
23505         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
23506         (dwarf2out_default_as_locview_support): New.
23507         (output_asm_line_debug_info): Use option variables.
23508         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
23509         (output_loc_list): Likewise.
23510         (add_high_low_attributes): Check option variables.
23511         Don't output entry view attribute in strict mode.
23512         (gen_inlined_subroutine_die): Check option variables.
23513         (dwarf2out_inline_entry): Likewise.
23514         (init_sections_and_labels): Likewise.
23515         (dwarf2out_early_finish): Likewise.
23516         (maybe_reset_location_view): New, from...
23517         (dwarf2out_var_location): ... here.  Call it.
23518         * debug.h (dwarf2out_default_as_loc_support): Declare.
23519         (dwarf2out_default_as_locview_support): Declare.
23520         * hooks.c (hook_int_rtx_insn_0): New.
23521         * hooks.h (hook_int_rtx_insn_0): Declare.
23522         * toplev.c (process_options): Take -gas-loc-support and
23523         -gas-locview-support from dwarf2out.  Enable
23524         -gvariable-location-views by default only with locview
23525         assembler support.  Enable -ginternal-reset-location-views by
23526         default only if the target defines the corresponding hook.
23527         Enable -ginline-points by default if location views are
23528         enabled; force it disabled if statement frontiers are
23529         disabled.
23530         * tree-inline.c (expand_call_inline): Check option variables.
23531         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
23533 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
23535         PR tree-optimization/84321
23536         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
23537         handling.  Also check whether the anti-range contains any values
23538         that satisfy the mask; switch to a VR_RANGE if not.
23540 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
23542         PR sanitizer/84340
23543         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
23545 2018-02-13  Martin Jambor  <mjambor@suse.cz>
23547         PR c++/83990
23548         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
23549         of call statements, also set location of a load to a temporary.
23551 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
23553         * config/rl78/rl78.c (add_vector_labels): New function.
23554         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
23555         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
23556         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
23557         which checks that no arguments are passed.
23558         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
23559         * doc/extend.texi: Documentation for the new attribute.
23561 2018-02-13  Andreas Schwab  <schwab@suse.de>
23563         * config/riscv/linux.h (CPP_SPEC): Define.
23565 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23567         PR target/84335
23568         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
23569         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
23570         OPTION_MASK_ISA_AES as first argument to def_builtin_const
23571         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
23572         instead of OPTION_MASK_ISA_PCLMUL as first argument to
23573         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
23574         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
23575         temporarily for AES and PCLMUL builtins.
23577         PR tree-optimization/84339
23578         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
23579         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
23580         Formatting fixes.
23582         PR middle-end/84309
23583         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
23584         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
23585         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
23586         inline function.
23587         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
23588         inline function.
23589         * omp-simd-clone.h: New file.
23590         * omp-simd-clone.c: Include omp-simd-clone.h.
23591         (expand_simd_clones): No longer static.
23592         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
23593         cgraph.h and omp-simd-clone.h.
23594         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
23595         (vect_recog_widen_shift_pattern): Formatting fix.
23596         (vect_pattern_recog_1): Don't check optab for calls.
23598         PR target/84336
23599         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
23600         operands[2] into a REG before using gen_lowpart on it.
23602 2018-02-12  Jeff Law  <law@redhat.com>
23604         PR target/83760
23605         * config/sh/sh.c (find_barrier): Consider a sibling call
23606         a barrier as well.
23608         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
23609         successfully back substituting a reg.
23611 2018-02-12  Richard Biener  <rguenther@suse.de>
23613         PR tree-optimization/84037
23614         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
23615         parameter, move visited init to caller.
23616         (vect_slp_analyze_operations): Separate cost from validity
23617         check, initialize visited once for all instances.
23618         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
23619         for all instances.
23620         * tree-vect-stmts.c (vect_model_simple_cost): Make early
23621         out an assert.
23622         (vect_model_promotion_demotion_cost): Likewise.
23623         (vectorizable_bswap): Guard cost modeling with !slp_node
23624         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23625         SLP stmts.
23626         (vectorizable_call): Likewise.
23627         (vectorizable_conversion): Likewise.
23628         (vectorizable_assignment): Likewise.
23629         (vectorizable_shift): Likewise.
23630         (vectorizable_operation): Likewise.
23631         (vectorizable_store): Likewise.
23632         (vectorizable_load): Likewise.
23633         (vectorizable_condition): Likewise.
23634         (vectorizable_comparison): Likewise.
23636 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
23638         PR sanitizer/84307
23639         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23640         (ASAN_MARK): Fix fnspec to account for return value, change pointer
23641         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23643 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23645         PR middle-end/83665
23646         * params.def (inline-min-speedup): Increase from 8 to 15.
23647         (max-inline-insns-auto): Decrease from 40 to 30.
23648         * ipa-split.c (consider_split): Add some buffer for function to
23649         be considered inlining candidate.
23650         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23651         default values.
23653 2018-02-12  Richard Biener  <rguenther@suse.de>
23655         PR tree-optimization/84037
23656         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23657         matched stmts if we cannot swap the non-matched ones.
23659 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23661         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23662         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23663         _mm_maskz_scalef_round_ss): New intrinsics.
23664         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23665         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23666         __builtin_ia32_scalefss_round): Remove.
23667         (__builtin_ia32_scalefsd_mask_round,
23668         __builtin_ia32_scalefss_mask_round): New intrinsics.
23669         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23670         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23671         ((match_operand:VF_128 2 "<round_nimm_predicate>"
23672         "<round_constraint>")): Changed to ...
23673         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23674         "<round_scalar_constraint>")): ... this.
23675         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23676         %0, %1, %2<round_op3>}"): Changed to ...
23677         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23678         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23679         %2<round_scalar_mask_op3>}"): ... this.
23680         * config/i386/subst.md (round_scalar_nimm_predicate): New.
23682 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23684         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23685         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23686         (_mm_maskz_sqrt_round_ss): New intrinsics.
23687         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23688         (__builtin_ia32_sqrtsd_mask_round)
23689         (__builtin_ia32_sqrtss_mask_round): New builtins.
23690         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23691         (__builtin_ia32_sqrtss_round): Remove.
23692         (__builtin_ia32_sqrtsd_mask_round)
23693         (__builtin_ia32_sqrtss_mask_round): New builtins.
23694         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23695         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23696         ((match_operand:VF_128 1 "vector_operand"
23697         "xBm,<round_constraint>")): Changed to ...
23698         ((match_operand:VF_128 1 "vector_operand"
23699         "xBm,<round_scalar_constraint>")): ... this.
23700         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23701         %0, %2, %<iptr>1<round_op3>}): Changed to ...
23702         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23703         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23704         %<iptr>1<round_scalar_mask_op3>}): ... this.
23705         ((set_attr "prefix" "<round_prefix>")): Changed to ...
23706         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23708 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
23710         PR target/84266
23711         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23712         Cast vec_cmpeq result to correct type.
23713         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23714         Cast vec_cmpgt result to correct type.
23716 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
23718         * final.c (final_scan_insn_1): Renamed from...
23719         (final_scan_insn): ... this.  New wrapper, to recover
23720         seen from the outermost call in recursive ones.
23721         * config/sparc/sparc.c (output_return): Drop seen from call.
23722         (output_sibcall): Likewise.
23723         * config/visium/visium.c (output_branch): Likewise.
23725 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
23727         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23728         function label.
23730 2018-02-10  Alan Modra  <amodra@gmail.com>
23732         PR target/84300
23733         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23734         Specify LR as an input.
23736 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
23738         PR sanitizer/83987
23739         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23740         remove_member_access_dummy_vars): New functions.
23741         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23742         lower_omp_1, execute_lower_omp): Use them.
23744         PR rtl-optimization/84308
23745         * shrink-wrap.c (spread_components): Release todo vector.
23747 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
23749         PR rtl-optimization/57193
23750         * ira-color.c (struct allocno_color_data): Add member
23751         conflict_allocno_hard_prefs.
23752         (update_conflict_allocno_hard_prefs): New.
23753         (bucket_allocno_compare_func): Add a preference based on
23754         conflict_allocno_hard_prefs.
23755         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23756         (color_allocnos): Remove a dead code.  Initiate
23757         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
23759 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23761         PR target/84226
23762         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23763         constraint from =wa to wa.  Avoid a subreg on the output operand,
23764         instead use a pseudo and subreg it in a move.
23765         (p9_xxbrd_<mode>): Changed to ...
23766         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23767         (p9_xxbrd_v2df): New expander.
23768         (p9_xxbrw_<mode>): Changed to ...
23769         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23770         (p9_xxbrw_v4sf): New expander.
23772 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23774         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23776 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
23778         PR target/83926
23779         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23780         multiply in 32-bit mode.
23781         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23782         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23783         mode.
23785 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23787         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23788         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23789         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23790         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23792 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
23794         PR lto/84213
23795         * dwarf2out.c (is_trivial_indirect_ref): New function.
23796         (dwarf2out_late_global_decl): Do not generate a location
23797         attribute for variables that have a non-trivial DECL_VALUE_EXPR
23798         and that are not defined in the current unit.
23800 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23802         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23803         instead of a libcall for UNORDERED.
23805 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
23807         PR target/82641
23808         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23809         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23811 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23813         PR target/PR84295
23814         * config/s390/s390.c (s390_set_current_function): Invoke
23815         s390_indirect_branch_settings also if fndecl didn't change.
23817 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23819         * config/rs6000/rs6000.md (blockage): Set length to zero.
23821 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23823         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23825 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23827         PR sanitizer/84285
23828         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23829         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23830         -static-lib*san.
23832         PR debug/84252
23833         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23834         PARALLEL incoming that failed vt_get_decl_and_offset check.
23836         PR middle-end/84237
23837         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23838         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23839         TREE_READONLY bit.
23840         (get_variable_section): For decls in named .bss* sections pass true as
23841         second argument to bss_initializer_p.
23843 2018-02-09  Marek Polacek  <polacek@redhat.com>
23844             Jakub Jelinek  <jakub@redhat.com>
23846         PR c++/83659
23847         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23848         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
23849         Sync some changes from cxx_fold_indirect_ref.
23851 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23853         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23854         markers.
23855         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23856         (BLOCK_INLINE_ENTRY_LABEL): New.
23857         (dwarf2out_var_location): Disregard inline entry markers.
23858         (inline_entry_data): New struct.
23859         (inline_entry_data_hasher): New hashtable type.
23860         (inline_entry_data_hasher::hash): New.
23861         (inline_entry_data_hasher::equal): New.
23862         (inline_entry_data_table): New variable.
23863         (add_high_low_attributes): Add DW_AT_entry_pc and
23864         DW_AT_GNU_entry_view attributes if a pending entry is found
23865         in inline_entry_data_table.  Add old entry_pc attribute only
23866         if debug nonbinding markers are disabled.
23867         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23868         markers are enabled.
23869         (block_within_block_p, dwarf2out_inline_entry): New.
23870         (dwarf2out_finish): Check that no entries remained in
23871         inline_entry_data_table.
23872         * final.c (reemit_insn_block_notes): Handle inline entry notes.
23873         (final_scan_insn, notice_source_line): Likewise.
23874         (rest_of_clean_state): Skip inline entry markers.
23875         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23876         markers.
23877         * gimple.c (gimple_build_debug_inline_entry): New.
23878         * gimple.h (enum gimple_debug_subcode): Add
23879         GIMPLE_DEBUG_INLINE_ENTRY.
23880         (gimple_build_debug_inline_entry): Declare.
23881         (gimple_debug_inline_entry_p): New.
23882         (gimple_debug_nonbind_marker_p): Adjust.
23883         * insn-notes.def (INLINE_ENTRY): New.
23884         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23885         inline entry marker notes.
23886         (print_insn): Likewise.
23887         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23888         (INSN_DEBUG_MARKER_KIND): Likewise.
23889         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23890         * tree-inline.c (expand_call_inline): Build and insert
23891         debug_inline_entry stmt.
23892         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23893         inline entry blocks early, if nonbind markers are enabled.
23894         (dump_scope_block): Dump fragment info.
23895         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23896         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23897         (gimple_build_debug_inline_entry): New.
23898         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23899         Enable/disable inline entry points too.
23900         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23901         (DEBUG_INSN): Describe inline entry markers.
23903         * common.opt (gvariable-location-views): New.
23904         (gvariable-location-views=incompat5): New.
23905         * config.in: Rebuilt.
23906         * configure: Rebuilt.
23907         * configure.ac: Test assembler for view support.
23908         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23909         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23910         * dwarf2out.c (var_loc_view): New typedef.
23911         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23912         (dwarf2out_locviews_in_attribute): New.
23913         (dwarf2out_locviews_in_loclist): New.
23914         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23915         (enum dw_line_info_opcode): Add LI_adv_address.
23916         (struct dw_line_info_table): Add view.
23917         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23918         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23919         (zero_view_p): New variable.
23920         (ZERO_VIEW_P): New macro.
23921         (output_asm_line_debug_info): New.
23922         (struct var_loc_node): Add view.
23923         (add_AT_view_list, AT_loc_list): New.
23924         (add_var_loc_to_decl): Add view param.  Test it against last.
23925         (new_loc_list): Add view params.  Record them.
23926         (AT_loc_list_ptr): Handle loc and view lists.
23927         (view_list_to_loc_list_val_node): New.
23928         (print_dw_val): Handle dw_val_class_view_list.
23929         (size_of_die): Likewise.
23930         (value_format): Likewise.
23931         (loc_list_has_views): New.
23932         (gen_llsym): Set vl_symbol too.
23933         (maybe_gen_llsym, skip_loc_list_entry): New.
23934         (dwarf2out_maybe_output_loclist_view_pair): New.
23935         (output_loc_list): Output view list or entries too.
23936         (output_view_list_offset): New.
23937         (output_die): Handle dw_val_class_view_list.
23938         (output_dwarf_version): New.
23939         (output_compilation_unit_header): Use it.
23940         (output_skeleton_debug_sections): Likewise.
23941         (output_rnglists, output_line_info): Likewise.
23942         (output_pubnames, output_aranges): Update version comments.
23943         (output_one_line_info_table): Output view numbers in asm comments.
23944         (dw_loc_list): Determine current endview, pass it to new_loc_list.
23945         Call maybe_gen_llsym.
23946         (loc_list_from_tree_1): Adjust.
23947         (add_AT_location_description): Create view list attribute if
23948         needed, check it's absent otherwise.
23949         (convert_cfa_to_fb_loc_list): Adjust.
23950         (maybe_emit_file): Call output_asm_line_debug_info for test.
23951         (dwarf2out_var_location): Reset views as needed.  Precompute
23952         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
23953         attribute.  Set view.
23954         (new_line_info_table): Reset next view.
23955         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23956         (dwarf2out_source_line): Likewise.  Output view resets and labels to
23957         the assembler, or select appropriate line info opcodes.
23958         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23959         (optimize_string_length): Catch it.  Adjust.
23960         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
23961         dw_val_class_view_list, and remove it if no longer needed.
23962         (hash_loc_list): Hash view numbers.
23963         (loc_list_hasher::equal): Compare them.
23964         (optimize_location_lists): Check whether a view list symbol is
23965         needed, and whether the locview attribute is present, and
23966         whether they match.  Remove the locview attribute if no longer
23967         needed.
23968         (index_location_lists): Call skip_loc_list_entry for test.
23969         (dwarf2out_finish): Call output_asm_line_debug_info for test.
23970         Use output_dwarf_version.
23971         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23972         (struct dw_val_node): Add val_view_list.
23973         * final.c (SEEN_NEXT_VIEW): New.
23974         (set_next_view_needed): New.
23975         (clear_next_view_needed): New.
23976         (maybe_output_next_view): New.
23977         (final_start_function): Rename to...
23978         (final_start_function_1): ... this.  Take pointer to FIRST,
23979         add SEEN parameter.  Emit param bindings in the initial view.
23980         (final_start_function): Reintroduce SEEN-less interface.
23981         (final): Rename to...
23982         (final_1): ... this.  Take SEEN parameter.  Output final pending
23983         next view at the end.
23984         (final): Reintroduce seen-less interface.
23985         (final_scan_insn): Output pending next view before switching
23986         sections or ending a block.  Mark the next view as needed when
23987         outputting variable locations.  Notify debug backend of section
23988         changes, and of location view changes.
23989         (rest_of_handle_final): Adjust.
23990         * toplev.c (process_options): Autodetect value for debug variable
23991         location views option.  Warn on incompat5 without -gdwarf-5.
23992         * doc/invoke.texi (gvariable-location-views): New.
23993         (gvariable-location-views=incompat5): New.
23994         (gno-variable-location-views): New.
23996 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
23998         PR tree-optimization/84136
23999         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
24000         that the result of find_edge is non-NULL.
24002 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
24004         PR target/83008
24005         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
24006         storing integer register in SImode.  Fix cost of 256 and 512
24007         byte aligned SSE register store.
24009 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
24011         * config/i386/i386.c (ix86_multiplication_cost): Fix
24012         multiplication cost for TARGET_AVX512DQ.
24014 2018-02-08  Marek Polacek  <polacek@redhat.com>
24016         PR tree-optimization/84238
24017         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
24018         get_range_strlen.
24020 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24022         PR tree-optimization/84265
24023         * tree-vect-stmts.c (vectorizable_store): Don't treat
24024         VMAT_CONTIGUOUS accesses as grouped.
24025         (vectorizable_load): Likewise.
24027 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24029         PR tree-optimization/81635
24030         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
24031         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
24032         (test_round_for_mask): New functions.
24033         (wide_int_cc_tests): Call test_round_for_mask.
24034         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
24035         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
24036         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
24037         range returned by get_range_info.
24039 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
24041         PR ipa/81360
24042         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
24043         * symtab.c: Include builtins.h
24044         (symtab_node::output_to_lto_symbol_table_p): Move here
24045         from lto-streamer-out.c:output_symbol_p.
24046         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
24047         (output_symbol_p): Move all logic to symtab.c
24048         (produce_symtab): Update.
24050 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24052         * config/s390/s390-opts.h (enum indirect_branch): Define.
24053         * config/s390/s390-protos.h (s390_return_addr_from_memory)
24054         (s390_indirect_branch_via_thunk)
24055         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
24056         (enum s390_indirect_branch_type): Define.
24057         * config/s390/s390.c (struct s390_frame_layout, struct
24058         machine_function): Remove.
24059         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
24060         (indirect_branch_table_label_no, indirect_branch_table_name):
24061         Define variables.
24062         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
24063         (enum s390_indirect_branch_option): Define.
24064         (s390_return_addr_from_memory): New function.
24065         (s390_handle_string_attribute): New function.
24066         (s390_attribute_table): Add new attribute handler.
24067         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
24068         (s390_indirect_branch_via_thunk): New function.
24069         (s390_indirect_branch_via_inline_thunk): New function.
24070         (s390_function_ok_for_sibcall): When jumping via thunk disallow
24071         sibling call optimization for non z10 compiles.
24072         (s390_emit_call): Force indirect branch target to be a single
24073         register.  Add r1 clobber for non-z10 compiles.
24074         (s390_emit_epilogue): Emit return jump via return_use expander.
24075         (s390_reorg): Handle JUMP_INSNs as execute targets.
24076         (s390_option_override_internal): Perform validity checks for the
24077         new command line options.
24078         (s390_indirect_branch_attrvalue): New function.
24079         (s390_indirect_branch_settings): New function.
24080         (s390_set_current_function): Invoke s390_indirect_branch_settings.
24081         (s390_output_indirect_thunk_function):  New function.
24082         (s390_code_end): Implement target hook.
24083         (s390_case_values_threshold): Implement target hook.
24084         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
24085         macros.
24086         * config/s390/s390.h (struct s390_frame_layout)
24087         (struct machine_function): Move here from s390.c.
24088         (TARGET_INDIRECT_BRANCH_NOBP_RET)
24089         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
24090         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
24091         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
24092         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
24093         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
24094         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
24095         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
24096         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
24097         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
24098         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
24099         (mnemonic attribute): Add values which aren't recognized
24100         automatically.
24101         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
24102         pattern for branch conversion.  Fix mnemonic attribute.
24103         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
24104         indirect branch via thunk if requested.
24105         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
24106         ("*indirect_jump"): Disable for branch conversion using out of
24107         line thunks.
24108         ("indirect_jump_via_thunk<mode>_z10")
24109         ("indirect_jump_via_thunk<mode>")
24110         ("indirect_jump_via_inlinethunk<mode>_z10")
24111         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
24112         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
24113         ("casesi_jump_via_inlinethunk<mode>_z10")
24114         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
24115         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
24116         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
24117         ("*indirect2_jump"): Disable for branch conversion.
24118         ("casesi_jump"): Turn into expander and expand patterns for branch
24119         conversion.
24120         ("return_use"): New expander.
24121         ("*return"): Emit return via thunk and rename it to ...
24122         ("*return<mode>"): ... this one.
24123         * config/s390/s390.opt: Add new options and and enum for the
24124         option values.
24126 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24128         * lra-constraints.c (match_reload): Unconditionally use
24129         gen_lowpart_SUBREG, rather than selecting between that
24130         and equivalent gen_rtx_SUBREG code.
24132 2018-02-08  Richard Biener  <rguenther@suse.de>
24134         PR tree-optimization/84233
24135         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
24136         changed flag instead of boguously re-using phi_inserted.
24138 2018-02-08  Martin Jambor  <mjambor@suse.cz>
24140         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
24141         static local variables.
24143 2018-02-08  Richard Biener  <rguenther@suse.de>
24145         PR tree-optimization/84278
24146         * tree-vect-stmts.c (vectorizable_store): When looking for
24147         smaller vector types to perform grouped strided loads/stores
24148         make sure the mode is supported by the target.
24149         (vectorizable_load): Likewise.
24151 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
24153         * config/aarch64/aarch64.c (aarch64_components_for_bb):
24154         Increase LDP/STP opportunities by adding adjacent callee-saves.
24156 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
24158         PR rtl-optimization/84068
24159         PR rtl-optimization/83459
24160         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
24162 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
24164         PR tree-optimization/84224
24165         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
24166         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
24167         non-zero arguments.
24169 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
24171         PR target/84113
24172         * config/rs6000/altivec.md (*restore_world): Remove LR use.
24173         * config/rs6000/predicates.md (restore_world_operation): Adjust op
24174         count, remove one USE.
24176 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24178         * doc/install.texi (Configuration): Document the
24179         --with-long-double-format={ibm,ieee} PowerPC configuration
24180         options.
24182         PR target/84154
24183         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
24184         Convert from define_expand to be define_insn_and_split.  Rework
24185         float/double/_Float128 conversions to QI/HI/SImode to work with
24186         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
24187         conversions to QI/HImode types did a store and then a load to
24188         truncate the value.  For conversions to VSX registers, don't split
24189         the insn, instead emit the code directly.  Use the code iterator
24190         any_fix to combine signed and unsigned conversions.
24191         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
24192         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
24193         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24194         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
24195         (fix_<mode>di2_hw): Likewise.
24196         (fixuns_<mode>di2_hw): Likewise.
24197         (fix_<mode>si2_hw): Likewise.
24198         (fixuns_<mode>si2_hw): Likewise.
24199         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
24200         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24201         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
24202         fix<uns>_trunc<SFDF:mode>si2_p8.
24203         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
24204         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
24205         (fix<uns>_<mode>_mem): Likewise.
24206         (fctiw<u>z_<mode>_mem): Likewise.
24207         (fix<uns>_<mode>_mem): Likewise.
24208         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
24209         the register allocator from doing a direct move to the GPRs to do
24210         a store, and instead use the ISA 3.0 store byte/half-word from
24211         vector register instruction.  For IEEE 128-bit floating point,
24212         also optimize stores of 32-bit ints.
24213         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
24215 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
24217         * genextract.c (push_pathstr_operand): New function to support
24218         [a-zA-Z].
24219         (walk_rtx): Call push_pathstr_operand.
24220         (print_path): Support [a-zA-Z].
24222 2018-02-07  Richard Biener  <rguenther@suse.de>
24224         PR tree-optimization/84037
24225         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
24226         (cse_and_gimplify_to_preheader): Declare.
24227         (vect_get_place_in_interleaving_chain): Likewise.
24228         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24229         ivexpr_map.
24230         (_loop_vec_info::~_loop_vec_info): Delete it.
24231         (cse_and_gimplify_to_preheader): New function.
24232         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
24233         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
24234         (vectorizable_load): Likewise.  For grouped stores always base
24235         the IV on the first element.
24236         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
24237         condition before gimplifying.
24239 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24241         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
24242         *DIV_EXPR and *MOD_EXPR.
24244 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
24246         PR target/84248
24247         * config/i386/i386.c (ix86_option_override_internal): Mask out
24248         the CF_SET bit when checking -fcf-protection.
24250 2018-02-07  Tom de Vries  <tom@codesourcery.com>
24252         PR libgomp/84217
24253         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
24254         enough.
24256 2018-02-07  Richard Biener  <rguenther@suse.de>
24258         PR tree-optimization/84204
24259         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
24260         this place.
24262         PR tree-optimization/84205
24263         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
24264         special-case isl_ast_op_zdiv_r.
24266         PR tree-optimization/84223
24267         * graphite-scop-detection.c (gather_bbs::before_dom_children):
24268         Only add conditions from within the region.
24269         (gather_bbs::after_dom_children): Adjust.
24271 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
24273         PR target/84209
24274         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
24275         * config/avr/avr.md: Only post-reload split REG-REG moves if
24276         either register is GENERAL_REG_P.
24278 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24280         PR tree-optimization/84235
24281         * tree-ssa-scopedtables.c
24282         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
24283         if the subtraction is performed in floating point type where NaNs are
24284         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
24285         build 1.  Formatting fix.
24287 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
24289         PR target/84146
24290         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
24291         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
24292         and skip it regardless of bb boundaries.  Use CALL_P macro,
24293         don't test INSN_P (insn) together with CALL_P or JUMP_P check
24294         unnecessarily, formatting fix.
24296 2018-02-06  Michael Collison  <michael.collison@arm.com>
24298         * config/arm/thumb2.md:
24299         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
24300         (*thumb_mov_notscc): Ditto.
24302 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
24304         PR target/84154
24305         * config/rs6000/rs6000.md (su code attribute): Use "u" for
24306         unsigned_fix, not "s".
24308 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24310         * configure.ac (gcc_fn_eh_frame_ro): New function.
24311         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
24312         correct .eh_frame permissions.
24313         * configure: Regenerate.
24315 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
24317         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
24318         irrelevant options.
24320 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24322         * config/rs6000/rs6000.c (rs6000_option_override_internal):
24323         Display warning message for -mno-speculate-indirect-jumps.
24325 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
24327         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
24328         Undocumented.
24329         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
24331 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
24333         PR tree-optimization/84225
24334         * tree-eh.c (find_trapping_overflow): Only call
24335         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
24337 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24339         PR target/84145
24340         * config/i386/i386.c: Reimplement the check of possible options
24341         -mibt/-mshstk conbination. Change error messages.
24342         * doc/invoke.texi: Fix a typo: remove extra '='.
24344 2018-02-06  Marek Polacek  <polacek@redhat.com>
24346         PR tree-optimization/84228
24347         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
24349 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
24351         PR target/82641
24352         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
24353         emitted arch directives.
24354         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
24355         __ARM_FEATURE_COPROC before changing architectures.
24357 2018-02-06  Richard Biener  <rguenther@suse.de>
24359         * config/i386/i386.c (print_reg): Fix typo.
24360         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
24362 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
24364         * configure: Regenerate.
24366 2018-02-05  Martin Sebor  <msebor@redhat.com>
24368         PR tree-optimization/83369
24369         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
24370         inlining context.
24372 2018-02-05  Martin Liska  <mliska@suse.cz>
24374         * doc/invoke.texi: Cherry-pick upstream r323995.
24376 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
24378         * ira.c (ira_init_register_move_cost): Adjust comment.
24380 2018-02-05  Martin Liska  <mliska@suse.cz>
24382         PR gcov-profile/84137
24383         * doc/gcov.texi: Fix typo in documentation.
24385 2018-02-05  Martin Liska  <mliska@suse.cz>
24387         PR gcov-profile/83879
24388         * doc/gcov.texi: Document necessity of --dynamic-list-data when
24389         using dlopen functionality.
24391 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
24393         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
24394         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
24395         _mm_maskz_range_ss, _mm_mask_range_round_ss,
24396         _mm_maskz_range_round_ss): New intrinsics.
24397         (__builtin_ia32_rangesd128_round)
24398         (__builtin_ia32_rangess128_round): Remove.
24399         (__builtin_ia32_rangesd128_mask_round,
24400         __builtin_ia32_rangess128_mask_round): New builtins.
24401         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
24402         __builtin_ia32_rangess128_round): Remove.
24403         (__builtin_ia32_rangesd128_mask_round,
24404         __builtin_ia32_rangess128_mask_round): New builtins.
24405         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
24406         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
24407         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
24408         "<round_saeonly_constraint>")): Changed to ...
24409         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
24410         "<round_saeonly_scalar_constraint>")): ... this.
24411         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
24412         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
24413         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
24414         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
24415         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
24417 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
24419         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
24420         options.
24421         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
24422         Remove all values except native, 8540 and 8548.
24424 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
24426         * config/i386/i386.c (ix86_output_function_return): Pass
24427         INVALID_REGNUM, instead of -1, as invalid register number to
24428         indirect_thunk_name and output_indirect_thunk.
24430 2018-02-02  Julia Koval  <julia.koval@intel.com>
24432         * config.gcc: Add -march=icelake.
24433         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
24434         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
24435         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
24436         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
24437         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
24438         (processor_target_table): Add icelake.
24439         (ix86_option_override_internal): Handle new PTAs.
24440         (get_builtin_code_for_version): Handle icelake.
24441         (M_INTEL_COREI7_ICELAKE): New.
24442         (fold_builtin_cpu): Handle icelake.
24443         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
24444         * doc/invoke.texi: Add -march=icelake.
24446 2018-02-02  Julia Koval  <julia.koval@intel.com>
24448         * config/i386/i386.c (ix86_option_override_internal): Change flags type
24449         to wide_int_bitmask.
24450         * wide-int-bitmask.h: New.
24452 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24454         PR target/84066
24455         * config/i386/i386.md: Replace Pmode with word_mode in
24456         builtin_setjmp_setup and builtin_longjmp to support x32.
24458 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
24460         PR target/56010
24461         PR target/83743
24462         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
24463         #include "opts.h".
24464         (rs6000_supported_cpu_names): New static variable.
24465         (linux_cpu_translation_table): Likewise.
24466         (elf_platform) <cpu>: Define new static variable and use it.
24467         Translate kernel AT_PLATFORM name to canonical name if needed.
24468         Error if platform name is unknown.
24470 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
24472         PR target/84089
24473         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
24475 2018-02-01  Jeff Law  <law@redhat.com>
24477         PR target/84128
24478         * config/i386/i386.c (release_scratch_register_on_entry): Add new
24479         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
24480         the scratch if RELEASE_VIA_POP is false.
24481         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
24482         If we have to save a temporary register, decrement SIZE appropriately.
24483         Pass new arguments to release_scratch_register_on_entry.
24484         (ix86_adjust_stack_and_probe): Likewise.
24485         (ix86_emit_probe_stack_range): Pass new arguments to
24486         release_scratch_register_on_entry.
24488 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
24490         PR rtl-optimization/84157
24491         * combine.c (change_zero_ext): Use REG_P predicate in
24492         front of HARD_REGISTER_P predicate.
24494 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
24496         * config/avr/avr.c (avr_option_override): Move disabling of
24497         -fdelete-null-pointer-checks to...
24498         * common/config/avr/avr-common.c (avr_option_optimization_table):
24499         ...here.
24501 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24503         PR tree-optimization/81635
24504         * tree-data-ref.c (split_constant_offset_1): For types that
24505         wrap on overflow, try to use range info to prove that wrapping
24506         cannot occur.
24508 2018-02-01  Renlin Li  <renlin.li@arm.com>
24510         PR target/83370
24511         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
24512         TAILCALL_ADDR_REGS.
24513         (aarch64_register_move_cost): Likewise.
24514         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
24515         TAILCALL_ADDR_REGS.
24516         (REG_CLASS_NAMES): Likewise.
24517         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
24518         TAILCALL_ADDR_REGS. Remove IP registers.
24519         * config/aarch64/aarch64.md (Ucs): Update register constraint.
24521 2018-02-01  Richard Biener  <rguenther@suse.de>
24523         * domwalk.h (dom_walker::dom_walker): Add additional constructor
24524         for specifying RPO order and allow NULL for that.
24525         * domwalk.c (dom_walker::dom_walker): Likewise.
24526         (dom_walker::walk): Handle NULL RPO order.
24527         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
24528         in RPO order.
24529         (rewrite_update_dom_walker): Likewise.
24530         (mark_def_dom_walker): Likewise.
24532 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24534         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
24535         (aarch64_maybe_expand_sve_subreg_move): Declare.
24536         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
24537         * config/aarch64/predicates.md (aarch64_any_register_operand): New
24538         predicate.
24539         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
24540         that are semantically a reverse operation.
24541         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
24542         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
24543         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
24544         functions.
24545         (aarch64_can_change_mode_class): For big-endian, forbid changes
24546         between two SVE modes if they have different element sizes.
24548 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24550         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
24551         the TImode handling for big-endian targets.
24553 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24555         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
24556         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
24557         not just bytes.
24558         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
24559         Remove BSWAP handing for big-endian targets and use the form of
24560         LD1RQ appropariate for the mode.
24562 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24564         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
24565         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
24566         duplicated element.
24568 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24570         PR tearget/83845
24571         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
24572         check for operands that need to go through aarch64_sve_reload_be.
24574 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
24576         PR tree-optimization/81661
24577         PR tree-optimization/84117
24578         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
24579         * tree-eh.c: Include gimplify.h.
24580         (find_trapping_overflow, replace_trapping_overflow,
24581         rewrite_to_non_trapping_overflow): New functions.
24582         * tree-vect-loop.c: Include tree-eh.h.
24583         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
24584         * tree-data-ref.c: Include tree-eh.h.
24585         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
24587 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
24589         PR rtl-optimization/84123
24590         * combine.c (change_zero_ext): Check if hard register satisfies
24591         can_change_dest_mode before calling gen_lowpart_SUBREG.
24593 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
24595         PR target/82444
24596         * ira.c (ira_init_register_move_cost): Remove assert.
24598 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24600         PR rtl-optimization/84071
24601         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
24602         * doc/tm.texi: Regenerate.
24604 2018-01-31  Richard Biener  <rguenther@suse.de>
24606         PR tree-optimization/84132
24607         * tree-data-ref.c (analyze_miv_subscript): Properly
24608         check whether evolution_function_is_affine_multivariate_p
24609         before calling gcd_of_steps_may_divide_p.
24611 2018-01-31  Julia Koval  <julia.koval@intel.com>
24613         PR target/83618
24614         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
24615         * config/i386/i386.md (rdpid_rex64) New.
24616         (rdpid): Make 32bit only.
24618 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
24620         PR lto/84105
24621         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
24622         an IDENTIFIER_NODE for FUNCTION_TYPE's.
24624 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24626         Revert
24627         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24629         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24631 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24633         PR rtl-optimization/84071
24634         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24635         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24637 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24639         * config/arc/arc.c (arc_handle_aux_attribute): New function.
24640         (arc_attribute_table): Add 'aux' attribute.
24641         (arc_in_small_data_p): Consider aux like variables.
24642         (arc_is_aux_reg_p): New function.
24643         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24644         (arc_get_aux_arg): New function.
24645         (prepare_move_operands): Handle aux-register access.
24646         (arc_handle_aux_attribute): New function.
24647         * doc/extend.texi (ARC Variable attributes): Add subsection.
24649 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24651         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24652         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24653         (arc_attribute_table): Add 'uncached' attribute.
24654         (arc_print_operand): Print '.di' flag for uncached memory
24655         accesses.
24656         (arc_in_small_data_p): Do not consider for small data the uncached
24657         types.
24658         (arc_is_uncached_mem_p): New function.
24659         * config/arc/predicates.md (compact_store_memory_operand): Check
24660         for uncached memory accesses.
24661         (nonvol_nonimm_operand): Likewise.
24662         * doc/extend.texi (ARC Type Attribute): New subsection.
24664 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
24666         PR c/84100
24667         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24668         falign-loops=): Add Optimization flag.
24670 2018-01-30  Jeff Law  <law@redhat.com>
24672         PR target/84064
24673         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24674         INT_REGISTERS_SAVED.  Check it prior to calling
24675         get_scratch_register_on_entry.
24676         (ix86_adjust_stack_and_probe): Similarly.
24677         (ix86_emit_probe_stack_range): Similarly.
24678         (ix86_expand_prologue): Corresponding changes.
24680 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24682         PR target/40411
24683         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24684         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24686 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
24688         PR target/84112
24689         * lra-constraints.c (curr_insn_transform): Process AND in the
24690         address.
24692 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
24694         PR rtl-optimization/83986
24695         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24696         dependence against last_pending_memory_flush in addition to
24697         pending_jump_insns.
24699 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
24701         PR tree-optimization/81611
24702         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24703         copies.
24705 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24707         PR target/83758
24708         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24709         a reg rtx.
24711 2018-01-30  Richard Biener  <rguenther@suse.de>
24712             Jakub Jelinek  <jakub@redhat.com>
24714         PR tree-optimization/84111
24715         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24716         inner loops added during recursion, as they don't have up-to-date
24717         SSA form.
24719 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24721         PR ipa/81360
24722         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24723         (can_inline_edge_by_limits_p): ... here.
24724         (can_early_inline_edge_p, check_callers,
24725         update_caller_keys, update_callee_keys, recursive_inlining,
24726         add_new_edges_to_heap, speculation_useful_p,
24727         inline_small_functions,
24728         inline_small_functions, flatten_function,
24729         inline_to_all_callers_1): Update.
24731 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24733         * profile-count.c (profile_count::combine_with_ipa_count): Handle
24734         zeros correctly.
24736 2018-01-30  Richard Biener  <rguenther@suse.de>
24738         PR tree-optimization/83008
24739         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24740         invariant and constant vector uses in stmts when they need
24741         more than one stmt.
24743 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24745         PR bootstrap/84017
24746         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24747         * configure: Regenerate.
24749 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24751         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24752         pattern.
24753         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24754         Use gen_rtx_REG rather than gen_lowpart.
24756 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24758         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24759         rather than 0 when creating partial subregs.
24761 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24763         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24764         of usage.
24766 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
24768         PR target/81550
24769         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24770         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24771         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24772         flags.  This restores the settings used before the 2017-07-24.
24773         Turning off pre increment/decrement/modify allows IVOPTS to
24774         optimize DF/SF loops where the index is an int.
24776 2018-01-29  Richard Biener  <rguenther@suse.de>
24777             Kelvin Nilsen  <kelvin@gcc.gnu.org>
24779         PR bootstrap/80867
24780         * tree-vect-stmts.c (vectorizable_call): Don't call
24781         targetm.vectorize_builtin_md_vectorized_function if callee is
24782         NULL.
24784 2018-01-22  Carl Love  <cel@us.ibm.com>
24786         * doc/extend.tex: Fix typo in second arg in
24787         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24789 2018-01-29  Richard Biener  <rguenther@suse.de>
24791         PR tree-optimization/84086
24792         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24793         (flush_ssaname_freelist): When SSA names were released reset
24794         the SCEV hash table.
24796 2018-01-29  Richard Biener  <rguenther@suse.de>
24798         PR tree-optimization/84057
24799         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24800         removed paths when removing edges.
24802 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
24804         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24805         -mfunction-return=@var{choice}.
24807 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
24809         PR diagnostic/84034
24810         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24811         Handle CR like TAB.
24812         (layout::print_source_line): Likewise.
24813         (test_get_line_width_without_trailing_whitespace): Add test cases.
24815 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
24817         PR middle-end/84040
24818         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24819         debug insns.
24821 2018-01-26  Jim Wilson  <jimw@sifive.com>
24823         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24825         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24826         specified.
24828 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24830         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24831         and CMP + SUB-immediate -> SUBS.
24833 2018-01-26  Martin Sebor  <msebor@redhat.com>
24835         PR tree-optimization/83896
24836         * tree-ssa-strlen.c (get_string_len): Rename...
24837         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
24838         Avoid assuming length is constant.
24839         (handle_char_store): Use HOST_WIDE_INT for string length.
24841 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
24843         PR target/81763
24844         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24845         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24847 2018-01-26  Richard Biener  <rguenther@suse.de>
24849         PR rtl-optimization/84003
24850         * dse.c (record_store): Only record redundant stores when
24851         the earlier store aliases at least all accesses the later one does.
24853 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
24855         PR rtl-optimization/83985
24856         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24857         REG_CFA_RESTORE insns.
24858         (delete_unmarked_insns): Don't ignore separate shrink wrapping
24859         REG_CFA_RESTORE insns here.
24861         PR c/83989
24862         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24863         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24865 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24867         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24868         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24869         (arc_init): Likewise.
24870         (arc_override_options): Likewise.
24871         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24872         value.
24873         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24874         support.
24875         * config/arc/arc.h (TARGET_DBNZ): Define.
24876         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24877         properly set the tune attribute.
24878         (dbnz): Use TARGET_DBNZ guard.
24879         * config/arc/arc.opt (mtune): Add core3 option.
24881 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24883         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24884         recognize new pic like addresses.
24885         (arc_delegitimize_address): Clean up.
24887 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24889         * config/arc/arc-arches.def: Option mrf16 valid for all
24890         architectures.
24891         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24892         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24893         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24894         * config/arc/arc-tables.opt: Regenerate.
24895         * config/arc/arc.c (arc_conditional_register_usage): Handle
24896         reduced register file case.
24897         (arc_file_start): Set must have build attributes.
24898         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24899         mrf16 option value.
24900         * config/arc/arc.opt (mrf16): Add new option.
24901         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24902         * config/arc/genmultilib.awk: Handle new mrf16 option.
24903         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24904         * config/arc/t-multilib: Regenerate.
24905         * doc/invoke.texi (ARC Options): Document mrf16 option.
24907 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24909         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24910         * config/arc/arc.c (arc_handle_secure_attribute): New function.
24911         (arc_attribute_table): Add 'secure_call' attribute.
24912         (arc_print_operand): Print secure call operand.
24913         (arc_function_ok_for_sibcall): Don't optimize tail calls when
24914         secure.
24915         (arc_is_secure_call_p): New function.  * config/arc/arc.md
24916         (call_i): Add support for sjli instruction.
24917         (call_value_i): Likewise.
24918         * config/arc/constraints.md (Csc): New constraint.
24920 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24921             John Eric Martin  <John.Martin@emmicro-us.com>
24923         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24924         * config/arc/arc.c (_arc_jli_section): New struct.
24925         (arc_jli_section): New type.
24926         (rc_jli_sections): New static variable.
24927         (arc_handle_jli_attribute): New function.
24928         (arc_attribute_table): Add jli_always and jli_fixed attribute.
24929         (arc_file_end): New function.
24930         (TARGET_ASM_FILE_END): Define.
24931         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24932         (arc_add_jli_section): New function.
24933         (jli_call_scan): Likewise.
24934         (arc_reorg): Call jli_call_scan.
24935         (arc_output_addsi): Remove 'S' from printing asm operand.
24936         (arc_is_jli_call_p): New function.
24937         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24938         operand.
24939         (movhi_insn): Likewise.
24940         (movsi_insn): Likewise.
24941         (movsi_set_cc_insn): Likewise.
24942         (loadqi_update): Likewise.
24943         (load_zeroextendqisi_update): Likewise.
24944         (load_signextendqisi_update): Likewise.
24945         (loadhi_update): Likewise.
24946         (load_zeroextendhisi_update): Likewise.
24947         (load_signextendhisi_update): Likewise.
24948         (loadsi_update): Likewise.
24949         (loadsf_update): Likewise.
24950         (movsicc_insn): Likewise.
24951         (bset_insn): Likewise.
24952         (bxor_insn): Likewise.
24953         (bclr_insn): Likewise.
24954         (bmsk_insn): Likewise.
24955         (bicsi3_insn): Likewise.
24956         (cmpsi_cc_c_insn): Likewise.
24957         (movsi_ne): Likewise.
24958         (movsi_cond_exec): Likewise.
24959         (clrsbsi2): Likewise.
24960         (norm_f): Likewise.
24961         (normw): Likewise.
24962         (swap): Likewise.
24963         (divaw): Likewise.
24964         (flag): Likewise.
24965         (sr): Likewise.
24966         (kflag): Likewise.
24967         (ffs): Likewise.
24968         (ffs_f): Likewise.
24969         (fls): Likewise.
24970         (call_i): Remove 'S' asm letter, add jli instruction.
24971         (call_value_i): Likewise.
24972         * config/arc/arc.op (mjli-always): New option.
24973         * config/arc/constraints.md (Cji): New constraint.
24974         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24975         operand.
24976         (subsf3_fpx): Likewise.
24977         (mulsf3_fpx): Likewise.
24978         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24979         asm operand.
24980         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24981         function attrbutes.
24982         * doc/invoke.texi (ARC): Document mjli-always option.
24984 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
24986         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24987         avoid addition with 0 and use incw and decw where possible.
24989 2018-01-26  Richard Biener  <rguenther@suse.de>
24991         PR tree-optimization/81082
24992         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24993         association if it requires casting to unsigned.
24994         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24995         from fold_plusminus_mult_expr to catch important cases late when
24996         range info is available.
24998 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25000         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
25001         * configure.ac (hidden_linkonce): New test.
25002         * configure: Regenerate.
25003         * config.in: Regenerate.
25005 2018-01-26  Julia Koval  <julia.koval@intel.com>
25007         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
25008         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
25009         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
25010         _mm_mask_bitshuffle_epi64_mask): Fix type.
25011         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
25012         USI_FTYPE_V4DI_V4DI_USI): Remove.
25013         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
25014         __builtin_ia32_vpshufbitqmb256_mask,
25015         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
25016         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
25017         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
25019 2018-01-26  Alan Modra  <amodra@gmail.com>
25021         PR target/84033
25022         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
25023         UNSPEC_VBPERMQ.  Sort other unspecs.
25025 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
25027         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
25029 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
25031         PR middle-end/83055
25032         * predict.c (drop_profile): Do not push/pop cfun; update also
25033         node->count.
25034         (handle_missing_profiles): Fix logic looking for zero profiles.
25036 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
25038         PR middle-end/83977
25039         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
25040         on functions with #pragma omp declare simd or functions with simd
25041         attribute.
25042         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
25043         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
25044         Remove trailing \n from warning_at calls.
25046 2018-01-25  Tom de Vries  <tom@codesourcery.com>
25048         PR target/84028
25049         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
25050         for neutered workers.
25052 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
25054         PR target/68467
25055         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
25056         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
25058 2018-01-24  Jeff Law  <law@redhat.com>
25060         PR target/83994
25061         * i386.c (get_probe_interval): Move to earlier point.
25062         (ix86_compute_frame_layout): If -fstack-clash-protection and
25063         the frame is larger than the probe interval, then use pushes
25064         to save registers rather than reg->mem moves.
25065         (ix86_expand_prologue): Remove conditional for int_registers_saved
25066         assertion.
25068 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
25070         PR target/84014
25071         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
25072         min/max for never referenced object.
25074 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
25076         PR middle-end/83977
25077         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
25078         here.
25079         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
25080         attributes from DECL_ATTRIBUTES (decl) without affecting
25081         DECL_ATTRIBUTES (current_function_decl).
25082         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
25083         functions with non-NULL DECL_ABSTRACT_ORIGIN.
25085 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
25087         PR tree-optimization/83979
25088         * fold-const.c (fold_comparison): Use constant_boolean_node
25089         instead of boolean_{true,false}_node.
25091 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
25093         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
25094         with zero counts.
25096 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25098         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25099         Simplify the clause that sets the length attribute.
25100         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25101         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
25102         clause that sets the length attribute.
25103         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25105 2018-01-24  Tom de Vries  <tom@codesourcery.com>
25107         PR target/83589
25108         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
25109         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
25110         Add strict parameter.
25111         (prevent_branch_around_nothing): Insert dummy insn between branch to
25112         label and label with no ptx insn inbetween.
25113         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
25115 2018-01-24  Tom de Vries  <tom@codesourcery.com>
25117         PR target/81352
25118         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
25119         for neutered threads in warp.
25120         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
25122 2018-01-24  Richard Biener  <rguenther@suse.de>
25124         PR tree-optimization/83176
25125         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
25126         operands.
25128 2018-01-24  Richard Biener  <rguenther@suse.de>
25130         PR tree-optimization/82819
25131         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
25132         code generating pluses that are no-ops in the target precision.
25134 2018-01-24  Richard Biener  <rguenther@suse.de>
25136         PR middle-end/84000
25137         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
25139 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25141         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
25142         to merge probabilities.
25143         * predict.c (probably_never_executed): Also mark as cold functions
25144         with global 0 profile and guessed local profile.
25145         * profile-count.c (profile_probability::combine_with_count): New
25146         member function.
25147         * profile-count.h (profile_probability::operator*,
25148         profile_probability::operator*=, profile_probability::operator/,
25149         profile_probability::operator/=): Reduce precision to adjusted
25150         and set value to guessed on contradictory divisions.
25151         (profile_probability::combine_with_freq): Remove.
25152         (profile_probability::combine_wiht_count): Declare.
25153         (profile_count::force_nonzero):: Set to adjusted.
25154         (profile_count::probability_in):: Set quality to adjusted.
25155         * tree-ssa-tail-merge.c (replace_block_by): Use
25156         combine_with_count.
25158 2018-01-23  Andrew Waterman  <andrew@sifive.com>
25159             Jim Wilson  <jimw@sifive.com>
25161         * config/riscv/riscv.c (riscv_stack_boundary): New.
25162         (riscv_option_override): Set riscv_stack_boundary.  Handle
25163         riscv_preferred_stack_boundary_arg.
25164         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
25165         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
25166         (STACK_BOUNDARY): Set to riscv_stack_boundary.
25167         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
25168         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
25169         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
25171 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
25173         PR target/83905
25174         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
25175         of struct ix86_frame.
25176         (ix86_expand_epilogue): Likewise.  Add a local variable for
25177         the reg_save_offset field in struct ix86_frame.
25179 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
25181         PR tree-optimization/82604
25182         * tree-loop-distribution.c (enum partition_kind): New enum item
25183         PKIND_PARTIAL_MEMSET.
25184         (partition_builtin_p): Support above new enum item.
25185         (generate_code_for_partition): Ditto.
25186         (compute_access_range): Differentiate cases that equality can be
25187         proven at all loops, the innermost loops or no loops.
25188         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
25189         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
25190         (finalize_partitions, distribute_loop): Don't fuse partition of
25191         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
25192         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
25193         parloop is enabled.
25195 2018-01-23  Martin Liska  <mliska@suse.cz>
25197         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
25198         order to ignore the predictor.
25199         (PRED_POLYMORPHIC_CALL): Likewise.
25200         (PRED_RECURSIVE_CALL): Likewise.
25202 2018-01-23  Martin Liska  <mliska@suse.cz>
25204         * tree-profile.c (tree_profiling): Print function header to
25205         aware reader which function we are working on.
25206         * value-prof.c (gimple_find_values_to_profile): Do not print
25207         not interesting value histograms.
25209 2018-01-23  Martin Liska  <mliska@suse.cz>
25211         * profile-count.h (enum profile_quality): Add
25212         profile_uninitialized as the first value. Do not number values
25213         as they are zero based.
25214         (profile_count::verify): Update sanity check.
25215         (profile_probability::verify): Likewise.
25217 2018-01-23  Nathan Sidwell  <nathan@acm.org>
25219         * doc/invoke.texi (ffor-scope): Deprecate.
25221 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25223         PR tree-optimization/83510
25224         * domwalk.c (set_all_edges_as_executable): New function.
25225         (dom_walker::dom_walker): Convert bool param
25226         "skip_unreachable_blocks" to enum reachability.  Move setup of
25227         edge flags to set_all_edges_as_executable and only do it when
25228         reachability is REACHABLE_BLOCKS.
25229         * domwalk.h (enum dom_walker::reachability): New enum.
25230         (dom_walker::dom_walker): Convert bool param
25231         "skip_unreachable_blocks" to enum reachability.
25232         (set_all_edges_as_executable): New decl.
25233         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
25234         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
25235         "reachability".
25236         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
25237         but converting true to REACHABLE_BLOCKS.
25238         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
25239         * tree-vrp.c
25240         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
25241         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
25242         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
25243         REACHABLE_BLOCKS.
25244         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
25245         if check_all_array_refs will be called.
25247 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25249         * tree.c (selftest::test_location_wrappers): Add more test
25250         coverage.
25252 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25254         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
25255         (selftest::test_bit_in_range): Likewise.
25257 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25259         PR testsuite/83888
25260         * doc/sourcebuild.texi (vect_float): Say that the selector
25261         only describes the situation when -funsafe-math-optimizations is on.
25262         (vect_float_strict): Document.
25264 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25266         PR tree-optimization/83965
25267         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
25268         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
25269         instead of checking only for a reduction.
25270         (vect_recog_widen_sum_pattern): Likewise.
25272 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25274         * predict.c (probably_never_executed): Only use precise profile info.
25275         (compute_function_frequency): Skip after inlining hack since we now
25276         have quality checking.
25278 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25280         * profile-count.h (profile_probability::very_unlikely,
25281         profile_probability::unlikely, profile_probability::even): Set
25282         precision to guessed.
25284 2018-01-23  Richard Biener  <rguenther@suse.de>
25286         PR tree-optimization/83963
25287         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
25288         Properly terminate dominator walk when crossing the exit edge not
25289         when visiting its source block.
25291 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
25293         PR c++/83918
25294         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
25295         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
25297 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
25299         PR tree-optimization/83957
25300         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
25301         semicolon after for body surrounded by braces.
25303         PR tree-optimization/83081
25304         * profile-count.h (profile_probability::split): New method.
25305         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
25306         Use profile_probability::split.
25307         (do_compare_rtx_and_jump): Fix adjustment of probabilities
25308         when splitting a single conditional jump into 2.
25310 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
25312         PR tree-optimization/69452
25313         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
25314         decl.
25316 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25318         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
25319         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
25320         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
25322 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25324         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
25325         declaration.
25326         * config/rl78/rl78.md (movdi): New define_expand.
25327         * config/rl78/rl78.c (rl78_split_movdi): New function.
25329 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
25331         PR target/83862
25332         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
25333         no longer used.
25334         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
25335         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
25336         128-bit to produce an UNSPEC move to get the double word with the
25337         signbit and then a shift directly to do signbit.
25338         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
25339         implementation with a new version that just does either a direct
25340         move or a regular move.  Move memory interface to separate insns.
25341         Move insns so they are next to the expander.
25342         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
25343         with signbit move.  Split big and little endian case.
25344         (signbit<mode>2_dm_mem_le): Likewise.
25345         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
25346         (signbit<mode>2_dm2): Likewise.
25348 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25350         * config/rl78/rl78.md (anddi3): New define_expand.
25352 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25354         * config/rl78/rl78.md (umindi3): New define_expand.
25356 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25358         * config/rl78/rl78.md (smindi3): New define_expand.
25360 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25362         * config/rl78/rl78.md (smaxdi3): New define_expand.
25364 2018-01-22  Carl Love  <cel@us.ibm.com>
25366         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
25367         LVX_V1TI): Add macro expansion.
25368         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
25369         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
25370         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
25371         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
25372         Change check to determine if the instruction is a byte reversing
25373         entry.  Fix typo in comment.
25374         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
25375         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
25376         Add def_builtin calls for new builtins.
25377         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
25378         Add define_insn expansion.
25380 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25382         * config/rl78/rl78.md (umaxdi3): New define_expand.
25384 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25386         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
25387         for non-QImode registers.
25389 2018-01-22  Richard Biener  <rguenther@suse.de>
25391         PR tree-optimization/83963
25392         * graphite-scop-detection.c (scop_detection::get_sese): Delay
25393         including the loop exit block.
25394         (scop_detection::merge_sese): Likewise.
25395         (scop_detection::add_scop): Do it here instead.
25397 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25399         * doc/sourcebuild.texi (arm_softfloat): Document.
25401 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
25403         PR gcc/77734
25404         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
25405         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
25406         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
25408 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25409             David Edelsohn  <dje.gcc@gmail.com>
25411         PR target/83946
25412         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25413         Change "crset eq" to "crset 2".
25414         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25415         (*call_indirect_aix<mode>_nospec): Likewise.
25416         (*call_value_indirect_aix<mode>_nospec): Likewise.
25417         (*call_indirect_elfv2<mode>_nospec): Likewise.
25418         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
25419         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
25420         change assembly output from . to $.
25421         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25422         (indirect_jump<mode>_nospec): Change assembly output from . to $.
25423         (*tablejump<mode>_internal1_nospec): Likewise.
25425 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
25427         PR target/80870
25428         * config/sh/sh_optimize_sett_clrt.cc:
25429         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
25431 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
25433         PR tree-optimization/83940
25434         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
25435         offset_dt to vect_constant_def rather than vect_unknown_def_type.
25436         (vect_check_load_store_mask): Add a mask_dt_out parameter and
25437         use it to pass back the definition type.
25438         (vect_check_store_rhs): Likewise rhs_dt_out.
25439         (vect_build_gather_load_calls): Add a mask_dt argument and use
25440         it instead of a call to vect_is_simple_use.
25441         (vectorizable_store): Update calls to vect_check_load_store_mask
25442         and vect_check_store_rhs.  Use the dt returned by the latter instead
25443         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
25444         instead of calls to vect_is_simple_use.  Pass the scalar rather
25445         than the vector operand to vect_is_simple_use when handling
25446         second and subsequent copies of an rhs value.
25447         (vectorizable_load): Update calls to vect_check_load_store_mask
25448         and vect_build_gather_load_calls.  Use the cached mask_dt and
25449         gs_info.offset_dt instead of calls to vect_is_simple_use.
25451 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
25453         PR middle-end/83945
25454         * tree-emutls.c: Include gimplify.h.
25455         (lower_emutls_2): New function.
25456         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
25457         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
25458         it before further processing.
25460         PR target/83930
25461         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
25462         UINTVAL (trueop1) instead of INTVAL (op1).
25464 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
25466         PR debug/81570
25467         PR debug/83728
25468         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
25469         INCOMING_FRAME_SP_OFFSET if not defined.
25470         (scan_trace): Add ENTRY argument.  If true and
25471         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
25472         emit a note to adjust the CFA offset.
25473         (create_cfi_notes): Adjust scan_trace callers.
25474         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
25475         INCOMING_FRAME_SP_OFFSET in the CIE.
25476         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
25477         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
25478         Likewise.
25479         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
25480         * doc/tm.texi: Regenerated.
25482 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25484         PR rtl-optimization/83147
25485         * lra-constraints.c (remove_inheritance_pseudos): Use
25486         lra_substitute_pseudo_within_insn.
25488 2018-01-19  Tom de Vries  <tom@codesourcery.com>
25489             Cesar Philippidis  <cesar@codesourcery.com>
25491         PR target/83920
25492         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
25494 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
25496         PR target/83790
25497         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
25498         spaces for function labels.
25500 2018-01-19  Martin Liska  <mliska@suse.cz>
25502         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
25503         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
25504         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
25505         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
25506         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
25507         (PRED_CONST_RETURN): Change from 69 to 65.
25508         (PRED_NULL_RETURN): Change from 91 to 71.
25509         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
25510         (PRED_LOOP_GUARD): Change from 66 to 73.
25512 2018-01-19  Martin Liska  <mliska@suse.cz>
25514         * predict.c (predict_insn_def): Add new assert.
25515         (struct branch_predictor): Change type to signed integer.
25516         (test_prediction_value_range): Amend test to cover
25517         PROB_UNINITIALIZED.
25518         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
25519         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
25520         (PRED_LOOP_ITERATIONS_MAX): Likewise.
25521         (PRED_LOOP_IV_COMPARE): Likewise.
25522         * predict.h (PROB_UNINITIALIZED): Define new constant.
25524 2018-01-19  Martin Liska  <mliska@suse.cz>
25526         * predict.c (dump_prediction): Add new format for
25527         analyze_brprob.py script which is enabled with -details
25528         suboption.
25529         * profile-count.h (precise_p): New function.
25531 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25533         PR tree-optimization/83922
25534         * tree-vect-loop.c (vect_verify_full_masking): Return false if
25535         there are no statements that need masking.
25536         (vect_active_double_reduction_p): New function.
25537         (vect_analyze_loop_operations): Use it when handling phis that
25538         are not in the loop header.
25540 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25542         PR tree-optimization/83914
25543         * tree-vect-loop.c (vectorizable_induction): Don't convert
25544         init_expr or apply the peeling adjustment for inductions
25545         that are nested within the vectorized loop.
25547 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25549         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
25550         instead of NEG.
25552 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25554         PR sanitizer/81715
25555         PR testsuite/83882
25556         * function.h (gimplify_parameters): Add gimple_seq * argument.
25557         * function.c: Include gimple.h and options.h.
25558         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
25559         for the added local temporaries if needed.
25560         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
25561         if there are any parameter cleanups, wrap whole body into a
25562         try/finally with the cleanups.
25564 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
25566         PR target/82964
25567         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
25568         Use GET_MODE_CLASS for scalar floating point.
25570 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25572         PR ipa/82256
25573         patch by PaX Team
25574         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
25575         Fix call of call_cgraph_insertion_hooks.
25577 2018-01-18  Martin Sebor  <msebor@redhat.com>
25579         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
25581 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25583         PR ipa/83619
25584         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
25585         frequencies.
25587 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
25589         PR other/70268
25590         * common.opt: (-ffile-prefix-map): New option.
25591         * opts.c (common_handle_option): Defer it.
25592         * opts-global.c (handle_common_deferred_options): Handle it.
25593         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
25594         * file-prefix-map.h: New file.
25595         (remap_debug_filename, add_debug_prefix_map): ...here.
25596         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
25597         * final.c (debug_prefix_map, add_debug_prefix_map
25598         remap_debug_filename): Move to...
25599         * file-prefix-map.c: New file.
25600         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
25601         generalize, get rid of alloca(), use strrchr() instead of strchr().
25602         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
25603         Implement in terms of add_prefix_map().
25604         (remap_macro_filename, remap_debug_filename): Implement in term of
25605         remap_filename().
25606         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
25607         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
25608         * dbxout.c: Include file-prefix-map.h.
25609         * varasm.c: Likewise.
25610         * vmsdbgout.c: Likewise.
25611         * xcoffout.c: Likewise.
25612         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
25613         * doc/cppopts.texi (-fmacro-prefix-map): Document.
25614         * doc/invoke.texi (-ffile-prefix-map): Document.
25615         (-fdebug-prefix-map): Update description.
25617 2018-01-18  Martin Liska  <mliska@suse.cz>
25619         * config/i386/i386.c (indirect_thunk_name): Document that also
25620         lfence is emitted.
25621         (output_indirect_thunk): Document why both instructions
25622         (pause and lfence) are generated.
25624 2018-01-18  Richard Biener  <rguenther@suse.de>
25626         PR tree-optimization/83887
25627         * graphite-scop-detection.c
25628         (scop_detection::get_nearest_dom_with_single_entry): Remove.
25629         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25630         (scop_detection::merge_sese): Re-implement with a flood-fill
25631         algorithm that properly finds a SESE region if it exists.
25633 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25635         PR c/61240
25636         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25637         pointer_diff optimizations use view_convert instead of convert.
25639 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25641         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25642         Generate different code for -mno-speculate-indirect-jumps.
25643         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25644         (*call_indirect_aix<mode>): Disable for
25645         -mno-speculate-indirect-jumps.
25646         (*call_indirect_aix<mode>_nospec): New define_insn.
25647         (*call_value_indirect_aix<mode>): Disable for
25648         -mno-speculate-indirect-jumps.
25649         (*call_value_indirect_aix<mode>_nospec): New define_insn.
25650         (*sibcall_nonlocal_sysv<mode>): Generate different code for
25651         -mno-speculate-indirect-jumps.
25652         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25654 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
25656         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25657         long double type, set the flags for noting the default long double
25658         type, even if we don't pass or return a long double type.
25660 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
25662         PR ipa/83051
25663         * ipa-inline.c (flatten_function): Do not overwrite final inlining
25664         failure.
25666 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
25668         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25669         support for merge[hl].
25670         (fold_mergehl_helper): New helper function.
25671         (tree-vector-builder.h): New #include for tree_vector_builder usage.
25672         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25673         (altivec_vmrglw_direct): Add xxmrglw insn.
25675 2018-01-17  Andrew Waterman  <andrew@sifive.com>
25677         * config/riscv/riscv.c (riscv_conditional_register_usage): If
25678         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25680 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
25682         PR lto/83121
25683         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25684         call the lto_location_cache before reading the
25685         DECL_SOURCE_LOCATION of the types.
25687 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
25688             Richard Sandiford  <richard.sandiford@linaro.org>
25690         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25691         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25692         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25693         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25694         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25695         Add declaration.
25696         * config/aarch64/constraints.md (aarch64_movti_operand):
25697         Limit immediates.
25698         * config/aarch64/predicates.md (Uti): Add new constraint.
25700 2018-01-17  Carl Love  <cel@us.ibm.com>
25702         * config/rs6000/vsx.md (define_expand xl_len_r,
25703         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25704         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25705         lxvll.
25706         (define_expand, define_insn): Move the shift left from  the
25707         define_insn to the define_expand for lxvl and stxvl instructions.
25708         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25709         and XL_LEN_R definitions to PURE.
25711 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
25713         * config/i386/i386.c (indirect_thunk_name): Declare regno
25714         as unsigned int.  Compare regno with INVALID_REGNUM.
25715         (output_indirect_thunk): Ditto.
25716         (output_indirect_thunk_function): Ditto.
25717         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
25718         in the call to output_indirect_thunk_function.
25720 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
25722         PR middle-end/83884
25723         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25724         rather than the size of inner_type to determine the stack slot size
25725         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25727 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
25729         PR target/83546
25730         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25731         to PTA_SILVERMONT.
25733 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
25735         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25736         endian Linux systems to optionally enable multilibs for selecting
25737         the long double type if the user configured an explicit type.
25738         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25739         have no long double multilibs if not defined.
25740         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25741         warn if the user used -mabi={ieee,ibm}longdouble and we built
25742         multilibs for long double.
25743         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25744         appropriate multilib option.
25745         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25746         multilib options.
25747         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25748         for building long double multilibs.
25749         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25751 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
25753         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25754         copies.
25756         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25757         64 bits.
25758         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25759         128 bits.
25761         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25762         variables.
25764         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25765         return value.
25767 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25769         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25770         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25772 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25774         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25775         different rtl trees depending on TARGET_64BIT.
25776         (rs6000_gen_lvx): Likewise.
25778 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25780         * config/visium/visium.md (nop): Tweak comment.
25781         (hazard_nop): Likewise.
25783 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25785         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25786         -mspeculate-indirect-jumps.
25787         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25788         for -mno-speculate-indirect-jumps.
25789         (*call_indirect_elfv2<mode>_nospec): New define_insn.
25790         (*call_value_indirect_elfv2<mode>): Disable for
25791         -mno-speculate-indirect-jumps.
25792         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25793         (indirect_jump): Emit different RTL for
25794         -mno-speculate-indirect-jumps.
25795         (*indirect_jump<mode>): Disable for
25796         -mno-speculate-indirect-jumps.
25797         (*indirect_jump<mode>_nospec): New define_insn.
25798         (tablejump): Emit different RTL for
25799         -mno-speculate-indirect-jumps.
25800         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25801         (tablejumpsi_nospec): New define_expand.
25802         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25803         (tablejumpdi_nospec): New define_expand.
25804         (*tablejump<mode>_internal1): Disable for
25805         -mno-speculate-indirect-jumps.
25806         (*tablejump<mode>_internal1_nospec): New define_insn.
25807         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25808         option.
25810 2018-01-16  Artyom Skrobov tyomitch@gmail.com
25812         * caller-save.c (insert_save): Drop unnecessary parameter.  All
25813         callers updated.
25815 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25816             Richard Biener  <rguenth@suse.de>
25818         PR libgomp/83590
25819         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25820         return early, inline manually is_gimple_sizepos.  Make sure if we
25821         call gimplify_expr we don't end up with a gimple constant.
25822         * tree.c (variably_modified_type_p): Don't return true for
25823         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
25824         * gimplify.h (is_gimple_sizepos): Remove.
25826 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25828         PR tree-optimization/83857
25829         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25830         vectorizable_live_operation for pure SLP statements.
25831         (vectorizable_live_operation): Handle PHIs.
25833 2018-01-16  Richard Biener  <rguenther@suse.de>
25835         PR tree-optimization/83867
25836         * tree-vect-stmts.c (vect_transform_stmt): Precompute
25837         nested_in_vect_loop_p since the scalar stmt may get invalidated.
25839 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25841         PR c/83844
25842         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25843         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25844         If off is not INTEGER_CST, issue a may not be aligned warning
25845         rather than isn't aligned.  Use isn%'t rather than isn't.
25846         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25847         into MULT_EXPR.
25848         <case MULT_EXPR>: Improve the case when bottom and one of the
25849         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25850         operand, in that case check if the other operand is multiple of
25851         bottom divided by the INTEGER_CST operand.
25853 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25855         PR target/83858
25856         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25857         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25858         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25859         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25860         * config/pa/pa.c (pa_function_arg_advance): Likewise.
25861         (pa_function_arg, pa_arg_partial_bytes): Likewise.
25862         (pa_function_arg_size): New function.
25864 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25866         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25867         in a separate statement.
25869 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25871         PR tree-optimization/83847
25872         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25873         group gathers and scatters.
25875 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25877         PR rtl-optimization/86620
25878         * params.def (max-sched-ready-insns): Bump minimum value to 1.
25880         PR rtl-optimization/83213
25881         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25882         to last if both are JUMP_INSNs.
25884         PR tree-optimization/83843
25885         * gimple-ssa-store-merging.c
25886         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25887         store_immediate_info for bswap/nop orig_stores.
25889 2018-01-15  Andrew Waterman  <andrew@sifive.com>
25891         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25892         !TARGET_MUL.
25893         <UDIV>: Increase cost if !TARGET_DIV.
25895 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
25897         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25898         (define_attr "cr_logical_3op"): New.
25899         (cceq_ior_compare): Adjust.
25900         (cceq_ior_compare_complement): Adjust.
25901         (*cceq_rev_compare): Adjust.
25902         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25903         (is_cracked_insn): Adjust.
25904         (insn_must_be_first_in_group): Adjust.
25905         * config/rs6000/40x.md: Adjust.
25906         * config/rs6000/440.md: Adjust.
25907         * config/rs6000/476.md: Adjust.
25908         * config/rs6000/601.md: Adjust.
25909         * config/rs6000/603.md: Adjust.
25910         * config/rs6000/6xx.md: Adjust.
25911         * config/rs6000/7450.md: Adjust.
25912         * config/rs6000/7xx.md: Adjust.
25913         * config/rs6000/8540.md: Adjust.
25914         * config/rs6000/cell.md: Adjust.
25915         * config/rs6000/e300c2c3.md: Adjust.
25916         * config/rs6000/e500mc.md: Adjust.
25917         * config/rs6000/e500mc64.md: Adjust.
25918         * config/rs6000/e5500.md: Adjust.
25919         * config/rs6000/e6500.md: Adjust.
25920         * config/rs6000/mpc.md: Adjust.
25921         * config/rs6000/power4.md: Adjust.
25922         * config/rs6000/power5.md: Adjust.
25923         * config/rs6000/power6.md: Adjust.
25924         * config/rs6000/power7.md: Adjust.
25925         * config/rs6000/power8.md: Adjust.
25926         * config/rs6000/power9.md: Adjust.
25927         * config/rs6000/rs64.md: Adjust.
25928         * config/rs6000/titan.md: Adjust.
25930 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25932         * config/i386/predicates.md (indirect_branch_operand): Rewrite
25933         ix86_indirect_branch_register logic.
25935 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25937         * config/i386/constraints.md (Bs): Update
25938         ix86_indirect_branch_register check.  Don't check
25939         ix86_indirect_branch_register with GOT_memory_operand.
25940         (Bw): Likewise.
25941         * config/i386/predicates.md (GOT_memory_operand): Don't check
25942         ix86_indirect_branch_register here.
25943         (GOT32_symbol_operand): Likewise.
25945 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25947         * config/i386/predicates.md (constant_call_address_operand):
25948         Rewrite ix86_indirect_branch_register logic.
25949         (sibcall_insn_operand): Likewise.
25951 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25953         * config/i386/constraints.md (Bs): Replace
25954         ix86_indirect_branch_thunk_register with
25955         ix86_indirect_branch_register.
25956         (Bw): Likewise.
25957         * config/i386/i386.md (indirect_jump): Likewise.
25958         (tablejump): Likewise.
25959         (*sibcall_memory): Likewise.
25960         (*sibcall_value_memory): Likewise.
25961         Peepholes of indirect call and jump via memory: Likewise.
25962         * config/i386/i386.opt: Likewise.
25963         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25964         (GOT_memory_operand): Likewise.
25965         (call_insn_operand): Likewise.
25966         (sibcall_insn_operand): Likewise.
25967         (GOT32_symbol_operand): Likewise.
25969 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25971         PR middle-end/83837
25972         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25973         type rather than type addr's type points to.
25974         (expand_omp_atomic_mutex): Likewise.
25975         (expand_omp_atomic): Likewise.
25977 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25979         PR target/83839
25980         * config/i386/i386.c (output_indirect_thunk_function): Use
25981         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25982         for  __x86_return_thunk.
25984 2018-01-15  Richard Biener  <rguenther@suse.de>
25986         PR middle-end/83850
25987         * expmed.c (extract_bit_field_1): Fix typo.
25989 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25991         PR target/83687
25992         * config/arm/iterators.md (VF): New mode iterator.
25993         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25994         Remove integer-related logic from pattern.
25995         (neon_vabd<mode>_3): Likewise.
25997 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25999         PR middle-end/82694
26000         * common.opt (fstrict-overflow): No longer an alias.
26001         (fwrapv-pointer): New option.
26002         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
26003         also for pointer types based on flag_wrapv_pointer.
26004         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
26005         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
26006         opts->x_flag_wrapv got set.
26007         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
26008         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
26009         POINTER_TYPE_OVERFLOW_UNDEFINED.
26010         * match.pd: Likewise in address comparison pattern.
26011         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
26013 2018-01-15  Richard Biener  <rguenther@suse.de>
26015         PR lto/83804
26016         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
26017         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
26018         Reset type names to their identifier if their TYPE_DECL doesn't
26019         have linkage (and thus is used for ODR and devirt).
26020         (save_debug_info_for_decl): Remove.
26021         (save_debug_info_for_type): Likewise.
26022         (add_tree_to_fld_list): Adjust.
26023         * tree-pretty-print.c (dump_generic_node): Make dumping of
26024         type names more robust.
26026 2018-01-15  Richard Biener  <rguenther@suse.de>
26028         * BASE-VER: Bump to 8.0.1.
26030 2018-01-14  Martin Sebor  <msebor@redhat.com>
26032         PR other/83508
26033         * builtins.c (check_access): Avoid warning when the no-warning bit
26034         is set.
26036 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
26038         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
26039         * ira-color (allocno_hard_regs_compare): Likewise.
26041 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
26043         PR target/83013
26044         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
26045         Use .pushsection/.popsection.
26047 2018-01-14  Martin Sebor  <msebor@redhat.com>
26049         PR c++/81327
26050         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
26052 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
26054         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
26055         entry from extra_headers.
26056         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
26057         extra_headers, make the list bitwise identical to the i?86-*-* one.
26059 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26061         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
26062         -mcmodel=large with -mindirect-branch=thunk,
26063         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
26064         -mfunction-return=thunk-extern.
26065         * doc/invoke.texi: Document -mcmodel=large is incompatible with
26066         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
26067         -mfunction-return=thunk and -mfunction-return=thunk-extern.
26069 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26071         * config/i386/i386.c (print_reg): Print the name of the full
26072         integer register without '%'.
26073         (ix86_print_operand): Handle 'V'.
26074         * doc/extend.texi: Document 'V' modifier.
26076 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26078         * config/i386/constraints.md (Bs): Disallow memory operand for
26079         -mindirect-branch-register.
26080         (Bw): Likewise.
26081         * config/i386/predicates.md (indirect_branch_operand): Likewise.
26082         (GOT_memory_operand): Likewise.
26083         (call_insn_operand): Likewise.
26084         (sibcall_insn_operand): Likewise.
26085         (GOT32_symbol_operand): Likewise.
26086         * config/i386/i386.md (indirect_jump): Call convert_memory_address
26087         for -mindirect-branch-register.
26088         (tablejump): Likewise.
26089         (*sibcall_memory): Likewise.
26090         (*sibcall_value_memory): Likewise.
26091         Disallow peepholes of indirect call and jump via memory for
26092         -mindirect-branch-register.
26093         (*call_pop): Replace m with Bw.
26094         (*call_value_pop): Likewise.
26095         (*sibcall_pop_memory): Replace m with Bs.
26096         * config/i386/i386.opt (mindirect-branch-register): New option.
26097         * doc/invoke.texi: Document -mindirect-branch-register option.
26099 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26101         * config/i386/i386-protos.h (ix86_output_function_return): New.
26102         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
26103         set function_return_type.
26104         (indirect_thunk_name): Add ret_p to indicate thunk for function
26105         return.
26106         (output_indirect_thunk_function): Pass false to
26107         indirect_thunk_name.
26108         (ix86_output_indirect_branch_via_reg): Likewise.
26109         (ix86_output_indirect_branch_via_push): Likewise.
26110         (output_indirect_thunk_function): Create alias for function
26111         return thunk if regno < 0.
26112         (ix86_output_function_return): New function.
26113         (ix86_handle_fndecl_attribute): Handle function_return.
26114         (ix86_attribute_table): Add function_return.
26115         * config/i386/i386.h (machine_function): Add
26116         function_return_type.
26117         * config/i386/i386.md (simple_return_internal): Use
26118         ix86_output_function_return.
26119         (simple_return_internal_long): Likewise.
26120         * config/i386/i386.opt (mfunction-return=): New option.
26121         (indirect_branch): Mention -mfunction-return=.
26122         * doc/extend.texi: Document function_return function attribute.
26123         * doc/invoke.texi: Document -mfunction-return= option.
26125 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26127         * config/i386/i386-opts.h (indirect_branch): New.
26128         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
26129         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
26130         with local indirect jump when converting indirect call and jump.
26131         (ix86_set_indirect_branch_type): New.
26132         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
26133         (indirectlabelno): New.
26134         (indirect_thunk_needed): Likewise.
26135         (indirect_thunk_bnd_needed): Likewise.
26136         (indirect_thunks_used): Likewise.
26137         (indirect_thunks_bnd_used): Likewise.
26138         (INDIRECT_LABEL): Likewise.
26139         (indirect_thunk_name): Likewise.
26140         (output_indirect_thunk): Likewise.
26141         (output_indirect_thunk_function): Likewise.
26142         (ix86_output_indirect_branch_via_reg): Likewise.
26143         (ix86_output_indirect_branch_via_push): Likewise.
26144         (ix86_output_indirect_branch): Likewise.
26145         (ix86_output_indirect_jmp): Likewise.
26146         (ix86_code_end): Call output_indirect_thunk_function if needed.
26147         (ix86_output_call_insn): Call ix86_output_indirect_branch if
26148         needed.
26149         (ix86_handle_fndecl_attribute): Handle indirect_branch.
26150         (ix86_attribute_table): Add indirect_branch.
26151         * config/i386/i386.h (machine_function): Add indirect_branch_type
26152         and has_local_indirect_jump.
26153         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
26154         to true.
26155         (tablejump): Likewise.
26156         (*indirect_jump): Use ix86_output_indirect_jmp.
26157         (*tablejump_1): Likewise.
26158         (simple_return_indirect_internal): Likewise.
26159         * config/i386/i386.opt (mindirect-branch=): New option.
26160         (indirect_branch): New.
26161         (keep): Likewise.
26162         (thunk): Likewise.
26163         (thunk-inline): Likewise.
26164         (thunk-extern): Likewise.
26165         * doc/extend.texi: Document indirect_branch function attribute.
26166         * doc/invoke.texi: Document -mindirect-branch= option.
26168 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
26170         PR ipa/83051
26171         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
26173 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
26175         * ipa-inline.c (want_inline_small_function_p): Return false if
26176         inlining has already failed with CIF_FINAL_ERROR.
26177         (update_caller_keys): Call want_inline_small_function_p before
26178         can_inline_edge_p.
26179         (update_callee_keys): Likewise.
26181 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26183         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
26184         New function.
26185         (rs6000_quadword_masked_address_p): Likewise.
26186         (quad_aligned_load_p): Likewise.
26187         (quad_aligned_store_p): Likewise.
26188         (const_load_sequence_p): Add comment to describe the outer-most loop.
26189         (mimic_memory_attributes_and_flags): New function.
26190         (rs6000_gen_stvx): Likewise.
26191         (replace_swapped_aligned_store): Likewise.
26192         (rs6000_gen_lvx): Likewise.
26193         (replace_swapped_aligned_load): Likewise.
26194         (replace_swapped_load_constant): Capitalize argument name in
26195         comment describing this function.
26196         (rs6000_analyze_swaps): Add a third pass to search for vector loads
26197         and stores that access quad-word aligned addresses and replace
26198         with stvx or lvx instructions when appropriate.
26199         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
26200         New function prototype.
26201         (rs6000_quadword_masked_address_p): Likewise.
26202         (rs6000_gen_lvx): Likewise.
26203         (rs6000_gen_stvx): Likewise.
26204         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
26205         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
26206         when memory address is aligned.
26207         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
26208         this split to select lvx instruction when memory address is aligned.
26209         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
26210         instruction when memory address is aligned.
26211         (*vsx_le_perm_load_v16qi): Likewise.
26212         (four unnamed splitters): Modify to select the stvx instruction
26213         when memory is aligned.
26215 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
26217         * predict.c (determine_unlikely_bbs): Handle correctly BBs
26218         which appears in the queue multiple times.
26220 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26221             Alan Hayward  <alan.hayward@arm.com>
26222             David Sherwood  <david.sherwood@arm.com>
26224         * tree-vectorizer.h (vec_lower_bound): New structure.
26225         (_loop_vec_info): Add check_nonzero and lower_bounds.
26226         (LOOP_VINFO_CHECK_NONZERO): New macro.
26227         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
26228         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
26229         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
26230         fields.  Make seg_len the distance travelled, not including the
26231         access size.
26232         (dr_direction_indicator): Declare.
26233         (dr_zero_step_indicator): Likewise.
26234         (dr_known_forward_stride_p): Likewise.
26235         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
26236         tree-ssanames.h.
26237         (runtime_alias_check_p): Allow runtime alias checks with
26238         variable strides.
26239         (operator ==): Compare access_size and align.
26240         (prune_runtime_alias_test_list): Rework for new distinction between
26241         the access_size and seg_len.
26242         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
26243         segment lengths.
26244         (get_segment_min_max): New function.
26245         (create_intersect_range_checks): Use it.
26246         (dr_step_indicator): New function.
26247         (dr_direction_indicator): Likewise.
26248         (dr_zero_step_indicator): Likewise.
26249         (dr_known_forward_stride_p): Likewise.
26250         * tree-loop-distribution.c (data_ref_segment_size): Return
26251         DR_STEP * (niters - 1).
26252         (compute_alias_check_pairs): Update call to the dr_with_seg_len
26253         constructor.
26254         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
26255         (vect_preserves_scalar_order_p): New function, split out from...
26256         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
26257         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
26258         (vect_vfa_access_size): New function.
26259         (vect_vfa_align): Likewise.
26260         (vect_compile_time_alias): Take access_size_a and access_b arguments.
26261         (dump_lower_bound): New function.
26262         (vect_check_lower_bound): Likewise.
26263         (vect_small_gap_p): Likewise.
26264         (vectorizable_with_step_bound_p): Likewise.
26265         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
26266         depencies if the vectorization factor is 1.  Convert the checks
26267         for nonzero steps into checks on the bounds of DR_STEP.  Try using
26268         a bunds check for variable steps if the minimum required step is
26269         relatively small. Update calls to the dr_with_seg_len
26270         constructor and to vect_compile_time_alias.
26271         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
26272         function.
26273         (vect_loop_versioning): Call it.
26274         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
26275         when retrying.
26276         (vect_estimate_min_profitable_iters): Account for any bounds checks.
26278 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26279             Alan Hayward  <alan.hayward@arm.com>
26280             David Sherwood  <david.sherwood@arm.com>
26282         * doc/sourcebuild.texi (vect_scatter_store): Document.
26283         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
26284         optabs.
26285         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
26286         Document.
26287         * genopinit.c (main): Add supports_vec_scatter_store and
26288         supports_vec_scatter_store_cached to target_optabs.
26289         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
26290         IFN_MASK_SCATTER_STORE.
26291         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
26292         functions.
26293         * internal-fn.h (internal_store_fn_p): Declare.
26294         (internal_fn_stored_value_index): Likewise.
26295         * internal-fn.c (scatter_store_direct): New macro.
26296         (expand_scatter_store_optab_fn): New function.
26297         (direct_scatter_store_optab_supported_p): New macro.
26298         (internal_store_fn_p): New function.
26299         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
26300         IFN_MASK_SCATTER_STORE.
26301         (internal_fn_mask_index): Likewise.
26302         (internal_fn_stored_value_index): New function.
26303         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
26304         for scatter stores.
26305         * optabs-query.h (supports_vec_scatter_store_p): Declare.
26306         * optabs-query.c (supports_vec_scatter_store_p): New function.
26307         * tree-vectorizer.h (vect_get_store_rhs): Declare.
26308         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
26309         true for scatter stores.
26310         (vect_gather_scatter_fn_p): Handle scatter stores too.
26311         (vect_check_gather_scatter): Consider using scatter stores if
26312         supports_vec_scatter_store_p.
26313         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
26314         scatter stores too.
26315         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26316         internal_fn_stored_value_index.
26317         (check_load_store_masking): Handle scatter stores too.
26318         (vect_get_store_rhs): Make public.
26319         (vectorizable_call): Use internal_store_fn_p.
26320         (vectorizable_store): Handle scatter store internal functions.
26321         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
26322         when deciding whether the end of the group has been reached.
26323         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
26324         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
26325         (mask_scatter_store<mode>): New insns.
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         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
26332         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
26333         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
26334         function.
26335         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
26336         Use vect_truncate_gather_scatter_offset if we can't treat the
26337         operation as a normal gather load or scatter store.
26338         (get_group_load_store_type): Take the gather_scatter_info
26339         as argument.  Try using a gather load or scatter store for
26340         single-element groups.
26341         (get_load_store_type): Update calls to get_group_load_store_type
26342         and vect_use_strided_gather_scatters_p.
26344 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26345             Alan Hayward  <alan.hayward@arm.com>
26346             David Sherwood  <david.sherwood@arm.com>
26348         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
26349         optional tree argument.
26350         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
26351         null target hooks.
26352         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
26353         but continue to use the current value as a fallback.
26354         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
26355         to compare the updates.
26356         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
26357         (get_load_store_type): Use it when handling a strided access.
26358         (vect_get_strided_load_store_ops): New function.
26359         (vect_get_data_ptr_increment): Likewise.
26360         (vectorizable_load): Handle strided gather loads.  Always pass
26361         a step to vect_create_data_ref_ptr and bump_vector_ptr.
26363 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26364             Alan Hayward  <alan.hayward@arm.com>
26365             David Sherwood  <david.sherwood@arm.com>
26367         * doc/md.texi (gather_load@var{m}): Document.
26368         (mask_gather_load@var{m}): Likewise.
26369         * genopinit.c (main): Add supports_vec_gather_load and
26370         supports_vec_gather_load_cached to target_optabs.
26371         * optabs-tree.c (init_tree_optimization_optabs): Use
26372         ggc_cleared_alloc to allocate target_optabs.
26373         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
26374         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
26375         functions.
26376         * internal-fn.h (internal_load_fn_p): Declare.
26377         (internal_gather_scatter_fn_p): Likewise.
26378         (internal_fn_mask_index): Likewise.
26379         (internal_gather_scatter_fn_supported_p): Likewise.
26380         * internal-fn.c (gather_load_direct): New macro.
26381         (expand_gather_load_optab_fn): New function.
26382         (direct_gather_load_optab_supported_p): New macro.
26383         (direct_internal_fn_optab): New function.
26384         (internal_load_fn_p): Likewise.
26385         (internal_gather_scatter_fn_p): Likewise.
26386         (internal_fn_mask_index): Likewise.
26387         (internal_gather_scatter_fn_supported_p): Likewise.
26388         * optabs-query.c (supports_at_least_one_mode_p): New function.
26389         (supports_vec_gather_load_p): Likewise.
26390         * optabs-query.h (supports_vec_gather_load_p): Declare.
26391         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
26392         and memory_type field.
26393         (NUM_PATTERNS): Bump to 15.
26394         * tree-vect-data-refs.c: Include internal-fn.h.
26395         (vect_gather_scatter_fn_p): New function.
26396         (vect_describe_gather_scatter_call): Likewise.
26397         (vect_check_gather_scatter): Try using internal functions for
26398         gather loads.  Recognize existing calls to a gather load function.
26399         (vect_analyze_data_refs): Consider using gather loads if
26400         supports_vec_gather_load_p.
26401         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
26402         (vect_get_gather_scatter_offset_type): Likewise.
26403         (vect_convert_mask_for_vectype): Likewise.
26404         (vect_add_conversion_to_patterm): Likewise.
26405         (vect_try_gather_scatter_pattern): Likewise.
26406         (vect_recog_gather_scatter_pattern): New pattern recognizer.
26407         (vect_vect_recog_func_ptrs): Add it.
26408         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26409         internal_fn_mask_index and internal_gather_scatter_fn_p.
26410         (check_load_store_masking): Take the gather_scatter_info as an
26411         argument and handle gather loads.
26412         (vect_get_gather_scatter_ops): New function.
26413         (vectorizable_call): Check internal_load_fn_p.
26414         (vectorizable_load): Likewise.  Handle gather load internal
26415         functions.
26416         (vectorizable_store): Update call to check_load_store_masking.
26417         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
26418         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
26419         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
26420         (aarch64_gather_scale_operand_d): New predicates.
26421         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
26422         (mask_gather_load<mode>): New insns.
26424 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26425             Alan Hayward  <alan.hayward@arm.com>
26426             David Sherwood  <david.sherwood@arm.com>
26428         * optabs.def (fold_left_plus_optab): New optab.
26429         * doc/md.texi (fold_left_plus_@var{m}): Document.
26430         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
26431         * internal-fn.c (fold_left_direct): Define.
26432         (expand_fold_left_optab_fn): Likewise.
26433         (direct_fold_left_optab_supported_p): Likewise.
26434         * fold-const-call.c (fold_const_fold_left): New function.
26435         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
26436         * tree-parloops.c (valid_reduction_p): New function.
26437         (gather_scalar_reductions): Use it.
26438         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
26439         (vect_finish_replace_stmt): Declare.
26440         * tree-vect-loop.c (fold_left_reduction_fn): New function.
26441         (needs_fold_left_reduction_p): New function, split out from...
26442         (vect_is_simple_reduction): ...here.  Accept reductions that
26443         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
26444         (vect_force_simple_reduction): Also store the reduction type in
26445         the assignment's STMT_VINFO_REDUC_TYPE.
26446         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
26447         (merge_with_identity): New function.
26448         (vect_expand_fold_left): Likewise.
26449         (vectorize_fold_left_reduction): Likewise.
26450         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
26451         scalar phi in place for it.  Check for target support and reject
26452         cases that would reassociate the operation.  Defer the transform
26453         phase to vectorize_fold_left_reduction.
26454         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
26455         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
26456         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
26458 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26460         * tree-if-conv.c (predicate_mem_writes): Remove redundant
26461         call to ifc_temp_var.
26463 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26464             Alan Hayward  <alan.hayward@arm.com>
26465             David Sherwood  <david.sherwood@arm.com>
26467         * target.def (legitimize_address_displacement): Take the original
26468         offset as a poly_int.
26469         * targhooks.h (default_legitimize_address_displacement): Update
26470         accordingly.
26471         * targhooks.c (default_legitimize_address_displacement): Likewise.
26472         * doc/tm.texi: Regenerate.
26473         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
26474         as an argument, moving assert of ad->disp == ad->disp_term to...
26475         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
26476         Try calling targetm.legitimize_address_displacement before expanding
26477         the address rather than afterwards, and adjust for the new interface.
26478         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
26479         Match the new hook interface.  Handle SVE addresses.
26480         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
26481         new hook interface.
26483 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26485         * Makefile.in (OBJS): Add early-remat.o.
26486         * target.def (select_early_remat_modes): New hook.
26487         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
26488         * doc/tm.texi: Regenerate.
26489         * targhooks.h (default_select_early_remat_modes): Declare.
26490         * targhooks.c (default_select_early_remat_modes): New function.
26491         * timevar.def (TV_EARLY_REMAT): New timevar.
26492         * passes.def (pass_early_remat): New pass.
26493         * tree-pass.h (make_pass_early_remat): Declare.
26494         * early-remat.c: New file.
26495         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
26496         function.
26497         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
26499 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26500             Alan Hayward  <alan.hayward@arm.com>
26501             David Sherwood  <david.sherwood@arm.com>
26503         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
26504         vfm1 with a bound_epilog parameter.
26505         (vect_do_peeling): Update calls accordingly, and move the prologue
26506         call earlier in the function.  Treat the base bound_epilog as 0 for
26507         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
26508         this base when peeling for gaps.
26509         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
26510         with fully-masked loops.
26511         (vect_estimate_min_profitable_iters): Handle the single peeled
26512         iteration in that case.
26514 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26515             Alan Hayward  <alan.hayward@arm.com>
26516             David Sherwood  <david.sherwood@arm.com>
26518         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
26519         single-element interleaving even if the size is not a power of 2.
26520         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
26521         accesses for single-element interleaving if the group size is
26522         not a power of 2.
26524 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26525             Alan Hayward  <alan.hayward@arm.com>
26526             David Sherwood  <david.sherwood@arm.com>
26528         * doc/md.texi (fold_extract_last_@var{m}): Document.
26529         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
26530         * optabs.def (fold_extract_last_optab): New optab.
26531         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
26532         * internal-fn.c (fold_extract_direct): New macro.
26533         (expand_fold_extract_optab_fn): Likewise.
26534         (direct_fold_extract_optab_supported_p): Likewise.
26535         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
26536         * tree-vect-loop.c (vect_model_reduction_cost): Handle
26537         EXTRACT_LAST_REDUCTION.
26538         (get_initial_def_for_reduction): Do not create an initial vector
26539         for EXTRACT_LAST_REDUCTION reductions.
26540         (vectorizable_reduction): Leave the scalar phi in place for
26541         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
26542         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
26543         epilogue code for EXTRACT_LAST_REDUCTION and defer the
26544         transform phase to vectorizable_condition.
26545         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
26546         split out from...
26547         (vect_finish_stmt_generation): ...here.
26548         (vect_finish_replace_stmt): New function.
26549         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
26550         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
26551         pattern.
26552         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
26554 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26555             Alan Hayward  <alan.hayward@arm.com>
26556             David Sherwood  <david.sherwood@arm.com>
26558         * doc/md.texi (extract_last_@var{m}): Document.
26559         * optabs.def (extract_last_optab): New optab.
26560         * internal-fn.def (EXTRACT_LAST): New internal function.
26561         * internal-fn.c (cond_unary_direct): New macro.
26562         (expand_cond_unary_optab_fn): Likewise.
26563         (direct_cond_unary_optab_supported_p): Likewise.
26564         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
26565         loops using EXTRACT_LAST.
26566         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
26567         (extract_last_<mode>): ...this optab.
26568         (vec_extract<mode><Vel>): Update accordingly.
26570 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26571             Alan Hayward  <alan.hayward@arm.com>
26572             David Sherwood  <david.sherwood@arm.com>
26574         * target.def (empty_mask_is_expensive): New hook.
26575         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
26576         * doc/tm.texi: Regenerate.
26577         * targhooks.h (default_empty_mask_is_expensive): Declare.
26578         * targhooks.c (default_empty_mask_is_expensive): New function.
26579         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
26580         if the target says that empty masks are expensive.
26581         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
26582         New function.
26583         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
26585 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26586             Alan Hayward  <alan.hayward@arm.com>
26587             David Sherwood  <david.sherwood@arm.com>
26589         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
26590         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
26591         (vect_use_loop_mask_for_alignment_p): New function.
26592         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
26593         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
26594         niters_skip argument.  Make sure that the first niters_skip elements
26595         of the first iteration are inactive.
26596         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
26597         Update call to vect_set_loop_masks_directly.
26598         (get_misalign_in_elems): New function, split out from...
26599         (vect_gen_prolog_loop_niters): ...here.
26600         (vect_update_init_of_dr): Take a code argument that specifies whether
26601         the adjustment should be added or subtracted.
26602         (vect_update_init_of_drs): Likewise.
26603         (vect_prepare_for_masked_peels): New function.
26604         (vect_do_peeling): Skip prologue peeling if we're using a mask
26605         instead.  Update call to vect_update_inits_of_drs.
26606         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26607         mask_skip_niters.
26608         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
26609         alignment.  Do not include the number of peeled iterations in
26610         the minimum threshold in that case.
26611         (vectorizable_induction): Adjust the start value down by
26612         LOOP_VINFO_MASK_SKIP_NITERS iterations.
26613         (vect_transform_loop): Call vect_prepare_for_masked_peels.
26614         Take the number of skipped iterations into account when calculating
26615         the loop bounds.
26616         * tree-vect-stmts.c (vect_gen_while_not): New function.
26618 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26619             Alan Hayward  <alan.hayward@arm.com>
26620             David Sherwood  <david.sherwood@arm.com>
26622         * doc/sourcebuild.texi (vect_fully_masked): Document.
26623         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26624         default value to 0.
26625         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26626         split out from...
26627         (vect_analyze_loop_2): ...here. Don't check the vectorization
26628         factor against the number of loop iterations if the loop is
26629         fully-masked.
26631 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26632             Alan Hayward  <alan.hayward@arm.com>
26633             David Sherwood  <david.sherwood@arm.com>
26635         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26636         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26637         (dump_groups): Update accordingly.
26638         (iv_use::mem_type): New member variable.
26639         (address_p): New function.
26640         (record_use): Add a mem_type argument and initialize the new
26641         mem_type field.
26642         (record_group_use): Add a mem_type argument.  Use address_p.
26643         Remove obsolete null checks of base_object.  Update call to record_use.
26644         (find_interesting_uses_op): Update call to record_group_use.
26645         (find_interesting_uses_cond): Likewise.
26646         (find_interesting_uses_address): Likewise.
26647         (get_mem_type_for_internal_fn): New function.
26648         (find_address_like_use): Likewise.
26649         (find_interesting_uses_stmt): Try find_address_like_use before
26650         calling find_interesting_uses_op.
26651         (addr_offset_valid_p): Use the iv mem_type field as the type
26652         of the addressed memory.
26653         (add_autoinc_candidates): Likewise.
26654         (get_address_cost): Likewise.
26655         (split_small_address_groups_p): Use address_p.
26656         (split_address_groups): Likewise.
26657         (add_iv_candidate_for_use): Likewise.
26658         (autoinc_possible_for_pair): Likewise.
26659         (rewrite_groups): Likewise.
26660         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26661         (determine_group_iv_cost): Update after split of USE_ADDRESS.
26662         (get_alias_ptr_type_for_ptr_address): New function.
26663         (rewrite_use_address): Rewrite address uses in calls that were
26664         identified by find_address_like_use.
26666 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26667             Alan Hayward  <alan.hayward@arm.com>
26668             David Sherwood  <david.sherwood@arm.com>
26670         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26671         TARGET_MEM_REFs.
26672         * gimple-expr.h (is_gimple_addressable: Likewise.
26673         * gimple-expr.c (is_gimple_address): Likewise.
26674         * internal-fn.c (expand_call_mem_ref): New function.
26675         (expand_mask_load_optab_fn): Use it.
26676         (expand_mask_store_optab_fn): Likewise.
26678 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26679             Alan Hayward  <alan.hayward@arm.com>
26680             David Sherwood  <david.sherwood@arm.com>
26682         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26683         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26684         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26685         (cond_umax@var{mode}): Document.
26686         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26687         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26688         (cond_umin_optab, cond_umax_optab): New optabs.
26689         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26690         (COND_IOR, COND_XOR): New internal functions.
26691         * internal-fn.h (get_conditional_internal_fn): Declare.
26692         * internal-fn.c (cond_binary_direct): New macro.
26693         (expand_cond_binary_optab_fn): Likewise.
26694         (direct_cond_binary_optab_supported_p): Likewise.
26695         (get_conditional_internal_fn): New function.
26696         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26697         Cope with reduction statements that are vectorized as calls rather
26698         than assignments.
26699         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26700         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26701         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26702         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26703         (UNSPEC_COND_EOR): New unspecs.
26704         (optab): Add mappings for them.
26705         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26706         (sve_int_op, sve_fp_op): New int attributes.
26708 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26709             Alan Hayward  <alan.hayward@arm.com>
26710             David Sherwood  <david.sherwood@arm.com>
26712         * optabs.def (while_ult_optab): New optab.
26713         * doc/md.texi (while_ult@var{m}@var{n}): Document.
26714         * internal-fn.def (WHILE_ULT): New internal function.
26715         * internal-fn.h (direct_internal_fn_supported_p): New override
26716         that takes two types as argument.
26717         * internal-fn.c (while_direct): New macro.
26718         (expand_while_optab_fn): New function.
26719         (convert_optab_supported_p): Likewise.
26720         (direct_while_optab_supported_p): New macro.
26721         * wide-int.h (wi::udiv_ceil): New function.
26722         * tree-vectorizer.h (rgroup_masks): New structure.
26723         (vec_loop_masks): New typedef.
26724         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26725         and fully_masked_p.
26726         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26727         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26728         (vect_max_vf): New function.
26729         (slpeel_make_loop_iterate_ntimes): Delete.
26730         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26731         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26732         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26733         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26734         internal-fn.h, stor-layout.h and optabs-query.h.
26735         (vect_set_loop_mask): New function.
26736         (add_preheader_seq): Likewise.
26737         (add_header_seq): Likewise.
26738         (interleave_supported_p): Likewise.
26739         (vect_maybe_permute_loop_masks): Likewise.
26740         (vect_set_loop_masks_directly): Likewise.
26741         (vect_set_loop_condition_masked): Likewise.
26742         (vect_set_loop_condition_unmasked): New function, split out from
26743         slpeel_make_loop_iterate_ntimes.
26744         (slpeel_make_loop_iterate_ntimes): Rename to..
26745         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
26746         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26747         (vect_do_peeling): Update call accordingly.
26748         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26749         loops.
26750         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26751         mask_compare_type, can_fully_mask_p and fully_masked_p.
26752         (release_vec_loop_masks): New function.
26753         (_loop_vec_info): Use it to free the loop masks.
26754         (can_produce_all_loop_masks_p): New function.
26755         (vect_get_max_nscalars_per_iter): Likewise.
26756         (vect_verify_full_masking): Likewise.
26757         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26758         retries, and free the mask rgroups before retrying.  Check loop-wide
26759         reasons for disallowing fully-masked loops.  Make the final decision
26760         about whether use a fully-masked loop or not.
26761         (vect_estimate_min_profitable_iters): Do not assume that peeling
26762         for the number of iterations will be needed for fully-masked loops.
26763         (vectorizable_reduction): Disable fully-masked loops.
26764         (vectorizable_live_operation): Likewise.
26765         (vect_halve_mask_nunits): New function.
26766         (vect_double_mask_nunits): Likewise.
26767         (vect_record_loop_mask): Likewise.
26768         (vect_get_loop_mask): Likewise.
26769         (vect_transform_loop): Handle the case in which the final loop
26770         iteration might handle a partial vector.  Call vect_set_loop_condition
26771         instead of slpeel_make_loop_iterate_ntimes.
26772         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26773         (check_load_store_masking): New function.
26774         (prepare_load_store_mask): Likewise.
26775         (vectorizable_store): Handle fully-masked loops.
26776         (vectorizable_load): Likewise.
26777         (supportable_widening_operation): Use vect_halve_mask_nunits for
26778         booleans.
26779         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26780         (vect_gen_while): New function.
26781         * config/aarch64/aarch64.md (umax<mode>3): New expander.
26782         (aarch64_uqdec<mode>): New insn.
26784 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26785             Alan Hayward  <alan.hayward@arm.com>
26786             David Sherwood  <david.sherwood@arm.com>
26788         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26789         (reduc_xor_scal_optab): New optabs.
26790         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26791         (reduc_xor_scal_@var{m}): Document.
26792         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26793         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26794         internal functions.
26795         * fold-const-call.c (fold_const_call): Handle them.
26796         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26797         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26798         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26799         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26800         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26801         (UNSPEC_XORV): New unspecs.
26802         (optab): Add entries for them.
26803         (BITWISEV): New int iterator.
26804         (bit_reduc_op): New int attributes.
26806 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26807             Alan Hayward  <alan.hayward@arm.com>
26808             David Sherwood  <david.sherwood@arm.com>
26810         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26811         * internal-fn.def (VEC_SHL_INSERT): New internal function.
26812         * optabs.def (vec_shl_insert_optab): New optab.
26813         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26814         (duplicate_and_interleave): Likewise.
26815         * tree-vect-loop.c: Include internal-fn.h.
26816         (neutral_op_for_slp_reduction): New function, split out from
26817         get_initial_defs_for_reduction.
26818         (get_initial_def_for_reduction): Handle option 2 for variable-length
26819         vectors by loading the neutral value into a vector and then shifting
26820         the initial value into element 0.
26821         (get_initial_defs_for_reduction): Replace the code argument with
26822         the neutral value calculated by neutral_op_for_slp_reduction.
26823         Use gimple_build_vector for constant-length vectors.
26824         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26825         but the first group_size elements have a neutral value.
26826         Use duplicate_and_interleave otherwise.
26827         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26828         Update call to get_initial_defs_for_reduction.  Handle SLP
26829         reductions for variable-length vectors by creating one vector
26830         result for each scalar result, with the elements associated
26831         with other scalar results stubbed out with the neutral value.
26832         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26833         Require IFN_VEC_SHL_INSERT for double reductions on
26834         variable-length vectors, or SLP reductions that have
26835         a neutral value.  Require can_duplicate_and_interleave_p
26836         support for variable-length unchained SLP reductions if there
26837         is no neutral value, such as for MIN/MAX reductions.  Also require
26838         the number of vector elements to be a multiple of the number of
26839         SLP statements when doing variable-length unchained SLP reductions.
26840         Update call to vect_create_epilog_for_reduction.
26841         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26842         and remove initial values.
26843         (duplicate_and_interleave): Make public.
26844         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26845         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26847 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26848             Alan Hayward  <alan.hayward@arm.com>
26849             David Sherwood  <david.sherwood@arm.com>
26851         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26852         (can_duplicate_and_interleave_p): New function.
26853         (vect_get_and_check_slp_defs): Take the vector of statements
26854         rather than just the current one.  Remove excess parentheses.
26855         Restriction rejectinon of vect_constant_def and vect_external_def
26856         for variable-length vectors to boolean types, or types for which
26857         can_duplicate_and_interleave_p is false.
26858         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26859         (duplicate_and_interleave): New function.
26860         (vect_get_constant_vectors): Use gimple_build_vector for
26861         constant-length vectors and suitable variable-length constant
26862         vectors.  Use duplicate_and_interleave for other variable-length
26863         vectors.  Don't defer the update when inserting new statements.
26865 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26866             Alan Hayward  <alan.hayward@arm.com>
26867             David Sherwood  <david.sherwood@arm.com>
26869         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26870         min_profitable_iters doesn't go negative.
26872 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26873             Alan Hayward  <alan.hayward@arm.com>
26874             David Sherwood  <david.sherwood@arm.com>
26876         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26877         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26878         * optabs.def (vec_mask_load_lanes_optab): New optab.
26879         (vec_mask_store_lanes_optab): Likewise.
26880         * internal-fn.def (MASK_LOAD_LANES): New internal function.
26881         (MASK_STORE_LANES): Likewise.
26882         * internal-fn.c (mask_load_lanes_direct): New macro.
26883         (mask_store_lanes_direct): Likewise.
26884         (expand_mask_load_optab_fn): Handle masked operations.
26885         (expand_mask_load_lanes_optab_fn): New macro.
26886         (expand_mask_store_optab_fn): Handle masked operations.
26887         (expand_mask_store_lanes_optab_fn): New macro.
26888         (direct_mask_load_lanes_optab_supported_p): Likewise.
26889         (direct_mask_store_lanes_optab_supported_p): Likewise.
26890         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26891         parameter.
26892         (vect_load_lanes_supported): Likewise.
26893         * tree-vect-data-refs.c (strip_conversion): New function.
26894         (can_group_stmts_p): Likewise.
26895         (vect_analyze_data_ref_accesses): Use it instead of checking
26896         for a pair of assignments.
26897         (vect_store_lanes_supported): Take a masked_p parameter.
26898         (vect_load_lanes_supported): Likewise.
26899         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26900         vect_store_lanes_supported and vect_load_lanes_supported.
26901         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26902         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26903         parameter.  Don't allow gaps for masked accesses.
26904         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
26905         and vect_load_lanes_supported.
26906         (get_load_store_type): Take a masked_p parameter and update
26907         call to get_group_load_store_type.
26908         (vectorizable_store): Update call to get_load_store_type.
26909         Handle IFN_MASK_STORE_LANES.
26910         (vectorizable_load): Update call to get_load_store_type.
26911         Handle IFN_MASK_LOAD_LANES.
26913 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26914             Alan Hayward  <alan.hayward@arm.com>
26915             David Sherwood  <david.sherwood@arm.com>
26917         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26918         modes for SVE.
26919         * config/aarch64/aarch64-protos.h
26920         (aarch64_sve_struct_memory_operand_p): Declare.
26921         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26922         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26923         (VPRED, vpred): Handle SVE structure modes.
26924         * config/aarch64/constraints.md (Utx): New constraint.
26925         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26926         (aarch64_sve_struct_nonimmediate_operand): New predicates.
26927         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26928         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26929         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26930         structure modes.  Split into pieces after RA.
26931         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26932         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26933         New patterns.
26934         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26935         SVE structure modes.
26936         (aarch64_classify_address): Likewise.
26937         (sizetochar): Move earlier in file.
26938         (aarch64_print_operand): Handle SVE register lists.
26939         (aarch64_array_mode): New function.
26940         (aarch64_sve_struct_memory_operand_p): Likewise.
26941         (TARGET_ARRAY_MODE): Redefine.
26943 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26944             Alan Hayward  <alan.hayward@arm.com>
26945             David Sherwood  <david.sherwood@arm.com>
26947         * target.def (array_mode): New target hook.
26948         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26949         * doc/tm.texi: Regenerate.
26950         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26951         * hooks.c (hook_optmode_mode_uhwi_none): New function.
26952         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26953         targetm.array_mode.
26954         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
26955         type sizes.
26957 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26958             Alan Hayward  <alan.hayward@arm.com>
26959             David Sherwood  <david.sherwood@arm.com>
26961         * fold-const.c (fold_binary_loc): Check the argument types
26962         rather than the result type when testing for a vector operation.
26964 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26966         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26967         * doc/tm.texi: Regenerate.
26969 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26970             Alan Hayward  <alan.hayward@arm.com>
26971             David Sherwood  <david.sherwood@arm.com>
26973         * doc/invoke.texi (-msve-vector-bits=): Document new option.
26974         (sve): Document new AArch64 extension.
26975         * doc/md.texi (w): Extend the description of the AArch64
26976         constraint to include SVE vectors.
26977         (Upl, Upa): Document new AArch64 predicate constraints.
26978         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26979         enum.
26980         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26981         (msve-vector-bits=): New option.
26982         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26983         SVE when these are disabled.
26984         (sve): New extension.
26985         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26986         modes.  Adjust their number of units based on aarch64_sve_vg.
26987         (MAX_BITSIZE_MODE_ANY_MODE): Define.
26988         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26989         aarch64_addr_query_type.
26990         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26991         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26992         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26993         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26994         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26995         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26996         (aarch64_simd_imm_zero_p): Delete.
26997         (aarch64_check_zero_based_sve_index_immediate): Declare.
26998         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26999         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
27000         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
27001         (aarch64_sve_float_mul_immediate_p): Likewise.
27002         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
27003         rather than an rtx.
27004         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
27005         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
27006         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
27007         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
27008         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
27009         (aarch64_regmode_natural_size): Likewise.
27010         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
27011         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
27012         left one place.
27013         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
27014         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
27015         for VG and the SVE predicate registers.
27016         (V_ALIASES): Add a "z"-prefixed alias.
27017         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
27018         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
27019         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
27020         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
27021         (REG_CLASS_NAMES): Add entries for them.
27022         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
27023         and the predicate registers.
27024         (aarch64_sve_vg): Declare.
27025         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
27026         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
27027         (REGMODE_NATURAL_SIZE): Define.
27028         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
27029         SVE macros.
27030         * config/aarch64/aarch64.c: Include cfgrtl.h.
27031         (simd_immediate_info): Add a constructor for series vectors,
27032         and an associated step field.
27033         (aarch64_sve_vg): New variable.
27034         (aarch64_dbx_register_number): Handle VG and the predicate registers.
27035         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
27036         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
27037         (VEC_ANY_DATA, VEC_STRUCT): New constants.
27038         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
27039         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
27040         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
27041         (aarch64_get_mask_mode): New functions.
27042         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
27043         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
27044         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
27045         predicate modes and predicate registers.  Explicitly restrict
27046         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
27047         to store a vector mode if it is recognized by
27048         aarch64_classify_vector_mode.
27049         (aarch64_regmode_natural_size): New function.
27050         (aarch64_hard_regno_caller_save_mode): Return the original mode
27051         for predicates.
27052         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
27053         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
27054         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
27055         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
27056         functions.
27057         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
27058         does not overlap dest if the function is frame-related.  Handle
27059         SVE constants.
27060         (aarch64_split_add_offset): New function.
27061         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
27062         them aarch64_add_offset.
27063         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
27064         and update call to aarch64_sub_sp.
27065         (aarch64_add_cfa_expression): New function.
27066         (aarch64_expand_prologue): Pass extra temporary registers to the
27067         functions above.  Handle the case in which we need to emit new
27068         DW_CFA_expressions for registers that were originally saved
27069         relative to the stack pointer, but now have to be expressed
27070         relative to the frame pointer.
27071         (aarch64_output_mi_thunk): Pass extra temporary registers to the
27072         functions above.
27073         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
27074         IP0 and IP1 values for SVE frames.
27075         (aarch64_expand_vec_series): New function.
27076         (aarch64_expand_sve_widened_duplicate): Likewise.
27077         (aarch64_expand_sve_const_vector): Likewise.
27078         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
27079         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
27080         into the register, rather than emitting a SET directly.
27081         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
27082         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
27083         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
27084         (offset_9bit_signed_scaled_p): New functions.
27085         (aarch64_replicate_bitmask_imm): New function.
27086         (aarch64_bitmask_imm): Use it.
27087         (aarch64_cannot_force_const_mem): Reject expressions involving
27088         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
27089         (aarch64_classify_index): Handle SVE indices, by requiring
27090         a plain register index with a scale that matches the element size.
27091         (aarch64_classify_address): Handle SVE addresses.  Assert that
27092         the mode of the address is VOIDmode or an integer mode.
27093         Update call to aarch64_classify_symbol.
27094         (aarch64_classify_symbolic_expression): Update call to
27095         aarch64_classify_symbol.
27096         (aarch64_const_vec_all_in_range_p): New function.
27097         (aarch64_print_vector_float_operand): Likewise.
27098         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
27099         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
27100         and the FP immediates 1.0 and 0.5.
27101         (aarch64_print_address_internal): Handle SVE addresses.
27102         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
27103         (aarch64_regno_regclass): Handle predicate registers.
27104         (aarch64_secondary_reload): Handle big-endian reloads of SVE
27105         data modes.
27106         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
27107         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
27108         (aarch64_convert_sve_vector_bits): New function.
27109         (aarch64_override_options): Use it to handle -msve-vector-bits=.
27110         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
27111         rather than an rtx.
27112         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
27113         Handle SVE vector and predicate modes.  Accept VL-based constants
27114         that need only one temporary register, and VL offsets that require
27115         no temporary registers.
27116         (aarch64_conditional_register_usage): Mark the predicate registers
27117         as fixed if SVE isn't available.
27118         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
27119         Return true for SVE vector and predicate modes.
27120         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
27121         rather than an unsigned int.  Handle SVE modes.
27122         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
27123         SVE modes.
27124         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
27125         if SVE is enabled.
27126         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
27127         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
27128         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
27129         (aarch64_sve_float_mul_immediate_p): New functions.
27130         (aarch64_sve_valid_immediate): New function.
27131         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
27132         Explicitly reject structure modes.  Check for INDEX constants.
27133         Handle PTRUE and PFALSE constants.
27134         (aarch64_check_zero_based_sve_index_immediate): New function.
27135         (aarch64_simd_imm_zero_p): Delete.
27136         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
27137         vector modes.  Accept constants in the range of CNT[BHWD].
27138         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
27139         ask for an Advanced SIMD mode.
27140         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
27141         (aarch64_simd_vector_alignment): Handle SVE predicates.
27142         (aarch64_vectorize_preferred_vector_alignment): New function.
27143         (aarch64_simd_vector_alignment_reachable): Use it instead of
27144         the vector size.
27145         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
27146         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
27147         functions.
27148         (MAX_VECT_LEN): Delete.
27149         (expand_vec_perm_d): Add a vec_flags field.
27150         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
27151         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
27152         (aarch64_evpc_ext): Don't apply a big-endian lane correction
27153         for SVE modes.
27154         (aarch64_evpc_rev): Rename to...
27155         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
27156         (aarch64_evpc_rev_global): New function.
27157         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
27158         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
27159         MAX_VECT_LEN.
27160         (aarch64_evpc_sve_tbl): New function.
27161         (aarch64_expand_vec_perm_const_1): Update after rename of
27162         aarch64_evpc_rev.  Handle SVE permutes too, trying
27163         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
27164         than aarch64_evpc_tbl.
27165         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
27166         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
27167         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
27168         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
27169         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
27170         (aarch64_expand_sve_vcond): New functions.
27171         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
27172         of aarch64_vector_mode_p.
27173         (aarch64_dwarf_poly_indeterminate_value): New function.
27174         (aarch64_compute_pressure_classes): Likewise.
27175         (aarch64_can_change_mode_class): Likewise.
27176         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
27177         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
27178         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
27179         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
27180         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
27181         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
27182         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
27183         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
27184         constraints.
27185         (Dn, Dl, Dr): Accept const as well as const_vector.
27186         (Dz): Likewise.  Compare against CONST0_RTX.
27187         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
27188         of "vector" where appropriate.
27189         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
27190         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
27191         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
27192         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
27193         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
27194         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
27195         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
27196         (v_int_equiv): Extend to SVE modes.
27197         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
27198         mode attributes.
27199         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
27200         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
27201         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
27202         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
27203         (SVE_COND_FP_CMP): New int iterators.
27204         (perm_hilo): Handle the new unpack unspecs.
27205         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
27206         attributes.
27207         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
27208         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
27209         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
27210         (aarch64_equality_operator, aarch64_constant_vector_operand)
27211         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
27212         (aarch64_sve_nonimmediate_operand): Likewise.
27213         (aarch64_sve_general_operand): Likewise.
27214         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
27215         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
27216         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
27217         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
27218         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
27219         (aarch64_sve_float_arith_immediate): Likewise.
27220         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
27221         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
27222         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
27223         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
27224         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
27225         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
27226         (aarch64_sve_float_arith_operand): Likewise.
27227         (aarch64_sve_float_arith_with_sub_operand): Likewise.
27228         (aarch64_sve_float_mul_operand): Likewise.
27229         (aarch64_sve_vec_perm_operand): Likewise.
27230         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
27231         (aarch64_mov_operand): Accept const_poly_int and const_vector.
27232         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
27233         as well as const_vector.
27234         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
27235         in file.  Use CONST0_RTX and CONSTM1_RTX.
27236         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
27237         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
27238         Use aarch64_simd_imm_zero.
27239         * config/aarch64/aarch64-sve.md: New file.
27240         * config/aarch64/aarch64.md: Include it.
27241         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
27242         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
27243         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
27244         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
27245         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
27246         (sve): New attribute.
27247         (enabled): Disable instructions with the sve attribute unless
27248         TARGET_SVE.
27249         (movqi, movhi): Pass CONST_POLY_INT operaneds through
27250         aarch64_expand_mov_immediate.
27251         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
27252         CNT[BHSD] immediates.
27253         (movti): Split CONST_POLY_INT moves into two halves.
27254         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
27255         Split additions that need a temporary here if the destination
27256         is the stack pointer.
27257         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
27258         (*add<mode>3_poly_1): New instruction.
27259         (set_clobber_cc): New expander.
27261 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27263         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
27264         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
27265         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
27266         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
27267         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
27268         Change innermode from fixed_mode_size to machine_mode.
27269         (simplify_subreg): Update call accordingly.  Handle a constant-sized
27270         subreg of a variable-length CONST_VECTOR.
27272 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27273             Alan Hayward  <alan.hayward@arm.com>
27274             David Sherwood  <david.sherwood@arm.com>
27276         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
27277         (add_offset_to_base): New function, split out from...
27278         (create_mem_ref): ...here.  When handling a scale other than 1,
27279         check first whether the address is valid without the offset.
27280         Add it into the base if so, leaving the index and scale as-is.
27282 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27284         PR c++/83778
27285         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
27286         fold_for_warn before checking if arg2 is INTEGER_CST.
27288 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
27290         * config/rs6000/predicates.md (load_multiple_operation): Delete.
27291         (store_multiple_operation): Delete.
27292         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
27293         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
27294         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
27295         guarded by TARGET_STRING.
27296         (rs6000_output_load_multiple): Delete.
27297         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
27298         OPTION_MASK_STRING / TARGET_STRING handling.
27299         (print_operand) <'N', 'O'>: Add comment that these are unused now.
27300         (const rs6000_opt_masks) <"string">: Change mask to 0.
27301         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
27302         (MASK_STRING): Delete.
27303         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
27304         parts.  Simplify.
27305         (load_multiple): Delete.
27306         (*ldmsi8): Delete.
27307         (*ldmsi7): Delete.
27308         (*ldmsi6): Delete.
27309         (*ldmsi5): Delete.
27310         (*ldmsi4): Delete.
27311         (*ldmsi3): Delete.
27312         (store_multiple): Delete.
27313         (*stmsi8): Delete.
27314         (*stmsi7): Delete.
27315         (*stmsi6): Delete.
27316         (*stmsi5): Delete.
27317         (*stmsi4): Delete.
27318         (*stmsi3): Delete.
27319         (movmemsi_8reg): Delete.
27320         (corresponding unnamed define_insn): Delete.
27321         (movmemsi_6reg): Delete.
27322         (corresponding unnamed define_insn): Delete.
27323         (movmemsi_4reg): Delete.
27324         (corresponding unnamed define_insn): Delete.
27325         (movmemsi_2reg): Delete.
27326         (corresponding unnamed define_insn): Delete.
27327         (movmemsi_1reg): Delete.
27328         (corresponding unnamed define_insn): Delete.
27329         * config/rs6000/rs6000.opt (mno-string): New.
27330         (mstring): Replace by deprecation warning stub.
27331         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
27333 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27335         * regrename.c (regrename_do_replace): If replacing the same
27336         reg multiple times, try to reuse last created gen_raw_REG.
27338         PR debug/81155
27339         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
27340         main to workaround a bug in GDB.
27342 2018-01-12  Tom de Vries  <tom@codesourcery.com>
27344         PR target/83737
27345         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
27347 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
27349         PR rtl-optimization/80481
27350         * ira-color.c (get_cap_member): New function.
27351         (allocnos_conflict_by_live_ranges_p): Use it.
27352         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
27353         (setup_slot_coalesced_allocno_live_ranges): Ditto.
27355 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
27357         PR target/83628
27358         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
27359         (*saddl_se_1): Ditto.
27360         (*ssubsi_1): Ditto.
27361         (*ssubl_se_1): Ditto.
27363 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27365         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
27366         rather than wi::to_widest for DR_INITs.
27367         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
27368         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
27369         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
27370         INTEGER_CSTs.
27371         (vect_analyze_group_access_1): Note that here.
27373 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27375         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
27376         polynomial type sizes.
27378 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27380         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
27381         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
27382         (gimple_add_tmp_var): Likewise.
27384 2018-01-12  Martin Liska  <mliska@suse.cz>
27386         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
27387         (gimple_alloc_sizes): Likewise.
27388         (dump_gimple_statistics): Use PRIu64 in printf format.
27389         * gimple.h: Change uint64_t to int.
27391 2018-01-12  Martin Liska  <mliska@suse.cz>
27393         * tree-core.h: Use uint64_t instead of int.
27394         * tree.c (tree_node_counts): Likewise.
27395         (tree_node_sizes): Likewise.
27396         (dump_tree_statistics): Use PRIu64 in printf format.
27398 2018-01-12  Martin Liska  <mliska@suse.cz>
27400         * Makefile.in: As qsort_chk is implemented in vec.c, add
27401         vec.o to linkage of gencfn-macros.
27402         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
27403         passing the info to record_node_allocation_statistics.
27404         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
27405         and pass the info.
27406         * ggc-common.c (struct ggc_usage): Add operator== and use
27407         it in operator< and compare function.
27408         * mem-stats.h (struct mem_usage): Likewise.
27409         * vec.c (struct vec_usage): Remove operator< and compare
27410         function. Can be simply inherited.
27412 2018-01-12  Martin Jambor  <mjambor@suse.cz>
27414         PR target/81616
27415         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
27416         * tree-ssa-math-opts.c: Include domwalk.h.
27417         (convert_mult_to_fma_1): New function.
27418         (fma_transformation_info): New type.
27419         (fma_deferring_state): Likewise.
27420         (cancel_fma_deferring): New function.
27421         (result_of_phi): Likewise.
27422         (last_fma_candidate_feeds_initial_phi): Likewise.
27423         (convert_mult_to_fma): Added deferring logic, split actual
27424         transformation to convert_mult_to_fma_1.
27425         (math_opts_dom_walker): New type.
27426         (math_opts_dom_walker::after_dom_children): New method, body moved
27427         here from pass_optimize_widening_mul::execute, added deferring logic
27428         bits.
27429         (pass_optimize_widening_mul::execute): Moved most of code to
27430         math_opts_dom_walker::after_dom_children.
27431         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
27432         * config/i386/i386.c (ix86_option_override_internal): Added
27433         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
27435 2018-01-12  Richard Biener  <rguenther@suse.de>
27437         PR debug/83157
27438         * dwarf2out.c (gen_variable_die): Do not reset old_die for
27439         inline instance vars.
27441 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
27443         PR target/81819
27444         * config/rx/rx.c (rx_is_restricted_memory_address):
27445         Handle SUBREG case.
27447 2018-01-12  Richard Biener  <rguenther@suse.de>
27449         PR tree-optimization/80846
27450         * target.def (split_reduction): New target hook.
27451         * targhooks.c (default_split_reduction): New function.
27452         * targhooks.h (default_split_reduction): Declare.
27453         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
27454         target requests first reduce vectors by combining low and high
27455         parts.
27456         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
27457         (get_vectype_for_scalar_type_and_size): Export.
27458         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
27459         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
27460         * doc/tm.texi: Regenerate.
27461         * config/i386/i386.c (ix86_split_reduction): Implement
27462         TARGET_VECTORIZE_SPLIT_REDUCTION.
27464 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27466         PR target/83368
27467         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
27468         in PIC mode except for TARGET_VXWORKS_RTP.
27469         * config/sparc/sparc.c: Include cfgrtl.h.
27470         (TARGET_INIT_PIC_REG): Define.
27471         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
27472         (sparc_pic_register_p): New predicate.
27473         (sparc_legitimate_address_p): Use it.
27474         (sparc_legitimize_pic_address): Likewise.
27475         (sparc_delegitimize_address): Likewise.
27476         (sparc_mode_dependent_address_p): Likewise.
27477         (gen_load_pcrel_sym): Remove 4th parameter.
27478         (load_got_register): Adjust call to above.  Remove obsolete stuff.
27479         (sparc_expand_prologue): Do not call load_got_register here.
27480         (sparc_flat_expand_prologue): Likewise.
27481         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
27482         (sparc_use_pseudo_pic_reg): New function.
27483         (sparc_init_pic_reg): Likewise.
27484         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27485         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
27487 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
27489         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
27490         Add item for branch_cost.
27492 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27494         PR rtl-optimization/83565
27495         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
27496         not extend the result to a larger mode for rotate operations.
27497         (num_sign_bit_copies1): Likewise.
27499 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27501         PR target/40411
27502         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
27503         -symbolic.
27504         Use values-Xc.o for -pedantic.
27505         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
27507 2018-01-12  Martin Liska  <mliska@suse.cz>
27509         PR ipa/83054
27510         * ipa-devirt.c (final_warning_record::grow_type_warnings):
27511         New function.
27512         (possible_polymorphic_call_targets): Use it.
27513         (ipa_devirt): Likewise.
27515 2018-01-12  Martin Liska  <mliska@suse.cz>
27517         * profile-count.h (enum profile_quality): Use 0 as invalid
27518         enum value of profile_quality.
27520 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
27522         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
27523         -mext-string options.
27525 2018-01-12  Richard Biener  <rguenther@suse.de>
27527         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
27528         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
27529         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
27530         Likewise.
27531         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
27533 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
27535         * configure.ac (--with-long-double-format): Add support for the
27536         configuration option to change the default long double format on
27537         PowerPC systems.
27538         * config.gcc (powerpc*-linux*-*): Likewise.
27539         * configure: Regenerate.
27540         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
27541         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
27542         used without modification.
27544 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27546         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
27547         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
27548         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
27549         MISC_BUILTIN_SPEC_BARRIER.
27550         (rs6000_init_builtins): Likewise.
27551         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
27552         enum value.
27553         (speculation_barrier): New define_insn.
27554         * doc/extend.texi: Document __builtin_speculation_barrier.
27556 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27558         PR target/83203
27559         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
27560         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
27561         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
27562         iterators.
27563         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
27564         integral modes instead of "ss" and "sd".
27565         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
27566         vectors with 32-bit and 64-bit elements.
27567         (vecdupssescalarmodesuffix): New mode attribute.
27568         (vec_dup<mode>): Use it.
27570 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
27572         PR target/83330
27573         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
27574         frame if argument is passed on stack.
27576 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27578         PR target/82682
27579         * ree.c (combine_reaching_defs): Optimize also
27580         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
27581         reg2=any_extend(exp); reg1=reg2;, formatting fix.
27583 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27585         PR middle-end/83189
27586         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
27588 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27590         PR middle-end/83718
27591         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
27592         after they are computed.
27594 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
27596         PR tree-optimization/83695
27597         * gimple-loop-linterchange.cc
27598         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
27599         reset cached scev information after interchange.
27600         (pass_linterchange::execute): Remove call to scev_reset_htab.
27602 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27604         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
27605         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
27606         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
27607         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
27608         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
27609         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
27610         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
27611         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
27612         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
27613         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
27614         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
27615         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
27616         (V_lane_reg): Likewise.
27617         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
27618         New define_expand.
27619         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
27620         (vfmal_lane_low<mode>_intrinsic,
27621         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
27622         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27623         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27624         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27625         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27626         vfmsl_lane_high<mode>_intrinsic): New define_insns.
27628 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27630         * config/arm/arm-cpus.in (fp16fml): New feature.
27631         (ALL_SIMD): Add fp16fml.
27632         (armv8.2-a): Add fp16fml as an option.
27633         (armv8.3-a): Likewise.
27634         (armv8.4-a): Add fp16fml as part of fp16.
27635         * config/arm/arm.h (TARGET_FP16FML): Define.
27636         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27637         when appropriate.
27638         * config/arm/arm-modes.def (V2HF): Define.
27639         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27640         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27641         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27642         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27643         vfmsl_low, vfmsl_high): New set of builtins.
27644         * config/arm/iterators.md (PLUSMINUS): New code iterator.
27645         (vfml_op): New code attribute.
27646         (VFMLHALVES): New int iterator.
27647         (VFML, VFMLSEL): New mode attributes.
27648         (V_reg): Define mapping for V2HF.
27649         (V_hi, V_lo): New mode attributes.
27650         (VF_constraint): Likewise.
27651         (vfml_half, vfml_half_selector): New int attributes.
27652         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27653         define_expand.
27654         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27655         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27656         New define_insn.
27657         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27658         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27659         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27660         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
27661         documentation.
27662         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27663         Document new effective target and option set.
27665 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27667         * config/arm/arm-cpus.in (armv8_4): New feature.
27668         (ARMv8_4a): New fgroup.
27669         (armv8.4-a): New arch.
27670         * config/arm/arm-tables.opt: Regenerate.
27671         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27672         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27673         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27674         Add matching rules for -march=armv8.4-a and extensions.
27675         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27677 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
27679         PR target/81821
27680         * config/rx/rx.md (BW): New mode attribute.
27681         (sync_lock_test_and_setsi): Add mode suffix to insn output.
27683 2018-01-11  Richard Biener  <rguenther@suse.de>
27685         PR tree-optimization/83435
27686         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27687         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27688         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27690 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27691             Alan Hayward  <alan.hayward@arm.com>
27692             David Sherwood  <david.sherwood@arm.com>
27694         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27695         field.
27696         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
27697         (aarch64_print_address_internal): Use it to check for a zero offset.
27699 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27700             Alan Hayward  <alan.hayward@arm.com>
27701             David Sherwood  <david.sherwood@arm.com>
27703         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27704         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27705         Return a poly_int64 rather than a HOST_WIDE_INT.
27706         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27707         rather than a HOST_WIDE_INT.
27708         * config/aarch64/aarch64.h (aarch64_frame): Protect with
27709         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
27710         hard_fp_offset, frame_size, initial_adjust, callee_offset and
27711         final_offset from HOST_WIDE_INT to poly_int64.
27712         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27713         to_constant when getting the number of units in an Advanced SIMD
27714         mode.
27715         (aarch64_builtin_vectorized_function): Check for a constant number
27716         of units.
27717         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27718         GET_MODE_SIZE.
27719         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27720         attribute instead of GET_MODE_NUNITS.
27721         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27722         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27723         GET_MODE_SIZE for fixed-size registers.
27724         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27725         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27726         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27727         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27728         (aarch64_print_operand, aarch64_print_address_internal)
27729         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27730         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27731         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27732         Handle polynomial GET_MODE_SIZE.
27733         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
27734         wider than SImode without modification.
27735         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27736         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27737         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27738         passing and returning SVE modes.
27739         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27740         rather than GEN_INT.
27741         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27742         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27743         (aarch64_allocate_and_probe_stack_space): Likewise.
27744         (aarch64_layout_frame): Cope with polynomial offsets.
27745         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27746         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
27747         polynomial offsets.
27748         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27749         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27750         poly_int64 rather than a HOST_WIDE_INT.
27751         (aarch64_get_separate_components, aarch64_process_components)
27752         (aarch64_expand_prologue, aarch64_expand_epilogue)
27753         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27754         (aarch64_anchor_offset): New function, split out from...
27755         (aarch64_legitimize_address): ...here.
27756         (aarch64_builtin_vectorization_cost): Handle polynomial
27757         TYPE_VECTOR_SUBPARTS.
27758         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27759         GET_MODE_NUNITS.
27760         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27761         number of elements from the PARALLEL rather than the mode.
27762         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27763         rather than GET_MODE_BITSIZE.
27764         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27765         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27766         (aarch64_expand_vec_perm_const_1): Handle polynomial
27767         d->perm.length () and d->perm elements.
27768         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
27769         Apply to_constant to d->perm elements.
27770         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27771         polynomial CONST_VECTOR_NUNITS.
27772         (aarch64_move_pointer): Take amount as a poly_int64 rather
27773         than an int.
27774         (aarch64_progress_pointer): Avoid temporary variable.
27775         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27776         the mode attribute instead of GET_MODE.
27778 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27779             Alan Hayward  <alan.hayward@arm.com>
27780             David Sherwood  <david.sherwood@arm.com>
27782         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27783         x exists before using it.
27784         (aarch64_add_constant_internal): Rename to...
27785         (aarch64_add_offset_1): ...this.  Replace regnum with separate
27786         src and dest rtxes.  Handle the case in which they're different,
27787         including when the offset is zero.  Replace scratchreg with an rtx.
27788         Use 2 additions if there is no spare register into which we can
27789         move a 16-bit constant.
27790         (aarch64_add_constant): Delete.
27791         (aarch64_add_offset): Replace reg with separate src and dest
27792         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
27793         Use aarch64_add_offset_1.
27794         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27795         an rtx rather than an int.  Take the delta as a poly_int64
27796         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
27797         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27798         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27799         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27800         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27801         and aarch64_add_sp.
27802         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27803         aarch64_add_constant.
27805 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27807         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27808         Use scalar_float_mode.
27810 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27812         * config/aarch64/aarch64-simd.md
27813         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27814         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27815         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27816         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27817         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27818         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27819         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27820         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27821         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27822         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27824 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27826         PR target/83514
27827         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27828         targ_options->x_arm_arch_string is non NULL.
27830 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
27832         * config/aarch64/aarch64.h
27833         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
27835 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
27837         PR target/82096
27838         * expmed.c (emit_store_flag_force): Swap if const op0
27839         and change VOIDmode to mode of op0.
27841 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27843         PR rtl-optimization/83761
27844         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27845         than bytes to mode_for_size.
27847 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27849         PR middle-end/83189
27850         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27851         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27852         profile.
27854 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27856         PR middle-end/83575
27857         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27858         when in layout mode.
27859         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27860         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27861         partition fixup.
27863 2018-01-10  Michael Collison  <michael.collison@arm.com>
27865         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27866         * config/aarch64/aarch64-option-extension.def: Add
27867         AARCH64_OPT_EXTENSION of 'fp16fml'.
27868         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27869         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27870         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27871         * config/aarch64/constraints.md (Ui7): New constraint.
27872         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27873         (VFMLA_SEL_W): Ditto.
27874         (f16quad): Ditto.
27875         (f16mac1): Ditto.
27876         (VFMLA16_LOW): New int iterator.
27877         (VFMLA16_HIGH): Ditto.
27878         (UNSPEC_FMLAL): New unspec.
27879         (UNSPEC_FMLSL): Ditto.
27880         (UNSPEC_FMLAL2): Ditto.
27881         (UNSPEC_FMLSL2): Ditto.
27882         (f16mac): New code attribute.
27883         * config/aarch64/aarch64-simd-builtins.def
27884         (aarch64_fmlal_lowv2sf): Ditto.
27885         (aarch64_fmlsl_lowv2sf): Ditto.
27886         (aarch64_fmlalq_lowv4sf): Ditto.
27887         (aarch64_fmlslq_lowv4sf): Ditto.
27888         (aarch64_fmlal_highv2sf): Ditto.
27889         (aarch64_fmlsl_highv2sf): Ditto.
27890         (aarch64_fmlalq_highv4sf): Ditto.
27891         (aarch64_fmlslq_highv4sf): Ditto.
27892         (aarch64_fmlal_lane_lowv2sf): Ditto.
27893         (aarch64_fmlsl_lane_lowv2sf): Ditto.
27894         (aarch64_fmlal_laneq_lowv2sf): Ditto.
27895         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27896         (aarch64_fmlalq_lane_lowv4sf): Ditto.
27897         (aarch64_fmlsl_lane_lowv4sf): Ditto.
27898         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27899         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27900         (aarch64_fmlal_lane_highv2sf): Ditto.
27901         (aarch64_fmlsl_lane_highv2sf): Ditto.
27902         (aarch64_fmlal_laneq_highv2sf): Ditto.
27903         (aarch64_fmlsl_laneq_highv2sf): Ditto.
27904         (aarch64_fmlalq_lane_highv4sf): Ditto.
27905         (aarch64_fmlsl_lane_highv4sf): Ditto.
27906         (aarch64_fmlalq_laneq_highv4sf): Ditto.
27907         (aarch64_fmlsl_laneq_highv4sf): Ditto.
27908         * config/aarch64/aarch64-simd.md:
27909         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27910         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27911         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27912         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27913         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27914         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27915         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27916         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27917         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27918         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27919         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27920         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27921         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27922         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27923         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27924         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27925         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27926         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27927         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27928         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27929         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27930         (vfmlsl_low_u32): Ditto.
27931         (vfmlalq_low_u32): Ditto.
27932         (vfmlslq_low_u32): Ditto.
27933         (vfmlal_high_u32): Ditto.
27934         (vfmlsl_high_u32): Ditto.
27935         (vfmlalq_high_u32): Ditto.
27936         (vfmlslq_high_u32): Ditto.
27937         (vfmlal_lane_low_u32): Ditto.
27938         (vfmlsl_lane_low_u32): Ditto.
27939         (vfmlal_laneq_low_u32): Ditto.
27940         (vfmlsl_laneq_low_u32): Ditto.
27941         (vfmlalq_lane_low_u32): Ditto.
27942         (vfmlslq_lane_low_u32): Ditto.
27943         (vfmlalq_laneq_low_u32): Ditto.
27944         (vfmlslq_laneq_low_u32): Ditto.
27945         (vfmlal_lane_high_u32): Ditto.
27946         (vfmlsl_lane_high_u32): Ditto.
27947         (vfmlal_laneq_high_u32): Ditto.
27948         (vfmlsl_laneq_high_u32): Ditto.
27949         (vfmlalq_lane_high_u32): Ditto.
27950         (vfmlslq_lane_high_u32): Ditto.
27951         (vfmlalq_laneq_high_u32): Ditto.
27952         (vfmlslq_laneq_high_u32): Ditto.
27953         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27954         (AARCH64_FL_FOR_ARCH8_4): New.
27955         (AARCH64_ISA_F16FML): New ISA flag.
27956         (TARGET_F16FML): New feature flag for fp16fml.
27957         (doc/invoke.texi): Document new fp16fml option.
27959 2018-01-10  Michael Collison  <michael.collison@arm.com>
27961         * config/aarch64/aarch64-builtins.c:
27962         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27963         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27964         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27965         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27966         (AARCH64_ISA_SHA3): New ISA flag.
27967         (TARGET_SHA3): New feature flag for sha3.
27968         * config/aarch64/iterators.md (sha512_op): New int attribute.
27969         (CRYPTO_SHA512): New int iterator.
27970         (UNSPEC_SHA512H): New unspec.
27971         (UNSPEC_SHA512H2): Ditto.
27972         (UNSPEC_SHA512SU0): Ditto.
27973         (UNSPEC_SHA512SU1): Ditto.
27974         * config/aarch64/aarch64-simd-builtins.def
27975         (aarch64_crypto_sha512hqv2di): New builtin.
27976         (aarch64_crypto_sha512h2qv2di): Ditto.
27977         (aarch64_crypto_sha512su0qv2di): Ditto.
27978         (aarch64_crypto_sha512su1qv2di): Ditto.
27979         (aarch64_eor3qv8hi): Ditto.
27980         (aarch64_rax1qv2di): Ditto.
27981         (aarch64_xarqv2di): Ditto.
27982         (aarch64_bcaxqv8hi): Ditto.
27983         * config/aarch64/aarch64-simd.md:
27984         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27985         (aarch64_crypto_sha512su0qv2di): Ditto.
27986         (aarch64_crypto_sha512su1qv2di): Ditto.
27987         (aarch64_eor3qv8hi): Ditto.
27988         (aarch64_rax1qv2di): Ditto.
27989         (aarch64_xarqv2di): Ditto.
27990         (aarch64_bcaxqv8hi): Ditto.
27991         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27992         (vsha512h2q_u64): Ditto.
27993         (vsha512su0q_u64): Ditto.
27994         (vsha512su1q_u64): Ditto.
27995         (veor3q_u16): Ditto.
27996         (vrax1q_u64): Ditto.
27997         (vxarq_u64): Ditto.
27998         (vbcaxq_u16): Ditto.
27999         * config/arm/types.md (crypto_sha512): New type attribute.
28000         (crypto_sha3): Ditto.
28001         (doc/invoke.texi): Document new sha3 option.
28003 2018-01-10  Michael Collison  <michael.collison@arm.com>
28005         * config/aarch64/aarch64-builtins.c:
28006         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
28007         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28008         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
28009         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
28010         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
28011         (AARCH64_ISA_SM4): New ISA flag.
28012         (TARGET_SM4): New feature flag for sm4.
28013         * config/aarch64/aarch64-simd-builtins.def
28014         (aarch64_sm3ss1qv4si): Ditto.
28015         (aarch64_sm3tt1aq4si): Ditto.
28016         (aarch64_sm3tt1bq4si): Ditto.
28017         (aarch64_sm3tt2aq4si): Ditto.
28018         (aarch64_sm3tt2bq4si): Ditto.
28019         (aarch64_sm3partw1qv4si): Ditto.
28020         (aarch64_sm3partw2qv4si): Ditto.
28021         (aarch64_sm4eqv4si): Ditto.
28022         (aarch64_sm4ekeyqv4si): Ditto.
28023         * config/aarch64/aarch64-simd.md:
28024         (aarch64_sm3ss1qv4si): Ditto.
28025         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
28026         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
28027         (aarch64_sm4eqv4si): Ditto.
28028         (aarch64_sm4ekeyqv4si): Ditto.
28029         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
28030         (sm3part_op): Ditto.
28031         (CRYPTO_SM3TT): Ditto.
28032         (CRYPTO_SM3PART): Ditto.
28033         (UNSPEC_SM3SS1): New unspec.
28034         (UNSPEC_SM3TT1A): Ditto.
28035         (UNSPEC_SM3TT1B): Ditto.
28036         (UNSPEC_SM3TT2A): Ditto.
28037         (UNSPEC_SM3TT2B): Ditto.
28038         (UNSPEC_SM3PARTW1): Ditto.
28039         (UNSPEC_SM3PARTW2): Ditto.
28040         (UNSPEC_SM4E): Ditto.
28041         (UNSPEC_SM4EKEY): Ditto.
28042         * config/aarch64/constraints.md (Ui2): New constraint.
28043         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
28044         * config/arm/types.md (crypto_sm3): New type attribute.
28045         (crypto_sm4): Ditto.
28046         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
28047         (vsm3tt1aq_u32): Ditto.
28048         (vsm3tt1bq_u32): Ditto.
28049         (vsm3tt2aq_u32): Ditto.
28050         (vsm3tt2bq_u32): Ditto.
28051         (vsm3partw1q_u32): Ditto.
28052         (vsm3partw2q_u32): Ditto.
28053         (vsm4eq_u32): Ditto.
28054         (vsm4ekeyq_u32): Ditto.
28055         (doc/invoke.texi): Document new sm4 option.
28057 2018-01-10  Michael Collison  <michael.collison@arm.com>
28059         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
28060         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
28061         (AARCH64_FL_FOR_ARCH8_4): New.
28062         (AARCH64_FL_V8_4): New flag.
28063         (doc/invoke.texi): Document new armv8.4-a option.
28065 2018-01-10  Michael Collison  <michael.collison@arm.com>
28067         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28068         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
28069         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
28070         * config/aarch64/aarch64-option-extension.def: Add
28071         AARCH64_OPT_EXTENSION of 'sha2'.
28072         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
28073         (crypto): Disable sha2 and aes if crypto disabled.
28074         (crypto): Enable aes and sha2 if enabled.
28075         (simd): Disable sha2 and aes if simd disabled.
28076         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
28077         New flags.
28078         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
28079         (TARGET_SHA2): New feature flag for sha2.
28080         (TARGET_AES): New feature flag for aes.
28081         * config/aarch64/aarch64-simd.md:
28082         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
28083         conditional on TARGET_AES.
28084         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
28085         (aarch64_crypto_sha1hsi): Make pattern conditional
28086         on TARGET_SHA2.
28087         (aarch64_crypto_sha1hv4si): Ditto.
28088         (aarch64_be_crypto_sha1hv4si): Ditto.
28089         (aarch64_crypto_sha1su1v4si): Ditto.
28090         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
28091         (aarch64_crypto_sha1su0v4si): Ditto.
28092         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
28093         (aarch64_crypto_sha256su0v4si): Ditto.
28094         (aarch64_crypto_sha256su1v4si): Ditto.
28095         (doc/invoke.texi): Document new aes and sha2 options.
28097 2018-01-10  Martin Sebor  <msebor@redhat.com>
28099         PR tree-optimization/83781
28100         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
28101         as string arrays.
28103 2018-01-11  Martin Sebor  <msebor@gmail.com>
28104             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28106         PR tree-optimization/83501
28107         PR tree-optimization/81703
28109         * tree-ssa-strlen.c (get_string_cst): Rename...
28110         (get_string_len): ...to this.  Handle global constants.
28111         (handle_char_store): Adjust.
28113 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
28114             Jim Wilson  <jimw@sifive.com>
28116         * config/riscv/riscv-protos.h (riscv_output_return): New.
28117         * config/riscv/riscv.c (struct machine_function): New naked_p field.
28118         (riscv_attribute_table, riscv_output_return),
28119         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
28120         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
28121         (riscv_compute_frame_info): Only compute frame->mask if not a naked
28122         function.
28123         (riscv_expand_prologue): Add early return for naked function.
28124         (riscv_expand_epilogue): Likewise.
28125         (riscv_function_ok_for_sibcall): Return false for naked function.
28126         (riscv_set_current_function): New.
28127         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
28128         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
28129         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
28130         * doc/extend.texi (RISC-V Function Attributes): New.
28132 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
28134         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
28135         check for 128-bit long double before checking TCmode.
28136         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
28137         128-bit long doubles before checking TFmode or TCmode.
28138         (FLOAT128_IBM_P): Likewise.
28140 2018-01-10  Martin Sebor  <msebor@redhat.com>
28142         PR tree-optimization/83671
28143         * builtins.c (c_strlen): Unconditionally return zero for the empty
28144         string.
28145         Use -Warray-bounds for warnings.
28146         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
28147         for non-constant array indices with COMPONENT_REF, arrays of
28148         arrays, and pointers to arrays.
28149         (gimple_fold_builtin_strlen): Determine and set length range for
28150         non-constant character arrays.
28152 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
28154         PR middle-end/81897
28155         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
28156         empty blocks.
28158 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
28160         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
28162 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
28164         PR target/83399
28165         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
28166         VECTOR_MEM_ALTIVEC_OR_VSX_P.
28167         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
28168         indexed_or_indirect_operand predicate.
28169         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
28170         (*vsx_le_perm_load_v8hi): Likewise.
28171         (*vsx_le_perm_load_v16qi): Likewise.
28172         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
28173         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
28174         (*vsx_le_perm_store_v8hi): Likewise.
28175         (*vsx_le_perm_store_v16qi): Likewise.
28176         (eight unnamed splitters): Likewise.
28178 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
28180         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
28181         * config/rs6000/emmintrin.h: Likewise.
28182         * config/rs6000/mmintrin.h: Likewise.
28183         * config/rs6000/xmmintrin.h: Likewise.
28185 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
28187         PR c++/43486
28188         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
28189         "public_flag".
28190         * tree.c (tree_nop_conversion): Return true for location wrapper
28191         nodes.
28192         (maybe_wrap_with_location): New function.
28193         (selftest::check_strip_nops): New function.
28194         (selftest::test_location_wrappers): New function.
28195         (selftest::tree_c_tests): Call it.
28196         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
28197         (maybe_wrap_with_location): New decl.
28198         (EXPR_LOCATION_WRAPPER_P): New macro.
28199         (location_wrapper_p): New inline function.
28200         (tree_strip_any_location_wrapper): New inline function.
28202 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
28204         PR target/83735
28205         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
28206         stack_realign_offset for the largest alignment of stack slot
28207         actually used.
28208         (ix86_find_max_used_stack_alignment): New function.
28209         (ix86_finalize_stack_frame_flags): Use it.  Set
28210         max_used_stack_alignment if we don't realign stack.
28211         * config/i386/i386.h (machine_function): Add
28212         max_used_stack_alignment.
28214 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
28216         * config/arm/arm.opt (-mbranch-cost): New option.
28217         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
28218         account.
28220 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
28222         PR target/83629
28223         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
28224         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
28226 2018-01-10  Richard Biener  <rguenther@suse.de>
28228         PR debug/83765
28229         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
28230         early out so it also covers the case where we have a non-NULL
28231         origin.
28233 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28235         PR tree-optimization/83753
28236         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
28237         for non-strided grouped accesses if the number of elements is 1.
28239 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28241         PR target/81616
28242         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
28243         * i386.h (TARGET_USE_GATHER): Define.
28244         * x86-tune.def (X86_TUNE_USE_GATHER): New.
28246 2018-01-10  Martin Liska  <mliska@suse.cz>
28248         PR bootstrap/82831
28249         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
28250         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
28251         partitioning.
28252         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
28253         CLEANUP_NO_PARTITIONING is not set.
28255 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28257         * doc/rtl.texi: Remove documentation of (const ...) wrappers
28258         for vectors, as a partial revert of r254296.
28259         * rtl.h (const_vec_p): Delete.
28260         (const_vec_duplicate_p): Don't test for vector CONSTs.
28261         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
28262         * expmed.c (make_tree): Likewise.
28264         Revert:
28265         * common.md (E, F): Use CONSTANT_P instead of checking for
28266         CONST_VECTOR.
28267         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
28268         checking for CONST_VECTOR.
28270 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28272         PR middle-end/83575
28273         * predict.c (force_edge_cold): Handle in more sane way edges
28274         with no prediction.
28276 2018-01-09  Carl Love  <cel@us.ibm.com>
28278         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
28279         V4SI, V4SF types.
28280         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
28281         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
28282         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
28283         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
28284         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
28285         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
28286         * config/rs6000/rs6000-protos.h: Add extern defition for
28287         rs6000_generate_float2_double_code.
28288         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
28289         function.
28290         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
28291         (float2_v2df): Add define_expand.
28293 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
28295         PR target/83628
28296         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
28297         op_mode in the force_to_mode call.
28299 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28301         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
28302         instead of checking each element individually.
28303         (aarch64_evpc_uzp): Likewise.
28304         (aarch64_evpc_zip): Likewise.
28305         (aarch64_evpc_ext): Likewise.
28306         (aarch64_evpc_rev): Likewise.
28307         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
28308         instead of checking each element individually.  Return true without
28309         generating rtl if
28310         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
28311         whether all selected elements come from the same input, instead of
28312         checking each element individually.  Remove calls to gen_rtx_REG,
28313         start_sequence and end_sequence and instead assert that no rtl is
28314         generated.
28316 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28318         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
28319         order of HIGH and CONST checks.
28321 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28323         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
28324         if the destination isn't an SSA_NAME.
28326 2018-01-09  Richard Biener  <rguenther@suse.de>
28328         PR tree-optimization/83668
28329         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
28330         move prologue...
28331         (canonicalize_loop_form): ... here, renamed from ...
28332         (canonicalize_loop_closed_ssa_form): ... this and amended to
28333         swap successor edges for loop exit blocks to make us use
28334         the RPO order we need for initial schedule generation.
28336 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
28338         PR tree-optimization/64811
28339         * match.pd: When optimizing comparisons with Inf, avoid
28340         introducing or losing exceptions from comparisons with NaN.
28342 2018-01-09  Martin Liska  <mliska@suse.cz>
28344         PR sanitizer/82517
28345         * asan.c (shadow_mem_size): Add gcc_assert.
28347 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
28349         Don't save registers in main().
28351         PR target/83738
28352         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
28353         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
28354         * config/avr/avr.c (avr_set_current_function): Don't error if
28355         naked, OS_task or OS_main are specified at the same time.
28356         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
28357         OS_main.
28358         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
28359         attribute.
28360         * common/config/avr/avr-common.c (avr_option_optimization_table):
28361         Switch on -mmain-is-OS_task for optimizing compilations.
28363 2018-01-09  Richard Biener  <rguenther@suse.de>
28365         PR tree-optimization/83572
28366         * graphite.c: Include cfganal.h.
28367         (graphite_transform_loops): Connect infinite loops to exit
28368         and remove fake edges at the end.
28370 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28372         * ipa-inline.c (edge_badness): Revert accidental checkin.
28374 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28376         PR ipa/80763
28377         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
28378         symbols; not inline clones.
28380 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
28382         PR target/83507
28383         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
28384         hard registers.  Formatting fixes.
28386         PR preprocessor/83722
28387         * gcc.c (try_generate_repro): Pass
28388         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
28389         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
28390         do_report_bug.
28392 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
28393             Kito Cheng  <kito.cheng@gmail.com>
28395         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
28396         (riscv_leaf_function_p): Delete.
28397         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
28399 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28401         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
28402         function.
28403         (do_ifelse): New function.
28404         (do_isel): New function.
28405         (do_sub3): New function.
28406         (do_add3): New function.
28407         (do_load_mask_compare): New function.
28408         (do_overlap_load_compare): New function.
28409         (expand_compare_loop): New function.
28410         (expand_block_compare): Call expand_compare_loop() when appropriate.
28411         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
28412         option description.
28413         (-mblock-compare-inline-loop-limit): New option.
28415 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28417         PR target/83677
28418         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
28419         Reverse order of second and third operands in first alternative.
28420         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
28421         of first and second elements in UNSPEC_VPERMR vector.
28422         (altivec_expand_vec_perm_le): Likewise.
28424 2018-01-08  Jeff Law  <law@redhat.com>
28426         PR rtl-optimizatin/81308
28427         * tree-switch-conversion.c (cfg_altered): New file scoped static.
28428         (process_switch): If group_case_labels makes a change, then set
28429         cfg_altered.
28430         (pass_convert_switch::execute): If a switch is converted, then
28431         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
28433         PR rtl-optimization/81308
28434         * recog.c (split_all_insns): Conditionally cleanup the CFG after
28435         splitting insns.
28437 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
28439         PR target/83663 - Revert r255946
28440         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
28441         generation for cases where splatting a value is not useful.
28442         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
28443         across a vec_duplicate and a paradoxical subreg forming a vector
28444         mode to a vec_concat.
28446 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28448         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
28449         -march=armv8.3-a variants.
28450         * config/arm/t-multilib: Likewise.
28451         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
28453 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28455         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
28456         to generate rtl.
28457         (cceq_ior_compare_complement): Give it a name so I can use it, and
28458         change boolean_or_operator predicate to boolean_operator so it can
28459         be used to generate a crand.
28460         (eqne): New code iterator.
28461         (bd/bd_neg): New code_attrs.
28462         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
28463         a single define_insn.
28464         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
28465         decrement (bdnzt/bdnzf/bdzt/bdzf).
28466         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
28467         with the new names of the branch decrement patterns, and added the
28468         names of the branch decrement conditional patterns.
28470 2018-01-08  Richard Biener  <rguenther@suse.de>
28472         PR tree-optimization/83563
28473         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
28474         cache.
28476 2018-01-08  Richard Biener  <rguenther@suse.de>
28478         PR middle-end/83713
28479         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
28481 2018-01-08  Richard Biener  <rguenther@suse.de>
28483         PR tree-optimization/83685
28484         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
28485         references to abnormals.
28487 2018-01-08  Richard Biener  <rguenther@suse.de>
28489         PR lto/83719
28490         * dwarf2out.c (output_indirect_strings): Handle empty
28491         skeleton_debug_str_hash.
28492         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
28494 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28496         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
28497         (emit_store_direct): Likewise.
28498         (arc_trampoline_adjust_address): Likewise.
28499         (arc_asm_trampoline_template): New function.
28500         (arc_initialize_trampoline): Use asm_trampoline_template.
28501         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
28502         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
28503         * config/arc/arc.md (flush_icache): Delete pattern.
28505 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28507         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
28508         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
28509         munaligned-access.
28511 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28513         PR target/83681
28514         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
28515         by not USED_FOR_TARGET.
28516         (make_pass_resolve_sw_modes): Likewise.
28518 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28520         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
28521         USED_FOR_TARGET.
28523 2018-01-08  Richard Biener  <rguenther@suse.de>
28525         PR middle-end/83580
28526         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
28528 2018-01-08  Richard Biener  <rguenther@suse.de>
28530         PR middle-end/83517
28531         * match.pd ((t * 2) / 2) -> t): Add missing :c.
28533 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
28535         PR middle-end/81897
28536         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
28537         basic blocks with a small number of successors.
28538         (convert_control_dep_chain_into_preds): Improve handling of
28539         forwarder blocks.
28540         (dump_predicates): Split apart into...
28541         (dump_pred_chain): ...here...
28542         (dump_pred_info): ...and here.
28543         (can_one_predicate_be_invalidated_p): Add debugging printfs.
28544         (can_chain_union_be_invalidated_p): Improve check for invalidation
28545         of paths.
28546         (uninit_uses_cannot_happen): Avoid unnecessary if
28547         convert_control_dep_chain_into_preds yielded nothing.
28549 2018-01-06  Martin Sebor  <msebor@redhat.com>
28551         PR tree-optimization/83640
28552         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
28553         subtracting negative offset from size.
28554         (builtin_access::overlap): Adjust offset bounds of the access to fall
28555         within the size of the object if possible.
28557 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
28559         PR rtl-optimization/83699
28560         * expmed.c (extract_bit_field_1): Restrict the vector usage of
28561         extract_bit_field_as_subreg to cases in which the extracted
28562         value is also a vector.
28564         * lra-constraints.c (process_alt_operands): Test for the equivalence
28565         substitutions when detecting a possible reload cycle.
28567 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
28569         PR debug/83480
28570         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
28571         by default if flag_selective_schedling{,2}.  Formatting fixes.
28573         PR rtl-optimization/83682
28574         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
28575         if it has non-VECTOR_MODE element mode.
28576         (vec_duplicate_p): Likewise.
28578         PR middle-end/83694
28579         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
28580         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
28582 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28584         PR target/83604
28585         * config/i386/i386-builtin.def
28586         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
28587         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
28588         Require also OPTION_MASK_ISA_AVX512F in addition to
28589         OPTION_MASK_ISA_GFNI.
28590         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
28591         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
28592         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
28593         to OPTION_MASK_ISA_GFNI.
28594         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
28595         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
28596         OPTION_MASK_ISA_AVX512BW.
28597         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
28598         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
28599         addition to OPTION_MASK_ISA_GFNI.
28600         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
28601         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
28602         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
28603         to OPTION_MASK_ISA_GFNI.
28604         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
28605         a requirement for all ISAs rather than any of them with a few
28606         exceptions.
28607         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
28608         processing.
28609         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
28610         bitmasks to be enabled with 3 exceptions, instead of requiring any
28611         enabled ISA with lots of exceptions.
28612         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
28613         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
28614         Change avx512bw in isa attribute to avx512f.
28615         * config/i386/sgxintrin.h: Add license boilerplate.
28616         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
28617         to __AVX512F__ and __AVX512VL to __AVX512VL__.
28618         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
28619         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
28620         defined.
28621         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
28622         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28623         temporarily sse2 rather than sse if not enabled already.
28625         PR target/83604
28626         * config/i386/sse.md (VI248_VLBW): Rename to ...
28627         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
28628         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28629         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28630         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28631         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28632         mode iterator instead of VI248_VLBW.
28634 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
28636         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28637         (record_modified): Skip clobbers; add debug output.
28638         (param_change_prob): Use sreal frequencies.
28640 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28642         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28643         punt for user-aligned variables.
28645 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28647         * tree-chrec.c (chrec_contains_symbols): Return true for
28648         POLY_INT_CST.
28650 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
28652         PR target/82439
28653         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28654         of (x|y) == x for BICS pattern.
28656 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28658         PR tree-optimization/83605
28659         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28660         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28661         can throw.
28663 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28665         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28666         * config/epiphany/rtems.h: New file.
28668 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28669             Uros Bizjak  <ubizjak@gmail.com>
28671         PR target/83554
28672         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28673         QIreg_operand instead of register_operand predicate.
28674         * config/i386/i386.c (ix86_rop_should_change_byte_p,
28675         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28676         comments instead of -fmitigate[-_]rop.
28678 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28680         PR bootstrap/81926
28681         * cgraphunit.c (symbol_table::compile): Switch to text_section
28682         before calling assembly_start debug hook.
28683         * run-rtl-passes.c (run_rtl_passes): Likewise.
28684         Include output.h.
28686 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28688         * tree-vrp.c (extract_range_from_binary_expr_1): Check
28689         range_int_cst_p rather than !symbolic_range_p before calling
28690         extract_range_from_multiplicative_op_1.
28692 2018-01-04  Jeff Law  <law@redhat.com>
28694         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28695         redundant test in assertion.
28697 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28699         * doc/rtl.texi: Document machine_mode wrapper classes.
28701 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28703         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28704         using tree_to_uhwi.
28706 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28708         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28709         the VEC_PERM_EXPR fold to fail.
28711 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28713         PR debug/83585
28714         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28715         to switched_sections.
28717 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28719         PR target/83680
28720         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28721         test for d.testing.
28723 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
28725         PR target/83387
28726         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28727         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28729 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28731         PR debug/83666
28732         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28733         is BLKmode and bitpos not zero or mode change is needed.
28735 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28737         PR target/83675
28738         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28739         TARGET_VIS2.
28741 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
28743         PR target/83628
28744         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28745         instead of MULT rtx.  Update all corresponding splitters.
28746         (*saddl_se): Ditto.
28747         (*ssub<modesuffix>): Ditto.
28748         (*ssubl_se): Ditto.
28749         (*cmp_sadd_di): Update split patterns.
28750         (*cmp_sadd_si): Ditto.
28751         (*cmp_sadd_sidi): Ditto.
28752         (*cmp_ssub_di): Ditto.
28753         (*cmp_ssub_si): Ditto.
28754         (*cmp_ssub_sidi): Ditto.
28755         * config/alpha/predicates.md (const23_operand): New predicate.
28756         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28757         Look for ASHIFT, not MULT inner operand.
28758         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28760 2018-01-04  Martin Liska  <mliska@suse.cz>
28762         PR gcov-profile/83669
28763         * gcov.c (output_intermediate_file): Add version to intermediate
28764         gcov file.
28765         * doc/gcov.texi: Document new field 'version' in intermediate
28766         file format. Fix location of '-k' option of gcov command.
28768 2018-01-04  Martin Liska  <mliska@suse.cz>
28770         PR ipa/82352
28771         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28773 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28775         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28777 2018-01-03  Martin Sebor  <msebor@redhat.com>
28779         PR tree-optimization/83655
28780         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28781         checking calls with invalid arguments.
28783 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28785         * tree-vect-stmts.c (vect_get_store_rhs): New function.
28786         (vectorizable_mask_load_store): Delete.
28787         (vectorizable_call): Return false for masked loads and stores.
28788         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
28789         instead of gimple_assign_rhs1.
28790         (vectorizable_load): Handle IFN_MASK_LOAD.
28791         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28793 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28795         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28796         split out from..,
28797         (vectorizable_mask_load_store): ...here.
28798         (vectorizable_load): ...and here.
28800 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28802         * tree-vect-stmts.c (vect_build_all_ones_mask)
28803         (vect_build_zero_merge_argument): New functions, split out from...
28804         (vectorizable_load): ...here.
28806 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28808         * tree-vect-stmts.c (vect_check_store_rhs): New function,
28809         split out from...
28810         (vectorizable_mask_load_store): ...here.
28811         (vectorizable_store): ...and here.
28813 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28815         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28816         split out from...
28817         (vectorizable_mask_load_store): ...here.
28819 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28821         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28822         (vect_model_store_cost): Take a vec_load_store_type instead of a
28823         vect_def_type.
28824         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28825         (vect_model_store_cost): Take a vec_load_store_type instead of a
28826         vect_def_type.
28827         (vectorizable_mask_load_store): Update accordingly.
28828         (vectorizable_store): Likewise.
28829         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28831 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28833         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28834         IFN_MASK_LOAD calls here rather than...
28835         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28837 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28838             Alan Hayward  <alan.hayward@arm.com>
28839             David Sherwood  <david.sherwood@arm.com>
28841         * expmed.c (extract_bit_field_1): For vector extracts,
28842         fall back to extract_bit_field_as_subreg if vec_extract
28843         isn't available.
28845 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28846             Alan Hayward  <alan.hayward@arm.com>
28847             David Sherwood  <david.sherwood@arm.com>
28849         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28850         they are variable or constant sized.
28851         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28852         slots for constant-sized data.
28854 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28855             Alan Hayward  <alan.hayward@arm.com>
28856             David Sherwood  <david.sherwood@arm.com>
28858         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28859         handling COND_EXPRs with boolean comparisons, try to find a better
28860         basis for the mask type than the boolean itself.
28862 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28864         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28865         is calculated and how it can be overridden.
28866         * genmodes.c (max_bitsize_mode_any_mode): New variable.
28867         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28868         if defined.
28869         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28870         if nonzero.
28872 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28873             Alan Hayward  <alan.hayward@arm.com>
28874             David Sherwood  <david.sherwood@arm.com>
28876         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28877         Remove the mode argument.
28878         (aarch64_simd_valid_immediate): Remove the mode and inverse
28879         arguments.
28880         * config/aarch64/iterators.md (bitsize): New iterator.
28881         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28882         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28883         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28884         aarch64_simd_valid_immediate.
28885         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28886         (aarch64_reg_or_bic_imm): Likewise.
28887         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28888         with an insn_type enum and msl with a modifier_type enum.
28889         Replace element_width with a scalar_mode.  Change the shift
28890         to unsigned int.  Add constructors for scalar_float_mode and
28891         scalar_int_mode elements.
28892         (aarch64_vect_float_const_representable_p): Delete.
28893         (aarch64_can_const_movi_rtx_p)
28894         (aarch64_simd_scalar_immediate_valid_for_move)
28895         (aarch64_simd_make_constant): Update call to
28896         aarch64_simd_valid_immediate.
28897         (aarch64_advsimd_valid_immediate_hs): New function.
28898         (aarch64_advsimd_valid_immediate): Likewise.
28899         (aarch64_simd_valid_immediate): Remove mode and inverse
28900         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
28901         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28902         and aarch64_float_const_representable_p on the result.
28903         (aarch64_output_simd_mov_immediate): Remove mode argument.
28904         Update call to aarch64_simd_valid_immediate and use of
28905         simd_immediate_info.
28906         (aarch64_output_scalar_simd_mov_immediate): Update call
28907         accordingly.
28909 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28910             Alan Hayward  <alan.hayward@arm.com>
28911             David Sherwood  <david.sherwood@arm.com>
28913         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28914         (mode_nunits): Likewise CONST_MODE_NUNITS.
28915         * machmode.def (ADJUST_NUNITS): Document.
28916         * genmodes.c (mode_data::need_nunits_adj): New field.
28917         (blank_mode): Update accordingly.
28918         (adj_nunits): New variable.
28919         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28920         parameter.
28921         (emit_mode_size_inline): Set need_bytesize_adj for all modes
28922         listed in adj_nunits.
28923         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28924         listed in adj_nunits.  Don't emit case statements for such modes.
28925         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28926         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
28927         nothing if adj_nunits is nonnull.
28928         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28929         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28930         (emit_mode_fbit): Update use of print_maybe_const_decl.
28931         (emit_move_size): Likewise.  Treat the array as non-const
28932         if adj_nunits.
28933         (emit_mode_adjustments): Handle adj_nunits.
28935 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28937         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28938         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28939         (VECTOR_MODES): Use it.
28940         (make_vector_modes): Take the prefix as an argument.
28942 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28943             Alan Hayward  <alan.hayward@arm.com>
28944             David Sherwood  <david.sherwood@arm.com>
28946         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28947         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28948         for MODE_VECTOR_BOOL.
28949         * machmode.def (VECTOR_BOOL_MODE): Document.
28950         * genmodes.c (VECTOR_BOOL_MODE): New macro.
28951         (make_vector_bool_mode): New function.
28952         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28953         MODE_VECTOR_BOOL.
28954         * lto-streamer-in.c (lto_input_mode_table): Likewise.
28955         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28956         Likewise.
28957         * stor-layout.c (int_mode_for_mode): Likewise.
28958         * tree.c (build_vector_type_for_mode): Likewise.
28959         * varasm.c (output_constant_pool_2): Likewise.
28960         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28961         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
28962         for MODE_VECTOR_BOOL.
28963         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28964         of mode class checks.
28965         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28966         instead of a list of mode class checks.
28967         (expand_vector_scalar_condition): Likewise.
28968         (type_for_widest_vector_mode): Handle BImode as an inner mode.
28970 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28971             Alan Hayward  <alan.hayward@arm.com>
28972             David Sherwood  <david.sherwood@arm.com>
28974         * machmode.h (mode_size): Change from unsigned short to
28975         poly_uint16_pod.
28976         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28977         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28978         or if measurement_type is not polynomial.
28979         (fixed_size_mode::includes_p): Check for constant-sized modes.
28980         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28981         return a poly_uint16 rather than an unsigned short.
28982         (emit_mode_size): Change the type of mode_size from unsigned short
28983         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
28984         (emit_mode_adjustments): Cope with polynomial vector sizes.
28985         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28986         for GET_MODE_SIZE.
28987         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28988         for GET_MODE_SIZE.
28989         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28990         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28991         * caller-save.c (setup_save_areas): Likewise.
28992         (replace_reg_with_saved_mem): Likewise.
28993         * calls.c (emit_library_call_value_1): Likewise.
28994         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28995         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28996         (gen_lowpart_for_combine): Likewise.
28997         * convert.c (convert_to_integer_1): Likewise.
28998         * cse.c (equiv_constant, cse_insn): Likewise.
28999         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
29000         (cselib_subst_to_values): Likewise.
29001         * dce.c (word_dce_process_block): Likewise.
29002         * df-problems.c (df_word_lr_mark_ref): Likewise.
29003         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
29004         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
29005         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
29006         (rtl_for_decl_location): Likewise.
29007         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
29008         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
29009         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
29010         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
29011         (expand_expr_real_1): Likewise.
29012         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
29013         (pad_below): Likewise.
29014         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29015         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
29016         * ira.c (get_subreg_tracking_sizes): Likewise.
29017         * ira-build.c (ira_create_allocno_objects): Likewise.
29018         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
29019         (ira_sort_regnos_for_alter_reg): Likewise.
29020         * ira-costs.c (record_operand_costs): Likewise.
29021         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
29022         (resolve_simple_move): Likewise.
29023         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
29024         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
29025         (lra_constraints): Likewise.
29026         (CONST_POOL_OK_P): Reject variable-sized modes.
29027         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
29028         (add_pseudo_to_slot, lra_spill): Likewise.
29029         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29030         * optabs-query.c (get_best_extraction_insn): Likewise.
29031         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29032         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
29033         (expand_mult_highpart, valid_multiword_target_p): Likewise.
29034         * recog.c (offsettable_address_addr_space_p): Likewise.
29035         * regcprop.c (maybe_mode_change): Likewise.
29036         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
29037         * regrename.c (build_def_use): Likewise.
29038         * regstat.c (dump_reg_info): Likewise.
29039         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
29040         (find_reloads, find_reloads_subreg_address): Likewise.
29041         * reload1.c (eliminate_regs_1): Likewise.
29042         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
29043         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
29044         (simplify_binary_operation_1, simplify_subreg): Likewise.
29045         * targhooks.c (default_function_arg_padding): Likewise.
29046         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
29047         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
29048         (verify_gimple_assign_ternary): Likewise.
29049         * tree-inline.c (estimate_move_cost): Likewise.
29050         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29051         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
29052         (get_address_cost_ainc): Likewise.
29053         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
29054         (vect_supportable_dr_alignment): Likewise.
29055         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29056         (vectorizable_reduction): Likewise.
29057         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
29058         (vectorizable_operation, vectorizable_load): Likewise.
29059         * tree.c (build_same_sized_truth_vector_type): Likewise.
29060         * valtrack.c (cleanup_auto_inc_dec): Likewise.
29061         * var-tracking.c (emit_note_insn_var_location): Likewise.
29062         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
29063         (ADDR_VEC_ALIGN): Likewise.
29065 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29066             Alan Hayward  <alan.hayward@arm.com>
29067             David Sherwood  <david.sherwood@arm.com>
29069         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
29070         unsigned short.
29071         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
29072         or if measurement_type is polynomial.
29073         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
29074         * combine.c (make_extraction): Likewise.
29075         * dse.c (find_shift_sequence): Likewise.
29076         * dwarf2out.c (mem_loc_descriptor): Likewise.
29077         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
29078         (extract_bit_field, extract_low_bits): Likewise.
29079         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
29080         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
29081         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
29082         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
29083         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29084         * reload.c (find_reloads): Likewise.
29085         * reload1.c (alter_reg): Likewise.
29086         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
29087         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
29088         * tree-if-conv.c (predicate_mem_writes): Likewise.
29089         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
29090         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
29091         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
29092         * valtrack.c (dead_debug_insert_temp): Likewise.
29093         * varasm.c (mergeable_constant_section): Likewise.
29094         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
29096 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29097             Alan Hayward  <alan.hayward@arm.com>
29098             David Sherwood  <david.sherwood@arm.com>
29100         * expr.c (expand_assignment): Cope with polynomial mode sizes
29101         when assigning to a CONCAT.
29103 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29104             Alan Hayward  <alan.hayward@arm.com>
29105             David Sherwood  <david.sherwood@arm.com>
29107         * machmode.h (mode_precision): Change from unsigned short to
29108         poly_uint16_pod.
29109         (mode_to_precision): Return a poly_uint16 rather than an unsigned
29110         short.
29111         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
29112         or if measurement_type is not polynomial.
29113         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
29114         in which the mode is already known to be a scalar_int_mode.
29115         * genmodes.c (emit_mode_precision): Change the type of mode_precision
29116         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
29117         initializer.
29118         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29119         for GET_MODE_PRECISION.
29120         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29121         for GET_MODE_PRECISION.
29122         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
29123         as polynomial.
29124         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
29125         (expand_field_assignment, make_extraction): Likewise.
29126         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
29127         (get_last_value): Likewise.
29128         * convert.c (convert_to_integer_1): Likewise.
29129         * cse.c (cse_insn): Likewise.
29130         * expr.c (expand_expr_real_1): Likewise.
29131         * lra-constraints.c (simplify_operand_subreg): Likewise.
29132         * optabs-query.c (can_atomic_load_p): Likewise.
29133         * optabs.c (expand_atomic_load): Likewise.
29134         (expand_atomic_store): Likewise.
29135         * ree.c (combine_reaching_defs): Likewise.
29136         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
29137         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
29138         * tree.h (type_has_mode_precision_p): Likewise.
29139         * ubsan.c (instrument_si_overflow): Likewise.
29141 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29142             Alan Hayward  <alan.hayward@arm.com>
29143             David Sherwood  <david.sherwood@arm.com>
29145         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
29146         polynomial numbers of units.
29147         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
29148         (valid_vector_subparts_p): New function.
29149         (build_vector_type): Remove temporary shim and take the number
29150         of units as a poly_uint64 rather than an int.
29151         (build_opaque_vector_type): Take the number of units as a
29152         poly_uint64 rather than an int.
29153         * tree.c (build_vector_from_ctor): Handle polynomial
29154         TYPE_VECTOR_SUBPARTS.
29155         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
29156         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
29157         (build_vector_from_val): If the number of units is variable,
29158         use build_vec_duplicate_cst for constant operands and
29159         VEC_DUPLICATE_EXPR otherwise.
29160         (make_vector_type): Remove temporary is_constant ().
29161         (build_vector_type, build_opaque_vector_type): Take the number of
29162         units as a poly_uint64 rather than an int.
29163         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
29164         VECTOR_CST_NELTS.
29165         * cfgexpand.c (expand_debug_expr): Likewise.
29166         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
29167         (store_constructor, expand_expr_real_1): Likewise.
29168         (const_scalar_mask_from_tree): Likewise.
29169         * fold-const-call.c (fold_const_reduction): Likewise.
29170         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
29171         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
29172         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
29173         (fold_relational_const): Likewise.
29174         (native_interpret_vector): Likewise.  Change the size from an
29175         int to an unsigned int.
29176         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
29177         TYPE_VECTOR_SUBPARTS.
29178         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
29179         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
29180         duplicating a non-constant operand into a variable-length vector.
29181         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
29182         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
29183         * ipa-icf.c (sem_variable::equals): Likewise.
29184         * match.pd: Likewise.
29185         * omp-simd-clone.c (simd_clone_subparts): Likewise.
29186         * print-tree.c (print_node): Likewise.
29187         * stor-layout.c (layout_type): Likewise.
29188         * targhooks.c (default_builtin_vectorization_cost): Likewise.
29189         * tree-cfg.c (verify_gimple_comparison): Likewise.
29190         (verify_gimple_assign_binary): Likewise.
29191         (verify_gimple_assign_ternary): Likewise.
29192         (verify_gimple_assign_single): Likewise.
29193         * tree-pretty-print.c (dump_generic_node): Likewise.
29194         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29195         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
29196         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
29197         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
29198         (vect_shift_permute_load_chain): Likewise.
29199         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
29200         (expand_vector_condition, optimize_vector_constructor): Likewise.
29201         (lower_vec_perm, get_compute_type): Likewise.
29202         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29203         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
29204         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
29205         (vect_recog_mask_conversion_pattern): Likewise.
29206         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
29207         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
29208         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29209         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
29210         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
29211         (vectorizable_shift, vectorizable_operation, vectorizable_store)
29212         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
29213         (supportable_widening_operation): Likewise.
29214         (supportable_narrowing_operation): Likewise.
29215         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
29216         Likewise.
29217         * varasm.c (output_constant): Likewise.
29219 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29220             Alan Hayward  <alan.hayward@arm.com>
29221             David Sherwood  <david.sherwood@arm.com>
29223         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
29224         so that both the length == 3 and length != 3 cases set up their
29225         own permute vectors.  Add comments explaining why we know the
29226         number of elements is constant.
29227         (vect_permute_load_chain): Likewise.
29229 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29230             Alan Hayward  <alan.hayward@arm.com>
29231             David Sherwood  <david.sherwood@arm.com>
29233         * machmode.h (mode_nunits): Change from unsigned char to
29234         poly_uint16_pod.
29235         (ONLY_FIXED_SIZE_MODES): New macro.
29236         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
29237         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
29238         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
29239         New typedefs.
29240         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
29241         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
29242         or if measurement_type is not polynomial.
29243         * genmodes.c (ZERO_COEFFS): New macro.
29244         (emit_mode_nunits_inline): Make mode_nunits_inline return a
29245         poly_uint16.
29246         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
29247         Use ZERO_COEFFS when emitting initializers.
29248         * data-streamer.h (bp_pack_poly_value): New function.
29249         (bp_unpack_poly_value): Likewise.
29250         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29251         for GET_MODE_NUNITS.
29252         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29253         for GET_MODE_NUNITS.
29254         * tree.c (make_vector_type): Remove temporary shim and make
29255         the real function take the number of units as a poly_uint64
29256         rather than an int.
29257         (build_vector_type_for_mode): Handle polynomial nunits.
29258         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
29259         * emit-rtl.c (const_vec_series_p_1): Likewise.
29260         (gen_rtx_CONST_VECTOR): Likewise.
29261         * fold-const.c (test_vec_duplicate_folding): Likewise.
29262         * genrecog.c (validate_pattern): Likewise.
29263         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
29264         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29265         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
29266         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
29267         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
29268         * rtlanal.c (subreg_get_info): Likewise.
29269         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29270         (vect_grouped_load_supported): Likewise.
29271         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
29272         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
29273         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
29274         (simplify_const_unary_operation, simplify_binary_operation_1)
29275         (simplify_const_binary_operation, simplify_ternary_operation)
29276         (test_vector_ops_duplicate, test_vector_ops): Likewise.
29277         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
29278         instead of CONST_VECTOR_NUNITS.
29279         * varasm.c (output_constant_pool_2): Likewise.
29280         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
29281         explicit-encoded elements in the XVEC for variable-length vectors.
29283 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29285         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
29287 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29288             Alan Hayward  <alan.hayward@arm.com>
29289             David Sherwood  <david.sherwood@arm.com>
29291         * coretypes.h (fixed_size_mode): Declare.
29292         (fixed_size_mode_pod): New typedef.
29293         * builtins.h (target_builtins::x_apply_args_mode)
29294         (target_builtins::x_apply_result_mode): Change type to
29295         fixed_size_mode_pod.
29296         * builtins.c (apply_args_size, apply_result_size, result_vector)
29297         (expand_builtin_apply_args_1, expand_builtin_apply)
29298         (expand_builtin_return): Update accordingly.
29300 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29302         * cse.c (hash_rtx_cb): Hash only the encoded elements.
29303         * cselib.c (cselib_hash_rtx): Likewise.
29304         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
29305         CONST_VECTOR encoding.
29307 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29308             Jeff Law  <law@redhat.com>
29310         PR target/83641
29311         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
29312         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
29313         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
29314         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
29316         PR target/83641
29317         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
29318         explicitly probe *sp in a noreturn function if there were any callee
29319         register saves or frame pointer is needed.
29321 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29323         PR debug/83621
29324         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
29325         BLKmode for ternary, binary or unary expressions.
29327         PR debug/83645
29328         * var-tracking.c (delete_vta_debug_insn): New inline function.
29329         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
29330         insns from get_insns () to NULL instead of each bb separately.
29331         Use delete_vta_debug_insn.  No longer static.
29332         (vt_debug_insns_local, variable_tracking_main_1): Adjust
29333         delete_vta_debug_insns callers.
29334         * rtl.h (delete_vta_debug_insns): Declare.
29335         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
29336         instead of variable_tracking_main.
29338 2018-01-03  Martin Sebor  <msebor@redhat.com>
29340         PR tree-optimization/83603
29341         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
29342         arguments past the endof the argument list in functions declared
29343         without a prototype.
29344         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
29345         Avoid checking when arguments are null.
29347 2018-01-03  Martin Sebor  <msebor@redhat.com>
29349         PR c/83559
29350         * doc/extend.texi (attribute const): Fix a typo.
29351         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
29352         issuing -Wsuggest-attribute for void functions.
29354 2018-01-03  Martin Sebor  <msebor@redhat.com>
29356         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
29357         offset_int::from instead of wide_int::to_shwi.
29358         (maybe_diag_overlap): Remove assertion.
29359         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
29360         * gimple-ssa-sprintf.c (format_directive): Same.
29361         (parse_directive): Same.
29362         (sprintf_dom_walker::compute_format_length): Same.
29363         (try_substitute_return_value): Same.
29365 2018-01-03  Jeff Law  <law@redhat.com>
29367         PR middle-end/83654
29368         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
29369         non-constant residual for zero at runtime and avoid probing in
29370         that case.  Reorganize code for trailing problem to mirror handling
29371         of the residual.
29373 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29375         PR tree-optimization/83501
29376         * tree-ssa-strlen.c (get_string_cst): New.
29377         (handle_char_store): Call get_string_cst.
29379 2018-01-03  Martin Liska  <mliska@suse.cz>
29381         PR tree-optimization/83593
29382         * tree-ssa-strlen.c: Include tree-cfg.h.
29383         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
29384         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
29385         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
29386         to false.
29387         (strlen_dom_walker::before_dom_children): Call
29388         gimple_purge_dead_eh_edges. Dump tranformation with details
29389         dump flags.
29390         (strlen_dom_walker::before_dom_children): Update call by adding
29391         new argument cleanup_eh.
29392         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
29394 2018-01-03  Martin Liska  <mliska@suse.cz>
29396         PR ipa/83549
29397         * cif-code.def (VARIADIC_THUNK): New enum value.
29398         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
29399         thunks.
29401 2018-01-03  Jan Beulich  <jbeulich@suse.com>
29403         * sse.md (mov<mode>_internal): Tighten condition for when to use
29404         vmovdqu<ssescalarsize> for TI and OI modes.
29406 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29408         Update copyright years.
29410 2018-01-03  Martin Liska  <mliska@suse.cz>
29412         PR ipa/83594
29413         * ipa-visibility.c (function_and_variable_visibility): Skip
29414         functions with noipa attribure.
29416 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29418         * gcc.c (process_command): Update copyright notice dates.
29419         * gcov-dump.c (print_version): Ditto.
29420         * gcov.c (print_version): Ditto.
29421         * gcov-tool.c (print_version): Ditto.
29422         * gengtype.c (create_file): Ditto.
29423         * doc/cpp.texi: Bump @copying's copyright year.
29424         * doc/cppinternals.texi: Ditto.
29425         * doc/gcc.texi: Ditto.
29426         * doc/gccint.texi: Ditto.
29427         * doc/gcov.texi: Ditto.
29428         * doc/install.texi: Ditto.
29429         * doc/invoke.texi: Ditto.
29431 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29433         * vector-builder.h (vector_builder::m_full_nelts): Change from
29434         unsigned int to poly_uint64.
29435         (vector_builder::full_nelts): Update prototype accordingly.
29436         (vector_builder::new_vector): Likewise.
29437         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
29438         (vector_builder::operator ==): Likewise.
29439         (vector_builder::finalize): Likewise.
29440         * int-vector-builder.h (int_vector_builder::int_vector_builder):
29441         Take the number of elements as a poly_uint64 rather than an
29442         unsigned int.
29443         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
29444         from unsigned int to poly_uint64.
29445         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
29446         (vec_perm_indices::new_vector): Likewise.
29447         (vec_perm_indices::length): Likewise.
29448         (vec_perm_indices::nelts_per_input): Likewise.
29449         (vec_perm_indices::input_nelts): Likewise.
29450         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
29451         number of elements per input as a poly_uint64 rather than an
29452         unsigned int.  Use the original encoding for variable-length
29453         vectors, rather than clamping each individual element.
29454         For the second and subsequent elements in each pattern,
29455         clamp the step and base before clamping their sum.
29456         (vec_perm_indices::series_p): Handle polynomial element counts.
29457         (vec_perm_indices::all_in_range_p): Likewise.
29458         (vec_perm_indices_to_tree): Likewise.
29459         (vec_perm_indices_to_rtx): Likewise.
29460         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
29461         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
29462         (tree_vector_builder::new_binary_operation): Handle polynomial
29463         element counts.  Return false if we need to know the number
29464         of elements at compile time.
29465         * fold-const.c (fold_vec_perm): Punt if the number of elements
29466         isn't known at compile time.
29468 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29470         * vec-perm-indices.h (vec_perm_builder): Change element type
29471         from HOST_WIDE_INT to poly_int64.
29472         (vec_perm_indices::element_type): Update accordingly.
29473         (vec_perm_indices::clamp): Handle polynomial element_types.
29474         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29475         (vec_perm_indices::all_in_range_p): Likewise.
29476         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
29477         than shwi trees.
29478         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
29479         polynomial vec_perm_indices element types.
29480         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
29481         * fold-const.c (fold_vec_perm): Likewise.
29482         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
29483         * tree-vect-generic.c (lower_vec_perm): Likewise.
29484         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29485         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
29486         element type to HOST_WIDE_INT.
29488 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29489             Alan Hayward  <alan.hayward@arm.com>
29490             David Sherwood  <david.sherwood@arm.com>
29492         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
29493         rather than an int.  Use plus_constant.
29494         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
29495         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
29497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29498             Alan Hayward  <alan.hayward@arm.com>
29499             David Sherwood  <david.sherwood@arm.com>
29501         * calls.c (emit_call_1, expand_call): Change struct_value_size from
29502         a HOST_WIDE_INT to a poly_int64.
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         * calls.c (load_register_parameters): Cope with polynomial
29509         mode sizes.  Require a constant size for BLKmode parameters
29510         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
29511         forces a parameter to be padded at the lsb end in order to
29512         fill a complete number of words, require the parameter size
29513         to be ordered wrt UNITS_PER_WORD.
29515 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29516             Alan Hayward  <alan.hayward@arm.com>
29517             David Sherwood  <david.sherwood@arm.com>
29519         * reload1.c (spill_stack_slot_width): Change element type
29520         from unsigned int to poly_uint64_pod.
29521         (alter_reg): Treat mode sizes as polynomial.
29523 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29524             Alan Hayward  <alan.hayward@arm.com>
29525             David Sherwood  <david.sherwood@arm.com>
29527         * reload.c (complex_word_subreg_p): New function.
29528         (reload_inner_reg_of_subreg, push_reload): Use it.
29530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29531             Alan Hayward  <alan.hayward@arm.com>
29532             David Sherwood  <david.sherwood@arm.com>
29534         * lra-constraints.c (process_alt_operands): Reject matched
29535         operands whose sizes aren't ordered.
29536         (match_reload): Refer to this check here.
29538 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29539             Alan Hayward  <alan.hayward@arm.com>
29540             David Sherwood  <david.sherwood@arm.com>
29542         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
29543         that the mode size is in the set {1, 2, 4, 8, 16}.
29545 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29546             Alan Hayward  <alan.hayward@arm.com>
29547             David Sherwood  <david.sherwood@arm.com>
29549         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
29550         Use plus_constant instead of gen_rtx_PLUS.
29552 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29553             Alan Hayward  <alan.hayward@arm.com>
29554             David Sherwood  <david.sherwood@arm.com>
29556         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
29557         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
29558         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
29559         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
29560         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
29561         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
29562         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
29563         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
29564         * config/i386/i386.c (ix86_push_rounding): ...this new function.
29565         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29566         a poly_int64.
29567         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
29568         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
29569         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
29570         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
29571         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
29572         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
29573         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
29574         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
29575         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
29576         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
29577         function.
29578         * expr.c (emit_move_resolve_push): Treat the input and result
29579         of PUSH_ROUNDING as a poly_int64.
29580         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
29581         (emit_push_insn): Likewise.
29582         * lra-eliminations.c (mark_not_eliminable): Likewise.
29583         * recog.c (push_operand): Likewise.
29584         * reload1.c (elimination_effects): Likewise.
29585         * rtlanal.c (nonzero_bits1): Likewise.
29586         * calls.c (store_one_arg): Likewise.  Require the padding to be
29587         known at compile time.
29589 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29590             Alan Hayward  <alan.hayward@arm.com>
29591             David Sherwood  <david.sherwood@arm.com>
29593         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
29594         Use plus_constant instead of gen_rtx_PLUS.
29596 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29597             Alan Hayward  <alan.hayward@arm.com>
29598             David Sherwood  <david.sherwood@arm.com>
29600         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
29601         rather than an int.
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         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
29608         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
29609         via stack temporaries.  Treat the mode size as polynomial too.
29611 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29612             Alan Hayward  <alan.hayward@arm.com>
29613             David Sherwood  <david.sherwood@arm.com>
29615         * expr.c (expand_expr_real_2): When handling conversions involving
29616         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
29617         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
29618         as a poly_uint64 too.
29620 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29621             Alan Hayward  <alan.hayward@arm.com>
29622             David Sherwood  <david.sherwood@arm.com>
29624         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
29626 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29627             Alan Hayward  <alan.hayward@arm.com>
29628             David Sherwood  <david.sherwood@arm.com>
29630         * combine.c (can_change_dest_mode): Handle polynomial
29631         REGMODE_NATURAL_SIZE.
29632         * expmed.c (store_bit_field_1): Likewise.
29633         * expr.c (store_constructor): Likewise.
29634         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29635         and polynomial REGMODE_NATURAL_SIZE.
29636         (gen_lowpart_common): Likewise.
29637         * reginfo.c (record_subregs_of_mode): Likewise.
29638         * rtlanal.c (read_modify_subreg_p): Likewise.
29640 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29641             Alan Hayward  <alan.hayward@arm.com>
29642             David Sherwood  <david.sherwood@arm.com>
29644         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29645         numbers of elements.
29647 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29648             Alan Hayward  <alan.hayward@arm.com>
29649             David Sherwood  <david.sherwood@arm.com>
29651         * match.pd: Cope with polynomial numbers of vector elements.
29653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29654             Alan Hayward  <alan.hayward@arm.com>
29655             David Sherwood  <david.sherwood@arm.com>
29657         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29658         in a POINTER_PLUS_EXPR.
29660 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29661             Alan Hayward  <alan.hayward@arm.com>
29662             David Sherwood  <david.sherwood@arm.com>
29664         * omp-simd-clone.c (simd_clone_subparts): New function.
29665         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29666         (ipa_simd_modify_function_body): Likewise.
29668 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29669             Alan Hayward  <alan.hayward@arm.com>
29670             David Sherwood  <david.sherwood@arm.com>
29672         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29673         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29674         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29675         (expand_vector_condition, vector_element): Likewise.
29676         (subparts_gt): New function.
29677         (get_compute_type): Use subparts_gt.
29678         (count_type_subparts): Delete.
29679         (expand_vector_operations_1): Use subparts_gt instead of
29680         count_type_subparts.
29682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29683             Alan Hayward  <alan.hayward@arm.com>
29684             David Sherwood  <david.sherwood@arm.com>
29686         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29687         (vect_compile_time_alias): ...this new function.  Do the calculation
29688         on poly_ints rather than trees.
29689         (vect_prune_runtime_alias_test_list): Update call accordingly.
29691 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29692             Alan Hayward  <alan.hayward@arm.com>
29693             David Sherwood  <david.sherwood@arm.com>
29695         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29696         numbers of units.
29697         (vect_schedule_slp_instance): Likewise.
29699 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29700             Alan Hayward  <alan.hayward@arm.com>
29701             David Sherwood  <david.sherwood@arm.com>
29703         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29704         constant and extern definitions for variable-length vectors.
29705         (vect_get_constant_vectors): Note that the number of units
29706         is known to be constant.
29708 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29709             Alan Hayward  <alan.hayward@arm.com>
29710             David Sherwood  <david.sherwood@arm.com>
29712         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29713         of units as polynomial.  Choose between WIDE and NARROW based
29714         on multiple_p.
29716 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29717             Alan Hayward  <alan.hayward@arm.com>
29718             David Sherwood  <david.sherwood@arm.com>
29720         * tree-vect-stmts.c (simd_clone_subparts): New function.
29721         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
29723 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29724             Alan Hayward  <alan.hayward@arm.com>
29725             David Sherwood  <david.sherwood@arm.com>
29727         * tree-vect-stmts.c (vectorizable_call): Treat the number of
29728         vectors as polynomial.  Use build_index_vector for
29729         IFN_GOMP_SIMD_LANE.
29731 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29732             Alan Hayward  <alan.hayward@arm.com>
29733             David Sherwood  <david.sherwood@arm.com>
29735         * tree-vect-stmts.c (get_load_store_type): Treat the number of
29736         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29737         for variable-length vectors.
29738         (vectorizable_mask_load_store): Treat the number of units as
29739         polynomial, asserting that it is constant if the condition has
29740         already been enforced.
29741         (vectorizable_store, vectorizable_load): Likewise.
29743 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29744             Alan Hayward  <alan.hayward@arm.com>
29745             David Sherwood  <david.sherwood@arm.com>
29747         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29748         of units as polynomial.  Punt if we can't tell at compile time
29749         which vector contains the final result.
29751 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29752             Alan Hayward  <alan.hayward@arm.com>
29753             David Sherwood  <david.sherwood@arm.com>
29755         * tree-vect-loop.c (vectorizable_induction): Treat the number
29756         of units as polynomial.  Punt on SLP inductions.  Use an integer
29757         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
29758         cast of such a series for variable-length floating-point
29759         reductions.
29761 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29762             Alan Hayward  <alan.hayward@arm.com>
29763             David Sherwood  <david.sherwood@arm.com>
29765         * tree.h (build_index_vector): Declare.
29766         * tree.c (build_index_vector): New function.
29767         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29768         of units as polynomial, forcibly converting it to a constant if
29769         vectorizable_reduction has already enforced the condition.
29770         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
29771         to create a {1,2,3,...} vector.
29772         (vectorizable_reduction): Treat the number of units as polynomial.
29773         Choose vectype_in based on the largest scalar element size rather
29774         than the smallest number of units.  Enforce the restrictions
29775         relied on above.
29777 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29778             Alan Hayward  <alan.hayward@arm.com>
29779             David Sherwood  <david.sherwood@arm.com>
29781         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29782         number of units as polynomial.
29784 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29785             Alan Hayward  <alan.hayward@arm.com>
29786             David Sherwood  <david.sherwood@arm.com>
29788         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29789         * target.def (autovectorize_vector_sizes): Return the vector sizes
29790         by pointer, using vector_sizes rather than a bitmask.
29791         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29792         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29793         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29794         Likewise.
29795         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29796         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29797         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29798         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29799         * omp-general.c (omp_max_vf): Likewise.
29800         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29801         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29802         * tree-vect-loop.c (vect_analyze_loop): Likewise.
29803         * tree-vect-slp.c (vect_slp_bb): Likewise.
29804         * doc/tm.texi: Regenerate.
29805         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29806         to a poly_uint64.
29807         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29808         the vector size as a poly_uint64 rather than an unsigned int.
29809         (current_vector_size): Change from an unsigned int to a poly_uint64.
29810         (get_vectype_for_scalar_type): Update accordingly.
29811         * tree.h (build_truth_vector_type): Take the size and number of
29812         units as a poly_uint64 rather than an unsigned int.
29813         (build_vector_type): Add a temporary overload that takes
29814         the number of units as a poly_uint64 rather than an unsigned int.
29815         * tree.c (make_vector_type): Likewise.
29816         (build_truth_vector_type): Take the number of units as a poly_uint64
29817         rather than an unsigned int.
29819 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29820             Alan Hayward  <alan.hayward@arm.com>
29821             David Sherwood  <david.sherwood@arm.com>
29823         * target.def (get_mask_mode): Take the number of units and length
29824         as poly_uint64s rather than unsigned ints.
29825         * targhooks.h (default_get_mask_mode): Update accordingly.
29826         * targhooks.c (default_get_mask_mode): Likewise.
29827         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29828         * doc/tm.texi: Regenerate.
29830 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29831             Alan Hayward  <alan.hayward@arm.com>
29832             David Sherwood  <david.sherwood@arm.com>
29834         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29835         * omp-general.c (omp_max_vf): Likewise.
29836         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29837         (expand_omp_simd): Handle polynomial safelen.
29838         * omp-low.c (omplow_simd_context): Add a default constructor.
29839         (omplow_simd_context::max_vf): Change from int to poly_uint64.
29840         (lower_rec_simd_input_clauses): Update accordingly.
29841         (lower_rec_input_clauses): Likewise.
29843 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29844             Alan Hayward  <alan.hayward@arm.com>
29845             David Sherwood  <david.sherwood@arm.com>
29847         * tree-vectorizer.h (vect_nunits_for_cost): New function.
29848         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29849         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29850         (vect_analyze_slp_cost): Likewise.
29851         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29852         (vect_model_load_cost): Likewise.
29854 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29855             Alan Hayward  <alan.hayward@arm.com>
29856             David Sherwood  <david.sherwood@arm.com>
29858         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29859         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29860         from an unsigned int * to a poly_uint64_pod *.
29861         (calculate_unrolling_factor): New function.
29862         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
29864 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29865             Alan Hayward  <alan.hayward@arm.com>
29866             David Sherwood  <david.sherwood@arm.com>
29868         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29869         from an unsigned int to a poly_uint64.
29870         (_loop_vec_info::slp_unrolling_factor): Likewise.
29871         (_loop_vec_info::vectorization_factor): Change from an int
29872         to a poly_uint64.
29873         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29874         (vect_get_num_vectors): New function.
29875         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29876         (vect_get_num_copies): Use vect_get_num_vectors.
29877         (vect_analyze_data_ref_dependences): Change max_vf from an int *
29878         to an unsigned int *.
29879         (vect_analyze_data_refs): Change min_vf from an int * to a
29880         poly_uint64 *.
29881         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29882         than an unsigned HOST_WIDE_INT.
29883         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29884         (vect_analyze_data_ref_dependence): Change max_vf from an int *
29885         to an unsigned int *.
29886         (vect_analyze_data_ref_dependences): Likewise.
29887         (vect_compute_data_ref_alignment): Handle polynomial vf.
29888         (vect_enhance_data_refs_alignment): Likewise.
29889         (vect_prune_runtime_alias_test_list): Likewise.
29890         (vect_shift_permute_load_chain): Likewise.
29891         (vect_supportable_dr_alignment): Likewise.
29892         (dependence_distance_ge_vf): Take the vectorization factor as a
29893         poly_uint64 rather than an unsigned HOST_WIDE_INT.
29894         (vect_analyze_data_refs): Change min_vf from an int * to a
29895         poly_uint64 *.
29896         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29897         vfm1 as a poly_uint64 rather than an int.  Make the same change
29898         for the returned bound_scalar.
29899         (vect_gen_vector_loop_niters): Handle polynomial vf.
29900         (vect_do_peeling): Likewise.  Update call to
29901         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29902         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29903         be constant.
29904         * tree-vect-loop.c (vect_determine_vectorization_factor)
29905         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29906         (vect_get_known_peeling_cost): Likewise.
29907         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29908         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29909         (vect_transform_loop): Likewise.  Use the lowest possible VF when
29910         updating the upper bounds of the loop.
29911         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
29912         rather than an int.
29913         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29914         polynomial unroll factors.
29915         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29916         (vect_make_slp_decision): Likewise.
29917         (vect_supported_load_permutation_p): Likewise, and polynomial
29918         vf too.
29919         (vect_analyze_slp_cost): Handle polynomial vf.
29920         (vect_slp_analyze_node_operations): Likewise.
29921         (vect_slp_analyze_bb_1): Likewise.
29922         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29923         than an unsigned HOST_WIDE_INT.
29924         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29925         (vectorizable_load): Handle polynomial vf.
29926         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29927         a poly_uint64.
29928         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29930 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29931             Alan Hayward  <alan.hayward@arm.com>
29932             David Sherwood  <david.sherwood@arm.com>
29934         * match.pd: Handle bit operations involving three constants
29935         and try to fold one pair.
29937 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29939         * tree-vect-loop-manip.c: Include gimple-fold.h.
29940         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29941         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
29942         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29943         Add a path that uses a step of VF instead of 1, but disable it
29944         for now.
29945         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29946         and niters_no_overflow parameters.  Update calls to
29947         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29948         Create a new SSA name if the latter choses to use a ste other
29949         than zero, and return it via niters_vector_mult_vf_var.
29950         * tree-vect-loop.c (vect_transform_loop): Update calls to
29951         vect_do_peeling, vect_gen_vector_loop_niters and
29952         slpeel_make_loop_iterate_ntimes.
29953         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29954         (vect_gen_vector_loop_niters): Update declarations after above changes.
29956 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
29958         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29959         128-bit round to integer instructions.
29960         (ceil<mode>2): Likewise.
29961         (btrunc<mode>2): Likewise.
29962         (round<mode>2): Likewise.
29964 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29966         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29967         unaligned VSX load/store on P8/P9.
29968         (expand_block_clear): Allow the use of unaligned VSX
29969         load/store on P8/P9.
29971 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29973         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29974         New function.
29975         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29976         swap associated with both a load and a store.
29978 2018-01-02  Andrew Waterman  <andrew@sifive.com>
29980         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29981         * config/riscv/riscv.md (clear_cache): Use it.
29983 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
29985         * web.c: Remove out-of-date comment.
29987 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29989         * expr.c (fixup_args_size_notes): Check that any existing
29990         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29991         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29992         (emit_single_push_insn): ...here.
29994 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29996         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29997         (const_vector_encoded_nelts): New function.
29998         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29999         (const_vector_int_elt, const_vector_elt): Declare.
30000         * emit-rtl.c (const_vector_int_elt_1): New function.
30001         (const_vector_elt): Likewise.
30002         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
30003         of CONST_VECTOR_ELT.
30005 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30007         * expr.c: Include rtx-vector-builder.h.
30008         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
30009         directly on the tree encoding.
30010         (const_vector_from_tree): Likewise.
30011         * optabs.c: Include rtx-vector-builder.h.
30012         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
30013         sequence of "u" values.
30014         * vec-perm-indices.c: Include rtx-vector-builder.h.
30015         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
30016         directly on the vec_perm_indices encoding.
30018 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30020         * doc/rtl.texi (const_vector): Describe new encoding scheme.
30021         * Makefile.in (OBJS): Add rtx-vector-builder.o.
30022         * rtx-vector-builder.h: New file.
30023         * rtx-vector-builder.c: Likewise.
30024         * rtl.h (rtx_def::u2): Add a const_vector field.
30025         (CONST_VECTOR_NPATTERNS): New macro.
30026         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
30027         (CONST_VECTOR_DUPLICATE_P): Likewise.
30028         (CONST_VECTOR_STEPPED_P): Likewise.
30029         (CONST_VECTOR_ENCODED_ELT): Likewise.
30030         (const_vec_duplicate_p): Check for a duplicated vector encoding.
30031         (unwrap_const_vec_duplicate): Likewise.
30032         (const_vec_series_p): Check for a non-duplicated vector encoding.
30033         Say that the function only returns true for integer vectors.
30034         * emit-rtl.c: Include rtx-vector-builder.h.
30035         (gen_const_vec_duplicate_1): Delete.
30036         (gen_const_vector): Call gen_const_vec_duplicate instead of
30037         gen_const_vec_duplicate_1.
30038         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
30039         (gen_const_vec_duplicate): Use rtx_vector_builder.
30040         (gen_const_vec_series): Likewise.
30041         (gen_rtx_CONST_VECTOR): Likewise.
30042         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
30043         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30044         Build a new vector rather than modifying a CONST_VECTOR in-place.
30045         (handle_special_swappables): Update call accordingly.
30046         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
30047         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30048         Build a new vector rather than modifying a CONST_VECTOR in-place.
30049         (handle_special_swappables): Update call accordingly.
30051 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30053         * simplify-rtx.c (simplify_const_binary_operation): Use
30054         CONST_VECTOR_ELT instead of XVECEXP.
30056 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30058         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
30059         the selector elements to be different from the data elements
30060         if the selector is a VECTOR_CST.
30061         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
30062         ssizetype for the selector.
30064 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30066         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
30067         before testing each element individually.
30068         * tree-vect-generic.c (lower_vec_perm): Likewise.
30070 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30072         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
30073         * selftest-run-tests.c (selftest::run_tests): Call it.
30074         * vector-builder.h (vector_builder::operator ==): New function.
30075         (vector_builder::operator !=): Likewise.
30076         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
30077         (vec_perm_indices::all_from_input_p): New function.
30078         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30079         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
30080         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
30081         instead of reading the VECTOR_CST directly.  Detect whether both
30082         vector inputs are the same before constructing the vec_perm_indices,
30083         and update the number of inputs argument accordingly.  Use the
30084         utility functions added above.  Only construct sel2 if we need to.
30086 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30088         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
30089         the broadcast of the low byte.
30090         (expand_mult_highpart): Use an explicit encoding for the permutes.
30091         * optabs-query.c (can_mult_highpart_p): Likewise.
30092         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
30093         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30094         (vectorizable_bswap): Likewise.
30095         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
30096         explicit encoding for the power-of-2 permutes.
30097         (vect_permute_store_chain): Likewise.
30098         (vect_grouped_load_supported): Likewise.
30099         (vect_permute_load_chain): Likewise.
30101 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30103         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
30104         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
30105         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
30106         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
30107         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
30108         (vect_gen_perm_mask_any): Likewise.
30110 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30112         * int-vector-builder.h: New file.
30113         * vec-perm-indices.h: Include int-vector-builder.h.
30114         (vec_perm_indices): Redefine as an int_vector_builder.
30115         (auto_vec_perm_indices): Delete.
30116         (vec_perm_builder): Redefine as a stand-alone class.
30117         (vec_perm_indices::vec_perm_indices): New function.
30118         (vec_perm_indices::clamp): Likewise.
30119         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
30120         (vec_perm_indices::new_vector): New function.
30121         (vec_perm_indices::new_expanded_vector): Update for new
30122         vec_perm_indices class.
30123         (vec_perm_indices::rotate_inputs): New function.
30124         (vec_perm_indices::all_in_range_p): Operate directly on the
30125         encoded form, without computing elided elements.
30126         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
30127         encoding.  Update for new vec_perm_indices class.
30128         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
30129         the given vec_perm_builder.
30130         (expand_vec_perm_var): Update vec_perm_builder constructor.
30131         (expand_mult_highpart): Use vec_perm_builder instead of
30132         auto_vec_perm_indices.
30133         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
30134         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
30135         or double series encoding as appropriate.
30136         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
30137         vec_perm_indices instead of auto_vec_perm_indices.
30138         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30139         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30140         (vect_permute_store_chain): Likewise.
30141         (vect_grouped_load_supported): Likewise.
30142         (vect_permute_load_chain): Likewise.
30143         (vect_shift_permute_load_chain): Likewise.
30144         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30145         (vect_transform_slp_perm_load): Likewise.
30146         (vect_schedule_slp_instance): Likewise.
30147         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30148         (vectorizable_mask_load_store): Likewise.
30149         (vectorizable_bswap): Likewise.
30150         (vectorizable_store): Likewise.
30151         (vectorizable_load): Likewise.
30152         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
30153         vec_perm_indices instead of auto_vec_perm_indices.  Use
30154         tree_to_vec_perm_builder to read the vector from a tree.
30155         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
30156         vec_perm_builder instead of a vec_perm_indices.
30157         (have_whole_vector_shift): Use vec_perm_builder and
30158         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
30159         truncation to calc_vec_perm_mask_for_shift.
30160         (vect_create_epilog_for_reduction): Likewise.
30161         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
30162         from auto_vec_perm_indices to vec_perm_indices.
30163         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30164         instead of changing individual elements.
30165         (aarch64_vectorize_vec_perm_const): Use new_vector to install
30166         the vector in d.perm.
30167         * config/arm/arm.c (expand_vec_perm_d::perm): Change
30168         from auto_vec_perm_indices to vec_perm_indices.
30169         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30170         instead of changing individual elements.
30171         (arm_vectorize_vec_perm_const): Use new_vector to install
30172         the vector in d.perm.
30173         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
30174         Update vec_perm_builder constructor.
30175         (rs6000_expand_interleave): Likewise.
30176         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
30177         (rs6000_expand_interleave): Likewise.
30179 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30181         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
30182         to qimode could truncate the indices.
30183         * optabs.c (expand_vec_perm_var): Likewise.
30185 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30187         * Makefile.in (OBJS): Add vec-perm-indices.o.
30188         * vec-perm-indices.h: New file.
30189         * vec-perm-indices.c: Likewise.
30190         * target.h (vec_perm_indices): Replace with a forward class
30191         declaration.
30192         (auto_vec_perm_indices): Move to vec-perm-indices.h.
30193         * optabs.h: Include vec-perm-indices.h.
30194         (expand_vec_perm): Delete.
30195         (selector_fits_mode_p, expand_vec_perm_var): Declare.
30196         (expand_vec_perm_const): Declare.
30197         * target.def (vec_perm_const_ok): Replace with...
30198         (vec_perm_const): ...this new hook.
30199         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
30200         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
30201         * doc/tm.texi: Regenerate.
30202         * optabs.def (vec_perm_const): Delete.
30203         * doc/md.texi (vec_perm_const): Likewise.
30204         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
30205         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
30206         expand_vec_perm for constant permutation vectors.  Assert that
30207         the mode of variable permutation vectors is the integer equivalent
30208         of the mode that is being permuted.
30209         * optabs-query.h (selector_fits_mode_p): Declare.
30210         * optabs-query.c: Include vec-perm-indices.h.
30211         (selector_fits_mode_p): New function.
30212         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
30213         is defined, instead of checking whether the vec_perm_const_optab
30214         exists.  Use targetm.vectorize.vec_perm_const instead of
30215         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
30216         fit in the vector mode before using a variable permute.
30217         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
30218         vec_perm_indices instead of an rtx.
30219         (expand_vec_perm): Replace with...
30220         (expand_vec_perm_const): ...this new function.  Take the selector
30221         as a vec_perm_indices rather than an rtx.  Also take the mode of
30222         the selector.  Update call to shift_amt_for_vec_perm_mask.
30223         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
30224         Use vec_perm_indices::new_expanded_vector to expand the original
30225         selector into bytes.  Check whether the indices fit in the vector
30226         mode before using a variable permute.
30227         (expand_vec_perm_var): Make global.
30228         (expand_mult_highpart): Use expand_vec_perm_const.
30229         * fold-const.c: Includes vec-perm-indices.h.
30230         * tree-ssa-forwprop.c: Likewise.
30231         * tree-vect-data-refs.c: Likewise.
30232         * tree-vect-generic.c: Likewise.
30233         * tree-vect-loop.c: Likewise.
30234         * tree-vect-slp.c: Likewise.
30235         * tree-vect-stmts.c: Likewise.
30236         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
30237         Delete.
30238         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
30239         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
30240         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
30241         (aarch64_vectorize_vec_perm_const): ...this new function.
30242         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30243         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30244         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
30245         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
30246         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30247         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30248         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
30249         into...
30250         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
30251         check for NEON modes.
30252         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
30253         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
30254         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
30255         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
30256         into...
30257         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
30258         the old VEC_PERM_CONST conditions.
30259         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
30260         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
30261         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
30262         (ia64_vectorize_vec_perm_const_ok): Merge into...
30263         (ia64_vectorize_vec_perm_const): ...this new function.
30264         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
30265         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
30266         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
30267         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
30268         * config/mips/mips.c (mips_expand_vec_perm_const)
30269         (mips_vectorize_vec_perm_const_ok): Merge into...
30270         (mips_vectorize_vec_perm_const): ...this new function.
30271         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
30272         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
30273         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
30274         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
30275         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
30276         (rs6000_expand_vec_perm_const): Delete.
30277         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
30278         Delete.
30279         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30280         (altivec_expand_vec_perm_const_le): Take each operand individually.
30281         Operate on constant selectors rather than rtxes.
30282         (altivec_expand_vec_perm_const): Likewise.  Update call to
30283         altivec_expand_vec_perm_const_le.
30284         (rs6000_expand_vec_perm_const): Delete.
30285         (rs6000_vectorize_vec_perm_const_ok): Delete.
30286         (rs6000_vectorize_vec_perm_const): New function.
30287         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30288         an element count and rtx array.
30289         (rs6000_expand_extract_even): Update call accordingly.
30290         (rs6000_expand_interleave): Likewise.
30291         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
30292         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
30293         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
30294         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
30295         (rs6000_expand_vec_perm_const): Delete.
30296         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30297         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30298         (altivec_expand_vec_perm_const_le): Take each operand individually.
30299         Operate on constant selectors rather than rtxes.
30300         (altivec_expand_vec_perm_const): Likewise.  Update call to
30301         altivec_expand_vec_perm_const_le.
30302         (rs6000_expand_vec_perm_const): Delete.
30303         (rs6000_vectorize_vec_perm_const_ok): Delete.
30304         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
30305         reference to the SPE evmerge intructions.
30306         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30307         an element count and rtx array.
30308         (rs6000_expand_extract_even): Update call accordingly.
30309         (rs6000_expand_interleave): Likewise.
30310         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
30311         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
30312         new function.
30313         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30315 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30317         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
30318         vector mode and that that mode matches the mode of the data
30319         being permuted.
30320         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
30321         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
30322         directly using expand_vec_perm_1 when forcing selectors into
30323         registers.
30324         (expand_vec_perm_var): New function, split out from expand_vec_perm.
30326 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30328         * optabs-query.h (can_vec_perm_p): Delete.
30329         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
30330         * optabs-query.c (can_vec_perm_p): Split into...
30331         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
30332         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
30333         particular selector is valid.
30334         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30335         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30336         (vect_grouped_load_supported): Likewise.
30337         (vect_shift_permute_load_chain): Likewise.
30338         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30339         (vect_transform_slp_perm_load): Likewise.
30340         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30341         (vectorizable_bswap): Likewise.
30342         (vect_gen_perm_mask_checked): Likewise.
30343         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
30344         implementations of variable permutation vectors into account
30345         when deciding which selector to use.
30346         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
30347         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
30348         with a false third argument.
30349         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
30350         to test whether the constant selector is valid and can_vec_perm_var_p
30351         to test whether a variable selector is valid.
30353 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30355         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
30356         * optabs-query.c (can_vec_perm_p): Likewise.
30357         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
30358         instead of vec_perm_indices.
30359         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
30360         (vect_gen_perm_mask_checked): Likewise,
30361         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
30362         (vect_gen_perm_mask_checked): Likewise,
30364 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30366         * optabs-query.h (qimode_for_vec_perm): Declare.
30367         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
30368         (qimode_for_vec_perm): ...this new function.
30369         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
30371 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30373         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
30374         does not have a conditional at the top.
30376 2018-01-02  Richard Biener  <rguenther@suse.de>
30378         * ipa-inline.c (big_speedup_p): Fix expression.
30380 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30382         PR target/81616
30383         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
30384         for generic 4->6.
30386 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30388         PR target/81616
30389         Generic tuning.
30390         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
30391         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
30392         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
30393         cond_taken_branch_cost 3->4.
30395 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
30397         PR tree-optimization/83581
30398         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
30399         TODO_cleanup_cfg if any changes have been made.
30401         PR middle-end/83608
30402         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
30403         convert_modes if target mode has the right side, but different mode
30404         class.
30406         PR middle-end/83609
30407         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
30408         last argument when extracting from CONCAT.  If either from_real or
30409         from_imag is NULL, use expansion through memory.  If result is not
30410         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
30411         the parts directly to inner mode, if even that fails, use expansion
30412         through memory.
30414         PR middle-end/83623
30415         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
30416         check for bswap in mode rather than HImode and use that in expand_unop
30417         too.
30419 Copyright (C) 2018 Free Software Foundation, Inc.
30421 Copying and distribution of this file, with or without modification,
30422 are permitted in any medium without royalty provided the copyright
30423 notice and this notice are preserved.