[PR87874] avoid const-wide-int subreg in LRA
[official-gcc.git] / gcc / ChangeLog
blobd1fde63c79c1fe4843e10eaed8c0bbd573168938
1 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
3         PR rtl-optimization/87874
4         * lra.c (lra_substitute_pseudo): Do not create a subreg for
5         const wide ints.
7 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
9         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
10         if not in indexed or indirect form.
11         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
12         (bswap<mode>2_store): Ditto.
14 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
16         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
17         the UNSPEC.
19 2018-11-06  Richard Biener  <rguenther@suse.de>
21         PR tree-optimization/86850
22         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
23         instead of src.m_vec.
25 2018-11-06  Jan Hubicka  <jh@suse.cz>
27         * tree.c (fld_simplified_type_name): Break out form ...
28         (free_lang_data_in_type): ... here.
29         (fld_type_variant_equal_p): Use it.
31 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33         * config/default-d.c: Include memmodel.h.
35         * config/sol2-d.c: New file.
36         * config/t-sol2 (sol2-d.o): New rule.
37         * config.gcc <*-*-solaris2*>: Set d_target_objs,
38         target_has_targetdm.
40 2018-11-06  Jan Hubicka  <jh@suse.cz>
42         * tree.c (fld_type_variant): Also copy alignment; be sure that
43         new variant is equal.
45 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
47         PR target/87762
48         * config/s390/s390.md: Add relative_long attribute.
50 2018-11-06  Jan Hubicka  <jh@suse.cz>
52         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
53         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
54         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
55         stream TYPE_NEEDS_CONSTRUCTING.
56         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
57         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
59 2018-11-06  Richard Biener  <rguenther@suse.de>
61         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
62         dump-scope ...
63         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
65 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
67         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
68         REG_CLASS_CONTENTS[GEN_REGS].
69         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
71 2018-11-06  Jan Hubicka  <jh@suse.cz>
73          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
75 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
77         PR sanitizer/80953
78         * config/sol2.h (ASAN_CC1_SPEC): Define.
79         (LD_WHOLE_ARCHIVE_OPTION): Define.
80         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
81         (ASAN_REJECT_SPEC): Provide default.
82         (LIBASAN_EARLY_SPEC): Define.
83         (LIBTSAN_EARLY_SPEC): Define.
84         (LIBLSAN_EARLY_SPEC): Define.
85         * config/i386/sol2.h (CC1_SPEC): Redefine.
86         (ASAN_REJECT_SPEC): Define.
88         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
89         (TARGET_ASAN_SHADOW_OFFSET): Define.
90         (sparc_asan_shadow_offset): New function.
91         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
92         (ASAN_REJECT_SPEC): Define.
94 2018-11-06  Jan Hubicka  <jh@suse.cz>
96         * tree.c (fld_type_variant): Copy canonical type.
97         (fld_incomplete_type_of): Check that canonical types looks sane;
98         copy canonical type.
99         (verify_type): Accept when incomplete type has complete canonical type.
101 2018-11-06  Jan Hubicka  <jh@suse.cz>
103         * tree.c (free_lang_data): Reset overwite_assembler_name,
104         print_xnode, print_decl, print_type and print_identifier of
105         langhooks.
107 2018-11-06  Richard Biener  <rguenther@suse.de>
109         PR tree-optimization/87889
110         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
111         Do nothing if old and new arg are the same
113 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
115         PR target/87723
116         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
117         attributes for operands 3 and 4.
119 2018-11-06  Richard Biener  <rguenther@suse.de>
121         PR middle-end/18041
122         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
123         matching bitfield insertion.
125 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
127         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
128         (reg_next_debug_use): New.
129         (attempt_change): Propagate adjusted expression into affected
130         debug insns.
131         (merge_in_block): Track uses in debug insns.
132         (pass_inc_dec::execute): Allocate and release
133         reg_next_debug_use.
135 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
137         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
138         (_mm512_fixupimm_round_pd): Update parameters and builtin.
139         (_mm512_maskz_fixupimm_round_pd): Ditto.
140         (_mm512_fixupimm_round_ps): Ditto.
141         (_mm512_maskz_fixupimm_round_ps): Ditto.
142         (_mm_fixupimm_round_sd): Ditto.
143         (_mm_maskz_fixupimm_round_sd): Ditto.
144         (_mm_fixupimm_round_ss): Ditto.
145         (_mm_maskz_fixupimm_round_ss): Ditto.
146         (_mm512_fixupimm_pd): Ditto.
147         (_mm512_maskz_fixupimm_pd): Ditto.
148         (_mm512_fixupimm_ps): Ditto.
149         (_mm512_maskz_fixupimm_ps): Ditto.
150         (_mm_fixupimm_sd): Ditto.
151         (_mm_maskz_fixupimm_sd): Ditto.
152         (_mm_fixupimm_ss): Ditto.
153         (_mm_maskz_fixupimm_ss): Ditto.
154         (_mm512_mask_fixupimm_round_pd): Update builtin.
155         (_mm512_mask_fixupimm_round_ps): Ditto.
156         (_mm_mask_fixupimm_round_sd): Ditto.
157         (_mm_mask_fixupimm_round_ss): Ditto.
158         (_mm512_mask_fixupimm_pd): Ditto.
159         (_mm512_mask_fixupimm_ps): Ditto.
160         (_mm_mask_fixupimm_sd): Ditto.
161         (_mm_mask_fixupimm_ss): Ditto.
162         * config/i386/avx512vlintrin.h:
163         (_mm256_fixupimm_pd): Update parameters and builtin.
164         (_mm256_maskz_fixupimm_pd): Ditto.
165         (_mm256_fixupimm_ps): Ditto.
166         (_mm256_maskz_fixupimm_ps): Ditto.
167         (_mm_fixupimm_pd): Ditto.
168         (_mm_maskz_fixupimm_pd): Ditto.
169         (_mm_fixupimm_ps): Ditto.
170         (_mm_maskz_fixupimm_ps): Ditto.
171         (_mm256_mask_fixupimm_pd): Update builtin.
172         (_mm256_mask_fixupimm_ps): Ditto.
173         (_mm_mask_fixupimm_pd): Ditto.
174         (_mm_mask_fixupimm_ps): Ditto.
175         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
176         * config/i386/i386-builtin.def: Update builtin definitions.
177         * config/i386/i386.c: Handle new builtin types and remove useless ones.
178         * config/i386/sse.md: Update VFIXUPIMM* patterns.
179         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
180         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
181         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
182         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
183         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
184         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
185         * config/i386/subst.md:
186         (round_saeonly_sd_mask_operand4): Add new subst_attr.
187         (round_saeonly_sd_mask_op4): Ditto.
188         (round_saeonly_expand_operand5): Ditto.
189         (round_saeonly_expand): Update.
191 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
193         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
195 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
197         PR rtl-optimization/87871
198         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
200 2018-11-05  Paul Koning  <ni1d@arrl.net>
202         * doc/sourcebuild.texi (target attributes): Document new "inf"
203         effective target keyword.
205 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
207         * config/s390/s390.c (s390_register_move_cost): Increase costs for
208         moves involving the CC reg.
210 2018-11-05  Richard Biener  <rguenther@suse.de>
212         PR tree-optimization/87873
213         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
214         argument.
215         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
216         * tree-vect-loop.c (vect_transform_loop): When splitting the
217         loop exit also create forwarder PHIs for constants.
218         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
219         Handle constant to_arg, add extra checking we match up the correct
220         PHIs.
222 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
224         * config/s390/s390.md: QImode and HImode for load on condition.
226 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
228         * config/s390/predicates.md: Fix typo.
229         * config/s390/s390.md: Allow immediates for load on condition.
231 2018-11-05  Martin Liska  <mliska@suse.cz>
233         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
234         * bitmap.h (struct bitmap_usage): Likewise.
235         * ggc-common.c (SCALE): Remove.
236         (LABEL): Likewise.
237         (struct ggc_usage): Use SIZE_AMOUNT. And update
238         compare method.
239         * ggc-page.c (SCALE): Remove.
240         (STAT_LABEL): Remove.
241         (ggc_print_statistics): Use SIZE_AMOUNT.
242         * gimple.h (SCALE): Remove.
243         (LABEL): Likewise.
244         * input.c (ONE_K): Remove.
245         (ONE_M): Likewise.
246         (SCALE): Likewise.
247         (STAT_LABEL): Likewise.
248         (FORMAT_AMOUNT): Likewise.
249         (dump_line_table_statistics): Use SIZE_AMOUNT.
250         * mem-stats.h (struct mem_usage): Likewise.
251         * rtl.c (dump_rtx_statistics): Likewise.
252         (rtx_alloc_counts): Change type to size_t.
253         (rtx_alloc_sizes): Likewise.
254         (rtx_count_cmp): New.
255         (dump_rtx_statistics): Sort first based on counts.
256         * tree.c (tree_nodes_cmp): New.
257         (tree_codes_cmp): New.
258         (dump_tree_statistics): Sort first based on counts.
259         * system.h (ONE_K): New.
260         (ONE_M): Likewise.
261         (SIZE_SCALE): Likewise.
262         (SIZE_LABEL): Likewise.
263         (SIZE_AMOUNT): Likewise.
264         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
265         * tree-dfa.c (dump_dfa_stats): Likewise.
266         * tree-phinodes.c (phinodes_print_statistics): Likewise.
267         * tree-ssanames.c (ssanames_print_statistics): Likewise.
268         * tree.c (dump_tree_statistics): Likewise.
269         * vec.c (struct vec_usage): Likewise.
270         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
271         trigger a -Werror=format-overflow with
272         --enable-gather-detailed-stats.
274 2018-11-05  Martin Liska  <mliska@suse.cz>
276         * mem-stats.h (mem_alloc_description::release_instance_overhead):
277         Return T *.
278         * vec.c (struct vec_usage): Register m_element_size.
279         (vec_prefix::register_overhead): New arguments: elements and
280         element_size.
281         (vec_prefix::release_overhead): Subtract elements.
282         * vec.h (struct vec_prefix): Change signature.
283         (va_heap::reserve): Pass proper arguments.
284         (va_heap::release): Likewise.
286 2018-11-05  Martin Liska  <mliska@suse.cz>
288         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
289         style.
290         * vec.c: Likewise.
292 2018-11-05  Richard Biener  <rguenther@suse.de>
294         * tree-scalar-evolution.h (final_value_replacement_loop): Update
295         prototype.
296         * tree-scalar-evolution.c (final_value_replacement_loop): Return
297         whether anything was done.
298         (scev_const_prop): Remove constant propagation part, fold
299         remains into ...
300         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
301         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
302         conditionally.
304 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
306         PR tree-optimization/87859
307         * gimple-ssa-store-merging.c (struct merged_store_group): Add
308         only_constants and first_nonmergeable_order members.
309         (merged_store_group::merged_store_group): Initialize them.
310         (merged_store_group::do_merge): Clear only_constants member if
311         adding something other than INTEGER_CST store.
312         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
313         stores with order >= first_nonmergeable_order.  Use
314         merged_store->only_constants instead of always recomputing it.
315         Set merged_store->first_nonmergeable_order if we've skipped any
316         stores.  Attempt to merge overlapping INTEGER_CST stores that
317         we would otherwise skip.
319         PR sanitizer/87837
320         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
322 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
324         PR target/87853
325         * config/i386/emmintrin.h (__v16qs): New to cope with option
326         -funsigned-char.
327         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
328         (_mm_cmplt_epi8): Likewise.
329         (_mm_cmpgt_epi8): Likewise.
331 2018-11-05  Richard Biener  <rguenther@suse.de>
333         PR rtl-optimization/87852
334         * fwprop.c (use_killed_between): Only consider single-defs of the
335         use whose definition statement dominates the use.
337 2018-11-05  Martin Liska  <mliska@suse.cz>
339         PR web/87829
340         * doc/invoke.texi: Remove options that are
341         not disabled with -Os.
343 2018-11-05  Martin Liska  <mliska@suse.cz>
345         PR c/87811
346         * doc/extend.texi: Update constrain about the last argument
347         of __builtin_expect_with_probability.
349 2018-11-05  Martin Liska  <mliska@suse.cz>
351         PR c/87811
352         * predict.c (expr_expected_value_1): Verify
353         that last argument is a real constants and emit
354         error.
356 2018-11-05  Martin Liska  <mliska@suse.cz>
358         PR gcov-profile/77698
359         * ipa-profile.c (ipa_profile): Adjust hotness threshold
360         only in LTO mode.
362 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
364         PR tree-optimization/86572
365         * builtins.c (c_strlen): Handle negative offsets in a safe way.
367 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
369         PR tree-optimization/87672
370         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
371         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
373 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
375         PR middle-end/58372
376         * cfgexpand.c (pass_expand::execute): Move the call to
377         finish_eh_generation in front of the call to expand_stack_alignment.
379 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
381         * common/config/i386/i386-common.c (processor_alias_table): Add
382         znver2 entry.
383         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
384         (case ${target}): Add znver2.
385         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
386         -march=native recognize znver2 processors.
387         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
388         * config/i386/i386.c (m_znver2): New definition.
389         (m_ZNVER): New definition.
390         (m_AMD_MULTIPLE): Includes m_znver2.
391         (processor_cost_table): Add znver2 entry.
392         (processor_target_table): Add znver2 entry.
393         (get_builtin_code_for_version): Set priority for
394         PROCESSOR_ZNVER2.
395         (processor_model): Add M_AMDFAM17H_ZNVER2.
396         (arch_names_table): Ditto.
397         (ix86_reassociation_width): Include znver2.
398         * config/i386/i386.h (TARGET_znver2): New definition.
399         (struct ix86_size_cost): Add TARGET_ZNVER2.
400         (enum processor_type): Add PROCESSOR_ZNVER2.
401         * config/i386/i386.md (define_attr "cpu"): Add znver2.
402         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
403         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
404         (ix86_adjust_cost): Add znver2.
405         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
406         * gcc/doc/extend.texi: Add details about znver2.
407         * gcc/doc/invoke.texi: Add details about znver2.
409 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
411         PR target/87079
413         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
414         pattern.
416 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
418         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
419         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
421 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
423         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
424         shrink-wrapping when -mtrack-speculation.
426 2018-11-02  Richard Biener  <rguenther@suse.de>
428         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
429         (pop_cost_one_pair): Do not free pair.
430         (pop_best_coalesce): Likewise.
431         (create_coalesce_list): Initialize obstack.
432         (delete_coalesce_list): Free obstack.
433         (find_coalesce_pair): Obstack-allocate coalesce pairs.
434         (add_cost_one_coalesce): Likewise.
435         (struct live_track): Remove bitmap pointer indirections.
436         (new_live_track): Adjust.
437         (delete_live_track): Likewise.
438         (live_track_remove_partition): Likewise.
439         (live_track_add_partition): Likewise.
440         (live_track_live_p): Likewise.
441         (live_track_process_def): Likewise.
442         (live_track_clear_base_vars): Likewise.
444 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
446         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
447         * configure: Regenerate.
448         * config.in: Regenerate.
449         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
450         HAVE_AS_REGISTER_PSEUDO_OP guard.
451         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
452         !HAVE_AS_REGISTER_PSEUDO_OP support.
453         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
454         guard.
456 2018-11-02  Richard Biener  <rguenther@suse.de>
458         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
459         with is_type_die.
461 2018-11-02  Richard Biener  <rguenther@suse.de>
463         PR tree-optimization/87776
464         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
465         executable when iterating but running into rpo-vn-max-loop-depth
466         and not eliding the iteration.
468 2018-11-30  Jan Hubicka  <jh@suse.cz>
470         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
471         TYPE_DECL.
473 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
475         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
476         prototype.
477         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
478         Combine with rs6000_address_for_fpconvert.
479         (rs6000_address_for_fpconvert): Combine with
480         rs6000_force_indexed_or_indirect_mem.
481         (rs6000_expand_vector_init): Change function call from
482         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
483         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
484         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
485         (floatsi<mode>2_lfiwax_mem): Ditto.
486         (floatunssi<mode>2_lfiwzx): Ditto.
487         (floatunssi<mode>2_lfiwzx_mem): Ditto.
488         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
489         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
490         (fix_trunc<mode>si2_stfiwx): Ditto.
491         (fixuns_trunc<mode>si2_stfiwx): Ditto.
492         (float_<mode>si2_hw): Ditto.
493         (floatuns_<mode>si2_hw): Ditto.
494         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
495         (vsx_splat_<mode>): Ditto.
497 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
499         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
500         at end of assembler input text.
501         * configure: Regenerate.
503 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
505         PR tree-optimization/87826
506         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
507         negative or larger or equal to type's precision.
509 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
511         * opts.c (default_options_table): Do not enable
512         OPT_fdelayed_branch at -Og.
513         * doc/invoke.texi (-fdelayed-branch): Document it.
515 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
517         * optabs-libfuncs.c (build_libfunc_function_visibility):
518         New, split out from...
519         (build_libfunc_function): ... here.
520         (init_one_libfunc_visibility): New, split out from ...
521         (init_one_libfunc): ... here.
523         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
524         scratch register need not be early-clobber.  Document the reason
525         why we cannot use ST<OP>.
527 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
529         PR bootstrap/82856
530         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
531         line for second argument of AC_DEFINE_UNQUOTED.
532         * doc/install.texi (Tools/packages necessary for modifying GCC):
533         Update to autoconf 2.69 and automake 1.15.1.
534         * aclocal.m4, config.in, configure: Regenerate.
536 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
538         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
539         initialization.
541 2018-10-31  Martin Liska  <mliska@suse.cz>
543         PR driver/83193
544         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
545         Add new argument invalid_extension.
546         (aarch64_get_all_extension_candidates): New function.
547         (aarch64_rewrite_selected_cpu): Add NULL to function call.
548         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
549         new argument.
550         (aarch64_get_all_extension_candidates): New function.
551         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
552         argument invalid_extension.
553         (aarch64_parse_cpu): Likewise.
554         (aarch64_print_hint_for_extensions): New function.
555         (aarch64_validate_mcpu): Provide hint about invalid extension.
556         (aarch64_validate_march): Likewise.
557         (aarch64_handle_attr_arch): Pass new argument.
558         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
559         (aarch64_handle_attr_isa_flags): Likewise.
561 2018-10-31  Richard Biener  <rguenther@suse.de>
563         PR middle-end/70359
564         PR middle-end/86270
565         * tree-outof-ssa.c (insert_backedge_copies): Restrict
566         copy generation to useful cases.  Place the copy before
567         the definition of the backedge value when possible.
569 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
571         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
572         * config/arc/arc.c (arc_active_insn): New function.
573         (check_store_cacheline_hazard): Likewise.
574         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
575         (arc_override_options): Disable delay slot scheduler for older
576         A7.
577         (arc_store_addr_hazard_p): New implementation, old one renamed to
578         ...
579         (arc_store_addr_hazard_internal_p): Renamed.
580         (arc_reorg): Don't combine into brcc instructions which are part
581         of hardware hazard solution.
582         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
583         (tune_arc700): Likewise.
584         * config/arc/arc.opt (arc7xx): New tune value.
585         * config/arc/arc700.md: Improve A7 scheduler.
587 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
589         * config/arc/arc.c (arc_override_options): Remove
590         TARGET_COMPACT_CASESI.
591         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
592         (CASE_VECTOR_MODE): Likewise.
593         (CASE_VECTOR_PC_RELATIVE): Likewise.
594         (CASE_VECTOR_SHORTEN_MODE): Likewise.
595         (CASE_VECTOR_SHORTEN_MODE1): Delete.
596         (ADDR_VEC_ALIGN): Update.
597         (ASM_OUTPUT_CASE_LABEL): Undefine.
598         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
599         (TARGET_BI_BIH): Define.
600         (DEFAULT_BRANCH_INDEX): Likewise.
601         * config/arc/arc.md (casesi): Rework to accept BI/BIH
602         instructions, remove compact_casesi use case.
603         (casesi_compact_jump): Remove.
604         (casesi_dispatch): New pattern.
605         * config/arc/arc.opt: Add mbranch-index option. Deprecate
606         compact_casesi option.
607         * doc/invoke.texi: Document mbranch-index option.
609 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
611         * config/arc/arc.c (arc_get_tp): Remove function.
612         (arc_emit_call_tls_get_addr): Likewise.
613         (arc_call_tls_get_addr): New function.
614         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
615         * config/arc/arc.md (tls_load_tp_soft): Remove.
616         (tls_gd_get_addr): Likewise.
618 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
620         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
621         (mulsi3_600_lib): Remove pattern.
622         (umulsi3_highpart_600_lib_le): Likewise.
623         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
624         (umulsidi3): Remove call to umulsidi3_600_lib.
625         (umulsidi3_600_lib): Remove pattern.
626         (peephole2): Remove peephole using the above deprecated patterns.
628 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
630         PR target/87374
631         * config/arm/arm.c (arm_option_check_internal): Disable the combined
632         use of -mslow-flash-data and -mword-relocations.
633         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
634         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
635         flag_pic.
636         * doc/invoke.texi (-mword-relocations): Mention conflict with
637         -mslow-flash-data.
638         (-mslow-flash-data): Reciprocally.
640 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
642         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
643         16-byte modes held in GP registers to use an even regno.
645         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
646         (aarch64_atomic_ldop_supported_p): Remove.
647         (aarch64_gen_atomic_ldop): Remove.
648         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
649         Fully expand LSE operations here.
650         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
651         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
652         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
653         and use ATOMIC_LDOP instead; use register_operand for the input;
654         drop the split and emit insns directly.
655         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
656         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
657         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
659         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
660         (aarch64_gen_atomic_ldop): Don't call it.
661         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
662         Use aarch64_reg_or_zero.
663         (aarch64_atomic_exchange<ALLI>): Likewise.
664         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
665         operand 0; use aarch64_reg_or_zero for input; merge ...
666         (@aarch64_atomic_swp<ALLI>): ... this and remove.
668         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
669         (aarch64_split_compare_and_swap): Use it.
670         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
671         test oldval against the proper predicate.
672         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
673         Use nonmemory_operand for expected.
674         (cas_short_expected_pred): New.
675         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
676         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
677         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
678         (aarch64_plushi_operand): New.
680         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
681         Force oldval into the rval register for TARGET_LSE; emit the compare
682         during initial expansion so that it may be deleted if unused.
683         (aarch64_gen_atomic_cas): Remove.
684         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
685         Change =&r to +r for operand 0; use match_dup for operand 2;
686         remove is_weak and mod_f operands as unused.  Drop the split
687         and merge with...
688         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
689         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
690         (@aarch64_atomic_cas<GPI>): Similarly.
692 2018-10-31  Richard Biener  <rguenther@suse.de>
694         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
695         using ABSU_EXPR.
697 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
699         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
700         * config/aarch64/aarch64.md: Include saphira.md
701         * config/aarch64/saphira.md: New file for pipeline description.
703 2018-10-30  Martin Sebor  <msebor@redhat.com>
705         PR middle-end/87041
706         * gimple-ssa-sprintf.c (format_directive): Use %G to include
707         inlining context.
708         (sprintf_dom_walker::compute_format_length):
709         Avoid setting POSUNDER4K here.
710         (get_destination_size): Handle null argument values.
711         (get_user_idx_format): New function.
712         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
713         functions, including user-defined with attribute format printf.
714         Use %G to include inlining context.
715         Set POSUNDER4K here.
717 2018-10-30  Jan Hubicka  <jh@suse.cz>
719         * params.def (lto-partitions): Bump from 32 to 128.
721 2018-10-30  Jan Hubicka  <jh@suse.cz>
723         * tree.c
724         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
725         head in file.
726         (free_lang_data_in_type): Forward declare.
727         (fld_type_variant_equal_p): New function.
728         (fld_type_variant): New function
729         (fld_incomplete_types): New hash.
730         (fld_incomplete_type_of): New function
731         (fld_simplfied-type): New function.
732         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
733         (free_lang_data): Allocate and free fld_incomplete_type; update call
734         of free_lang_data_in_decl.
736 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
738         * gcov.c (output_lines): Remove duplicate line.
740 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
742         * config/rs6000/rs6000.md (bswapdi2): Force address into register
743         if not in indexed or indirect form.
744         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
745         (bswapdi2_store): Ditto.
746         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
747         helper function.
748         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
749         Prototype for helper function.
751 2018-10-30  Martin Sebor  <msebor@redhat.com>
753         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
754         (target, pragma GCC optimize, pragma GCC target): Ditto.
756 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
758         * doc/extend.texi: Fix prototype and description of
759         __builtin_expect_with_probability.
761 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
763         * cgraph.h (clone_function_name_1): Replaced by new
764           clone_function_name_numbered that takes name as string; for
765           privatize_symbol_name_1 use only.
766           (clone_function_name): Renamed to
767           clone_function_name_numbered to be explicit about numbering.
768           (clone_function_name): New two-argument function that does
769           not number its output.
770           (clone_function_name): New three-argument function that
771           takes a number to append to its output.
772         * cgraphclones.c (duplicate_thunk_for_node):
773           (clone_function_name_1): Renamed.
774           (clone_function_name_numbered): Two new functions.
775           (clone_function_name): Improved documentation.
776           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
777         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
778         * final.c (final_scan_insn_1): Use the new clone_function_name
779           without numbering.
780         * multiple_target.c (create_dispatcher_calls): Ditto.
781           (create_target_clone): Ditto.
782         * omp-expand.c (grid_expand_target_grid_body): Ditto.
783         * omp-low.c (create_omp_child_function_name): Ditto.
784         * omp-simd-clone.c (simd_clone_create): Ditto.
785         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
786           new clone_function_name without numbering.
788 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
790         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
791         Assert that the allocation size is not zero.
793 2018-10-30  Richard Biener  <rguenther@suse.de>
795         PR tree-optimization/87800
796         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
797         non-induction or reduction PHIs.
799 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
801         * config/aarch64/falkor-tag-collision-avoidance.c
802         (execute_tag_collision_avoidance): Call df_note_add_problem.
804 2018-10-30  Martin Liska  <mliska@suse.cz>
806         * doc/extend.texi: Fix typo in documentation
807         of __builtin_expect_with_probability.
809 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
811         PR c++/87721
812         * input.c (get_substring_ranges_for_loc): Detect if
813         linemap_resolve_location gives us a NULL map, and reject
814         this case.
816 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
818         * config.gcc (xstormy16-*-elf): Set tm_d_file.
820 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
821             Martin Sebor  <msebor@redhat.com>
822             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
824         * doc/ux.texi (Quoting): New subsection, adapted from material at
825         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
826         MartinSebor and ManuelLopezIbanez.
827         (Fix-it hints): Note that fix-it hints shouldn't be marked for
828         translation.
830 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
832         PR middle-end/87469
833         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
834         max value.
836 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
838         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
840 2018-10-29  Paul Koning  <ni1d@arrl.net>
842         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
844 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
846         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
847         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
848         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
849         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
850         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
851         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
852         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
853         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
854         Change 'vector' to '__vector'.
855         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
856         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
857         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
858         _mm_avg_pu16): Likewise.  And, whitespace corrections.
860 2018-10-29  Richard Biener  <rguenther@suse.de>
862         PR tree-optimization/87785
863         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
864         internal defs.
866 2018-10-29  Olivier Hainque  <hainque@adacore.com>
868         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
870 2018-10-29  Olivier Hainque  <hainque@adacore.com>
872         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
873         ports configuration macro, defaults to "gnu".
874         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
875         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
877 2018-10-29  Olivier Hainque  <hainque@adacore.com>
879         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
881 2018-10-29  Richard Biener  <rguenther@suse.de>
883         PR tree-optimization/87790
884         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
885         (vect_make_slp_decision): Adjust.
886         (vect_slp_analyze_bb_1): Likewise.
887         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
888         edges.
890 2018-10-29  Richard Biener  <rguenther@suse.de>
892         PR tree-optimization/87785
893         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
894         and processing.
895         (vect_build_slp_tree): Likewise.
896         (vect_gather_slp_loads): New function.
897         (vect_analyze_slp_instance): Gather loads separately from the
898         SLP tree build.
900 2018-10-29  Martin Liska  <mliska@suse.cz>
902         * Makefile.in: Make dependency to json.o.
903         * doc/gcov.texi: Document new JSON format, remove
904         old intermediate format documentation.
905         * gcov.c (struct function_info): Come up with m_name and
906         m_demangled_name.
907         (function_info::function_info): Initialize it.
908         (function_info::~function_info): Release it.
909         (main): Rename flag_intermediate_format to flag_json_format.
910         (print_usage): Describe --json-format.
911         (process_args): Set flag_json_format.
912         (output_intermediate_line): Remove.
913         (output_intermediate_json_line): Likewise.
914         (get_gcov_intermediate_filename): Return new extension
915         ".gcov.json.gz".
916         (output_intermediate_file): Implement JSON emission.
917         (output_json_intermediate_file): Implement JSON emission.
918         (generate_results): Use ::get_name for function name.
919         Handle JSON output file.
920         (read_graph_file): Use ::get_name instead of cplus_demangle.
921         (read_count_file): Likewise.
922         (solve_flow_graph): Likewise.
923         (add_line_counts): Likewise.
924         (accumulate_line_counts): Use new flag_json_format.
925         (output_function_details): Use ::get_name instead of cplus_demangle.
926         (output_lines): Likewise.
927         * json.cc (test_writing_literals): Add new tests.
928         * json.h (class literal): Add new boolean constructor.
930 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
932         PR rtl-optimization/87701
933         PR rtl-optimization/87780
934         * combine.c (make_more_copies): Rewrite.
936 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
938         * doc/generic.texi (ABSU_EXPR): Document.
939         * match.pd (absu(x)*absu(x) -> x*x): Handle.
940         (absu(absu(X)) -> absu(X)): Likewise.
941         (absu(-X) -> absu(X)): Likewise.
942         (absu(X)  where X is nonnegative -> X): Likewise.
944 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
946         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
947         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
948         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
949         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
950         (s-tm-texi): Also check timestamp on d-target.def.
951         (generated_files): Add TM_D_H and d-target-hooks-def.h.
952         (build/genhooks.o): Also depend on D_TARGET_DEF.
953         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
954         variables.
955         * config/aarch64/aarch64-d.c: New file.
956         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
957         Define.
958         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
959         prototype.
960         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
961         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
962         * config/arm/arm-d.c: New file.
963         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
964         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
965         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
966         * config/arm/t-arm (arm-d.o): New rule.
967         * config/default-d.c: New file.
968         * config/glibc-d.c: New file.
969         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
970         * config/i386/i386-d.c: New file.
971         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
972         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
973         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
974         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
975         * config/i386/t-i386 (i386-d.o): New rule.
976         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
977         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
978         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
979         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
980         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
981         * config/mips/mips-d.c: New file.
982         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
983         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
984         * config/mips/t-mips (mips-d.o): New rule.
985         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
986         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
987         * config/powerpcspe/powerpcspe-d.c: New file.
988         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
989         New prototype.
990         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
991         Support GNU D by using 0 as the language type.
992         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
993         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
994         * config/riscv/riscv-d.c: New file.
995         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
996         prototype.
997         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
998         * config/riscv/t-riscv (riscv-d.o): New rule.
999         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1000         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1001         * config/rs6000/rs6000-d.c: New file.
1002         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
1003         prototype.
1004         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1005         Support GNU D by using 0 as the language type.
1006         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
1007         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
1008         * config/s390/s390-d.c: New file.
1009         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
1010         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
1011         * config/s390/t-s390 (s390-d.o): New rule.
1012         * config/sparc/sparc-d.c: New file.
1013         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
1014         prototype.
1015         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
1016         * config/sparc/t-sparc (sparc-d.o): New rule.
1017         * config/t-glibc (glibc-d.o): New rule.
1018         * configure: Regenerated.
1019         * configure.ac (tm_d_file): New variable.
1020         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
1021         * doc/contrib.texi (Contributors): Add self for the D frontend.
1022         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
1023         * doc/install.texi (Configuration): Mention libphobos as an option for
1024         --enable-shared.  Mention d as an option for --enable-languages.
1025         (Testing): Mention check-d as a target.
1026         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
1027         name suffixes.  Mention d as a -x option.
1028         * doc/sourcebuild.texi (Top Level): Mention libphobos.
1029         * doc/standards.texi (Standards): Add section on D language.
1030         * doc/tm.texi: Regenerated.
1031         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
1032         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
1033         * dwarf2out.c (is_dlang): New function.
1034         (gen_compile_unit_die): Use DW_LANG_D for D.
1035         (declare_in_namespace): Return module die for D, instead of adding
1036         extra declarations into the namespace.
1037         (gen_namespace_die): Generate DW_TAG_module for D.
1038         (gen_decl_die): Handle CONST_DECLSs for D.
1039         (dwarf2out_decl): Likewise.
1040         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
1041         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
1042         kinds of aggregates.
1043         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
1044         * genhooks.c: Include d/d-target.def.
1046 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
1048         PR target/85669
1049         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
1050         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
1051         (STACK_DYNAMIC_OFFSET): Likewise.
1053 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
1055         PR target/80024
1056         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
1057         error message.
1059 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
1061         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
1062         return type and other typos.
1064 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
1066         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
1067         a shorter sequence with fewer branches.
1068         (emit_final_str_compare_gpr): Ditto.
1070 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1072         * config/rs6000/tmmintrin.h: New file.
1073         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
1075 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1077         * config/rs6000/mmintrin.h: Enable 32bit compilation.
1078         * config/rs6000/xmmintrin.h: Likewise.
1080 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1082         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
1084 2018-10-26  Richard Biener  <rguenther@suse.de>
1086         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
1087         and wrapper.
1088         (vect_mark_slp_stmts_relevant): Likewise.
1089         (vect_detect_hybrid_slp_stmts): Likewise.
1090         (vect_bb_slp_scalar_cost): Likewise.
1091         (vect_remove_slp_scalar_calls): Likewise.
1093 2018-10-26  Jan Hubicka  <jh@suse.cz>
1095         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
1096         (warn_types_mismatch): Fix walk of DECL_NAME.
1097         (odr_types_equivalent_p): Fix overactive assert.
1099 2018-10-26  Richard Biener  <rguenther@suse.de>
1101         PR tree-optimization/87105
1102         * tree-vectorizer.h (_slp_tree::refcnt): New member.
1103         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
1104         refcnt.
1105         (vect_create_new_slp_node): Initialize refcnt to one.
1106         (bst_traits): Move.
1107         (scalar_stmts_set_t, bst_fail): Remove.
1108         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
1109         (vect_build_slp_tree): Add bst_map argument and lookup
1110         already created SLP nodes.
1111         (vect_print_slp_tree): Handle a SLP graph, print SLP node
1112         addresses.
1113         (vect_slp_rearrange_stmts): Handle a SLP graph.
1114         (vect_analyze_slp_instance): Adjust and free SLP nodes from
1115         the CSE map.  Fix indenting.
1116         (vect_schedule_slp_instance): Add short-cut.
1118 2018-10-26  Martin Liska  <mliska@suse.cz>
1120         PR testsuite/86158
1121         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
1122         addr_expr and not with pointers.
1124 2018-10-26  Jan Hubicka  <jh@suse.cz>
1126         * tree.c (free_lang_data_in_type): Only check main variants.
1127         * ipa-devirt.c (warn_odr): Make static.
1128         (types_same_for_odr): Drop strict variant.
1129         (types_odr_comparable): Likewise.
1130         (odr_or_derived_type_p): Look for main variants.
1131         (odr_name_hasher::equal): Cleanup comment.
1132         (odr_subtypes_equivalent): Add warn and warned arguments; check main
1133         variants.
1134         (type_variants_equivalent_p): break out from ...
1135         (odr_types_equivalent): ... here; go for main variants where needed.
1136         (warn_odr): ... here; turn static.
1137         (warn_types_mismatch): Compare mangled names of main variants.
1138         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
1139         (type_with_linkage_p): Sanity check that we look at main variant.
1140         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
1141         * tree.h (types_same_for_odr): Drop strict argument.
1143 2018-10-26  Richard Biener  <rguenther@suse.de>
1145         PR tree-optimization/87746
1146         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1147         Simplify and fix WRT strided store groups with size not
1148         equal to step in element count.
1149         (vect_analyze_group_access_1): Dump the whole group.
1151 2018-10-25  Carl Love  <cel@us.ibm.com>
1153         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
1154         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
1155         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
1156         precicion entry for each overloaded builtin.
1157         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
1158         VSCEDPUO): Rename overloaded name.
1159         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
1160         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
1161         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
1162         define_expand for xscmpexqp instruction.
1163         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
1165 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1166             Jinsong Ji <jji@us.ibm.com>
1168         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
1169         function with vec_sl.
1170         (_mm_slli_epi32): Likewise.
1171         (_mm_slli_epi64): Likewise.
1172         (_mm_srai_epi16): Replace deprecated function with vec_sra.
1173         (_mm_srai_epi32): Likewise.
1174         (_mm_srli_epi16): Replace deprecated function with vec_sr.
1175         (_mm_srli_epi32): Likewise.
1176         (_mm_srli_epi64): Likewise.
1177         (_mm_sll_epi16): Replace deprecated function with vec_sl.
1178         (_mm_sll_epi32): Likewise.
1179         (_mm_sll_epi64): Likewise.
1180         (_mm_sra_epi16): Replace deprecated function with vec_sra.
1181         (_mm_sra_epi32): Likewise.
1182         (_mm_srl_epi16): Replace deprecated function with vec_sr.
1183         (_mm_srl_epi32): Likewise.
1184         (_mm_srl_epi64): Likewise.
1186 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1187             Jinsong Ji <jji@us.ibm.com>
1189         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
1190         comparison operators with vec_cmp* for compatibility due to
1191         unfortunate history; clean up formatting and use types more
1192         appropriately.
1193         (_mm_sll_epi32): Likewise.
1194         (_mm_sll_epi64): Likewise.
1195         (_mm_srl_epi16): Likewise.
1196         (_mm_srl_epi32): Likewise.
1197         (_mm_srl_epi64): Likewise.
1199 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
1200             Jinsong Ji <jji@us.ibm.com>
1202         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
1203         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
1204         __vector __bool int.  Use vec_cmpgt in preference to deprecated
1205         function vec_vcmpgtfp.
1206         (_mm_max_ps): Likewise.
1208 2018-10-25  Jeff Law  <law@redhat.com>
1210         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
1211         if returning false.
1213 2018-10-25  Martin Sebor  <msebor@redhat.com>
1215         * doc/extend.texi (aligned): Expand attribute description.
1216         (Alignment): Rename section.  Discuss function arguments.
1218 2018-10-25  Jan Hubicka  <jh@suse.cz>
1220         * ipa-devirt.c (main_odr_variant): Remove.
1221         (hash_odr_name, types_same_for_odr, types_odr_comparable,
1222         odr_name_hasher::equal, odr_subtypes_equivalent_p):
1223         Drop use of main_odr_variant.
1224         (add_type_duplicate): Silence confused warnings on integer types.
1225         (get_odr_type): Always look for main variant.
1226         (register_odr_type): Simplify.
1228 2018-10-25  Richard Biener  <rguenther@suse.de>
1230         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
1231         Initialize ng to silence error with release checking bootstrap.
1233 2018-10-25  Richard Biener  <rguenther@suse.de>
1235         * tree-if-conv.c: Include tree-ssa-sccvn.h.
1236         (tree_if_conversion): Run CSE on the if-converted loop body.
1238 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1240         * config/s390/constraints.md (ZL): New constraint.
1241         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
1242         operands.
1243         * config/s390/s390.md (movdi_larl): Remove.
1244         (movdi_64): Add the LARL alternative.
1246 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
1248         PR bootstrap/87747
1249         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
1250         (RTX_CODE_HWINT_P): New macro.
1251         (rtx_code_size): Use RTX_CODE_HWINT_P ().
1253 2018-10-25  Jan Hubicka  <jh@suse.cz>
1255         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
1256         is anonymous.
1258 2018-10-25  Richard Biener  <rguenther@suse.de>
1260         PR tree-optimization/87665
1261         PR tree-optimization/87745
1262         * tree-vectorizer.h (get_earlier_stmt): Remove.
1263         (get_later_stmt): Pick up UID from the original non-pattern stmt.
1265 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
1267         * options.texi (Deprecated): Move list to Var section.
1269 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
1270             Jinsong Ji <jji@us.ibm.com>
1272         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
1273         __vector long to __vector long long.
1274         (_mm_cvtpd_ps): Likewise.
1275         (_mm_cvttpd_epi32): Likewise.
1276         (_mm_cvtpi32_pd): Likewise.
1277         (_mm_unpackhi_epi64): Likewise.
1278         (_mm_unpacklo_epi64): Likewise.
1280 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
1282         PR rtl-optimization/87720
1283         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
1285 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
1287         * gimple-ssa-isolate-paths.c
1288         (find_implicit_erroneous_behavior): Do not change code if the
1289         pass is running for warnings only.
1290         (find_explicit_erroneous_behavior): Likewise.
1292 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
1294         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
1295         Define as rs6000_mangle_decl_assembler_name.
1296         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
1297         long double to IEEE long double, switch the names of the long
1298         double built-in functions to be <func>f128 instead of <func>l.
1300 2018-10-24  Martin Sebor  <msebor@redhat.com>
1302         * doc/extend.texi (nonnull): List no-argument form.  Reference
1303         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
1305 2018-10-24  Richard Biener  <rguenther@suse.de>
1307         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
1309 2018-10-24  Martin Liska  <mliska@suse.cz>
1311         PR tree-optimization/84436
1312         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
1313         Remove.
1314         (switch_conversion::contains_linear_function_p): New.
1315         (switch_conversion::build_one_array): Support linear
1316         transformation on input.
1317         * tree-switch-conversion.h (struct switch_conversion): Add
1318         contains_linear_function_p declaration.
1320 2018-10-24  Richard Biener  <rguenther@suse.de>
1322         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
1323         if its argument is CONSTANT_CLASS_P.
1325 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1327         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
1328         it is wrong for forward declarations.
1330 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
1332         * config/s390/s390.c (s390_check_qrst_address): Add the missing
1333         SYMBOL_REF_P () check.
1335 2018-10-24  Richard Biener  <rguenther@suse.de>
1337         PR tree-optimization/87105
1338         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
1339         dump classification.
1340         (vect_analyze_data_ref_accesses): Handle duplicate loads and
1341         stores by splitting the affected group after the fact.
1342         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
1343         fail the SLP build because of size constraints.
1345 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1347         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
1348         * configure: Regenerate.
1349         * config.in: Regenerate.
1350         * varasm.c (mergeable_string_section): Use readonly_data_section
1351         if linker doesn't support SHF_MERGE with alignment > 8.
1352         (mergeable_constant_section): Likewise.
1354 2018-10-24  Richard Biener  <rguenther@suse.de>
1356         PR tree-optimization/84013
1357         * tree-ssa-structalias.c (struct msdi_data): New struct for
1358         marshalling data to walk_stmt_load_store_ops.
1359         (maybe_set_dependence_info): Refactor as callback for
1360         walk_stmt_load_store_ops.
1361         (compute_dependence_clique): Set restrict info on all stmt kinds.
1363 2018-10-24  Martin Liska  <mliska@suse.cz>
1365         * cgraph.c (cgraph_node::dump):
1366         Remove reduntant dumps and make tp_first_run dump more compact.
1368 2018-10-24  Richard Biener  <rguenther@suse.de>
1370         PR tree-optimization/87665
1371         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
1372         to reflect reality.
1374 2018-10-12  Jeff Law  <law@redhat.com>
1376         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
1377         for H8/S.
1379 2018-10-23  Richard Biener  <rguenther@suse.de>
1381         * tree-vrp.c (add_assert_info): Guard dump_printf with
1382         dump_enabled_p.
1383         * gimple-ssa-evrp-analyze.c
1384         (evrp_range_analyzer::record_ranges_from_incoming_edge):
1385         Use value_range::ignore_equivs_equal_p.
1387 2018-10-23  Richard Biener  <rguenther@suse.de>
1389         PR tree-optimization/87105
1390         PR tree-optimization/87608
1391         * passes.def (pass_all_early_optimizations): Add early phi-opt
1392         after dce.
1393         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
1394         addition to debug stmts.
1395         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
1396         and abs replacement early.
1397         * tree-cfg.c (gimple_empty_block_p): Likewise.
1399 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
1401         PR target/86383
1402         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
1403         specified to configure.
1404         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
1406 2018-10-23  Richard Biener  <rguenther@suse.de>
1408         PR tree-optimization/87700
1409         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
1411 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
1413         PR target/87674
1414         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
1415         second argument from __mmask16 to __mmask8.
1416         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
1417         _mm_mask_packs_epi32): Likewise.
1418         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
1419         Likewise.
1420         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
1422 2018-10-23  Richard Biener  <rguenther@suse.de>
1424         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
1426 2018-10-23  Richard Biener  <rguenther@suse.de>
1428         PR tree-optimization/86144
1429         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
1430         over simd attribute.
1432 2018-10-23  Richard Biener  <rguenther@suse.de>
1434         PR tree-optimization/87693
1435         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
1436         the case we do not find the taken edge.
1438 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
1439             Jinsong Ji  <jji@us.ibm.com>
1441         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
1442         (_mm_store_pd): Use unaligned vector type for pointer cast.
1443         (_mm_maskmoveu_si128): Likewise.
1444         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
1445         (_mm_store_ps): Use unaligned vector type for pointer cast.
1447 2018-10-22  Paul Koning  <ni1d@arrl.net>
1449         * symtab.c (symtab_node::increase_alignment): Correct max
1450         alignment check.
1452 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
1454         PR tree-optimization/87633
1455         * match.pd: Do not generate unordered integer comparisons.
1457 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1459         PR rtl-optimization/87600
1460         * combine.c: Add include of expr.h.
1461         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
1462         register to a pseudo.
1463         (make_more_copies): New function, add a copy to a new pseudo after
1464         the moves from hard registers into pseudos.
1465         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
1466         later.  Call make_more_copies.
1468 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
1470         * lra-constraints.c (process_alt_operands): New local array,
1471         matching_early_clobber.  Check matching_early_clobber before
1472         decrementing reject, and set matching_early_clobber after.
1474 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
1476         PR target/87598
1477         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
1478         call output_addr_const and hope for the best.
1480 2018-10-22  Richard Biener  <rguenther@suse.de>
1482         * gimple-ssa-evrp-analyze.c
1483         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
1484         smarter about what ranges to use.
1485         * tree-vrp.c (add_assert_info): Dump here.
1486         (register_edge_assert_for_2): Instead of here at multiple but
1487         not all places.
1489         * gcc.dg/tree-ssa/evrp12.c: New testcase.
1490         * gcc.dg/predict-6.c: Adjust.
1491         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
1492         * gcc.dg/tree-ssa/vrp02.c: Likewise.
1493         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
1495 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
1496         Richard Biener  <rguenther@suse.de>
1498         * bitmap.h: Update data structure documentation, including a
1499         description of bitmap views as either linked-lists or splay trees.
1500         (struct bitmap_element_def): Update comments for splay tree bitmaps.
1501         (struct bitmap_head_def): Likewise.
1502         (bitmap_list_view, bitmap_tree_view): New prototypes.
1503         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
1504         tree_form fields.
1505         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
1506         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
1507         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
1508         released bitmap element here.
1509         (bitmap_element_free): Remove.
1510         (bitmap_elt_clear_from): Work on splay tree bitmaps.
1511         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
1512         this function similar ones such that linked-list bitmap implementation
1513         functions are grouped.
1514         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
1515         and moved for grouping.
1516         (bitmap_list_insert_element_after): Renamed from
1517         bitmap_elt_insert_after, and moved for grouping.
1518         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
1519         (bitmap_tree_link_left, bitmap_tree_link_right,
1520         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
1521         bitmap_tree_link_element, bitmap_tree_unlink_element,
1522         bitmap_tree_find_element): New functions for splay-tree bitmap
1523         implementation.
1524         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
1525         Renamed and moved, see above entries.
1526         (bitmap_tree_listify_from): New function to convert part of a splay
1527         tree bitmap to a linked-list bitmap.
1528         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
1529         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
1530         (bitmap_find_bit): Remove.
1531         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
1532         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
1533         Handle splay tree bitmaps.
1534         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
1535         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
1536         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
1537         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
1538         bitmap_intersect_compl_p, bitmap_ior_and_compl,
1539         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
1540         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
1541         corresponding changes to use linked-list specific bitmap_element
1542         manipulation functions as applicable for efficiency.
1543         (bitmap_tree_to_vec): New function.
1544         (debug_bitmap_elt_file): New function split out from ...
1545         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
1546         (bitmap_print): Likewise.
1548         PR tree-optimization/63155
1549         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
1550         SSA edge worklists.
1551         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
1552         in tree-view.
1554 2018-10-22  Martin Liska  <mliska@suse.cz>
1556         PR tree-optimization/87686
1557         Revert
1558         2018-08-29  Martin Liska  <mliska@suse.cz>
1560         * tree-switch-conversion.c (switch_conversion::expand):
1561         Strenghten assumption about gswitch statements.
1563 2018-10-22  Martin Liska  <mliska@suse.cz>
1565         * ipa-icf.c (sem_item::compare_attributes): Remove.
1566         (sem_item::compare_referenced_symbol_properties): Use
1567         attribute_list_equal instead.
1568         (sem_function::equals_wpa): Likewise.
1569         * ipa-icf.h: Remove compare_attributes.
1571 2018-10-22  Richard Biener  <rguenther@suse.de>
1573         PR middle-end/87682
1574         * mem-stats.h (mem_usage::operator==): Fix pasto.
1576 2018-10-22  Richard Biener  <rguenther@suse.de>
1578         PR tree-optimization/87640
1579         * tree-vrp.c (set_value_range_with_overflow): Decompose
1580         incomplete result.
1581         (extract_range_from_binary_expr_1): Adjust.
1583 2018-10-22  Martin Jambor  <mjambor@suse.cz>
1585         * tree-eh.h (stmt_could_throw_p): Add function parameter.
1586         (stmt_can_throw_external): Likewise.
1587         (stmt_can_throw_internal): Likewise.
1588         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
1589         (lower_eh_constructs_2): Likewise.
1590         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
1591         (stmt_can_throw_external): Likewise.
1592         (stmt_can_throw_internal): Likewise.
1593         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
1594         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
1595         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
1596         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
1597         (pass_lower_eh_dispatch::execute): Pass cfun to
1598         stmt_can_throw_external.
1599         (cleanup_empty_eh): Likewise.
1600         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
1601         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
1602         stmt_can_throw_external instead of pushing it to cfun.
1603         (symbol_table::create_edge): Likewise.
1604         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
1605         stmt_can_throw_internal.
1606         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
1607         to stmt_could_throw_p.
1608         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
1609         stmt_can_throw_internal.
1610         (pass_store_merging::execute): Likewise.
1611         * gimple-ssa-strength-reduction.c
1612         (find_candidates_dom_walker::before_dom_children): Pass cfun to
1613         stmt_could_throw_p.
1614         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
1615         stmt_can_throw_internal.
1616         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1617         to stmt_can_throw_external.
1618         (check_stmt): Pass cfun to stmt_could_throw_p.
1619         (check_stmt): Pass cfun to stmt_can_throw_external.
1620         (pass_nothrow::execute): Likewise.
1621         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1622         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1623         stmt_can_throw_internal.
1624         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1625         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1626         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1627         * tree-complex.c (expand_complex_libcall): Pass cfun to
1628         stmt_could_throw_p and to stmt_can_throw_internal.
1629         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1630         * tree-inline.c (copy_edges_for_bb): Likewise.
1631         (maybe_move_debug_stmts_to_successors): Likewise.
1632         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1633         stmt_could_throw_p.
1634         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1635         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1636         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1637         stmt_can_throw_internal.
1638         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1639         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1640         stmt_could_throw_p.
1641         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1642         stmt_can_throw_internal.
1643         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1644         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1645         stmt_could_throw_p.
1646         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1647         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1648         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1649         (convert_mult_to_fma_1): Likewise.
1650         (convert_to_divmod): Likewise.
1651         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1652         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1653         * tree-ssa-propagate.c
1654         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1655         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1656         (maybe_optimize_range_tests): Likewise.
1657         (linearize_expr_tree): Likewise.
1658         (reassociate_bb): Likewise.
1659         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1660         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1661         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1662         (handle_char_store): Likewise.
1663         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1664         stmt_can_throw_internal.
1665         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1666         stmt_could_throw_p.
1667         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1668         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1669         (vectorizable_simd_clone_call): Likewise.
1670         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1671         (gimple_stringop_fixed_value): Likewise.
1673 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
1675         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1676         literal pool references.
1677         (s390_check_qrst_address): Adapt to the new behavior of
1678         s390_loadrelative_operand_p ().
1680 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1682         PR target/72782
1683         * config/i386/sse.md (*andnot<mode>3_bcst): New.
1685 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1687         PR target/72782
1688         * config/i386/sse.md (*<code><mode>3_bcst): New.
1690 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1692         PR target/72782
1693         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1694         V4DI, V16SI and V8DI.
1695         (*sub<mode>3<mask_name>_bcst): New.
1696         (*add<mode>3<mask_name>_bcst): Likewise.
1698 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
1699             Jinsong Ji  <jji@us.ibm.com>
1701         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1702         __m64 with __vector unsigned long long for compatibility.
1703         (_mm_movemask_epi8): Likewise.
1704         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1705         (_mm_cvttps_pi32): Likewise.
1706         (_mm_cvtpi32_ps): Likewise.
1707         (_mm_cvtps_pi16): Likewise.
1708         (_mm_loadh_pi): Likewise.
1709         (_mm_storeh_pi): Likewise.
1710         (_mm_movehl_ps): Likewise.
1711         (_mm_movelh_ps): Likewise.
1712         (_mm_loadl_pi): Likewise.
1713         (_mm_storel_pi): Likewise.
1714         (_mm_movemask_ps): Likewise.
1715         (_mm_shuffle_pi16): Likewise.
1717 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1719         PR target/72782
1720         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1721         __builtin_ia32_vfnmsubpd512_mask.
1722         (_mm512_mask_fnmsub_round_pd): Likewise.
1723         (_mm512_fnmsub_pd): Likewise.
1724         (_mm512_mask_fnmsub_pd): Likewise.
1725         (_mm512_maskz_fnmsub_round_pd): Use
1726         __builtin_ia32_vfnmsubpd512_maskz.
1727         (_mm512_maskz_fnmsub_pd): Likewise.
1728         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1729         (_mm512_mask_fnmsub_round_ps): Likewise.
1730         (_mm512_fnmsub_ps): Likewise.
1731         (_mm512_mask_fnmsub_ps): Likewise.
1732         (_mm512_maskz_fnmsub_round_ps): Use
1733         __builtin_ia32_vfnmsubps512_maskz.
1734         (_mm512_maskz_fnmsub_ps): Likewise.
1735         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1736         __builtin_ia32_vfnmsubpd256_mask.
1737         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1738         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1739         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1740         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1741         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1742         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1743         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1744         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1745         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1746         __builtin_ia32_vfnmsubpd.
1747         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1748         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1749         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1750         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1751         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1752         * config/i386/i386-builtin.def: Add
1753         __builtin_ia32_vfnmsubpd256_mask,
1754         __builtin_ia32_vfnmsubpd256_maskz,
1755         __builtin_ia32_vfnmsubpd128_mask,
1756         __builtin_ia32_vfnmsubpd128_maskz,
1757         __builtin_ia32_vfnmsubps256_mask,
1758         __builtin_ia32_vfnmsubps256_maskz,
1759         __builtin_ia32_vfnmsubps128_mask,
1760         __builtin_ia32_vfnmsubps128_maskz,
1761         __builtin_ia32_vfnmsubpd512_mask,
1762         __builtin_ia32_vfnmsubpd512_maskz,
1763         __builtin_ia32_vfnmsubps512_mask,
1764         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1765         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1766         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1767         __builtin_ia32_vfnmsubpd256.
1768         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1769         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1770         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1771         Likewise.
1772         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1773         Likewise.
1774         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1775         Likewise.
1776         (fmai_vmfnmsub_<mode><round_name>): Likewise.
1778 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1780         PR target/72782
1781         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1782         __builtin_ia32_vfnmaddpd512_mask.
1783         (_mm512_mask_fnmadd_round_pd): Likewise.
1784         (_mm512_fnmadd_pd): Likewise.
1785         (_mm512_mask_fnmadd_pd): Likewise.
1786         (_mm512_maskz_fnmadd_round_pd): Use
1787         __builtin_ia32_vfnmaddpd512_maskz.
1788         (_mm512_maskz_fnmadd_pd): Likewise.
1789         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1790         (_mm512_mask_fnmadd_round_ps): Likewise.
1791         (_mm512_fnmadd_ps): Likewise.
1792         (_mm512_mask_fnmadd_ps): Likewise.
1793         (_mm512_maskz_fnmadd_round_ps): Use
1794         __builtin_ia32_vfnmaddps512_maskz.
1795         (_mm512_maskz_fnmadd_ps): Likewise.
1796         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1797         __builtin_ia32_vfnmaddpd256_mask.
1798         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1799         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1800         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1801         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1802         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1803         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1804         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1805         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1806         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1807         __builtin_ia32_vfnmaddpd.
1808         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1809         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1810         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1811         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1812         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1813         * config/i386/i386-builtin.def: Add
1814         __builtin_ia32_vfnmaddpd256_mask,
1815         __builtin_ia32_vfnmaddpd256_maskz,
1816         __builtin_ia32_vfnmaddpd128_mask,
1817         __builtin_ia32_vfnmaddpd128_maskz,
1818         __builtin_ia32_vfnmaddps256_mask,
1819         __builtin_ia32_vfnmaddps256_maskz,
1820         __builtin_ia32_vfnmaddps128_mask,
1821         __builtin_ia32_vfnmaddps128_maskz,
1822         __builtin_ia32_vfnmaddpd512_mask,
1823         __builtin_ia32_vfnmaddpd512_maskz,
1824         __builtin_ia32_vfnmaddps512_mask,
1825         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1826         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1827         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1828         __builtin_ia32_vfnmaddpd256.
1829         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1830         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1831         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1832         Likewise.
1833         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1834         Likewise.
1835         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1836         Likewise.
1837         (fmai_vmfnmadd_<mode><round_name>): Likewise.
1839 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1841         PR target/72782
1842         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1843         __builtin_ia32_vfmsubpd512_mask.
1844         (_mm512_mask_fmsub_round_pd): Likewise.
1845         (_mm512_fmsub_pd): Likewise.
1846         (_mm512_mask_fmsub_pd): Likewise.
1847         (_mm512_maskz_fmsub_round_pd): Use
1848         __builtin_ia32_vfmsubpd512_maskz.
1849         (_mm512_maskz_fmsub_pd): Likewise.
1850         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1851         (_mm512_mask_fmsub_round_ps): Likewise.
1852         (_mm512_fmsub_ps): Likewise.
1853         (_mm512_mask_fmsub_ps): Likewise.
1854         (_mm512_maskz_fmsub_round_ps): Use
1855         __builtin_ia32_vfmsubps512_maskz.
1856         (_mm512_maskz_fmsub_ps): Likewise.
1857         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1858         __builtin_ia32_vfmsubpd256_mask.
1859         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1860         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1861         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1862         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1863         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1864         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1865         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1866         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1867         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1868         __builtin_ia32_vfmsubpd.
1869         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1870         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1871         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1872         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1873         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1874         * config/i386/i386-builtin.def: Add
1875         __builtin_ia32_vfmsubpd256_mask,
1876         __builtin_ia32_vfmsubpd256_maskz,
1877         __builtin_ia32_vfmsubpd128_mask,
1878         __builtin_ia32_vfmsubpd128_maskz,
1879         __builtin_ia32_vfmsubps256_mask,
1880         __builtin_ia32_vfmsubps256_maskz,
1881         __builtin_ia32_vfmsubps128_mask,
1882         __builtin_ia32_vfmsubps128_maskz,
1883         __builtin_ia32_vfmsubpd512_mask,
1884         __builtin_ia32_vfmsubpd512_maskz,
1885         __builtin_ia32_vfmsubps512_mask,
1886         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1887         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1888         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1889         __builtin_ia32_vfmsubpd256.
1890         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1891         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1892         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1893         Likewise.
1894         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1895         Likewise.
1896         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1897         Likewise.
1898         (fmai_vmfmsub_<mode><round_name>): Likewise.
1900 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1902         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1903         Remove plus.  Renamed to ...
1904         (*sub<mode>3<mask_name>_bcst): This.
1905         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1906         (*add<mode>3<mask_name>_bcst): This.
1908 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1910         PR target/72782
1911         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1913 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1915         PR target/87662
1916         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1917         (_mm_or_epi32): Likewise.
1918         (_mm256_xor_epi32): Likewise.
1919         (_mm_xor_epi32): Likewise.
1920         (_mm256_or_epi64): Likewise.
1921         (_mm_or_epi64): Likewise.
1922         (_mm256_xor_epi64): Likewise.
1923         (_mm_xor_epi64): Likewise.
1925 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
1927         PR target/72782
1928         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1930 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
1932         PR middle-end/87647
1933         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1935 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
1937         * doc/ux.texi: Move @section directly after @node.
1939 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
1941         PR middle-end/85488
1942         PR middle-end/87649
1943         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1944         depend closely nested inside of loop with ordered clause with
1945         a parameter.
1947 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
1949         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1950         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1951         * doc/ux.texi: New file.
1953 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
1955         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1956         be the first CR field allocated.
1958 2018-10-19  Richard Biener  <rguenther@suse.de>
1960         PR target/87657
1961         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1962         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1964 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1966         PR target/72782
1967         * config/i386/sse.md
1968         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1969         (*add<mode>3<mask_name>_bcst_2): Likewise.
1971 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1973         * config/i386/sse.md
1974         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1975         Replace nonimmediate_operand with register_operand.
1976         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1977         Likewise.
1978         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1979         Likewise.
1981 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
1983         PR rtl-optimization/87596
1984         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1985         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1986         for instructions in FROM..TO range.
1988 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
1990         * cfgexpand.c (expand_one_var): Use specific wording in error message
1991         for non-local frame variables.
1992         * stor-layout.c (layout_decl): Do not issue a warning for them.
1994 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
1996         * haifa-sched.c (priority): Add force_recompute parameter.
1997         (apply_replacement): Call priority () with force_recompute = true.
1998         (restore_pattern): Likewise.
2000 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2002         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
2003         HOST_BITS_PER_WIDE_INT.
2004         (test_vector_ops_duplicate): Likewise.
2006 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2008         PR target/72782
2009         * config/i386/sse.md (VF_AVX512): New.
2010         (avx512bcst): Likewise.
2011         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
2012         Likewise.
2013         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
2014         Likewise.
2015         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
2016         Likewise.
2018 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
2020         * doc/invoke.texi (-dumpversion): Improve grammar.
2021         (-dumpfullversion): Make more consistent with -dumpversion.
2023 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
2025         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
2026         Set JUMP_LABEL to the jump insn.
2027         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
2028         Predict emitted jump and add label to jump insn.
2030 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
2032         PR tree-optimization/87562
2033         * input.c (get_substring_ranges_for_loc): Use
2034         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
2035         getting the linemap for the endpoint.  Verify that it's either
2036         in the same linemap as the start point's spelling location, or
2037         at least in the same file.
2039 2018-10-18  Richard Biener  <rguenther@suse.de>
2041         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
2042         feed width-specific load/store costs through ix86_vec_cost.
2043         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
2044         (k8_cost): Likewise.
2045         (bdver_cost): Likewise.
2046         (znver1_cost): Likewise.
2047         (btver1_cost): Likewise.
2048         (btver2_cost): Likewise.
2050 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2052         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
2053         to simplify subreg of vec_merge.
2055 2018-10-18  Richard Biener  <rguenther@suse.de>
2057         * config/i386/i386.c: Fix costing of vector FMA.
2059 2018-10-18  Richard Biener  <rguenther@suse.de>
2061         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
2062         and argument.
2063         (ix86_builtin_vectorization_cost): For vec_construct properly
2064         cost insertion into SSE regs.
2065         (...): Adjust calls to ix86_vec_cost.
2067 2018-10-18  Richard Biener  <rguenther@suse.de>
2069         PR middle-end/87087
2070         Revert
2071         2018-02-07  Richard Biener  <rguenther@suse.de>
2073         PR tree-optimization/84204
2074         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2075         this place.
2077 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2079         PR target/87537
2080         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
2081         of vec_duplicate.
2082         (test_vector_ops_duplicate): Add test for a scalar subreg of a
2083         VEC_MERGE of a VEC_DUPLICATE.
2085 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
2087         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
2088         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
2089         * doc/standards.texi (C Language): Document C2X.
2090         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
2091         (rl78_option_override): Handle "GNU C2X" language name.
2093 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
2095         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
2096         Document C17 as published in 2018.
2098 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
2100         PR middle-end/87623
2101         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
2102         bail out if both sides do not have the same storage order.
2104 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2106         * bitmap.c (bitmap_head::dump): New.
2107         * bitmap.h (bitmap_head): Add dump().
2108         * gimple-ssa-evrp-analyze.c
2109         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
2110         (evrp_range_analyzer::set_ssa_range_info): Same.
2111         (evrp_range_analyzer::record_ranges_from_phis): Same.
2112         (evrp_range_analyzer::record_ranges_from_stmt): Same.
2113         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
2114         * gimple-ssa-sprintf.c (get_int_range): Same.
2115         (format_integer): Same.
2116         (sprintf_dom_walker::handle_gimple_call): Same.
2117         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
2118         (ipcp_vr_lattice::top_p): Same.
2119         (ipcp_vr_lattice::bottom_p): Same.
2120         (ipcp_vr_lattice::set_to_bottom): Same.
2121         (ipa_vr_operation_and_type_effects): Same.
2122         (propagate_vr_across_jump_function): Same.
2123         (ipcp_store_vr_results): Same.
2124         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
2125         (ipa_print_node_jump_functions_for_edge): Same.
2126         (ipa_get_value_range): Same.
2127         (ipa_compute_jump_functions_for_edge): Same.
2128         (ipa_write_jump_function): Same.
2129         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
2130         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2131         Same.
2132         * vr-values.c (set_value_range_to_nonnegative): Same.
2133         (set_value_range_to_truthvalue): Same.
2134         (vr_values::get_value_range): Same.
2135         (vr_values::set_defs_to_varying): Same.
2136         (vr_values::update_value_range): Same.
2137         (symbolic_range_based_on_p): Same.
2138         (vr_values::op_with_boolean_value_range_p): Same.
2139         (vr_values::extract_range_for_var_from_comparison_expr): Same.
2140         (vr_values::extract_range_from_ssa_name): Same.
2141         (vr_values::extract_range_from_binary_expr): Same.
2142         (vr_values::extract_range_from_unary_expr): Same.
2143         (vr_values::extract_range_from_cond_expr): Same.
2144         (vr_values::extract_range_from_comparison): Same.
2145         (vr_values::check_for_binary_op_overflow): Same.
2146         (vr_values::extract_range_basic): Same.
2147         (vr_values::extract_range_from_assignment): Same.
2148         (compare_ranges): Same.
2149         (compare_range_with_value): Same.
2150         (vr_values::adjust_range_with_scev): Same.
2151         (vrp_valueize): Same.
2152         (vrp_valueize_1): Same.
2153         (vr_values::get_vr_for_comparison): Same.
2154         (vr_values::compare_name_with_value): Same.
2155         (vr_values::compare_names): Same.
2156         (vr_values::vrp_evaluate_conditional): Same.
2157         (find_case_label_ranges): Same.
2158         (vr_values::vrp_visit_switch_stmt): Same.
2159         (vr_values::extract_range_from_phi_node): Same.
2160         (vr_values::simplify_div_or_mod_using_ranges): Same.
2161         (vr_values::simplify_bit_ops_using_ranges): Same.
2162         (test_for_singularity): Same.
2163         (range_fits_type_p): Same.
2164         (vr_values::simplify_cond_using_ranges_1): Same.
2165         (vr_values::simplify_switch_using_ranges): Same.
2166         (vr_values::simplify_float_conversion_using_ranges): Same.
2167         (vr_values::two_valued_val_range_p): Same.
2168         (vr_values::add_equivalence): Move to value_range::equiv_add.
2169         * vr-values.h (vr_values::add_equivalence): Remove.
2170         (VR_INITIALIZER): Remove.
2171         * tree-vrp.c (value_range::set): New.
2172         (value_range::equiv_add): New.
2173         (value_range::value_range): New.
2174         (value_range::deep_copy): New.
2175         (value_range::check): New.
2176         (value_range::equal_p): New.
2177         (value_range::ignore_equivs_equal_p): New.
2178         (value_range::operator==): New.
2179         (value_range::operator!=): New.
2180         (value_range::symbolic_p): New.
2181         (value_range::numeric_p): New.
2182         (value_range::set_undefined): New.
2183         (value_range::set_varying): New.
2184         (value_range::may_contain_p): New.
2185         (value_range::equiv_clear): New.
2186         (value_range::singleton_p): New.
2187         (value_range::intersect): New.
2188         (value_range::dump): New.
2189         (value_range::set_and_canonicalize): New.
2190         (set_value_range): Adjust for value_range API.
2191         (set_value_range_to_undefined): Same.
2192         (set_value_range_to_varying): Same.
2193         (set_and_canonicalize_value_range): Same.
2194         (set_value_range_to_nonnull): Same.
2195         (set_value_range_to_null): Same.
2196         (range_is_null): Same.
2197         (range_is_nonnull): Same.
2198         (range_int_cst_p): Same.
2199         (range_int_cst_singleton_p): Same.
2200         (symbolic_range_p): Same.
2201         (range_includes_zero_p): Same.
2202         (value_range_constant_singleton): Same.
2203         (vrp_set_zero_nonzero_bits): Same.
2204         (ranges_from_anti_range): Same.
2205         (extract_range_into_wide_ints): Same.
2206         (extract_range_from_multiplicative_op): Same.
2207         (set_value_range_with_overflow): Same.
2208         (extract_range_from_binary_expr_1): Same.
2209         (extract_range_from_unary_expr): Same.
2210         (dump_value_range): Same.
2211         (debug_value_range): Same.
2212         (vrp_prop::check_array_ref): Same.
2213         (vrp_prop::check_mem_ref): Same.
2214         (vrp_prop::vrp_initialize): Same.
2215         (vrp_prop::visit_stmt): Same.
2216         (intersect_ranges): Same.
2217         (vrp_prop::visit_phi): Same.
2218         (vrp_prop::vrp_finalize): Same.
2219         (determine_value_range_1): Same.
2220         (determine_value_range): Same.
2221         (vrp_intersect_ranges_1): Rename to...
2222         (vrp_intersect_1): this.
2223         (vrp_intersect_ranges): Rename to...
2224         (value_range::intersect_helper): ...this.
2225         (vrp_meet_1): Rename to...
2226         (value_range::union_helper): ...this.
2227         (vrp_meet): Rename to...
2228         (value_range::union_): ...this.
2229         (copy_value_range): Remove.
2230         * tree-vrp.h (struct value_range): Rewrite into a proper class.
2231         (value_range::vrtype): New.
2232         (value_range::type): New.
2233         (value_range::equiv): New.
2234         (value_range::min): New.
2235         (value_range::max): New.
2236         (value_range::varying_p): New.
2237         (value_range::undefined_p): New.
2238         (value_range::null_p): New.
2239         (value_range::equiv_add): New.
2240         (copy_value_range): Remove.
2242 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
2244         * Makefile.in (SELFTEST_TARGETS): New.
2245         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
2246         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
2247         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
2248         c/Make-lang.in.
2249         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
2250         (selftest-c++-gdb, selftest-c++-valgrind): Move to
2251         cp/Make-lang.in.
2252         * configure: Regenerate.
2253         * configure.ac (selftest_languages): New.
2255 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2257         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
2258         overflow wraps argument.
2259         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
2260         wide_int_range_multiplicative_op.
2261         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
2262         overflow wraps argument.
2263         (wide_int_range_multiplicative_op): Same.
2264         (wide_int_range_lshift): Same.
2265         (wide_int_range_div): Same.
2266         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
2267         (wide_int_range_lshift): Same.
2268         (wide_int_range_div): Same.
2270 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
2272         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
2273         use sign as argument.
2274         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
2275         wide_int_range_shift_undefined_p.
2277 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2279         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
2280         Rename to...
2281         (@despeculate_copy<ALLI_TI:mode>): ... This.
2282         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
2283         switch statement.
2285 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2287         * config.gcc: Obsolete *-*-solaris2.10*.
2288         * doc/install.texi (Specific, *-*-solaris2*): Document it.
2290 2018-10-12  Jeff Law  <law@redhat.com>
2292         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
2293         reg + sym +- const_int addressing modes.
2295 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2297         * common.opt (fdiagnostics-minimum-margin-width=): New option.
2298         * diagnostic-show-locus.c (layout::layout): Apply the minimum
2299         margin width.
2300         (layout::start_annotation_line): Only print up to 3 of the
2301         margin character, to avoid touching the left-hand side.
2302         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
2303         minimum margin width, as set by test_diagnostic_context's ctor.
2304         (selftest::test_fixit_insert_containing_newline): Likewise.
2305         (selftest::test_fixit_insert_containing_newline_2): Likewise.
2306         (selftest::test_line_numbers_multiline_range): Clear
2307         dc.min_margin_width.
2308         * diagnostic.c (diagnostic_initialize): Initialize
2309         min_margin_width.
2310         * diagnostic.h (struct diagnostic_context): Add field
2311         "min_margin_width".
2312         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
2313         * opts.c (common_handle_option): Handle
2314         OPT_fdiagnostics_minimum_margin_width_.
2315         * selftest-diagnostic.c
2316         (selftest::test_diagnostic_context::test_diagnostic_context):
2317         Initialize min_margin_width to 6.
2318         * toplev.c (general_init): Initialize global_dc->min_margin_width.
2320 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
2322         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2323         Fix usage of "error_at_rich_loc" in the comment.
2325 2018-10-15  Renlin Li  <renlin.li@arm.com>
2327         PR target/87563
2328         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
2329         if-conversioned loop when it contains ifn with types not
2330         supported by backend.
2331         * internal-fn.c (expand_direct_optab_fn): Add an assert.
2332         (direct_internal_fn_supported_p): New helper function.
2333         * internal-fn.h (direct_internal_fn_supported_p): Declare.
2335 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
2337         PR target/87572
2338         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2339         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2340         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2342 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
2344         PR tree-optimization/87022
2345         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
2346         bits in dist vector rather than the first one.
2348 2018-10-15  Richard Biener  <rguenther@suse.de>
2350         PR middle-end/87610
2351         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2352         (visit_loadstore): When a used restrict tag escaped verify that
2353         the points-to solution of "other" pointers do not include
2354         escaped.
2355         (compute_dependence_clique): If a used restrict tag escaped
2356         communicated that down to visit_loadstore.
2358 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
2360         * config/s390/s390.c (s390_expand_vec_init): Force vector element
2361         into reg if it isn't a general operand.
2363 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2365         PR target/87599
2366         * config/i386/sse.md (*vec_dupv2di): Add register source to
2367         movddup.
2369 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
2371         PR target/87572
2372         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
2373         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
2374         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
2375         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
2377 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
2379         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
2380         (notice_args_size): Set it in the current trace if no insn that can
2381         throw internally has been seen yet.
2382         (connect_traces): When connecting args_size between traces, allow the
2383         incoming values not to match if there is an insn setting it before the
2384         first insn that can throw internally; in that case, force the creation
2385         of a CFI note on this latter insn.
2387 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
2389         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
2390         the base class, because using the injected-class-name was not clearly
2391         specified until DR 176.
2393 2018-10-12  Paul Koning  <ni1d@arrl.net>
2395         * config/pdp11/pdp11.md (doloop_end): New expander.
2396         (doloop_end_insn): renamed from "doloop_end".
2397         (addqi3): New pattern.
2398         (subqi3): New pattern.
2399         * config/pdp11/predicates.md (incdec_operand): New predicate.
2401 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
2403         PR middle-end/81376
2404         * real.c (format_helper::can_represent_integral_type_p): New function
2405         * real.h (format_helper::can_represent_integral_type_p): Ditto.
2406         * match.pd: New pattern.
2408 2018-10-12  Alexandre Oliva <oliva@adacore.com>
2410         * configure.ac: Introduce --enable-large-address-aware
2411         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
2412         * doc/install.texi: Document it.
2413         * configure, config.in: Rebuilt.
2414         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
2415         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
2416         (LINK_SPEC): Insert it.
2417         * config/i386/mingw-w64.h: Likewise.
2419         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
2421 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
2423         PR rtl-optimization/87600
2424         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
2426 2018-10-12  Paul Koning  <ni1d@arrl.net>
2428         * doc/md.texi (doloop_end): Document that the pattern code may
2429         need to check operand mode.
2431 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
2433         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
2434         to zero-extend between int and floating-point registers.
2435         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
2436         ldp into floating-point registers.  Add type and arch attributes.
2437         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
2438         Use f_loads for type attribute.
2440 2018-10-11  Martin Sebor  <msebor@redhat.com>
2442         * doc/extend.texi (attribute packed): Correct typos.
2444 2018-10-11  Martin Sebor  <msebor@redhat.com>
2446         * doc/extend.texi (attribute flatten): Mention interaction with
2447         noinline.
2449 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
2451         PR target/87156
2452         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2453         Set new_decl virtual flag to zero.
2455 2018-10-11  Martin Sebor  <msebor@redhat.com>
2457         PR middle-end/87593
2458         * doc/extend.texi (attribute format_arg): Discuss using multiple
2459         attributes on a single function.
2461 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
2463         PR tree-optimization/86829
2464         * match.pd (sin (atan (x))): New simplification rules.
2465         (cos (atan (x))): Likewise.
2466         * real.c (build_sinatan_real): New function.
2467         * real.h (build_sinatan_real): Prototype.
2469 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
2471         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
2472         function.
2473         (fold_mergeeo_helper): New helper function.
2474         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
2475         intrinsics.  Correct some whitespace indentation issues.
2477 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
2479         PR target/87511
2480         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2481         Use HOST_WIDE_INT_1U for shift.
2483 2018-10-11  Doug Rupp  <rupp@adacore.com>
2484             Olivier Hainque  <hainque@adacore.com>
2486         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
2487         Pass --relax to the linker for RTPs.
2488         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
2490 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
2491             Jan Hubicka  <jh@suse.cz>
2492             Martin Jambor  <mjambor@suse.cz>
2494         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
2495         the same elements are repeated rather than printing all of them.
2496         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
2497         "repeated" elements.
2498         * read-rtl-function.c (test_loading_repeat): New function.
2499         (read_rtl_function_c_tests): Call test_loading_repeat.
2500         * rtl-tests.c (test_dumping_repeat): New function.
2501         (rtl_tests_c_tests): Call test_dumping_repeat.
2503 2018-10-11  Richard Biener  <rguenther@suse.de>
2505         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
2506         bdver?_cost): Unify to ...
2507         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
2508         * config/i386/i386.c (processor_cost_table): Adjust.
2510 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
2512         PR middle-end/87574
2513         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
2514         the thunk when expanding to GIMPLE.
2516 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2518         * varasm.c (mergeable_string_section): Don't try to move zero-length
2519         strings to the merge section.
2521 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
2523         PR target/87573
2524         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
2526 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
2528         PR target/87550
2529         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2530         to special_args set.
2532 2018-10-10  Richard Biener  <rguenther@suse.de>
2534         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
2535         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
2536         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
2537         and recursing and pattern terminating the recursion on SSE
2538         vector width using ix86_expand_reduc.
2539         (reduc_sminmax_scal_<mode>): Split into part reducing to half
2540         width and recursing and SSE2 vector variant doing the final
2541         reduction with ix86_expand_reduc.
2542         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
2543         with terminating the recursion at AVX level, splitting that
2544         to SSE there.
2546 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
2548         * genmatch.c (error_cb): Rename to...
2549         (diagnostic_cb): ...this, converting int params to enums.
2550         (fatal_at): Update for renaming.
2551         (warning_at): Likewise.
2552         (main): Likewise.
2553         * input.c (selftest::ebcdic_execution_charset::apply):
2554         Update for renaming of...
2555         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
2556         to...
2557         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
2558         converting level and reason to enums.
2559         (class selftest::lexer_error_sink): Rename to...
2560         (class selftest::lexer_test_options): ...this, renaming field
2561         "m_errors" to "m_diagnostics".
2562         (selftest::lexer_test_options::apply): Update for renaming of...
2563         (selftest::lexer_test_options::on_error): ...this, renaming to...
2564         (selftest::lexer_test_options::on_diagnostic): ...this
2565         converting level and reason to enums.
2566         (selftest::test_lexer_string_locations_raw_string_unterminated):
2567         Update for renamings.
2568         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
2569         "reason".
2571 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
2573         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
2574         * config/rs6000/pmmintrin.h: New file.
2576 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
2578         PR tree-optimization/86659
2579         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
2581 2018-10-09  Richard Biener  <rguenther@suse.de>
2583         PR tree-optimization/63155
2584         * tree-ssa-structalias.c: Include tree-ssa.h.
2585         (get_constraint_for_ssa_var): For undefs return nothing_id.
2586         (find_func_aliases): Cleanup PHI handling.
2588 2018-10-09  Richard Biener  <rguenther@suse.de>
2590         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
2591         replacements.
2593 2018-10-09  Martin Liska  <mliska@suse.cz>
2595         * asan.c (asan_emit_stack_protection): If a stack variable
2596         is located in a same file as current function, then emit
2597         line info into variable definition string.
2599 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2601         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
2602         information.
2604 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2606         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
2607         on the thunk.
2609 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2611         PR tree-optimization/86659
2612         * gimple-match.h (struct gimple_match_op): Add reverse field.
2613         (gimple_match_op::set_op): New overloaded method.
2614         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
2615         the REF_REVERSE_STORAGE_ORDER flag on the value.
2616         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2617         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2618         
2619 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
2621         PR middle-end/63155
2622         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2623         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2625 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
2627         PR target/87517
2628         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2629         Defined with __builtin_ia32_vfmaddsubpd512_mask.
2631 2018-10-08  Richard Biener  <rguenther@suse.de>
2633         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2634         cost the same as AVX128 ones.
2636 2018-10-08  Paul Koning  <ni1d@arrl.net>
2638         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2639         (expand_block_move): New function.
2640         * config/pdp11/pdp11.c (output_block_move): Remove.
2641         (expand_block_move): New function.
2642         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2643         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2644         (*movmemhi1): Remove.
2646 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
2648         * config/s390/2827.md: Increase latencies for some FP instructions.
2650 2018-10-08  Richard Biener  <rguenther@suse.de>
2652         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2653         Open a dump scope.
2654         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2655         * tree-vectorizer.h (dump_stmt_cost): Adjust.
2656         (add_stmt_cost): Dump return value of the hook.
2658 2018-10-08  Richard Biener  <rguenther@suse.de>
2660         PR tree-optimization/63155
2661         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2662         (ssa_propagation_engine::ssa_propagate): Remove redundant
2663         bitmap bit clearing.
2665 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
2667         PR rtl-optimization/86939
2668         PR rtl-optimization/87479
2669         * ira.h (non_conflicting_reg_copy_p): New prototype.
2670         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2671         (make_hard_regno_dead): Don't add conflicts for register
2672         ignore_reg_for_conflicts.
2673         (make_object_dead): Likewise.
2674         (non_conflicting_reg_copy_p): New function.
2675         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2676         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2677         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2678         (make_hard_regno_dead): Don't add conflicts for register
2679         ignore_reg_for_conflicts.  Remove special conflict handling of
2680         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
2681         check_pic_pseudo_p and update callers.
2682         (mark_pseudo_dead): Don't add conflicts for register
2683         ignore_reg_for_conflicts.
2684         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2686 2018-10-05  Andrew Waterman  <andrew@sifive.com>
2687             Jim Wilson  <jimw@sifive.com>
2689         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2690         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
2691         new pattern using HONOR_SNANS that emits one extra instruction.
2693 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2695         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2696         patterns): Merge SI and DI patterns to a GPR pattern.
2697         (unnamed define_insn and define_split for record form of that): Merge
2698         to a single define_insn_and_split pattern.
2700 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
2702         PR c++/56856
2703         * input.c (expand_location_to_spelling_point): Add param "aspect"
2704         and use rather than hardcoding LOCATION_ASPECT_CARET.
2705         (get_substring_ranges_for_loc): Handle the case of a single token
2706         within a macro expansion.
2707         * input.h (expand_location_to_spelling_point): Add "aspect" param,
2708         defaulting to LOCATION_ASPECT_CARET.
2710 2018-10-05  Paul Koning  <ni1d@arrl.net>
2712         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2713         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2714         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2715         (pdp11_guard_type): New function.
2717 2018-10-05  Paul Koning  <ni1d@arrl.net>
2719         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2720         * config/pdp11/pdp11.opt (mfloat32): Remove.
2721         (mfloat64): Remove.
2722         * doc/invoke.texi (pdp11 -mfloat32): Remove:
2723         (pdp11 -mfloat64): Remove.
2725 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2727         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2728         (*cmp<mode>_cc_i387): Ditto.
2729         (*cmpu<mode>_cc_i387): Ditto.
2730         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2731         * config/i386/i386.c (ix86_expand_fp_compare): Remove
2732         "scratch" argument.
2733         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2734         Emit x86_sahf_1 pattern.
2735         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2736         (ix86_expand_carry_flag_compare): Ditto.
2738 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2740         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2741         to reg_or_0_operand.  Add "C" constraint.
2742         (*cmpxf_cc_i387): Ditto.
2743         (*cmp<mode>_i387): Change operand 2 predicate
2744         to nonimm_or_0_operand.  Add "C" constraint.
2745         (*cmp<mode>_cc_i387): Ditto.
2746         (*cmp<mode>_0_i387): Remove insn pattern.
2747         (*cmp<mode>_0_cc_i387): Ditto.
2749 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2751         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2752         * config/i386/predicates.md (nonimm_or_0_operand): Rename
2753         from vector_move_operand.  Update all uses.
2755 2018-10-05  Martin Sebor  <msebor@redhat.com>
2757         PR tree-optimization/87490
2758         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2759         consistently.
2761 2018-10-05  Richard Biener  <rguenther@suse.de>
2763         PR tree-optimization/63155
2764         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2765         vertical space in dumpfiles.
2766         * tree-ssa-propagate.h
2767         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2768         * tree-ssa-propagate.c (cfg_blocks_back): New global.
2769         (ssa_edge_worklist_back): Likewise.
2770         (curr_order): Likewise.
2771         (cfg_blocks_get): Remove abstraction.
2772         (cfg_blocks_add): Likewise.
2773         (cfg_blocks_empty_p): Likewise.
2774         (add_ssa_edge): Add to current or next worklist based on
2775         RPO index.
2776         (add_control_edge): Likewise.
2777         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2778         into ...
2779         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
2780         iteration from CFG and SSA edge worklist so we process
2781         everything in RPO order, prioritizing forward progress
2782         over iteration.
2783         (ssa_prop_init): Allocate new worklists, do not dump
2784         immediate uses.
2785         (ssa_prop_fini): Free new worklists.
2787 2018-10-05  Richard Biener  <rguenther@suse.de>
2789         * tree-core.h (tree_block::abstract_flag): Remove.
2790         (tree_block::block_num): Make full 32bits.
2791         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2792         * tree.h (BLOCK_ABSTRACT): Remove.
2793         * dwarf2out.c (gen_lexical_block_die): Remove dead code
2794         resulting from BLOCK_ABSTRACT being always false.
2795         (gen_inlined_subroutine_die): Likewise.
2796         (gen_block_die): Likewise.
2797         * tree.c (block_ultimate_origin): Likewise.
2798         * tree-pretty-print.c (dump_block_node): Remove code dealing
2799         with BLOCK_ABSTRACT.
2800         * tree-ssa-live.c (dump_scope_block): Likewise.
2801         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2802         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2804 2018-10-05   Richard Biener  <rguenther@suse.de>
2806         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2807         is asked for initialize mode to the component mode of the
2808         vector type.
2810 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
2812         PR target/87522
2813         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2814         assembler for -mavx.
2815         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2817 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2819         PR target/87509
2820         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2821         RS6000_BTM_DFP.
2822         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2823         to be DImode.  When using mffscrn, force the operand to a register.
2825 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
2827         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2828         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2829         X87MODEF mode iterator.
2830         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2831         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2832         X87MODEF mode iterator.
2834 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
2836         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2837         -Wno-prio-ctor-dtor.
2839 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2841         * Makefile.in (OBJS): Add opt-problem.o.
2842         * dump-context.h: Include "selftest.h.
2843         (selftest::temp_dump_context): New forward decl.
2844         (class dump_context): Make friend of class
2845         selftest::temp_dump_context.
2846         (dump_context::dump_loc_immediate): New decl.
2847         (class dump_pretty_printer): Move here from dumpfile.c.
2848         (class temp_dump_context): Move to namespace selftest.
2849         (temp_dump_context::temp_dump_context): Add param
2850         "forcibly_enable_dumping".
2851         (selftest::verify_dumped_text):
2852         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2853         (selftest::verify_item):
2854         (ASSERT_IS_TEXT): Move here from dumpfile.c.
2855         (ASSERT_IS_TREE): Likewise.
2856         (ASSERT_IS_GIMPLE): Likewise.
2857         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2858         to...
2859         (dump_context::dump_loc_immediate): ...this new function.
2860         (class dump_pretty_printer): Move to dump-context.h.
2861         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2862         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2863         (temp_dump_context::temp_dump_context): Move to "selftest"
2864         namespace.  Add param "forcibly_enable_dumping", and use it to
2865         conditionalize the use of m_pp;
2866         (selftest::verify_dumped_text): Make non-static.
2867         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2868         (selftest::verify_item): Make non-static.
2869         (ASSERT_IS_TEXT): Move to dump-context.h.
2870         (ASSERT_IS_TREE): Likewise.
2871         (ASSERT_IS_GIMPLE): Likewise.
2872         (selftest::test_capture_of_dump_calls): Pass "true" for new
2873         param of temp_dump_context.
2874         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2875         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
2876         TDF_COMPARE_DEBUG.
2877         * opt-problem.cc: New file.
2878         * opt-problem.h: New file.
2879         * optinfo-emit-json.cc
2880         (selftest::test_building_json_from_dump_calls): Pass "true" for
2881         new param of temp_dump_context.
2882         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2883         (optinfo::emit_for_opt_problem): New function.
2884         (optinfo::emit): Clarity which emit_item is used.
2885         * optinfo.h (optinfo::get_dump_location): New accessor.
2886         (optinfo::emit_for_opt_problem): New decl.
2887         (optinfo::emit): Make const.
2888         * selftest-run-tests.c (selftest::run_tests): Call
2889         selftest::opt_problem_cc_tests.
2890         * selftest.h (selftest::opt_problem_cc_tests): New decl.
2891         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2892         bool to opt_result, converting fprintf messages to
2893         opt_result::failure_at calls.  Add "stmt" param for use by the
2894         failure_at calls.
2895         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2896         (runtime_alias_check_p): Convert return type from bool to
2897         opt_result, converting dump_printf calls to
2898         opt_result::failure_at, using the statement DDR_A for their
2899         location.
2900         (find_data_references_in_stmt): Convert return type from bool to
2901         opt_result, converting "return false" to opt_result::failure_at
2902         with a new message.
2903         * tree-data-ref.h: Include "opt-problem.h".
2904         (dr_analyze_innermost): Convert return type from bool to opt_result,
2905         and add a const gimple * param.
2906         (find_data_references_in_stmt): Convert return type from bool to
2907         opt_result.
2908         (runtime_alias_check_p): Likewise.
2909         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2910         dr_analyze_innermost.
2911         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2912         Convert return type from bool to opt_result, adding a message for
2913         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2914         (vect_analyze_data_ref_dependence): Convert return type from bool
2915         to opt_result.  Change sense of return type from "false"
2916         effectively meaning "no problems" to "false" meaning a problem,
2917         so that "return false" becomes "return opt_result::success".
2918         Convert "return true" calls to opt_result::failure_at, using
2919         the location of statement A rather than vect_location.
2920         (vect_analyze_data_ref_dependences): Convert return type from bool
2921         to opt_result.
2922         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2923         calls to opt_result::failure_at, using the stmt location rather
2924         than vect_location.
2925         (vect_verify_datarefs_alignment): Convert return type from bool
2926         to opt_result.
2927         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
2928         into multiple more-tightly-scoped copies.
2929         (vect_analyze_data_refs_alignment): Convert return type from bool
2930         to opt_result.
2931         (vect_analyze_data_ref_accesses): Likewise, converting a
2932         "return false" to a "return opt_result::failure_at", adding a
2933         new message.
2934         (vect_prune_runtime_alias_test_list): Convert return type from
2935         bool to opt_result, converting dump_printf_loc to
2936         opt_result::failure_at.  Add a %G to show the pertinent statement,
2937         and use the stmt's location rather than vect_location.
2938         (vect_find_stmt_data_reference): Convert return type from
2939         bool to opt_result, converting dump_printf_loc to
2940         opt_result::failure_at, using stmt's location.
2941         (vect_analyze_data_refs):  Convert return type from bool to
2942         opt_result.  Convert "return false" to "return
2943         opt_result::failure_at", adding messages as needed.
2944         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2945         type from bool to opt_result.
2946         (vect_determine_vf_for_stmt): Likewise.
2947         (vect_determine_vectorization_factor): Likewise, converting
2948         dump_printf_loc to opt_result::failure_at, using location of phi
2949         rather than vect_location.
2950         (vect_analyze_loop_form_1): Convert return type from bool to
2951         opt_result, converting dump_printf_loc calls, retaining the use of
2952         vect_location.
2953         (vect_analyze_loop_form): Convert return type from loop_vec_info
2954         to opt_loop_vec_info.
2955         (vect_analyze_loop_operations): Convert return type from bool to
2956         opt_result, converting dump_printf_loc calls, using the location
2957         of phi/stmt rather than vect_location where available.  Convert
2958         various "return false" to "return opt_result::failure_at" with
2959         "unsupported phi" messages.
2960         (vect_get_datarefs_in_loop): Convert return type from bool to
2961         opt_result.  Add a message for the
2962         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2963         (vect_analyze_loop_2): Convert return type from bool to
2964         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
2965         each "goto again;", adding new messages where needed.
2966         Add "unsupported grouped {store|load}" messages.
2967         (vect_analyze_loop): Convert return type from loop_vec_info to
2968         opt_loop_vec_info.
2969         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2970         bool to opt_result.
2971         * tree-vect-stmts.c (process_use): Likewise, converting
2972         dump_printf_loc call and using stmt location, rather than
2973         vect_location.
2974         (vect_mark_stmts_to_be_vectorized): Likeise.
2975         (vect_analyze_stmt): Likewise, adding a %G.
2976         (vect_get_vector_types_for_stmt): Convert return type from bool to
2977         opt_result, converting dump_printf_loc calls and using stmt
2978         location, rather than vect_location.
2979         (vect_get_mask_type_for_stmt): Convert return type from tree to
2980         opt_tree, converting dump_printf_loc calls and using stmt location.
2981         * tree-vectorizer.c: Include "opt-problem.h.
2982         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2983         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
2984         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
2985         enabled, use it to report at the top level "couldn't vectorize
2986         loop" followed by the problem.
2987         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2988         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2989         to opt_result.
2990         (vect_analyze_stmt): Likewise.
2991         (vect_get_vector_types_for_stmt): Likewise.
2992         (tree vect_get_mask_type_for_stmt): Likewise.
2993         (vect_analyze_data_ref_dependences): Likewise.
2994         (vect_enhance_data_refs_alignment): Likewise.
2995         (vect_analyze_data_refs_alignment): Likewise.
2996         (vect_verify_datarefs_alignment): Likewise.
2997         (vect_analyze_data_ref_accesses): Likewise.
2998         (vect_prune_runtime_alias_test_list): Likewise.
2999         (vect_find_stmt_data_reference): Likewise.
3000         (vect_analyze_data_refs): Likewise.
3001         (vect_analyze_loop): Convert return type from loop_vec_info to
3002         opt_loop_vec_info.
3003         (vect_analyze_loop_form): Likewise.
3004         (vect_analyze_slp): Convert return type from bool to opt_result.
3006 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3008         * doc/invoke.texi (-fopt-info): Document new "internals"
3009         sub-option.
3010         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
3011         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
3012         MSG_ALL_KINDS.
3013         (optinfo_verbosity_options): Add "internals".
3014         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
3015         (dump_context::apply_dump_filter_p): New member function.
3016         (dump_context::dump_loc): Use apply_dump_filter_p rather than
3017         explicitly masking the dump_kind.
3018         (dump_context::begin_scope): Increment the scope depth first.  Use
3019         apply_dump_filter_p rather than explicitly masking the dump_kind.
3020         (dump_context::emit_item): Use apply_dump_filter_p rather than
3021         explicitly masking the dump_kind.
3022         (dump_dec): Likewise.
3023         (dump_hex): Likewise.
3024         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
3025         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
3026         (opt_info_switch_p): Update handling of default
3027         MSG_OPTIMIZED_LOCATIONS to cope with default of
3028         MSG_PRIORITY_USER_FACING.
3029         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
3030         masking the dump_kind.
3031         (selftest::test_capture_of_dump_calls): Update test_dump_context
3032         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
3033         than MSG_ALL.  Generalize scope test to be run at all four
3034         combinations of with/without MSG_PRIORITY_USER_FACING and
3035         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
3036         for each of the two values.
3037         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
3038         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
3039         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
3040         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
3041         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
3042         with MSG_PRIORITY_*.
3043         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
3044         dump messages as MSG_PRIORITY_USER_FACING.
3045         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
3046         about the interaction with MSG_PRIORITY_*.
3048 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3050         * varasm.c (output_constant): Add new parameter merge_strings.
3051         Make strings properly zero terminated in merge string sections.
3052         (mergeable_string_section): Don't fail if the last char is non-zero.
3053         (assemble_variable_contents): Handle merge string sections.
3054         (assemble_variable): Likewise.
3055         (assemble_constant_contents): Likewise.
3056         (output_constant_def_contents): Likewise.
3057         (output_constructor_array_range,
3058         output_constructor_regular_field): Adjust call to output_constant.
3059         (output_object_block): Adjust call to assemble_constant_contents
3060         and assemble_variable_contents.
3062 2018-10-04  Martin Liska  <mliska@suse.cz>
3064         PR c/87483
3065         * cgraphunit.c (process_function_and_variable_attributes):
3066         Warn about a function with alias attribute and a body.
3068 2018-10-04  Martin Liska  <mliska@suse.cz>
3070         PR ipa/82625
3071         * multiple_target.c (redirect_to_specific_clone): New function.
3072         (ipa_target_clone): Use it.
3073         * tree-inline.c: Fix comment.
3075 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3077         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
3078         fields.
3079         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
3080         (gcc::dump_manager::register_pass): New member function, adapted
3081         from loop body in gcc::pass_manager::register_pass, adding a
3082         call to update_dfi_for_opt_info.
3083         (gcc::dump_manager::opt_info_enable_passes): Store the
3084         -fopt-info options into the new fields.  Move the loop
3085         bodies into...
3086         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
3087         function.
3088         * dumpfile.h (struct opt_pass): New forward decl.
3089         (gcc::dump_manager::register_pass): New decl.
3090         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
3091         (class gcc::dump_manager): Add fields "m_optgroup_flags",
3092         "m_optinfo_flags", and "m_optinfo_filename".
3093         * passes.c (gcc::pass_manager::register_pass): Move all of the
3094         dump-handling code to gcc::dump_manager::register_pass.
3096 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
3098         PR rtl-optimization/87466
3099         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
3100         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
3101         * doc/tm.texi: Regenerate.
3102         * ira-lives.c (process_bb_node_lives): Use the new target hook.
3103         * lra-lives.c (process_bb_lives): Likewise.
3104         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
3105         Define.
3107 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
3109         * params.c (add_params): Fix initialization.
3111 2018-10-04  Martin Liska  <mliska@suse.cz>
3113         PR gcov-profile/84107
3114         * tree-profile.c (init_ic_make_global_vars):
3115         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
3116         Come up with new ic_tuple* variables.  Emit
3117         __gcov_indirect_call{,_topn} variables.
3118         (gimple_gen_ic_profiler): Access the variable
3119         and emit gimple.
3120         (gimple_gen_ic_func_profiler): Access
3121         __gcov_indirect_call.callee field.
3122         (gimple_init_gcov_profiler): Use ptr_type_node.
3123         * value-prof.c (gimple_ic): Use ptr_type_node.
3125 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3127         PR tree-optimization/85787
3128         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
3129         into this function and add support for detecting multiple phis.
3130         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
3132 2018-10-04  Martin Liska  <mliska@suse.cz>
3134         PR ipa/87491
3135         * ipa-inline.c (inline_to_all_callers_1):
3136         Call ultimate_alias_target for node being inlined.
3138 2018-10-03  Jeff Law  <law@redhat.com>
3140         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
3141         target's wchar_t.
3142         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
3143         * tree.h (get_typenode_from_name): Prototype.
3145 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
3147         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
3148         Change operand 2 predicate to nonimmediate_operand.
3149         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
3151 2018-10-03  Martin Sebor  <msebor@redhat.com>
3152             Jeff Law  <law@redhat.com>
3154         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
3155         initialize it.
3156         (get_string_length): Detect unterminated arrays.
3157         (format_string): Same.
3158         (format_directive): Warn about unterminated arrays.
3159         (handle_gimple_call): Mark statements with no_warning as needed.
3161 2018-10-03  Jim Wilson  <jimw@sifive.com>
3163         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
3164         also define __riscv_abi_rve.  Delete trailing white space.
3166 2018-10-03  Paul Koning  <ni1d@arrl.net>
3168         Enable LRA register allocator for PDP11.
3169         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
3170         (R): Likewise.
3171         (D): Likewise.
3172         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
3173         * config/pdp11/pdp11.opt (-mlra): New option.
3174         * doc/invoke.texi (PDP-11 Options): Document -mlra.
3176 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
3178         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
3179         (*<absneg:code>extend<mode>xf2): Ditto.
3181 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
3183         PR tree-optimization/87415
3184         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
3185         precision fields.
3187 2018-10-02  Jeff Law  <law@redhat.com>
3189         * gimple-fold.c (get_range_strlen): Only set *nonstr when
3190         an unterminated string is discovered.  Bubble up range
3191         even for unterminated strings.
3192         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
3193         indicates the string was not terminated via NONSTR.
3195 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
3197         * tree-vrp.c (extract_range_from_unary_expr): Special case all
3198         pointer conversions.
3199         Do not do anything special for anti-ranges.
3201 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
3203         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
3204         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
3206 2018-10-03  Martin Liska  <mliska@suse.cz>
3208         PR gcov-profile/86109
3209         * coverage.c (coverage_begin_function): Do not
3210         mark lambdas as artificial.
3211         * tree-core.h (struct GTY): Remove tm_clone_flag
3212         and introduce new lambda_function.
3213         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
3215 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3217         PR target/87474
3218         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
3219         P8_VECTOR and VSX are enabled.
3221 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3223         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
3224         0x3907 as CPU model number.
3226 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
3228         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
3229         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
3230         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
3231         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
3232         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
3233         * config/s390/s390.md: Likewise. Rename also the cpu attribute
3234         value from arch12 to z14.
3236 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3238         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
3239         (isinfxf2): Ditto.
3240         (isinf<mode>2): Ditto.
3242 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
3244         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
3245         before emitting fxam.  Perform calculations in XFmode.
3247 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3249         * match.pd (((X /[ex] A) +- B) * A): New transformation.
3251 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
3253         PR middle-end/87319
3254         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
3255         * tree.c (signed_or_unsigned_type_for): Handle complex.
3257 2018-10-02  Jeff Law  <law@redhat.com>
3259         * gimple-fold.c (get_range_strlen): Remove dead code.
3261 2018-10-02  Martin Sebor  <msebor@redhat.com>
3262             Jeff Law  <law@redhat.com>
3264         * builtins.c (unterminated_array): Add new arguments.
3265         If argument is not terminated, bubble up size and exact
3266         state to callers.
3267         (expand_builtin_strnlen): Detect, avoid expanding
3268         and diagnose unterminated arrays.
3269         (c_strlen): Fill in offset of start of unterminated strings.
3270         * builtins.h (unterminated_array): Update prototype.
3272 2018-10-02  Richard Biener  <rguenther@suse.de>
3274         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
3275         of haddv4df, first reduce to SSE width and exploit the fact
3276         that we only need element zero with the reduction result.
3277         (reduc_plus_scal_v2df): Likewise.
3279 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3281         * dojump.h (do_jump): Delete.
3282         (do_jump_1): Likewise.
3283         (split_comparison): Move around.
3284         * dojump.c (do_jump): Make static.
3285         (do_jump_1): Likewise.
3286         (jumpifnot): Move around.
3287         (jumpifnot_1): Likewise.
3288         (jumpif): Likewise.
3289         (jumpif_1): Likewise.
3290         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
3292 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
3294         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
3295         insns in the delay slot and add_insn_after for the jump insn.
3297 2018-10-02  Richard Biener  <rguenther@suse.de>
3299         * tree-inline.c (expand_call_inline): Use the location of
3300         the callee declaration for the inline-entry marker.
3301         * final.c (notice_source_line): Remove special-casing of
3302         NOTE_INSN_INLINE_ENTRY.
3304 2018-10-01  Carl Love  <cel@us.ibm.com>
3306         PR 69431
3307         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
3308         (__builtin_mtfsb0): New.
3309         (__builtin_mtfsb1): New.
3310         ( __builtin_set_fpscr_rn): New.
3311         (__builtin_set_fpscr_drn): New.
3312         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
3313         (rs6000_expand_set_fpscr_rn_builtin): Add.
3314         (rs6000_expand_set_fpscr_drn_builtin): Add.
3315         (rs6000_expand_builtin): Add case statement entries for
3316         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
3317         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
3318         RS6000_BUILTIN_MFFSL.
3319         (rs6000_init_builtins): Add ftype initialization and def_builtin
3320         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
3321         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
3322         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
3323         rs6000_mffscdrn): Add define_insn.
3324         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
3325         * doc/extend.texi: Add documentation for the builtins.
3327 2018-10-01  Richard Biener  <rguenther@suse.de>
3329         PR tree-optimization/87465
3330         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
3331         causing branch miscounts.
3333 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3335         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
3336         aarch64_option_default_param):  New.
3337         (params.h): Include.
3338         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
3339         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
3340         stack-clash protection validation code.
3342 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3344         * params.c (validate_param): New.
3345         (add_params): Use it.
3346         (set_param_value): Refactor param validation into validate_param.
3347         (diagnostic.h): Include.
3348         * diagnostic.h (diagnostic_ready_p): New.
3350 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3352         * params.c (set_param_value):
3353         Add index of parameter being validated.
3354         * common/common-target.def (option_validate_param): New.
3355         * common/common-targhooks.h (default_option_validate_param): New.
3356         * common/common-targhooks.c (default_option_validate_param): New.
3357         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
3358         * doc/tm.texi: Regenerate.
3360 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3362         PR target/86486
3363         * config/aarch64/aarch64.c (aarch64_override_options_internal):
3364         Add validation for stack-clash parameters and set defaults.
3366 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3368         PR target/86486
3369         * configure.ac: Add stack-clash-protection-guard-size.
3370         * doc/install.texi: Document it.
3371         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
3372         * params.def: Update comment for guard-size.
3373         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
3374         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
3375         * configure: Regenerate.
3377 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3379         PR target/86486
3380         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
3381         STACK_DYNAMIC_OFFSET): New.
3382         * config/aarch64/aarch64.c (aarch64_layout_frame):
3383         Update outgoing args size.
3384         (aarch64_stack_clash_protection_alloca_probe_range,
3385         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3387 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3389         PR target/86486
3390         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
3391         probe ranges.
3392         * target.def (stack_clash_protection_alloca_probe_range): New.
3393         (stack_clash_protection_final_dynamic_probe): Remove.
3394         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
3395         (default_stack_clash_protection_final_dynamic_probe): Remove.
3396         * targhooks.c: Likewise.
3397         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
3398         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
3399         * doc/tm.texi: Regenerate.
3401 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3403         PR target/86486
3404         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
3405         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
3406         aarch64_clamp_to_uimm12_shift): New.
3407         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
3408         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
3410 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
3412         PR target/86486
3413         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
3415 2018-10-01  Jeff Law  <law@redhat.com>
3416             Richard Sandiford <richard.sandiford@linaro.org>
3417             Tamar Christina  <tamar.christina@arm.com>
3419         PR target/86486
3420         * config/aarch64/aarch64.md
3421         (probe_stack_range): Add k (SP) constraint.
3422         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
3423         STACK_CLASH_MAX_UNROLL_PAGES): New.
3424         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
3425         stack probes for stack clash.
3426         (aarch64_allocate_and_probe_stack_space): New.
3427         (aarch64_expand_prologue): Use it.
3428         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
3429         (aarch64_sub_sp): Add emit_move_imm optional param.
3431 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
3433         PR tree-optimization/87261
3434         * match.pd: Remove trailing whitespace.
3435         Add (x & y) | ~(x | y) -> ~(x ^ y),
3436         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
3438 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3440         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
3441         constraints.
3442         (ashlsi3_insn): Update instruction constraints.
3443         (ashrsi3_insn): Likewise.
3444         (rotrsi3): Likewise.
3445         (add_shift): Likewise.
3446         * config/arc/constraints.md (Csz): New 32 bit constraint. It
3447         avoids placing in the limm field small constants which, otherwise,
3448         could end into a small instruction.
3450 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
3452         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
3453         destination register is not odd-even.
3454         (umaddsidi4_split): Likewise.
3456 2018-10-01  Richard Biener  <rguenther@suse.de>
3458         * tree-inline.c (expand_call_inline): Store origin of fn
3459         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
3460         * tree.c (block_ultimate_origin): Simplify and do some
3461         checking.
3463 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
3465         * config/i386/mmx.md (EMMS): New int iterator.
3466         (emms): New int attribute.
3467         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
3468         EMMS int iterator.  Explicitly declare clobbers.
3469         (mmx_emms): Remove expander.
3470         (mmx_femms): Ditto.
3471         * config/i386/predicates.md (emms_operation): Remove predicate.
3472         (vzeroall_pattern): New predicate.
3473         (vzeroupper_pattern): Rename from vzeroupper_operation.
3474         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
3475         vzeroupper_pattern and vzeroall_pattern predicates.
3477 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
3479         PR rtl-optimization/86939
3480         * ira-lives.c (make_hard_regno_born): Rename from this...
3481         (make_hard_regno_live): ... to this.  Remove update to conflict
3482         information.  Update function comment.
3483         (make_hard_regno_dead): Add conflict information update.  Update
3484         function comment.
3485         (make_object_born): Rename from this...
3486         (make_object_live): ... to this.  Remove update to conflict information.
3487         Update function comment.
3488         (make_object_dead):  Add conflict information update.  Update function
3489         comment.
3490         (mark_pseudo_regno_live): Call make_object_live.
3491         (mark_pseudo_regno_subword_live): Likewise.
3492         (mark_hard_reg_dead): Update function comment.
3493         (mark_hard_reg_live): Call make_hard_regno_live.
3494         (process_bb_node_lives): Likewise.
3495         * lra-lives.c (make_hard_regno_born): Rename from this...
3496         (make_hard_regno_live): ... to this.  Remove update to conflict
3497         information.  Remove now uneeded check_pic_pseudo_p argument.
3498         Update function comment.
3499         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
3500         to conflict information.  Update function comment.
3501         (mark_pseudo_live): Remove update to conflict information.  Update
3502         function comment.
3503         (mark_pseudo_dead): Add conflict information update.
3504         (mark_regno_live): Call make_hard_regno_live.
3505         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
3506         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
3508 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
3510         PR target/87370
3511         * config/i386/i386.c (construct_container): Use TImode for
3512         BLKmode values in 2 integer registers.
3514 2018-09-29  Jeff Law  <law@redhat.com>
3516         * builtins.c (unterminated_array): Pass in c_strlen_data * to
3517         c_strlen rather than just a tree *.
3518         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
3519         Update recursive calls appropriately.  If caller did not provide a
3520         suitable data pointer, create a local one.  When a non-terminated
3521         string is discovered, bubble up information about the string via the
3522         c_strlen_data object.
3523         * builtins.h (c_strlen): Update prototype.
3524         (c_strlen_data): New structure.
3525         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
3526         For a type 2 call, if c_strlen indicates a non-terminated string
3527         use the length of the non-terminated string.
3528         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
3530 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
3532         PR target/87467
3533         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
3534         __m512d type for __A argument rather than __m512.
3536 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
3538         * match.pd (simple_comparison): Don't optimize if either operand is
3539         a function pointer when target needs function pointer canonicalization.
3541 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3543         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
3544         power5 .. power9 to remove indirection.
3545         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
3546         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
3547         ASM_CPU_476_SPEC): Delete.
3548         (ASM_CPU_SPEC): Adjust.
3549         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
3550         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
3552 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3554         * config.in: Delete HAVE_AS_DCI.
3555         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
3556         * config/rs6000/rs6000.h: Ditto.
3557         * configure.ac: Delete HAVE_AS_DCI.
3558         * configure: Regenerate.
3560 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3562         * config.in (HAVE_AS_LWSYNC): Delete.
3563         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3564         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
3565         do it as a .long .
3566         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3567         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
3568         as a .long .
3569         * configure.ac: Delete HAVE_AS_LWSYNC.
3570         * configure: Regenerate.
3572 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3573             Pierre-Marie de Rodat  <derodat@adacore.com>
3575         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
3576         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
3577         (cgraph_node::create_thunk): Add indirect_offset parameter.
3578         (thunk_adjust): Likewise.
3579         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
3580         and initialize the corresponding field with it.
3581         (cgraph_node::dump): Dump indirect_offset field.
3582         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
3583         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
3584         (thunk_adjust): Add indirect_offset parameter and deal with it.
3585         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
3586         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
3587         or if the thunk is external or local.  Fix formatting.  Do not chain
3588         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
3589         if any, in the GIMPLE representation.
3590         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
3591         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
3592         (input_node): Read indirect_offset field.
3593         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
3594         call to thunk_adjust.
3595         * tree-nested.c (struct nesting_info): Add thunk_p field.
3596         (create_nesting_tree): Set it.
3597         (convert_all_function_calls): Copy static chain from targets to thunks.
3598         (finalize_nesting_tree_1): Return early for thunks.
3599         (unnest_nesting_tree_1): Do not finalize thunks.
3600         (gimplify_all_functions): Do not gimplify thunks.
3602 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3604         * opt-suggestions.c (option_proposer::build_option_suggestions):
3605         Release "option_values".
3607 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3609         * coverage.c (get_coverage_counts): Convert problem-reporting dump
3610         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
3611         * dumpfile.c (kind_as_string): New function.
3612         (dump_loc): Rather than a hardcoded prefix of "note: ", use
3613         kind_as_string to vary the prefix based on dump_kind.
3614         (selftest::test_capture_of_dump_calls): Update for above.
3616 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3618         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3619         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
3621 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3623         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3624         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3625         INVALID_REGNUM instead of FPSR_REG.
3626         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3627         * config/i386/i386.md: Update comment of FP compares.
3628         (fldenv): Do not clobber FPSR_REG.
3630 2018-09-28  Richard Biener  <rguenther@suse.de>
3632         * tree.h (BLOCK_ORIGIN): New.
3633         * omp-expand.c (grid_expand_target_grid_body): Assign
3634         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3635         * tree-inline.c (remap_block): Likewise.
3636         * auto-profile.c (get_function_decl_from_block): Simplify
3637         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3638         * langhooks.c (lhd_print_error_function): Likewise.
3639         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3640         Likewise.
3641         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3642         * tree.c (block_nonartificial_location): Likewise.
3643         (block_ultimate_origin): Likewise.
3644         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
3645         no longer needed LTO case.
3647 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
3648             Jan Hubicka  <jh@suse.cz>
3649             Martin Jambor  <mjambor@suse.cz>
3651         * simplify-rtx.c (simplify_merge_mask): New function.
3652         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3653         same masks are used in op1 or op2.
3654         (test_vec_merge): New function.
3655         (test_vector_ops): Call test_vec_merge.
3657 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3659         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3660         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3661         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3662         (cypress_costs): Set it.
3663         (supersparc_costs): Likewise.
3664         (hypersparc_costs): Likewise.
3665         (leon_cost): Likewise.
3666         (leon3_costs): Likewise.
3667         (sparclet_costs): Likewise.
3668         (ultrasparc_costs): Likewise.
3669         (ultrasparc_costs): Likewise.
3670         (niagara_costs): Likewise.
3671         (niagara2_costs): Likewise.
3672         (niagara3_costs): Likewise.
3673         (niagara4_costs): Likewise.
3674         (niagara7_costs): Likewise.
3675         (m8_costs): Likewise.
3676         (TARGET_CAN_FOLLOW_JUMP): Define.
3677         (pass_work_around_errata::gate): Minor tweak.
3678         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3679         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3680         Automaitcally clear MASK_FSMULD mask for V7 processors.
3681         (sparc_can_follow_jump): New static function.
3682         (output_ubranch): Deal with CROSSING_JUMP_P.
3683         (sparc_use_sched_lookahead): Rewrite using switch statement.
3684         (sparc_issue_rate): Reorder.
3685         (sparc_branch_cost): New function.
3687 2018-09-27  Martin Sebor  <msebor@redhat.com>
3689         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3690         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3691         (int_fits_type_p): Same.
3693 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
3695         * config/i386/i386.md (FPCR_REG): Remove.
3696         (UNSPEC_FLDCW): Remove.
3697         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3698         (x86_fldcw_1): Remove insn pattern.
3699         (fnstenv): Do not clobber FPCR_REG.
3700         (fldenv): Ditto.
3701         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3702         (CALL_USED_REGISTERS): Ditto.
3703         (REG_ALLOC_ORDER): Ditto.
3704         (REG_CLASS_CONTENTS): Ditto.
3705         (HI_REGISTER_NAMES): Ditto.
3706         (ADDITIONAL_REGISTER_NAMES): Use defines instead
3707         of numerical constants.
3708         * config/i386/i386.c (regclass_map): Remove fpsr register.
3709         (dbx_register_map): Ditto.
3710         (dbx64_register_map): Ditto.
3711         (svr4_dbx_register_map): Ditto.
3712         (print_reg): Do not handle FPCR_REG.
3714 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
3716         PR target/87149
3717         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3718         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3719         Delete, always treat as true.
3720         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3721         Ditto.  Simplify remaining code.
3722         * config/powerpcspe/powerpcspe.h: Ditto.
3723         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3724         Simplify remaining code.
3725         (rs6000_expand_builtin): Ditto.
3726         * config/rs6000/rs6000.h: Ditto.
3727         * configure.ac: Ditto.
3728         * configure: Regenerate.
3730 2018-09-27  Martin Liska  <mliska@suse.cz>
3732         * coverage.c (get_coverage_counts): Revert the formatting
3733         of missing profile opt info.
3735 2018-09-27  Richard Biener  <rguenther@suse.de>
3737         PR debug/37801
3738         PR debug/87440
3739         * dwarf2out.c (set_block_origin_self): Do not mark outermost
3740         block as we do not output that.
3741         (gen_inlined_subroutine_die): Elide the originally outermost
3742         block, matching what we do for concrete instances.
3743         (decls_for_scope): Add parameter specifying whether to recurse
3744         to subblocks.
3746 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
3747             Tom de Vries  <tom@codesourcery.com>
3749         PR 82089
3751         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3752         STORE_FLAG_VALUE == 1.
3754 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
3756         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3757         constant definitions.
3758         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3759         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3760         ("speculation_barrier"): New expander definition.
3762 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
3764         PR gcov-profile/86957
3765         * common.opt: New warning option -Wmissing-profile.
3766         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3767         * doc/invoke.texi: Document -Wmissing-profile.
3769 2018-09-26  Jim Wilson  <jimw@sifive.com>
3771         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3772         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3774 2018-09-26  Martin Sebor  <msebor@redhat.com>
3776         * tree.c (zerop): Change return type to bool.
3777         (integer_zerop, integer_onep, integer_each_onep): Same.
3778         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3779         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3780         (real_onep, real_minus_onep, chain_index): Same.
3781         (print_type_hash_statistics, type_list_equal): Same.
3782         * tree.h (zerop): Same.
3783         (zerop, 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.
3789 2018-09-26  Jim Wilson  <jimw@sifive.com>
3791         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3793 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
3795         PR target/87414
3796         * config/i386/i386.c: Include debug.h and dwarf2out.h.
3797         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3798         call.
3800 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
3802         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3804 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3806         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3807         and FP_SECOND_SSE_REGS.
3808         (REG_CLASS_NAMES): Ditto.
3809         (REG_CLASS_CONTENTS): Ditto.
3810         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3811         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3812         (ix86_preferred_output_reload_class): Ditto.
3813         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3814         clobber constraint to "=&f".
3815         (fix_truncdi_i387): Ditto.
3816         (lrintxfdi2): Ditto.
3817         (fistdi2_<rounding>): Ditto.
3818         (fpremxf4_i387): Change "=u" constraint to "=f".
3819         (fprem1xf4_i387): Ditto.
3820         (sincosxf3): Ditto.
3821         (fptanxf4_i387): Ditto.
3822         (fxtractxf3_i387): Ditto.
3823         (fscalexf4_i387): Ditto.
3824         (atan2xf3): Change "u" constraint to "f".
3825         (fyl2xxf3_i387): Ditto.
3826         (fyl2xp1xf3_i387): Ditto.
3828 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3830         PR target/87439
3831         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3832         for removed I387_MASK_PM entity.
3835 2018-09-26  Jeff Law  <law@redhat.com>
3836         Revert
3837         2018-09-26  Alexey Neyman  <stilor@att.net>
3839         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3840         headers are no longer pulled in by <isl/val.h>.
3842 2018-09-26  Richard Biener  <rguenther@suse.de>
3844         PR debug/87443
3845         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3846         or concrete instance DIE to the tree.  Create abstract origin
3847         attributes also for concrete instances.
3849 2018-09-26  Alexey Neyman  <stilor@att.net>
3851         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3852         headers are no longer pulled in by <isl/val.h>.
3854 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
3856         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3857         Use new helper functions.
3858         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3859         Use new helper functions.
3860         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3861         aarch_mm_needs_release): New declarations.
3862         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3863         aarch_mm_needs_release): New.
3865 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
3867         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3868         (arm32_output_mi_thunk): Deal with long calls.
3870 2018-09-26  Richard Biener  <rguenther@suse.de>
3872         PR debug/87428
3873         PR debug/87362
3874         * tree-inline.c (expand_call_inline): When the location
3875         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3876         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3877         the inserted BLOCK to make inlined_function_outer_scope_p
3878         recognize it.
3879         * dwarf2out.c (add_call_src_coords_attributes): Do not add
3880         coords for reserved locations.
3882 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3884         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3885         (*call_indirect_nonlocal_sysv<mode>): Ditto.
3886         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3887         (*sibcall_nonlocal_sysv<mode>): Ditto.
3888         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3889         (<bd>_<mode>): Ditto.
3890         (<bd>tf_<mode>): Ditto.
3892 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3894         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3895         control string as a list of templates instead of as C code.
3896         (*altivec_movti): Ditto.
3897         * config/rs6000/darwin.md (movdf_low_di): Ditto.
3899 2018-09-25  Jim Wilson  <jimw@sifive.com>
3901         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3902         when target symbol is weak.
3904 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3906         PR c/87387
3907         * builtins.c (unterminated_array): Simplify.
3908         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
3909         where pointer arithmetic is safe.
3911 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3913         PR target/86987
3914         * config/rs6000/altivec.md (altivec_vspltb): Use
3915         const_0_to_15_operand instead of u5bit_cint_operand.
3916         (*altivec_vspltb_internal): Ditto.
3917         (altivec_vspltb_direct): Ditto.
3918         (altivec_vsplth): Use const_0_to_7_operand instead of
3919         u5bit_cint_operand.
3920         (*altivec_vsplth_internal): Ditto.
3921         (altivec_vsplth_direct): Ditto.
3922         (altivec_vspltw): Use const_0_to_3_operand instead of
3923         u5bit_cint_operand.
3924         (*altivec_vspltw_internal): Ditto.
3925         (altivec_vspltw_direct): Ditto.
3926         (altivec_vspltsf): Ditto.
3927         (*altivec_vspltsf_internal): Ditto.
3928         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3929         various splats with the proper size immediate.  Reorder the various
3930         cases by ascending size of immediate, and put all such together.
3932 2018-09-25  Richard Biener  <rguenther@suse.de>
3934         PR debug/83941
3935         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3936         GC-ification.
3937         (maybe_create_die_with_external_ref): Do not create
3938         DW_TAG_imported_unit here.
3939         (add_abstract_origin_attribute): Handle external BLOCK refs.
3940         (dwarf2out_abstract_function): Simplify LTO case.
3941         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3942         rather than using maybe_create_die_with_external_ref.
3944 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
3946         PR target/71278
3947         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3948         (frndintxf2_mask_pm_i387): Ditto.
3949         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3950         Enable for !flag_trapping_math.
3951         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3952         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3953         Change operand 1 predicate to nonimmediate_operand.
3954         (attr "i387_cw"): Remove mask_pm.
3955         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3956         (enum ix86_entity): Remove I387_MASK_PM.
3957         * config/i386/i386.c (ix86_i387_mode_needed): Do not
3958         handle I387_MASK_PM.
3959         (ix86_mode_needed): Ditto.
3960         (ix86_mode_after): Ditto.
3961         (ix86_mode_entry): Ditto.
3962         (ix86_mode_exit): Ditto.
3963         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3965 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
3967         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3968         to_update_switch_stmts to vNULL instead of calling create on them
3969         immediately.
3971 2018-09-25  Richard Biener  <rguenther@suse.de>
3973         PR tree-optimization/87402
3974         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3975         (visit_phi): Re-instantiate handling of supposed to be VARYING
3976         but non-VARYING backedge value.
3978 2018-09-25  Richard Biener  <rguenther@suse.de>
3980         PR debug/83941
3981         * dwarf2out.c (struct sym_off_pair): New.
3982         (external_die_map): New global.
3983         (lookup_decl_die): When in LTO create DIEs lazily from the
3984         external_die_map.
3985         (lookup_block_die): New function, create DIEs lazily in LTO.
3986         (equate_block_to_die): New function.
3987         (dwarf2out_die_ref_for_decl): During WPA get the association
3988         from the external DIE map.
3989         (dwarf2out_register_external_die): Record mapping into the
3990         external DIE map.
3991         (maybe_create_die_with_external_ref): New function split out from
3992         DIE generation part of old dwarf2out_register_external_die.
3993         (add_abstract_origin_attribute): Do not return the DIE.  When
3994         in LTO reference externals directly.
3995         (dwarf2out_abstract_function): When in LTO ignore calls for
3996         decls with external DIEs (already present abstract instances).
3997         (gen_call_site_die): Adjust.
3998         (add_high_low_attributes): Likewise.
3999         (gen_lexical_block_die): Likewise.
4000         (gen_inlined_subroutine_die): Likewie.
4001         (gen_block_die): Likewise.
4002         (dwarf2out_inline_entry): Likewise.
4003         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
4004         DIEs.
4006 2018-09-25  Martin Liska  <mliska@suse.cz>
4008         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
4009         integers and not by a float value.
4011 2018-09-25  Martin Liska  <mliska@suse.cz>
4013         PR fortran/87394
4014         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
4015         instead of NULL.
4016         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
4017         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
4018         * godump.c (go_format_type): Remove extra parenthesis.
4020 2018-09-25  Martin Liska  <mliska@suse.cz>
4022         * alias.c (set_dest_equal_p): Remove unused function.
4023         * config/i386/i386.c (def_builtin_pure2): Likewise.
4024         * diagnostic-show-locus.c (class layout): Remove
4025         unused field.
4026         (layout::layout): Likewise here.
4027         * dump-context.h (class temp_dump_context): Likewise.
4028         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
4029         (add_AT_loclistsptr): Likewise.
4030         (add_AT_offset): Likewise.
4031         (get_AT_hi_pc): Likewise.
4032         (is_comdat_die): Likewise.
4033         (type_is_enum): Likewise.
4034         (ceiling): Likewise.
4035         (add_AT_vms_delta): Likewise.
4036         (is_class_die): Likewise.
4037         * edit-context.c (class line_event): Remove unused field.
4038         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
4039         unused function.
4040         * ipa-cp.c (ipa_get_vr_lat): Likewise.
4041         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
4042         (ok_for_base_p_nonstrict): Likewise.
4043         * tree-chrec.c (is_not_constant_evolution): Likewise.
4044         (chrec_fold_poly_cst): Likewise.
4045         * tree-if-conv.c (has_pred_critical_p): Likewise.
4046         * tree-ssa-coalesce.c (print_exprs): Likewise.
4047         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
4048         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
4049         * tree-vrp.c (value_ranges_intersect_p): Likewise.
4050         (value_range_nonnegative_p): Likewise.
4052 2018-09-25  Martin Liska  <mliska@suse.cz>
4054         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
4055         Do not handle "GNU Pascal".
4056         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
4057         Likewise.
4058         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
4059         from documentation. Likewise.
4060         * dbxout.c (dbxout_range_type): Likewise.
4061         * doc/cpp.texi: Likewise.
4062         * doc/extend.texi: Likewise.
4063         * doc/frontends.texi: Likewise.
4064         * doc/invoke.texi: Remove Pascal entry.
4065         * tree.def (CLEANUP_POINT_EXPR): Likewise.
4066         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
4068 2018-09-25  Martin Liska  <mliska@suse.cz>
4070         PR middle-end/86078
4071         * doc/invoke.texi: Document all parameters and remove default
4072         of the parameters.
4074 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
4076         PR bootstrap/87417
4077         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
4078         contains HOST_WIDE_INTs when computing its size.
4080 2018-09-24  Jim Wilson  <jimw@sifive.com>
4082         PR target/87391
4083         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
4084         not TARGET_RVE.
4085         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
4087 2018-09-24  Andrew Pinski  <apinski@marvell.com>
4089         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
4090         access prev before checking it for NULLness in the
4091         AARCH64_FUSE_CMP_BRANCH case.
4093 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
4095         PR target/82699
4096         * config/i386/i386.c (rest_of_insert_endbranch): Set
4097         endbr_queued_at_entrance to true and don't insert ENDBR if
4098         x86_function_profiler will be called.
4099         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
4100         is true.
4101         * config/i386/i386.h (machine_function): Add
4102         endbr_queued_at_entrance.
4104 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
4106         * genattrtab.c (mk_attr_alt): Use alternative_mask.
4107         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
4108         types.
4109         (check_attr_test): Use alternative_mask.
4110         (get_attr_value): Likewise.
4111         (compute_alternative_mask): Use alternative_mask and XWINT.
4112         (make_alternative_compare): Use alternative_mask.
4113         (attr_alt_subset_p): Use XWINT.
4114         (attr_alt_subset_of_compl_p): Likewise.
4115         (attr_alt_intersection): Use alternative_mask and XWINT.
4116         (attr_alt_union): Likewise.
4117         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
4118         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
4119         (simplify_test_exp): Use alternative_mask and XWINT.
4120         (write_test_expr): Use alternative_mask and XWINT, adjust bit
4121         number calculation to support 64 bits.  Generate code that
4122         checks 64-bit masks.
4123         (main): Use alternative_mask.
4124         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
4126 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
4128         PR target/80080
4129         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
4130         RETURN+USE when returning via %r14.
4132 2018-09-24  Martin Liska  <mliska@suse.cz>
4134         * gcov.c (output_lines): Print colorization legend
4135         for both flag_use_colors and flag_use_hotness_colors.
4136         Reword the help.
4138 2018-09-24  Martin Liska  <mliska@suse.cz>
4140         * coverage.c (get_coverage_counts): Use warning_at
4141         with current_function_decl location. Use %qD in warning
4142         message.
4144 2018-09-24  Martin Liska  <mliska@suse.cz>
4146         * memory-block.h (memory_block_pool::release): Annotate with
4147         valgrind that the memory is not accessible.
4149 2018-09-24  Martin Liska  <mliska@suse.cz>
4151         PR sanitizer/85774
4152         * asan.c: Make asan_handled_variables extern.
4153         * asan.h: Likewise.
4154         * cfgexpand.c (expand_stack_vars): Make sure
4155         a representative is unpoison if another
4156         variable in the partition is handled by
4157         use-after-scope sanitization.
4159 2018-09-24  Richard Biener  <rguenther@suse.de>
4161         PR tree-optimization/63155
4162         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
4163         the worklist when the edge of the respective argument isn't
4164         executable.
4166 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4168         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
4169         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
4170         (MASK_CLASS_P): Update for rename.
4171         (MAYBE_MASK_CLASS_P): Ditto.
4172         (REG_CLASS_NAMES): Update.
4173         (REG_CLASS_CONTENT): Update.
4174         * config/i386/i386.c (regclass_map): Update for MASK_REG
4175         and ALL_MASK_REGS rename.
4176         * config/i386/constraints.md (Yk): Update for rename.
4177         (k): Ditto.
4179 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4181         * config/i386/i386.h (enum reg_class): Remove
4182         EVEX_SSE_REGS and MOD4_SSE_REGS.
4183         (REG_CLASS_NAMES): Update.
4184         (REG_CLASS_CONTENT): Update.
4185         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
4186         registers as ALL_SSE_REGS.
4187         (ix86_additional_allocno_class_p): Remove.
4188         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
4189         (ix86_register_priority): Lower priority of EVEX SSE registers.
4190         Use IN_RANGE macro where appropriate.
4191         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
4192         AVX-5124VNNIW checks.
4193         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
4194         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
4195         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
4196         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
4197         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
4198         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
4199         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
4200         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
4201         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
4202         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
4203         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
4204         * config/i386/constraints.md (Yh): Remove.
4206 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
4208         * config/i386/i386.c (regclass_map): Declare integer REX registers
4209         as GENERAL_REGS.
4211 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
4213         * doc/service.texi (Service): Switch the fsf.org link to https.
4215 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
4217         PR target/86798
4218         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4219         Define to speculation_safe_value_not_needed.
4221 2018-09-21  Florian Weimer  <fweimer@redhat.com>
4223         PR middle-end/81035
4224         * doc/extend.texi (Common Function Attributes): Mention that
4225         noreturn suppresses tail call optimization.
4227 2018-09-21  Jeff Law  <law@redhat.com>
4229         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
4230         vr_values::cleanup_edges_and_switches.
4231         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
4232         vr_values class.
4233         (identify_jump_threads): Remove EDGE_IGNORE handling.
4234         (execute_vrp): Move handling of to_remove_edges and
4235         to_update_switch_stmts into vr_values class member functions.
4236         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
4237         (to_update_switch_stmts): Likewise.
4238         * vr-values.c: Include cfghooks.h.
4239         (vr_values::vr_values): Initialize to_remove_edges and
4240         to_update_switch_stmts.
4241         (vr_values::~vr_values): Verify to_remove_edges and
4242         to_update_switch_stmts are empty.
4243         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
4244         (vr_values::cleanup_edges_and_switches): New member function.
4245         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
4246         function.  Add new data members.
4248 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
4250         PR tree-optimization/87309
4251         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
4252         calls with pflags and alt_flags.
4253         (selftest::test_capture_of_dump_calls): Add test of interaction of
4254         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
4256 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4258         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
4260 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4262         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
4264 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4266         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
4267         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
4268         (VXWORKS_LIBS_RTP): Minor reordering.
4270 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4272         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
4273         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
4275 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4277         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
4278         (PTRDIFF_TYPE): Likewise.
4280 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4282         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
4283         triplet, similar to support for VxWorks7.
4284         * config/vxworks-dummy.h: Provide a default definition
4285         of TARGET_VXWORKS64 to 0.
4287 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4289         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
4290         * config/vxworks-dummy.h: here.
4292 2018-09-21  Olivier Hainque  <hainque@adacore.com>
4294         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
4296 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
4297             Bo Zhou  <zbo.zhou@hisilicon.com>
4299         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
4300         * config/aarch64/aarch64-tune.md: Regenerated.
4301         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
4302         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
4303         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
4305 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
4306             Julian Brown  <julian@codesourcery.com>
4308         * builtins.c (get_builtin_sync_mem): Handle address spaces.
4310 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4312         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
4313         if the call takes a static chain.
4315 2018-09-21  Martin Liska  <mliska@suse.cz>
4317         * auto-profile.c (autofdo_source_profile::read): Do not
4318         set sum_all.
4319         (read_profile): Do not add working sets.
4320         (read_autofdo_file): Remove sum_all.
4321         (afdo_callsite_hot_enough_for_early_inline): Remove const
4322         qualifier.
4323         * coverage.c (struct counts_entry): Remove gcov_summary.
4324         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
4325         do not support GCOV_TAG_PROGRAM_SUMMARY.
4326         (get_coverage_counts): Remove summary and expected
4327         arguments.
4328         * coverage.h (get_coverage_counts): Likewise.
4329         * doc/gcov-dump.texi: Remove -w option.
4330         * gcov-dump.c (dump_working_sets): Remove.
4331         (main): Do not support '-w' option.
4332         (print_usage): Likewise.
4333         (tag_summary): Likewise.
4334         * gcov-io.c (gcov_write_summary): Do not dump
4335         histogram.
4336         (gcov_read_summary): Likewise.
4337         (gcov_histo_index): Remove.
4338         (gcov_histogram_merge): Likewise.
4339         (compute_working_sets): Likewise.
4340         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
4341         it not obsolete.
4342         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
4343         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
4344         (GCOV_HISTOGRAM_SIZE): Remove.
4345         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
4346         (struct gcov_summary): Simplify rapidly just
4347         to runs and sum_max fields.
4348         (gcov_histo_index): Remove.
4349         (NUM_GCOV_WORKING_SETS): Likewise.
4350         (compute_working_sets): Likewise.
4351         * gcov-tool.c (print_overlap_usage_message): Remove
4352         trailing empty line.
4353         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
4354         (output_lines): Remove program related line.
4355         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
4356         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
4357         histogram.
4358         (input_profile_summary): Do not read it.
4359         (merge_profile_summaries): And do not merge it.
4360         (input_symtab): Do not call removed function.
4361         * modulo-sched.c (sms_schedule): Do not print sum_max.
4362         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
4363         removed when histogram method was invented.
4364         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
4365         mode.
4366         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
4367         GCOV coding style.
4368         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
4369         and dump selected value.
4370         * profile.c (add_working_set): Remove.
4371         (get_working_sets): Likewise.
4372         (find_working_set): Likewise.
4373         (get_exec_counts): Do not work with working sets.
4374         (read_profile_edge_counts): Do not inform as sum_max is removed.
4375         (compute_branch_probabilities): Likewise.
4376         (compute_value_histograms): Remove argument for call of
4377         get_coverage_counts.
4378         * profile.h: Do not make gcov_summary const.
4380 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
4382         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
4384 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
4386         PR tree-optimization/86990
4387         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
4388         Check that the entire merged store group is made of constants only for
4389         overlapping stores.
4391 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
4393         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
4394         (VTABLE_VERIFICATION_SPEC): Likewise.
4395         (SANITIZER_EARLY_SPEC): Likewise.
4396         (SANITIZER_SPEC): Likewise.
4397         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
4398         * doc/invoke.texi (Link Options): Document -r.
4400 2018-09-20  Richard Biener <rguenther@suse.de>
4402         PR middle-end/87054
4403         * gimplify.c (gimplify_expr): Retain alignment of
4404         addressable lvalue in dereference.
4406 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
4408         PR bootstrap/87013
4409         * configure.ac: Check for .loc is_stmt support.
4410         * configure, config.in: Rebuilt.
4411         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
4412         if not supported.
4414 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4416         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
4417         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
4418         -misel=no.
4420 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
4422         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4423         VECTOR_OTHER.
4424         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
4425         case VECTOR_OTHER.
4427 2018-09-20  Marek Polacek  <polacek@redhat.com>
4429         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
4431 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4433         PR tree-optimization/87288
4434         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
4435         into account when determining PEELING_FOR_NITERS.
4437 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
4439         PR tree-optimization/86877
4440         * tree-vect-loop.c (vect_analyze_loop_2): Call
4441         vect_verify_datarefs_alignment.
4443 2018-09-19  Marek Polacek  <polacek@redhat.com>
4445         * doc/invoke.texi: Document -Wclass-conversion.
4447 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
4449         * config/pa/pa.c (pa_adjust_priority): Delete.
4450         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
4452         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
4453         (atomic_storehi): Likewise.
4454         (atomic_storesi): Likewise.
4455         (atomic_loaddi): Restore compare and swap exchange loop code.
4457 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
4459         PR rtl-optimization/86902
4460         * combine.c (try_combine): When changing the CC mode used, don't change
4461         an unrelated mode in other_insn to that new CC mode.
4463 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
4465         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
4466         with %T in place of calls to dump_generic_expr.
4467         (prune_runtime_alias_test_list): Likewise.
4468         (create_runtime_alias_checks): Likewise.
4469         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
4470         (vect_analyze_data_ref_dependence): Likewise.
4471         (vect_slp_analyze_data_ref_dependence): Likewise.
4472         (vect_record_base_alignment): Likewise.  Use %G in place of call
4473         to dump_gimple_stmt.
4474         (vect_compute_data_ref_alignment): Likewise.
4475         (verify_data_ref_alignment): Likewise.
4476         (vect_find_same_alignment_drs): Likewise.
4477         (vect_analyze_group_access_1): Likewise.
4478         (vect_analyze_data_ref_accesses): Likewise.
4479         (dependence_distance_ge_vf): Likewise.
4480         (dump_lower_bound): Likewise.
4481         (vect_prune_runtime_alias_test_list): Likewise.
4482         (vect_find_stmt_data_reference): Likewise.
4483         (vect_analyze_data_refs): Likewise.
4484         (vect_create_addr_base_for_vector_ref): Likewise.
4485         (vect_create_data_ref_ptr): Likewise.
4486         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
4487         (vect_can_advance_ivs_p): Likewise.
4488         (vect_update_ivs_after_vectorizer): Likewise.
4489         (vect_gen_prolog_loop_niters): Likewise.
4490         (vect_prepare_for_masked_peels): Likewise.
4491         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4492         (vect_determine_vectorization_factor): Likewise.
4493         (vect_is_simple_iv_evolution): Likewise.
4494         (vect_analyze_scalar_cycles_1): Likewise.
4495         (vect_analyze_loop_operations): Likewise.
4496         (report_vect_op): Likewise.
4497         (vect_is_slp_reduction): Likewise.
4498         (check_reduction_path): Likewise.
4499         (vect_is_simple_reduction): Likewise.
4500         (vect_create_epilog_for_reduction): Likewise.
4501         (vect_finalize_reduction:): Likewise.
4502         (vectorizable_induction): Likewise.
4503         (vect_transform_loop_stmt): Likewise.
4504         (vect_transform_loop): Likewise.
4505         (optimize_mask_stores): Likewise.
4506         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
4507         (vect_split_statement): Likewise.
4508         (vect_recog_over_widening_pattern): Likewise.
4509         (vect_recog_average_pattern): Likewise.
4510         (vect_determine_min_output_precision_1): Likewise.
4511         (vect_determine_precisions_from_range): Likewise.
4512         (vect_determine_precisions_from_users): Likewise.
4513         (vect_mark_pattern_stmts): Likewise.
4514         (vect_pattern_recog_1): Likewise.
4515         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4516         (vect_record_max_nunits): Likewise.
4517         (vect_build_slp_tree_1): Likewise.
4518         (vect_build_slp_tree_2): Likewise.
4519         (vect_print_slp_tree): Likewise.
4520         (vect_analyze_slp_instance): Likewise.
4521         (vect_detect_hybrid_slp_stmts): Likewise.
4522         (vect_detect_hybrid_slp_1): Likewise.
4523         (vect_slp_analyze_operations): Likewise.
4524         (vect_slp_analyze_bb_1): Likewise.
4525         (vect_transform_slp_perm_load): Likewise.
4526         (vect_schedule_slp_instance): Likewise.
4527         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4528         (vect_mark_stmts_to_be_vectorized): Likewise.
4529         (vect_init_vector_1): Likewise.
4530         (vect_get_vec_def_for_operand): Likewise.
4531         (vect_finish_stmt_generation_1): Likewise.
4532         (vect_check_load_store_mask): Likewise.
4533         (vectorizable_call): Likewise.
4534         (vectorizable_conversion): Likewise.
4535         (vectorizable_operation): Likewise.
4536         (vectorizable_load): Likewise.
4537         (vect_analyze_stmt): Likewise.
4538         (vect_is_simple_use): Likewise.
4539         (vect_get_vector_types_for_stmt): Likewise.
4540         (vect_get_mask_type_for_stmt): Likewise.
4541         * tree-vectorizer.c (increase_alignment): Likewise.
4543 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
4545         * doc/rtl.texi: Adjust vec_select description.
4546         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
4547         non-constant selectors.
4549 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
4551         * config/aarch64/aarch64-protos.h
4552         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
4553         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
4554         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
4555         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
4556         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
4557         (AARCH64_FL_PROFILE): Move index so flags are ordered.
4558         (AARCH64_ISA_RCPC8_4): New flag.
4559         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
4560         to aarch64_offset_9bit_signed_unscaled_p.
4561         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
4562         and use stlur.
4563         * config/aarch64/constraints.md (Ust): New constraint.
4564         * config/aarch64/predicates.md.
4565         (aarch64_9bit_offset_memory_operand): New predicate.
4566         (aarch64_rcpc_memory_operand): New predicate.
4568 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
4570         PR rtl-optimization/87361
4571         * rtlanal.c (nonzero_bits1): Revert accidental change.
4573 2018-09-19  Richard Biener  <rguenther@suse.de>
4575         PR tree-optimization/87349
4576         PR tree-optimization/87342
4577         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
4579 2018-09-18  Marek Polacek  <polacek@redhat.com>
4581         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
4582         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
4584 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4586         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
4588 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4590         PR rtl-optimization/86882
4591         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
4593 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
4595         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
4596         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
4598 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
4600         PR other/87353
4601         * doc/invoke.texi (Link Options): Fix formatting and grammar.
4603 2018-09-18  Richard Biener  <rguenther@suse.de>
4605         PR middle-end/63155
4606         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
4607         (compute_samebase_partition_bases): Likewise.
4608         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
4609         (gimple_can_coalesce_p): Simplify.
4611 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
4613         Handle a library implementation of ffs calling __builtin_ffs.
4614         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
4615         (mmix_init_libfuncs): New function: make __builtin_ffs expand
4616         to __ffsdi2.
4618 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4620         * diagnostic-show-locus.c (class layout_range): Add field
4621         "m_original_idx".
4622         (layout_range::layout_range): Add "original_idx" param and use it
4623         to initialize new field.
4624         (make_range): Use 0 for original_idx.
4625         (layout::layout): Pass in index to calls to
4626         maybe_add_location_range.
4627         (layout::maybe_add_location_range): Add param "original_idx" and
4628         pass it on to layout_range.
4629         (layout::print_any_labels): Pass on range->m_original_idx to
4630         get_text call.
4631         (gcc_rich_location::add_location_if_nearby): Use 0 for
4632         original_idx.
4633         * gcc-rich-location.h (text_range_label::get_text): Update for new
4634         param.
4635         (range_label_for_type_mismatch::get_text): Likewise.
4637 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4639         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4641 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4643         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4644         format_string_diagnostic_t.
4645         (fmtwarn_n): Likewise.
4646         * substring-locations.c
4647         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4648         (format_warning_n_va): Convert to...
4649         (format_string_diagnostic_t::emit_warning_n_va): ...this.
4650         (format_warning_va): Convert to...
4651         (format_string_diagnostic_t::emit_warning_va): ...this.
4652         (format_warning_at_substring): Convert to...
4653         (format_string_diagnostic_t::emit_warning): ...this.
4654         (format_warning_at_substring_n): Convert to...
4655         (format_string_diagnostic_t::emit_warning_n): ...this.
4656         * substring-locations.h (class format_string_diagnostic_t): New
4657         class.
4658         (format_warning_va): Convert to
4659         format_string_diagnostic_t::emit_warning_va.
4660         (format_warning_n_va): Convert to
4661         format_string_diagnostic_t::emit_warning_n_va.
4662         (format_warning_at_substring): Convert to
4663         format_string_diagnostic_t::emit_warning.
4664         (format_warning_at_substring_n): Convert to
4665         format_string_diagnostic_t::emit_warning_n.
4667 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
4668             Bernd Schmidt <bernds_cb1@t-online.de>
4670         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4671         SImode args.
4673 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4675         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4676         operand 0 predicate to nonimmediate operand.
4677         (rint<mode>2_frndint): Remove insn pattern.
4678         (rint<mode>2): Change operand 1 predicate to general_operand.
4679         Extend operand 1 to XFmode and generate rintxf2 insn.
4680         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4681         Do not use X87MODEF mode macro.
4682         (frndintxf2_<rounding>_i387): Rename from
4683         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
4684         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4685         to XFmode and generate significandxf3 insn.
4687 2018-09-17  Richard Biener  <rguenther@suse.de>
4689         PR tree-optimization/87328
4690         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4691         visiting unexecutable backedges when not iterating.
4692         (do_rpo_vn): Mark all edges not executable even when not
4693         iterating.
4695 2018-09-17  Martin Jambor  <mjambor@suse.cz>
4697         PR c/63886
4698         * doc/invoke.texi (Warning Options): Likewise.
4700 2018-09-17  Richard Biener  <rguenther@suse.de>
4702         PR tree-optimization/87301
4703         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4704         clean EH info from leftover copy assignments.
4706 2018-09-17  Martin Liska  <mliska@suse.cz>
4708         PR gcov-profile/85871
4709         * gcov.c (output_intermediate_file): Fix out of bounds
4710         access.
4712 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
4714         * config/arc/arc.c: Object attributes for core4 not reflected
4715         correctly.
4716         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4717         core3).
4719 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
4721         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4723 2018-09-17  Martin Liska  <mliska@suse.cz>
4725         * doc/gcov.texi: Document new option --use-hotness-colors.
4726         * gcov.c (struct source_info): Declare new field.
4727         (source_info::source_info): Set default for maximum_count.
4728         (print_usage): Add new -q option.
4729         (process_args): Process it.
4730         (accumulate_line_info): Save src->maximum_count.
4731         (output_line_beginning): Make color line number if
4732         flag_use_hotness_colors is set.
4733         (output_line_details): Pass default argument value.
4734         (output_lines): Pass src->maximum_count.
4736 2018-09-17  Martin Liska  <mliska@suse.cz>
4738         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4739         Use processor_names table.
4740         * config/i386/i386.c (ix86_default_align): Use
4741         processor_cost_table for alignment values.
4742         (ix86_option_override_internal): Use processor_names.
4743         (ix86_function_specific_print): Likewise.
4744         * config/i386/i386.h (struct processor_costs):
4745         Add alignment values.
4746         (struct ptt): Remove and replace with const char *.
4747         * config/i386/x86-tune-costs.h (struct processor_costs):
4748         Declare default alignments for all costs.
4750 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
4752         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4753         symbolics or VR_VARYING ranges for ABS_EXPR.
4754         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4755         when range will wrap.
4757 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
4759         PR middle-end/86864
4760         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4761         before and after a JUMP_TABLE_DATA.
4763 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
4765         PR middle-end/87188
4766         * dojump.c (do_compare_and_jump): Canonicalize function pointers
4767         when one operand is a function pointer.  Use POINTER_TYPE_P and
4768         FUNC_OR_METHOD_TYPE_P.
4769         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4770         * fold-const.c (build_range_check): Likewise.
4771         * match.pd (simple_comparison): Likewise.
4773 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
4775         PR c/82967
4776         * spellcheck.c (get_edit_distance_cutoff): New function.
4777         (selftest::test_edit_distance_unit_test_oneway): Rename to...
4778         (selftest::test_get_edit_distance_one_way): ...this.
4779         (selftest::test_get_edit_distance_unit): Rename to...
4780         (selftest::test_get_edit_distance_both_ways): ...this.
4781         (selftest::test_edit_distances): Move tests to this new function,
4782         and test some more pairs of strings.  Update for above renaming.
4783         (selftest::get_old_cutoff): New function.
4784         (selftest::test_get_edit_distance_cutoff): New function.
4785         (selftest::assert_suggested_for): New function.
4786         (ASSERT_SUGGESTED_FOR): New macro.
4787         (selftest::assert_not_suggested_for): New function.
4788         (ASSERT_NOT_SUGGESTED_FOR): New macro.
4789         (selftest::test_suggestions): New function.
4790         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4791         tests to selftest::test_edit_distances and call it.  Add calls to
4792         selftest::test_get_edit_distance_cutoff and
4793         selftest::test_suggestions.
4794         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4795         (best_match::consider): Replace hard-coded cutoff calculation with
4796         a call to...
4797         (best_match::get_cutoff): New declaration.
4798         (best_match::get_best_meaningful_candidate): Likewise.
4800 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4802         * builtins.c (fold_builtin_strlen): Remove TODO comment.
4804 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4806         revert:
4807         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4809         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4810         terminated string literal.
4812 2018-09-14  Martin Sebor  <msebor@redhat.com>
4814         * builtins.c (unterminated_array): Handle ARRAY_REF.
4815         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4816         * builtins.h (unterminated_array): Declare extern.
4817         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4818         arrays.
4819         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4820         calls.
4822 2018-09-14  Martin Sebor  <msebor@redhat.com>
4823             Jeff Law  <law@redhat.com>
4825         * builtins.c (unterminated_array): New.
4826         (expand_builtin_strcpy): Adjust.
4827         (expand_builtin_strcpy_args): Detect unterminated arrays.
4828         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
4829         unterminated arrays.
4830         * gimple-fold.h (get_maxval_strlen): Add argument.
4831         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4833         * gimple-fold.c (get_range_strlen): Add argument.
4834         (get_maxval_strlen): Adjust.
4835         * gimple-fold.h (get_range_strlen): Add argument.
4837 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
4839         * config/i386/movdirintrin.h: Fix copyright year.
4841 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
4843         * reg-stack.c: Include regs.h.
4844         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4845         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4846         FIRST_STACK_REG, not DFmode.
4847         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4848         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
4849         (change stack): Default register mode to the reg_raw_mode of
4850         FIRST_STACK_REG, not DFmode.
4851         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4852         (*swapxf): Rename from swapxf.
4854 2018-09-14  Carl Love  <cel@us.ibm.com>
4856         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4857         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4859 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
4861         PR target/87224
4862         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4863         alternatives.
4865 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
4867         PR target/85628
4868         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4870 2018-09-14  Jason Merrill  <jason@redhat.com>
4872         Fix --enable-gather-detailed-mem-stats.
4873         * hash-table.c (hash_table_usage): Change from variable to function.
4874         * hash-table.h: Adjust.
4875         * Makefile.in: Add missing dependencies on hash-table.h.
4877 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4879         PR tree-optimization/87259
4880         PR lto/87283
4881         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4882         execute_cse_reciprocals_1 has tried transforming.
4884 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
4886         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4887         VR_VARYING for PLUS/MINUS_EXPR.
4889 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4891         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4892         formatting.
4894 2018-09-14  Richard Biener  <rguenther@suse.de>
4896         PR middle-end/63155
4897         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4898         bits for the merged partition.
4900 2018-09-13  Martin Sebor  <msebor@redhat.com>
4901             Bernd Edlinger  <bernd.edlinger@hotmail.de>
4903         * builtins.h (c_srlen): Add argument.
4904         * builtins.c (warn_string_no_nul): New function.
4905         (c_strlen): Add argument and use it.  Update recursive calls.
4906         Pass DECL argument to string_constant to get info on non
4907         terminated strings.  Update *NONSTR as needed.
4908         (fold_builtin_strlen): Add argument to calls to c_strlen.
4909         Warn for unterminated arrays.
4910         (warn_string_no_null): Add prototype.
4911         * expr.c (string_constant): Update arguments.  Update recursive
4912         calls appropriately.  Detect missing NUL terminator and outermost
4913         declaration its missing in.
4914         Improve checks for arrays with nonzero lower bound or elements
4915         that are not a single byte.  Simplify offset computation.
4916         Simplify checks for non-NUL terminated strings.
4917         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4918         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4920 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4922         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4923         correctly.
4924         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
4925         argument.  Fix range checks.
4926         * fold-const.h (c_getstr): Adjust protoype.
4927         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4928         string is constant but contains no NUL byte.
4930         * expr.c (string_constant): Adjust function comment.
4931         Remove bogus check for zero termination.
4933         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4935         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4936         (get_constant_size): Don't make STRING_CSTs larger than they are.
4937         (check_string_literal): New check function for STRING_CSTs.
4938         (output_constant): Use it.
4940 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
4942         PR target/86812
4943         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4945 2018-09-13  Richard Biener  <rguenther@suse.de>
4947         PR tree-optimization/87263
4948         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4949         (struct unwind_state): Add max_rpo field.
4950         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4951         Compute max_rpo, the max RPO number a block can be backwards reached
4952         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
4953         separating it from the iterating mode.
4955 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4957         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4958         (rfs_decision): New scheduling decision.
4960 2018-09-13  Richard Biener  <rguenther@suse.de>
4962         PR bootstrap/87134
4963         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4964         (vn_nary_op_insert_pieces_predicated): Do not write useless
4965         valid_dominated_by_p entry outside of the allocated storage.
4967 2018-09-13  Omar Sandoval  <osandov@osandov.com>
4968             Tom de Vries  <tdevries@suse.de>
4970         PR debug/86985
4971         * dwarf2out.c (is_c): New function.
4972         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4974 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
4976         PR target/85628
4977         * config/aarch64/aarch64.md (*aarch64_bfxil):
4978         Define.
4979         * config/aarch64/constraints.md (Ulc): Define.
4980         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4981         Define.
4982         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4983         New function.
4985 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4987         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4988         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4989         aarch64_layout_frame call.
4990         (aarch64_expand_epilogue): Likewise.
4991         (aarch64_initial_elimination_offset): Likewise.
4992         (aarch64_get_separate_components): Likewise.
4993         (aarch64_use_return_insn_p): Likewise.
4994         (aarch64_layout_frame): Remove unneeded check.
4996 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4998         * configure.ac: Only append
4999         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
5000         gcc_config_arguments if it was never reconfigured or last reconfigure
5001         was with different arguments.
5002         * configure: Regenerated.
5004 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5005             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5007         PR middle-end/87290
5008         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
5009         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
5011 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5013         PR tree-optimization/87287
5014         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
5015         X % C == 0 to X % (unsigned) C == 0 optimization to ...
5016         * match.pd (X % C == 0): ... here.  New optimization.
5018 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5020         PR middle-end/82853
5021         * expr.h (maybe_optimize_mod_cmp): Declare.
5022         * expr.c (mod_inv): New function.
5023         (maybe_optimize_mod_cmp): New function.
5024         (do_store_flag): Use it.
5025         * cfgexpand.c (expand_gimple_cond): Likewise.
5027 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
5028             Julian Brown  <julian@codesourcery.com>
5030         PR middle-end/86336
5031         * gimplify.c (gimplify_scan_omp_clauses): Set
5032         target_firstprivatize_array_bases in OpenACC parallel and kernels
5033         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
5034         OpenACC data regions.
5036 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
5038         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
5039         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
5041 2018-09-12  Richard Biener  <rguenther@suse.de>
5043         PR tree-optimization/87280
5044         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
5045         edge but unreachable target.
5046         (do_rpo_vn): For conservatively handling a PHI only mark
5047         the backedge executable but not the block reachable.
5049 2018-09-12  Richard Biener  <rguenther@suse.de>
5051         PR tree-optimization/87266
5052         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
5053         visited blocks.
5055 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
5057         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
5058         constants.
5059         ("trunc<BFP:mode><DFP_ALL:mode>2")
5060         ("trunc<DFP_ALL:mode><BFP:mode>2")
5061         ("extend<BFP:mode><DFP_ALL:mode>2")
5062         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
5063         according to the target operand type.
5065 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5066             Andreas Krebbel  <krebbel@linux.ibm.com>
5068         PR tree-optimization/86844
5069         * gimple-ssa-store-merging.c
5070         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
5071         there are any overlapping stores in between them, make sure they are
5072         also coalesced or we give up completely.
5074 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5076         PR middle-end/87248
5077         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
5078         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
5080 2018-09-12  Tom de Vries  <tdevries@suse.de>
5082         * common.opt (gdescribe-dies): Add option.
5083         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
5084         attribute for artifical and nameless decls.
5085         (dwarf2out_register_external_die): Add description attribute to
5086         external reference die.
5087         (add_desc_attribute): New functions.
5088         (gen_subprogram_die): Add description attribute to
5089         DW_TAG_call_site_parameter.
5090         * tree-pretty-print.c (print_generic_expr_to_str): New function.
5091         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
5092         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
5093         -gno-describe-dies.
5094         (@item -gdescribe-dies): Add.
5096 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
5098         * tree-vrp.c (vrp_shift_undefined_p): Remove.
5099         (extract_range_from_binary_expr_1: Call
5100         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
5101         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
5102         depend on sign.
5104 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
5106         * gimple-ssa-warn-alloca.c
5107         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
5108         field for ALLOCA_BOUND_*_LARGE.
5110 2018-09-11  Nathan Sidwell  <nathan@acm.org>
5112         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
5114 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
5116         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
5117         for clobbers.  Remove obsolete comment.
5119 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
5121         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
5122         mpxchk, mpxld and mpxst types.
5123         (define_attr length_immediate): Remove all processing of mpx types.
5124         (define_attr prefix_0f): Ditto.
5125         (define_attr memory): Ditto.
5127 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
5129         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
5130         (log<mode>2): Change operand 1 predicate to general_operand.
5131         Extend operand 1 to XFmode and generate logxf3 insn.
5132         (log10<mode>2): Change operand 1 predicate to general_operand.
5133         Extend operand 1 to XFmode and generate log10xf3 insn.
5134         (log2<mode>2): Change operand 1 predicate to general_operand.
5135         Extend operand 1 to XFmode and generate log2xf3 insn.
5136         (fyl2xp1_extend<mode>xf3_i387): Remove.
5137         (log1p<mode>2): Change operand 1 predicate to general_operand.
5138         Extend operand 1 to XFmode and generate log1pxf3 insn.
5139         (fxtract_extend<mode>xf3_i387): Remove.
5140         (logb<mode>2): Change operand 1 predicate to general_operand.
5141         Extend operand 1 to XFmode and generate logbxf3 insn.
5142         (ilogb<mode>2): Change operand 1 predicate to general_operand.
5143         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
5144         (significand<mode>2): Change operand 1 predicate to general_operand.
5145         Extend operand 1 to XFmode and generate significandxf3 insn.
5147 2018-09-11  Nathan Sidwell  <nathan@acm.org>
5149         * gcc.c (perror_with_name, pfatal_with_name): Delete.
5150         (load_specs): Use fatal_error.
5151         (DELETE_IF_ORDINARY, process_command): Use error.
5152         (execute, run_attempt): Use fatal_error.
5154 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
5156         * diagnostic-core.h (sorry_at): New prototype.
5157         * diagnostic.c (sorry_at): New function.
5159 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
5161         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
5162         by zero as VR_UNDEFINED.
5164 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
5166         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
5167         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
5168         (<sincos>mode2): New expander.
5169         (sincos_extend<mode>xf3_i387): Remove insn pattern.
5170         (sincos -> sin, cos splitters): Remove splitter patterns.
5171         (sincos<mode>3): Change operand 2 predicate to general_operand.
5172         Extend operand 2 to XFmode and generate sincosxf3 insn.
5173         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
5174         Change operand 3 predicate to const1_operand.
5175         (fptan_extend<mode>xf4_i387): Remove insn pattern.
5176         (tanxf2): Update operands in the call to fptanxf4_i387.
5177         (tan<mode>2): Change operand 1 predicate to general_operand.
5178         Extend operand 1 to XFmode and generate tanxf3 insn.
5179         (atan2xf3): Rename from *fpatanxf3_i387.
5180         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
5181         (atan2xf3): Remove expander.
5182         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
5183         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
5184         (atan<mode>2): Change operand 1 predicate to general_operand.
5185         Extend operand 1 to XFmode and generate atanxf3 insn.
5187 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
5189         * config/i386/i386.md (x87/SSE constant load splitter): Use
5190         memory_operand instead of nonimmediate_operand for input operand
5191         predicate.
5193 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
5195         * config/i386/i386.md (float partial SSE register stall splitter): Move
5196         splitter near its instruction pattern.
5197         (float_extend partial SSE register stall splitter): Ditto.
5198         (float_truncate partial SSE register stall splitter): Ditto.
5200 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
5202         PR target/86794
5203         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5204         to speculation_safe_value_not_needed.
5206         PR target/85666
5207         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
5208         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
5209         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
5210         leaf_function_p, instead use has_hard_reg_initial_val.
5212 2018-09-09  Nathan Sidwell  <nathan@acm.org>
5214         * gcc.h (pfatal_with_name): Don't declare here.
5215         * gcc.c (pfatal_with_name): Make static.
5217 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
5219         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
5220         earlyclobber.
5222 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
5224         PR rtl-optimization/85458
5225         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
5226         priority hook to reduce the priority of EXPR.
5228 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
5230         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
5231         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
5232         with X87_ENABLE_FLOAT.  Remove preparation code.
5233         (*float<SWI48:mode><MODEF:mode>2): Rename from
5234         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
5235         math using "enabled" attribute.
5236         (*floatdi<MODEF:mode>2_i387): Rename from
5237         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
5238         enable for 32bit targets only.
5239         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
5240         splitter.
5241         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
5242         as operand 1 predicate.  Rewrite as define_insn_and_split.
5243         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
5245 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5247         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
5248         to fallthru to FLOAT case.
5250 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5252         PR target/86731
5253         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
5254         around folding of vec_sl to handle out of range shift values.
5256 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
5258         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
5259         Update callers to gen_fix_trunc<mode>_i387_fisttp
5260         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
5261         nonimmediate_operand.
5262         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
5263         and corresponding splitters.
5264         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
5265         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
5266         (fix_truncdi_i387_with_temp): Remove insn pattern
5267         and corresponding splitters.
5268         (fix_trunc<mode>_i387): Change operand 0 predicate to
5269         nonimmediate_operand.
5270         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
5271         and corresponding splitters.
5272         (*fistdi2_1): Remove.
5273         (fistdi2): Ditto.
5274         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
5275         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
5276         (*fist<mode>2_1): Remove.
5277         (fist<mode>2): Ditto.
5278         (fist<mode>2_with_temp): Remove insn pattern and corresponding
5279         splitters.
5280         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
5281         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
5282         (fistdi2_<rounding>): Change operand 0 predicate to
5283         nonimmediate_operand.
5284         (fistdi2_<rounding>_with_temp): Remove insn pattern
5285         and corresponding splitters.
5286         (fist<mode>2_<rounding>): Change operand 0 predicate to
5287         nonimmediate_operand.
5288         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
5289         and corresponding splitters.
5291         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
5293 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5295         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
5296         the init value.
5298 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5300         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
5301         early gimple folding of vec_splat().
5302         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
5303         * gimple-fold.h: Add an extern define for tree_vec_extract().
5305 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
5307         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
5308         wrappers around TREE_TYPE comparisons.
5310 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5312         PR target/80080
5313         * config/s390/predicates.md: Add nonsym_memory_operand.
5314         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
5315         contains a SYMBOL_REF, load it into an intermediate pseudo.
5316         (s390_emit_compare_and_swap): Legitimize operand.
5317         * config/s390/s390.md: Use the new nonsym_memory_operand
5318         with UNSPECV_CAS patterns.
5320 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5322         PR target/80080
5323         * config/s390/s390-passes.def: New file.
5324         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
5325         declaration.
5326         (make_pass_s390_early_mach): Add declaration.
5327         * config/s390/s390.c (make_pass_s390_early_mach):
5328         (s390_option_override): Remove dynamic registration.
5329         * config/s390/t-s390: Add s390-passes.def.
5331 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
5333         * config/s390/s390.c (s390_decompose_constant_pool_ref):
5334         Remove UNSPEC_LTREL_BASE check.
5335         (annotate_constant_pool_refs): Likewise.
5336         (find_constant_pool_ref): Likewise.
5337         (find_ltrel_base): Removed.
5338         (replace_ltrel_base): Removed.
5339         (s390_mainpool_finish): Remove replace_ltrel_base call.
5340         (s390_chunkify_start): Remove pending LTREL_BASE logic.
5341         (s390_chunkify_finish): Remove replace_ltrel_base call.
5342         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
5344 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
5346         PR target/86779
5347         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5348         to speculation_safe_value_not_needed.
5350 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
5351             Bernd Schmidt  <bernds_cb1@t-online.de>
5353         * config/nvptx/nvptx-opts.h: New file.
5354         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
5355         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
5356         (ASM_SPEC): Define.
5357         (TARGET_SM35): New macro.
5358         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
5359         correct predicate.
5360         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
5361         values.
5362         (misa=): New option.
5363         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
5365 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
5367         * config/i386/i386.md (truncdfsf2): Remove expander.
5368         (truncdfsf2_with_temp): Ditto.
5369         (truncxf<mode>2): Ditto.
5370         (*truncdfsf_fast_mixed): Remove insn pattern.
5371         (*truncdfsf_fast_i387): Ditto.
5372         (*truncdfsf_mixed): Ditto.
5373         (*truncdfsf_i387): Ditto.
5374         (*truncdfsf2_i387_1): Ditto.
5375         (*truncxfsf2_mixed): Ditto.
5376         (*truncxfdf2_mixed): Ditto.
5377         (*truncxf<mode>2_i387_noop): Ditto. Update callers
5378         to call gen_truncxf<mode>2 instead.
5379         (*truncxf<mode>2_i387): Remove.
5380         (reg->reg splitters): Remove splitter pattern.
5381         (reg->mem splitters): Ditto.
5383         (truncdfsf2): New insn pattern.
5384         (truncxf<mode>2): Ditto.
5386 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5388         * tree-ssa-math-opts.c (is_mult_by): New function.
5389         (is_square_of): Use the above.
5390         (optimize_recip_sqrt): New function.
5391         (pass_cse_reciprocals::execute): Use the above.
5393 2018-09-05  Richard Biener  <rguenther@suse.de>
5395         PR bootstrap/87134
5396         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
5397         to zero-init the emplaced vec.
5399 2018-09-05  Martin Liska  <mliska@suse.cz>
5401         PR tree-optimization/87205
5402         * tree-switch-conversion.c (pass_lower_switch::execute):
5403         Group cases for switch statements.
5405 2018-09-05  Richard Biener  <rguenther@suse.de>
5407         PR tree-optimization/87217
5408         * tree-ssa-sccvn.c (vuse_valueize): New.
5409         (vn_reference_lookup_pieces): Use it.
5410         (vn_reference_lookup): Likewise.
5412 2018-09-05  Nathan Sidwell  <nathan@acm.org>
5414         PR c++/87137
5415         * stor-layout.c (place_field): Scan forwards to check last
5416         bitfield when ms_bitfield_placement is in effect.
5418 2018-09-05  Richard Biener  <rguenther@suse.de>
5420         PR bootstrap/87225
5421         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
5422         return.
5424 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
5425             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
5427         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
5428         * config.gcc (extra_objs): Build it.
5429         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
5430         Likewise.
5431         * config/aarch64/aarch64-passes.def
5432         (pass_tag_collision_avoidance): New pass.
5433         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
5434         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
5435         (aarch64_classify_address): Remove static qualifier.
5436         (aarch64_address_info, aarch64_address_type): Move to...
5437         * config/aarch64/aarch64-protos.h: ... here.
5438         (make_pass_tag_collision_avoidance): New function.
5439         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
5440         New tuning flag.
5442 2018-09-05  Martin Liska  <mliska@suse.cz>
5444         * doc/gcov.texi: Update documentation of humar
5445         readable mode.
5446         * gcov.c (format_count): Print one decimal place, it provides
5447         more fine number of situations like '1G' vs. '1.4G'.
5449 2018-09-05  Martin Liska  <mliska@suse.cz>
5451         PR target/87164
5452         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
5453         * optc-gen.awk: Allow 'Var' for Deprecated options in order
5454         to generate a MASK value.
5456 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
5458         PR debug/86593
5459         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
5460         if frame pointer isn't used.
5461         (compute_frame_pointer_to_fb_displacement): Likewise.
5463 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
5465         PR target/87198
5466         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
5467         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
5468         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
5469         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
5470         and OPTION_MASK_ISA_XSAVEC_UNSET.
5472 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
5474         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
5475         XOR operations in NAND case.
5477 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5479         * wide-int-range.cc (wide_int_range_convert): New.
5480         * wide-int-range.h (wide_int_range_convert): New.
5481         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
5482         code into wide_int_range_convert.
5483         (extract_range_into_wide_ints): Do not munge anti range constants
5484         into the entire domain.  Just return the range back.
5486 2018-09-04  Martin Liska  <mliska@suse.cz>
5488         * genmatch.c (output_line_directive): Add new argument
5489         fnargs.
5490         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
5492 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5494         * doc/invoke.texi (Option Summary): Add whitespace.
5496         * doc/invoke.texi (Option Summary): Add -Waligned-new.
5498 2018-09-04  Richard Biener  <rguenther@suse.de>
5500         PR tree-optimization/87211
5501         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
5502         backedge value we're supposed to treat as VARYING also number
5503         the PHI to VARYING in case it got a different value-number already.
5505 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
5507         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
5508         (extract_range_from_binary_expr_1): Do not call
5509         vrp_can_optimize_bit_op.
5510         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
5511         static.
5512         (wide_int_range_get_mask_and_bounds): New.
5513         (wide_int_range_optimize_bit_op): New.
5514         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
5515         (wide_int_range_bit_and): Same.
5516         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
5517         (wide_int_range_optimize_bit_op): New.
5518         (wide_int_range_get_mask_and_bounds): New.
5520 2018-09-04  Richard Biener  <rguenther@suse.de>
5522         PR tree-optimization/87176
5523         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
5524         variable.  When value-numbering a virtual PHI node make sure
5525         to not value-number to the backedge value.
5527 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
5529         * doc/extend.texi (Long Long, Hex Floats): Document support for
5530         long long and hex floats in more recent versions of ISO C++.
5532 2018-09-03  Richard Biener  <rguenther@suse.de>
5534         PR tree-optimization/87177
5535         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
5536         cleanup.
5538 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5540         * bb-reorder.c (edge_order): Convert to C-qsort-style
5541         tri-state comparator.
5542         (reorder_basic_blocks_simple): Change std::stable_sort to
5543         gcc_stablesort.
5545 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5547         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
5548         tri-state comparator.
5549         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
5551 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5553         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
5554         (mergesort): ... here as maximum count for using netsort.
5555         (gcc_qsort): Set nlim to 3 if stable sort is requested.
5556         (gcc_stablesort): New.
5557         * system.h (gcc_stablesort): Declare.
5559 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5561         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
5562         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
5563         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
5565 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
5567         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
5568         lxsdx and stxsdx alternatives.
5569         (*mov<mode>_hardfloat64): Ditto.
5570         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
5572 2018-09-03  Richard Biener  <rguenther@suse.de>
5574         PR tree-optimization/87200
5575         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
5576         simplify result.
5578 2018-09-03  Martin Liska  <mliska@suse.cz>
5580         PR tree-optimization/87201
5581         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5582         Fix parenthesis in an expression.
5584 2018-09-03  Richard Biener  <rguenther@suse.de>
5586         PR tree-optimization/87197
5587         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
5588         visited.  CSE the VN_INFO hashtable lookup.
5590         PR tree-optimization/87169
5591         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
5592         iterating make sure there's no extra backedges from irreducible
5593         regions feeding the header.  Mark the destination block
5594         executable.
5596 2018-09-03  Martin Liska  <mliska@suse.cz>
5598         PR driver/83193
5599         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
5600         * common/common-targhooks.c (default_get_valid_option_values):
5601         New function.
5602         * common/common-targhooks.h (default_get_valid_option_values):
5603         Likewise.
5604         * common/config/i386/i386-common.c: Move processor_target_table
5605         from i386.c.
5606         (ix86_get_valid_option_values): New function.
5607         (TARGET_GET_VALID_OPTION_VALUES): New macro.
5608         * config/i386/i386.c (struct ptt): Move to i386-common.c.
5609         (PTA_*): Move all defined masks into i386-common.c.
5610         (ix86_function_specific_restore): Use new processor_cost_table.
5611         * config/i386/i386.h (struct ptt): Moved from i386.c.
5612         (struct pta): Likewise.
5613         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
5614         * doc/tm.texi.in: Likewise.
5615         * opt-suggestions.c (option_proposer::suggest_option):
5616         Pass prefix to build_option_suggestions.
5617         (option_proposer::get_completions): Likewise.
5618         (option_proposer::build_option_suggestions): Use the new target
5619         hook.
5620         * opts.c (struct option_help_tuple): New struct.
5621         (print_filtered_help): Use the new target hook.
5623 2018-09-03  Martin Liska  <mliska@suse.cz>
5625         PR middle-end/59521
5626         * predict.c (set_even_probabilities): Add likely_edges
5627         argument and handle cases where we have precisely one
5628         likely edge.
5629         (combine_predictions_for_bb): Catch also likely_edges.
5630         (tree_predict_by_opcode): Handle gswitch statements.
5631         * tree-cfg.h (find_case_label_for_value): New declaration.
5632         (find_taken_edge_switch_expr): Likewise.
5633         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5634         Find pivot in decision tree based on probabily, not by number of
5635         nodes.
5637 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
5639         * doc/standards.texi (Standards): Update Objective-C reference.
5641 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5643         * doc/install.texi (Prerequisites): Update link for MPC.
5645 2018-09-01  Michael Matz  <matz@suse.de>
5647         PR tree-optimization/87074
5648         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5649         PHIs for outer-loop uses.
5651 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5653         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5654         * doc/invoke.texi (C Dialect Options): Ditto.
5656 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5658         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5660 2018-08-31  Richard Biener  <rguenther@suse.de>
5662         PR tree-optimization/87168
5663         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5664         (rpo_elim::eliminate_avail): When OP was not visited it must
5665         be available.
5667 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
5669         * tree-vrp.c (copy_value_range): Convert param "from" from
5670         "value_range *" to "const value_range *".
5671         (range_is_null): Likewise for param "vr".
5672         (range_int_cst_p): Likewise.
5673         (range_int_cst_singleton_p): Likewise.
5674         (symbolic_range_p): Likewise.
5675         (value_ranges_intersect_p): Likewise for both params.
5676         (value_range_nonnegative_p): Likewise for param "vr".
5677         (value_range_constant_singleton): Likewise.
5678         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5679         (extract_range_into_wide_ints): Likewise for param "vr".
5680         (extract_range_from_multiplicative_op): Likewise for params "vr0"
5681         and "vr1".
5682         (vrp_can_optimize_bit_op): Likewise.
5683         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5684         "vr1_".
5685         (extract_range_from_unary_expr): Likewise.
5686         (debug_value_range): Likewise for param "vr".
5687         (value_range::dump): Add "const" qualifier.
5688         (vrp_prop::check_array_ref): Convert local "vr" from
5689         "value_range *" to "const value_range *".
5690         (vrp_prop::check_mem_ref): Likewise.
5691         (vrp_prop::visit_stmt): Likewise for local "old_vr".
5692         (vrp_intersect_ranges_1): Likewise for param "vr_1".
5693         (vrp_intersect_ranges): Likewise.
5694         (simplify_stmt_for_jump_threading): Likewise for local "vr".
5695         (vrp_prop::vrp_finalize): Likewise.
5696         * tree-vrp.h (value_range::dump): Add "const" qualifier.
5697         (vrp_intersect_ranges): Add "const" qualifier to params as above.
5698         (extract_range_from_unary_expr): Likewise.
5699         (value_range_constant_singleton): Likewise.
5700         (symbolic_range_p): Likewise.
5701         (copy_value_range): Likewise.
5702         (extract_range_from_binary_expr_1): Likewise.
5703         (range_int_cst_p): Likewise.
5704         (vrp_set_zero_nonzero_bits): Likewise.
5705         (range_int_cst_singleton_p): Likewise.
5707 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
5709         * config/aarch64/arm_neon.h (vabsd_s64): New.
5710         (vnegd_s64): Likewise.
5712 2018-08-31  Martin Jambor  <mjambor@suse.cz>
5714         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5716 2018-08-31  Martin Liska  <mliska@suse.cz>
5718         * ipa-icf.c (sem_item::add_type): Use
5719         sem_item::m_type_hash_cache.
5720         * ipa-icf.h: Move the cache from sem_item_optimizer
5721         to sem_item.
5723 2018-08-31  Nathan Sidwell  <nathan@acm.org>
5725         * doc/extend.texi (Backwards Compatibility): Remove implicit
5726         extern C leeway of () being (...).
5728 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5730         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5732 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
5734         PR target/86684
5735         PR target/87149
5736         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5738 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
5740         PR middle-end/87138
5741         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5742         gen_int_mode.  Formatting fixes.
5744 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
5746         * target.def (custom_function_descriptors): Improve documentation.
5747         * doc/tm.texi.in (Trampolines): Expand discussion of function
5748         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5749         beginning of the section.
5750         * doc/tm.texi: Regenerated.
5752 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
5754         * cfg.h (class auto_edge_flag): Spell out the template-id of the
5755         base class in the initializer list.  This is a workaround for
5756         building with older GCC.
5757         (class auto_bb_flag): Likewise.
5759 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5761         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5762         (altivec_vcmpequ<VI_char>_p): Remove star.
5763         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5764         vector load modes.
5765         (expand_strncmp_vec_sequence): New function.
5766         (emit_final_str_compare_vec): New function.
5767         (expand_strn_compare): Add support for vector strncmp.
5768         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5769         length specification to bytes.
5770         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5771         (vcmpnezb_p): New pattern.
5772         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5773         for option -mstring-compare-inline-limit.
5775 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
5777         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5778         (PTA_SKYLAKE): Add PTA_AES.
5779         (PTA_GOLDMONT): Likewise.
5781 2018-08-29  Jan Hubicka  <jh@suse.cz>
5783         PR lto/86517
5784         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5785         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5787 2018-08-29  Jan Hubicka  <jh@suse.cz>
5789         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5790         TYPE_STUB_DECL.
5791         (hash_tree): Do not visit TYPE_STUB_DECL.
5792         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5793         stream TYPE_STUB_DECL.
5794         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5795         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5796         after free_lang_data.
5797         (type_in_anonymous_namespace_p): Likewise.
5799 2018-08-29  Jan Hubicka  <jh@suse.cz>
5801         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5802         comment that it has to be even number.
5803         (class sreal): Change m_sig type to int32_t.
5804         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5805         int64_t for temporary calculations.
5806         (sreal_verify_basics): Drop one bit from minimum and maximum.
5808 2018-08-30  Richard Biener  <rguenther@suse.de>
5810         PR tree-optimization/87147
5811         * tree-ssa-sccvn.c (SSA_VISITED): New function.
5812         (visit_phi): When the degenerate result is from the backedge and
5813         we didn't visit its definition yet drop to VARYING.
5814         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5816 2018-08-29  Jan Hubicka  <jh@suse.cz>
5818         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5819         DECL_VINDEX.
5820         (hash_tree): Likewise.
5822 2018-08-29  Jan Hubicka  <jh@suse.cz>
5824         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5825         and TYPE_NEXT_REF_TO.
5827 2018-08-29  Jan Hubicka  <jh@suse.cz>
5829         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5830         comment that it has to be even number.
5831         (class sreal): Change m_sig type to int32_t.
5832         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5833         int64_t for temporary calculations.
5834         (sreal_verify_basics): Drop one bit from minimum and maximum.
5836 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
5838         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5840 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
5842         PR middle-end/86995
5843         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5844         if to_add is negative.
5846 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5848         PR middle-end/87053
5849         * builtins.c (c_strlen): Improve range checks.
5851 2018-08-29  Martin Sebor  <msebor@redhat.com>
5852             Jeff Law  <law@redhat.com>
5854         PR tree-optimization/86714
5855         PR tree-optimization/86711
5856         * builtins.c (c_strlen): Add arguments to call to string_constant.
5857         * expr.c (string_constant): Add argument.  Detect missing nul
5858         terminator and outermost declaration it's missing in.
5859         * expr.h (string_constant): Add argument.
5860         * fold-const.c (read_from_constant_string): Add arguments to call to
5861         string_constant.
5862         (c_getstr): Likewise.
5863         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5864         to string_constant.
5865         * tree-ssa-strlen.c (get_stridx): Likewise.
5867 2018-08-29  Jan Hubicka  <jh@suse.cz>
5869         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5870         Do not stream DECL_VINDEX.
5871         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5872         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5873         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5875 2018-08-29  Richard Biener  <rguenther@suse.de>
5877         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5878         virtual operands that are not default defs to honor region
5879         boundaries.
5880         (rpo_vn_valueize): Remove ineffective code here.
5882 2018-08-29  Richard Biener  <rguenther@suse.de>
5884         PR tree-optimization/87132
5885         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5886         when skipping defs reachable over backedges.
5888 2018-08-29  Richard Biener  <rguenther@suse.de>
5890         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5891         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5892         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5893         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5894         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5895         (vn_reference_lookup_3): Remove use of const_parms.
5896         (free_rpo_vn): Do not free const_parms.
5897         (do_rpo_vn): Do not call init_const_parms.
5898         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5899         SSA_NAME_POINTS_TO_READONLY_MEMORY.
5900         (call_may_clobber_ref_p_1): Likewise.
5902 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
5904         PR other/86726
5905         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5906         (-O): Ditto.
5907         (-ftree-scev-cprop): Document.
5909 2018-08-29  Jan Hubicka  <jh@suse.cz>
5911         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5912         parameters.
5913         (sreal constructor): Update.
5914         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5915         sreal:operator/): Update.
5917 2018-08-29  Martin Liska  <mliska@suse.cz>
5919         * tree-switch-conversion.c (switch_conversion::expand):
5920         Strenghten assumption about gswitch statements.
5922 2018-08-29  Richard Biener  <rguenther@suse.de>
5924         PR tree-optimization/87117
5925         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5926         re-value-number released SSA VDEFs.
5928 2018-08-29  Richard Biener  <rguenther@suse.de>
5930         PR tree-optimization/87126
5931         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5933 2018-08-28  Jim Wilson  <jimw@sifive.com>
5935         * config/riscv/pic.md: Rewrite.
5936         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5937         invalid address.
5938         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5939         (SOFTF, default_load, softload, softstore): New.
5941 2018-08-28  Jeff Law  <law@redhat.com>
5943         * fold-const.c (fold_binary_loc): Remove recently added assert.
5945 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
5947         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5948         to OP parmeter of generated function.
5950 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
5952         PR tree-optimization/87009
5953         * match.pd: Add boolean optimizations.
5955 2018-08-28  Martin Sebor  <msebor@redhat.com>
5957         PR middle-end/86631
5958         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5959         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5960         (pass_walloca::gate): Use it.
5961         (alloca_call_type): Same.
5962         (pass_walloca::execute): Same.
5963         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5965 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
5967         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5968         GCC_VERSION for usage of "__gcc_dump_printf__" format from
5969         >= 3005 to >= 9000.
5971 2018-08-28  Richard Biener  <rguenther@suse.de>
5973         PR tree-optimization/87124
5974         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5975         constants before looking up avail.
5977 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
5979         PR middle-end/87099
5980         * calls.c (maybe_warn_nonstring_arg): Punt early if
5981         warn_stringop_overflow is zero.  Don't call get_range_strlen
5982         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5983         Swap comparison operands to have constants on rhs.  Only use
5984         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
5985         increment lenrng[0].
5987 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
5989         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5990         use of tree_to_shwi.  Remove duplicated test for the size being
5991         a whole number of bytes.
5993 2018-08-28  Richard Biener  <rguenther@suse.de>
5995         PR tree-optimization/87117
5996         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5997         Handle removed stmt without LHS (GIMPLE_NOP).
5999 2018-08-28  Richard Biener  <rguenther@suse.de>
6001         PR tree-optimization/87117
6002         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
6003         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
6005 2018-08-28  Richard Biener  <rguenther@suse.de>
6007         PR tree-optimization/87117
6008         * tree-ssa-pre.c (compute_avail): Do not make expressions
6009         with predicated values available.
6010         (get_expr_value_id): Assert we do not run into predicated value
6011         expressions.
6013 2018-08-28  Richard Biener  <rguenther@suse.de>
6015         PR tree-optimization/87117
6016         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
6017         get virtual operands.
6018         (get_expr_operands): Handle STRING_CST like other decls.
6020 2018-08-28  Martin Liska  <mliska@suse.cz>
6022         * tree.h: Update documentation of fndecl_built_in_p
6023         functions.
6026 2018-08-27  Jeff Law  <law@redhat.com>
6027         PR tree-optimization/87110
6028         * tree-ssa-dse.c (compute_trims): Handle non-constant
6029         TYPE_SIZE_UNIT.
6031 2018-08-27  Martin Sebor  <msebor@redhat.com>
6033         PR tree-optimization/86914
6034         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
6036 2018-08-27  Martin Sebor  <msebor@redhat.com>
6038         PR tree-optimization/87112
6039         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
6040         the type of the bound argument.
6042 2018-08-27  Jeff Law  <law@redhat.com>
6044         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
6045         type does not have a TYPE_SIZE_UNIT.
6047 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
6049         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
6050         with include of backend.h.
6052 2018-08-27  Richard Biener  <rguenther@suse.de>
6054         PR tree-optimization/86927
6055         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
6056         use const cond reduction code.
6058 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
6060         PR tree-optimization/85758
6061         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
6063 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
6065         PR c++/87091
6066         * diagnostic-show-locus.c (class layout_range): Update for
6067         conversion of show_caret_p to a tri-state.
6068         (layout_range::layout_range): Likewise.
6069         (make_range): Likewise.
6070         (layout::maybe_add_location_range): Likewise.
6071         (layout::should_print_annotation_line_p): Don't show annotation
6072         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
6073         (layout::get_state_at_point): Update for conversion of
6074         show_caret_p to a tri-state.  Bail out early for
6075         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
6076         underlining or source colorization.
6077         (gcc_rich_location::add_location_if_nearby): Update for conversion
6078         of show_caret_p to a tri-state.
6079         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
6080         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
6081         Likewise.
6082         (selftest::test_one_liner_labels): Likewise.
6083         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
6084         conversion of show_caret_p to a tri-state.
6085         * pretty-print.c (text_info::set_location): Likewise.
6086         * pretty-print.h (text_info::set_location): Likewise.
6087         * substring-locations.c (format_warning_n_va): Likewise.
6088         * tree-diagnostic.c (default_tree_printer): Likewise.
6089         * tree-pretty-print.c (newline_and_indent): Likewise.
6091 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
6093         PR c++/87091
6094         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
6095         line above for line-insertion fix-it hints.
6096         (selftest::test_fixit_insert_containing_newline): Update the
6097         expected results, and add a test with line-numbering enabled.
6099 2018-08-27  Martin Liska  <mliska@suse.cz>
6101         PR sanitizer/86962
6102         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
6103         params with DECL_HAS_VALUE_EXPR_P.
6105 2018-08-27  Martin Liska  <mliska@suse.cz>
6107         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
6108         selected expansion strategy.
6110 2018-08-27  Martin Liska  <mliska@suse.cz>
6112         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
6113         * builtins.c (is_builtin_fn): Likewise.
6114         * attribs.c (diag_attr_exclusions): Use new function
6115         fndecl_built_in_p and remove check for FUNCTION_DECL if
6116         possible.
6117         (builtin_mathfn_code): Likewise.
6118         (fold_builtin_expect): Likewise.
6119         (fold_call_expr): Likewise.
6120         (fold_builtin_call_array): Likewise.
6121         (fold_call_stmt): Likewise.
6122         (set_builtin_user_assembler_name): Likewise.
6123         (is_simple_builtin): Likewise.
6124         * calls.c (gimple_alloca_call_p): Likewise.
6125         (maybe_warn_nonstring_arg): Likewise.
6126         * cfgexpand.c (expand_call_stmt): Likewise.
6127         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
6128         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
6129         (cgraph_node::verify_node): Likewise.
6130         * cgraphclones.c (build_function_decl_skip_args): Likewise.
6131         (cgraph_node::create_clone): Likewise.
6132         * config/arm/arm.c (arm_insert_attributes): Likewise.
6133         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
6134         * dse.c (scan_insn): Likewise.
6135         * expr.c (expand_expr_real_1): Likewise.
6136         * fold-const.c (operand_equal_p): Likewise.
6137         (fold_binary_loc): Likewise.
6138         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
6139         * gimple-low.c (lower_stmt): Likewise.
6140         * gimple-pretty-print.c (dump_gimple_call): Likewise.
6141         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6142         Likewise.
6143         * gimple.c (gimple_build_call_from_tree): Likewise.
6144         (gimple_call_builtin_p): Likewise.
6145         (gimple_call_combined_fn): Likewise.
6146         * gimplify.c (gimplify_call_expr): Likewise.
6147         (gimple_boolify): Likewise.
6148         (gimplify_modify_expr): Likewise.
6149         (gimplify_addr_expr): Likewise.
6150         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
6151         * ipa-cp.c (determine_versionability): Likewise.
6152         * ipa-fnsummary.c (compute_fn_summary): Likewise.
6153         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
6154         * ipa-split.c (visit_bb): Likewise.
6155         (split_function): Likewise.
6156         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6157         * lto-cgraph.c (input_node): Likewise.
6158         * lto-streamer-out.c (write_symbol): Likewise.
6159         * omp-low.c (setjmp_or_longjmp_p): Likewise.
6160         (lower_omp_1): Likewise.
6161         * predict.c (strip_predict_hints): Likewise.
6162         * print-tree.c (print_node): Likewise.
6163         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
6164         * trans-mem.c (is_tm_irrevocable): Likewise.
6165         (is_tm_load): Likewise.
6166         (is_tm_simple_load): Likewise.
6167         (is_tm_store): Likewise.
6168         (is_tm_simple_store): Likewise.
6169         (is_tm_abort): Likewise.
6170         (tm_region_init_1): Likewise.
6171         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
6172         * tree-cfg.c (verify_gimple_call): Likewise.
6173         (move_stmt_r): Likewise.
6174         (stmt_can_terminate_bb_p): Likewise.
6175         * tree-eh.c (lower_eh_constructs_2): Likewise.
6176         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
6177         * tree-inline.c (remap_gimple_stmt): Likewise.
6178         (copy_bb): Likewise.
6179         (estimate_num_insns): Likewise.
6180         (fold_marked_statements): Likewise.
6181         * tree-sra.c (scan_function): Likewise.
6182         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
6183         (optimize_stack_restore): Likewise.
6184         (pass_fold_builtins::execute): Likewise.
6185         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
6186         (mark_all_reaching_defs_necessary_1): Likewise.
6187         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
6188         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
6189         (pass_forwprop::execute): Likewise.
6190         * tree-ssa-loop-im.c (stmt_cost): Likewise.
6191         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
6192         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
6193         * tree-ssa-strlen.c (get_string_length): Likewise.
6194         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
6195         (find_func_aliases_for_call): Likewise.
6196         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
6197         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
6198         * tree-tailcall.c (find_tail_calls): Likewise.
6199         * tree.c (need_assembler_name_p): Likewise.
6200         (free_lang_data_in_decl): Likewise.
6201         (get_call_combined_fn): Likewise.
6202         * ubsan.c (is_ubsan_builtin_p): Likewise.
6203         * varasm.c (incorporeal_function_p): Likewise.
6204         * tree.h (DECL_BUILT_IN): Remove and replace with
6205         fndecl_built_in_p.
6206         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
6207         (fndecl_built_in_p): New.
6209 2018-08-27  Martin Liska  <mliska@suse.cz>
6211         PR tree-optimization/86847
6212         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
6213         Dump also subtree probability.
6214         (switch_decision_tree::do_jump_if_equal): New function.
6215         (switch_decision_tree::emit_case_nodes): Handle special
6216         situations in balanced tree that can be emitted much simpler.
6217         Fix calculation of probabilities that happen in tree expansion.
6218         * tree-switch-conversion.h (struct cluster): Add
6219         is_single_value_p.
6220         (struct simple_cluster): Likewise.
6221         (struct case_tree_node): Add new function has_child.
6222         (do_jump_if_equal): New.
6224 2018-08-27  Martin Liska  <mliska@suse.cz>
6226         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
6227         Add new argument to bit_test_cluster constructor.
6228         (bit_test_cluster::emit): Set bits really number of values
6229         handlel by a test.
6230         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
6231         probability argument.
6232         * tree-switch-conversion.h (struct bit_test_cluster):
6233         Add m_handles_entire_switch.
6235 2018-08-27  Martin Liska  <mliska@suse.cz>
6237         PR tree-optimization/86702
6238         * tree-switch-conversion.c (jump_table_cluster::emit):
6239         Make probabilities even for values in jump table
6240         according to number of cases handled.
6241         (switch_decision_tree::compute_cases_per_edge): Pass
6242         argument to reset_out_edges_aux function.
6243         (switch_decision_tree::analyze_switch_statement): Likewise.
6244         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
6245         Make it static.
6247 2018-08-27  Martin Liska  <mliska@suse.cz>
6249         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
6250         cfun argument explicitly.
6251         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
6252         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
6253         function gimple_switch_default_bb.
6254         (convert_switch_statements):
6255         (expand_builtins):
6256         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
6257         * stmt.c (label_to_block_fn): Use label_to_block and pass
6258         cfun argument explicitly and use gimple_switch_label_bb.
6259         (expand_case): Likewise.
6260         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
6261         cfun argument explicitly. Likewise.
6262         (make_edges_bb): Likewise.
6263         (make_cond_expr_edges): Likewise.
6264         (get_cases_for_edge): Likewise.
6265         (make_gimple_switch_edges): Likewise.
6266         (label_to_block_fn): Likewise.
6267         (label_to_block): Likewise.
6268         (make_goto_expr_edges): Likewise.
6269         (make_gimple_asm_edges): Likewise.
6270         (main_block_label): Likewise.
6271         (group_case_labels_stmt): Likewise.
6272         (find_taken_edge_computed_goto): Likewise.
6273         (find_taken_edge_switch_expr): Likewise.
6274         (gimple_verify_flow_info): Likewise.
6275         (gimple_redirect_edge_and_branch): Likewise.
6276         (gimple_switch_label_bb): New function.
6277         (gimple_switch_default_bb): Likewise.
6278         (gimple_switch_edge): Likewise.
6279         (gimple_switch_default_edge): Likewise.
6280         * tree-cfg.h (label_to_block_fn): Remove and replace ...
6281         (label_to_block): ... with this.
6282         (gimple_switch_label_bb): New.
6283         (gimple_switch_default_bb): Likewise.
6284         (gimple_switch_edge): Likewise.
6285         (gimple_switch_default_edge): Likewise.
6286         * tree-cfgcleanup.c (convert_single_case_switch): Use
6287         new gimple functions and pass new argument to label_to_block.
6288         (cleanup_control_flow_bb):
6289         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
6290         cfun argument explicitly.
6291         (make_eh_edges): Likewise.
6292         (redirect_eh_dispatch_edge): Likewise.
6293         (lower_resx): Likewise.
6294         (lower_eh_dispatch): Likewise.
6295         (maybe_remove_unreachable_handlers): Likewise.
6296         (unsplit_eh): Likewise.
6297         (cleanup_empty_eh): Likewise.
6298         (verify_eh_edges): Likewise.
6299         (verify_eh_dispatch_edge): Likewise.
6300         * tree-ssa-dom.c (record_edge_info): Likewise.
6301         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
6302         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
6303         (thread_through_normal_block): Likewise.
6304         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
6305         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
6306         * tree-switch-conversion.c (switch_conversion::collect): Use new
6307         gimple functions.
6308         (switch_conversion::check_final_bb): Likewise.
6309         (switch_conversion::gather_default_values): Pass new argument
6310         to label_to_block.
6311         (switch_conversion::build_constructors): Likewise.
6312         (switch_decision_tree::compute_cases_per_edge): Use new
6313         gimple_switch_edge function.
6314         (switch_decision_tree::analyze_switch_statement): Pass new argument
6315         to label_to_block.
6316         (switch_decision_tree::try_switch_expansion): Use
6317         gimple_switch_default_edge.
6318         * tree-vrp.c (find_switch_asserts): Pass new argument
6319         to label_to_block.
6320         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
6321         (vr_values::simplify_switch_using_ranges): Likewise.
6323 2018-08-27  Richard Biener  <rguenther@suse.de>
6325         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
6326         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
6328         * tree-ssa-sccvn.h (struct vn_pval): New structure.
6329         (struct vn_nary_op_s): Add unwind_to member.  Add
6330         predicated_values flag and put result into a union together
6331         with a linked list of vn_pval.
6332         (struct vn_ssa_aux): Add name member to make maintaining
6333         a map of SSA name to vn_ssa_aux possible.  Remove no longer
6334         needed info, dfsnum, low, visited, on_sccstack, use_processed
6335         and range_info_anti_range_p members.
6336         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
6337         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
6338         New functions.
6339         (vn_valueize): New global.
6340         (vn_context_bb): Likewise.
6341         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
6342         VN_INFO_PTR_INFO): Remove.
6343         * tree-ssa-sccvn.c: ... (rewrite)
6344         (pass_fre::execute): For -O2+ initialize loops and run
6345         RPO VN in optimistic mode (iterating).  For -O1 and -Og
6346         run RPO VN in non-optimistic mode.
6347         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
6348         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
6349         * doc/invoke.texi (sccvn-max-scc-size): Remove.
6350         (rpo-vn-max-loop-depth): Document.
6351         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
6352         when valuezing the VUSE signals we walked out of the region.
6353         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
6354         (phi_translate): Set VN context block to use for availability
6355         lookup.
6356         (compute_avail): Likewise.
6357         (pre_valueize): New function.
6358         (pass_pre::execute): Adjust to the RPO VN API.
6360         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
6361         (propagate_constants_for_unrolling): Remove.
6362         (tree_unroll_loops_completely): Perform value-numbering
6363         on the unrolled bodies loop parent.
6365 2018-08-27  Richard Biener  <rguenther@suse.de>
6367         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
6368         for partial antic compute.
6370 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
6372         PR rtl-optimization/87065
6373         * combine.c (simplify_if_then_else): Formatting fix.
6374         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
6375         check.
6376         (known_cond): Don't return const_true_rtx for vector modes.  Use
6377         CONST0_RTX instead of const0_rtx.  Formatting fixes.
6379 2018-08-27  Martin Liska  <mliska@suse.cz>
6381         PR gcov-profile/87069
6382         * gcov.c (process_file): Record files already processed
6383         and warn about a file being processed multiple times.
6385 2018-08-27  Martin Liska  <mliska@suse.cz>
6387         PR driver/83193
6388         * config/aarch64/aarch64.c (aarch64_override_options_internal):
6389         Set default values for x_aarch64_*_string strings.
6390         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
6391         prefix.  For -mabi do not print '=ABI' in help and use
6392         <option_value> format for -msve-vector-bits and -moverride
6393         options.
6395 2018-08-26  Jeff Law  <law@redhat.com>
6397         * config/mips/frame-header-opt.c: Include "backend.h" rather than
6398         "cfg.h"
6400 2018-08-26  Marek Polacek  <polacek@redhat.com>
6402         PR c++/87029, Implement -Wredundant-move.
6403         * doc/invoke.texi: Document -Wredundant-move.
6405 2018-08-25  Martin Sebor  <msebor@redhat.com>
6407         PR tree-optimization/87059
6408         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
6409         to the same type as the other.
6410         * fold-const.c (fold_binary_loc): Assert expectation.
6412 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
6414         * config/darwin.c (machopic_legitimize_pic_address): Clean up
6415         extraneous parentheses, dead code section and formatting.
6417 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
6419         PR c++/87091
6420         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
6421         wide enough for jumps in the line-numbering to be visible.
6422         (layout::print_gap_in_line_numbering): New member function.
6423         (layout::calculate_line_spans): When using line numbering, merge
6424         line spans that are only 1 line apart.
6425         (diagnostic_show_locus): When printing line numbers, show gaps in
6426         line numbering directly, rather than printing headers.
6427         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
6428         line-numbering with multiple line spans.
6429         (selftest::test_fixit_insert_containing_newline_2): Add test of
6430         line-numbering, in which the spans are close enough to be merged.
6432 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6434         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
6435         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
6436         * tree-vrp.c (range_is_nonnull): Remove.
6437         (range_includes_zero_p): Accept value_range instead of min/max.
6438         (extract_range_from_binary_expr_1): Do not early bail on
6439         POINTER_PLUS_EXPR.
6440         Use range_includes_zero_p instead of range_is_nonnull.
6441         (extract_range_from_unary_expr): Use range_includes_zero_p instead
6442         of range_is_nonnull.
6443         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
6444         special case VR_ANTI_RANGE.
6445         (vrp_finalize): Same.
6446         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
6447         instead of min/max.
6448         (range_is_nonnull): Remove.
6449         * vr-values.c (vrp_stmt_computes_nonzero): Use
6450         range_includes_zero_p instead of range_is_nonnull.
6451         (extract_range_basic): Pass value_range to range_includes_zero_p
6452         instead of range_is_nonnull.
6454 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
6456         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
6457         * emit-rtl.h (rtl_data): Remove return_bnd.
6458         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
6459         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
6460         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
6461         (POINTER_BOUNDS_MODE): Remove definition.
6462         (make_pointer_bounds_mode): Remove.
6463         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
6464         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
6465         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
6466         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
6467         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
6468         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
6469         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
6471         * config/i386/i386-modes.def (BND32, BND64): Remove.
6472         * config/i386/i386.c (dbx_register_map): Remove bound registers.
6473         (dbx64_register_map): Ditto.
6474         (svr4_dbx_register_map): Ditto.
6475         (indirect_thunk_bnd_needed): Remove.
6476         (indirect_thunks_bnd_used): Ditto.
6477         (indirect_return_bnd_needed): Ditto.
6478         (indirect_return_via_cx_bnd): Ditto.
6479         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
6480         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
6481         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
6482         (output_indirect_thunk_function): Remove handling of
6483         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
6484         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
6485         (ix86_save_reg): Remove handling of crtl->return_bnd.
6486         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
6487         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
6488         and UNSPEC_BNDLX_ADDR.
6489         (ix86_output_indirect_branch_via_reg): Remove handling of
6490         indirect_thunk_prefix_bnd.
6491         (ix86_output_indirect_branch_via_push): Ditto.
6492         (ix86_output_function_return): Ditto.
6493         (ix86_output_indirect_function_return): Ditto.
6494         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
6495         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
6496         (CALL_USED_REGISTERS): Ditto.
6497         (REG_ALLOC_ORDER): Update for removal of bound registers.
6498         (HI_REGISTER_NAMES): Ditto.
6499         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
6500         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
6501         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
6502         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
6503         (FIRST_PSEUDO_REG): Update.
6504         (BND): Remove mode iterator.
6505         * config/i386/predicates.md (bnd_mem_operator): Remove.
6507 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6509         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
6510         vectors.
6512 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
6514         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
6515         the case in which the permute needs only a single element and
6516         repeats for every vector of the result.  Extend that case to
6517         handle variable-length vectors.
6518         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6520 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
6522         PR debug/79342
6523         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
6524         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
6526 2018-08-24  Richard Biener  <rguenther@suse.de>
6528         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
6529         bb_flags_allocated members.
6530         (auto_flag): New RAII class for allocating flags.
6531         (auto_edge_flag): New RAII class for allocating edge flags.
6532         (auto_bb_flag): New RAII class for allocating bb flags.
6533         * cfgloop.c (verify_loop_structure): Allocate temporary edge
6534         flag dynamically.
6535         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
6536         in favor of temporarily allocated BB flag.
6537         * hsa-brig.c: Re-order includes.
6538         * hsa-dump.c: Likewise.
6539         * hsa-regalloc.c: Likewise.
6540         * print-rtl.c: Likewise.
6541         * profile-count.c: Likewise.
6543 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
6545         PR target/86989
6546         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
6547         the TOC register.
6549 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
6551         PR 87073/bootstrap
6552         * wide-int-range.cc (wide_int_range_div): Do not ignore result
6553         from wide_int_range_multiplicative_op.
6555 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6557         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
6558         "permutaion".
6560 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
6562         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
6563         to 'expanded'.
6565 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
6567         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
6568         full GENERIC expression used for replacement.
6570 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
6572         * tree-vrp.c (abs_extent_range): Remove.
6573         (extract_range_into_wide_ints): Pass wide ints by reference.
6574         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
6575         Pass wide ints by reference in all calls to
6576         extract_range_into_wide_ints.
6577         * wide-int-range.cc (wide_int_range_div): New.
6578         * wide-int-range.h (wide_int_range_div): New.
6579         (wide_int_range_includes_zero_p): New.
6580         (wide_int_range_zero_p): New.
6582 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
6584         * config/aarch64/aarch64.md (arches): New enum.
6585         (arch): New enum attr.
6586         (arch_enabled): New attr.
6587         (enabled): Now uses arch_enabled only.
6588         (simd, sve, fp16): Removed attribute.
6589         (fp): Attr now defined in terms of 'arch'.
6590         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
6591         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
6592         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
6593         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
6594         attributes into 'arch'.
6595         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
6596         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
6597         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
6598         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
6599         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
6600         'simd' attribute into 'arch'.
6601         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
6602         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
6603         Convert use of 'fp' attribute to 'arch'.
6604         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
6605         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
6606         into 'arch'.
6607         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
6608         (different modes) Merge 'fp' and 'simd' into 'arch'.
6609         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
6610         'simd' into 'arch'.
6612 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
6614         PR rtl-optimization/87026
6615         * expmed.c (canonicalize_comparison): If we can no longer create
6616         pseudoregisters, don't.
6618 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
6620         PR target/86951
6621         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6622         prototype.
6623         * config/arm/arm.c (speculation_barrier_libfunc): New static
6624         variable.
6625         (arm_init_libfuncs): Initialize it.
6626         (arm_emit_speculation_barrier): New function.
6627         * config/arm/arm.md (speculation_barrier): Call
6628         arm_emit_speculation_barrier for architectures that do not have
6629         DSB or ISB.
6630         (speculation_barrier_insn): Only match on Armv7 or later.
6632 2018-08-23  Richard Biener  <rguenther@suse.de>
6634         PR middle-end/87024
6635         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6636         calls.
6638 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6640         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6641         of single-vector TBLs.
6642         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6643         one input is given.
6645 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6647         PR target/85910
6648         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6649         aarch64_evpc_tbl guard.
6651 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6653         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6654         behaviour.
6656 2018-08-22  Martin Sebor  <msebor@redhat.com>
6658         PR middle-end/87052
6659         * tree-pretty-print.c (pretty_print_string): Add argument.
6660         (dump_generic_node): Call to pretty_print_string with string size.
6662 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
6664         PR rtl-optimization/86771
6665         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6666         of two SETs into those two SETs, one to be placed at i2, if that SETs
6667         destination is modified between i2 and i3.
6669 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6671         PR tree-optimization/86725
6672         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6673         function.
6674         (vect_analyze_scalar_cycles_1): Check it.
6676 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6678         PR tree-optimization/86725
6679         * tree-vect-loop.c (vect_is_simple_reduction): When treating
6680         an outer loop phi as a double reduction, make sure that the
6681         single user of the phi result is an inner loop phi.
6683 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6685         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6686         grouped stores with gaps to a strided group.
6688 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6690         * tree-vect-stmts.c (get_group_load_store_type)
6691         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6692         first statement in a group.
6694 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6696         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6697         the sequence used in gcc/gcc.c.
6699 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6701         PR other/704
6702         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6703         building it.
6705 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6707         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6708         Darwin10-specific unwinder-shim.
6709         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6710         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6711         New to cater for Darwin10 Rosetta.
6713 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6715         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6716         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6718 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6720         PR bootstrap/81033
6721         PR target/81733
6722         PR target/52795
6723         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6724         (dwarf2out_switch_text_section): Generate a local label for the
6725         second function sub-section and apply it as the second FDE start
6726         label.
6727         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6728         second sub-section start.
6730 2018-08-22  Richard Biener  <rguenther@suse.de>
6732         PR tree-optimization/86988
6733         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6735 2018-08-22  Richard Biener  <rguenther@suse.de>
6737         PR tree-optimization/86945
6738         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6740 2018-08-22  Alexandre Oliva <oliva@adacore.com>
6742         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6743         a comment about how uses of r2 for .sdata2 come about.
6745 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
6747         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6749 2018-08-21  Marek Polacek  <polacek@redhat.com>
6751         PR c++/86981, Implement -Wpessimizing-move.
6752         * doc/invoke.texi: Document -Wpessimizing-move.
6754 2018-08-21  Jan Hubicka  <jh@suse.cz>
6756         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6757         * tree.h (is_redundant_typedef): Remove.
6758         * dwarf2out.c (is_redundant_typedef): Turn into static function.
6760 2018-08-21  Jan Hubicka  <jh@suse.cz>
6762         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6763         when possible.
6765 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
6767         * expmed.c (extract_low_bits): Reject invalid subregs early.
6769 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6771         PR middle-end/86121
6772         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6773         behaviour.
6775 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6777         * config/vxworks.h: Guard vxworks_asm_out_constructor and
6778         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6779         * config/vxworks.c: Likewise.
6781 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6783         * config/vxworks.c: Set targetm.have_ctors_dtors
6784         if HAVE_INITFINI_ARRAY_SUPPORT.
6785         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6786         if HAVE_INITFINI_ARRAY_SUPPORT.
6788 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6790         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6791         default search path for VxWorks < 7.
6793 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6795         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6796         (get_format_string): Refer to c_getstr.
6798 2018-08-21  Tom de Vries  <tdevries@suse.de>
6800         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6801         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6802         (debuginfo_early_stop): Declare.
6803         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6804         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6805         (debuginfo_early_stop): New function.
6806         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6807         and debuginfo_early_stop.
6808         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6809         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6810         (general_init): Call debuginfo_early_init.
6811         (finalize): Call debuginfo_fini.
6812         (do_compile): Call debuginfo_init.
6813         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6814         -fdump-early-debug.
6815         (@item -fdump-debug, @item -fdump-earlydebug): Add.
6817 2018-08-21  Tom de Vries  <tdevries@suse.de>
6819         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6820         flag_dump_noaddr and flag_dump_unnumbered.
6822 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
6824         * wide-int-range.cc (wide_int_range_abs): New.
6825         (wide_int_range_order_set): Rename from wide_int_range_min_max.
6826         * wide-int-range.h (wide_int_range_abs): New.
6827         (wide_int_range_min_max): New.
6828         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6829         case to call wide_int_range_abs.
6830         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6831         (extract_range_from_abs_expr): Delete.
6833 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
6835         PR target/87033
6836         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6837         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6838         for indexed loads.
6840 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6841             Jeff Law <law@redhat.com>
6843         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6844         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6846 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
6848         PR other/84889
6849         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6850         (decl_attributes): Likewise.
6851         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6852         instance.
6853         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6854         * diagnostic-core.h (class auto_diagnostic_group): New class.
6855         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6856         (diagnostic_report_diagnostic): Handle the first diagnostics within
6857         a group.
6858         (emit_diagnostic): Add auto_diagnostic_group instance.
6859         (inform): Likewise.
6860         (inform_n): Likewise.
6861         (warning): Likewise.
6862         (warning_at): Likewise.
6863         (warning_n): Likewise.
6864         (pedwarn): Likewise.
6865         (permerror): Likewise.
6866         (error): Likewise.
6867         (error_n): Likewise.
6868         (error_at): Likewise.
6869         (sorry): Likewise.
6870         (fatal_error): Likewise.
6871         (internal_error): Likewise.
6872         (internal_error_no_backtrace): Likewise.
6873         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6874         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6875         * diagnostic.h (struct diagnostic_context): Add fields
6876         "diagnostic_group_nesting_depth",
6877         "diagnostic_group_emission_count", "begin_group_cb",
6878         "end_group_cb".
6879         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6880         Add auto_diagnostic_group instance(s).
6881         (find_explicit_erroneous_behavior): Likewise.
6882         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6883         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6884         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6885         (gimplify_va_arg_expr): Likewise.
6886         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6887         (HSA_SORRY_AT): Likewise.
6888         * ipa-devirt.c (compare_virtual_tables): Likewise.
6889         (warn_odr): Likewise.
6890         * multiple_target.c (expand_target_clones): Likewise.
6891         * opts-common.c (cmdline_handle_error): Likewise.
6892         * reginfo.c (globalize_reg): Likewise.
6893         * substring-locations.c (format_warning_n_va): Likewise.
6894         * tree-inline.c (expand_call_inline): Likewise.
6895         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6896         * tree-ssa-loop-niter.c
6897         (do_warn_aggressive_loop_optimizations): Likewise.
6898         * tree-ssa-uninit.c (warn_uninit): Likewise.
6899         * tree.c (warn_deprecated_use): Likewise.
6901 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
6903         PR target/87014
6904         * config/i386/i386.md (eh_return): Always update EH return
6905         address in word_mode.
6907 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
6909         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6910         TARGET_SPLIT_COMPLEX_ARG is defined.
6912 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6914         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6916 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6918         PR target/86984
6919         * expr.c (expand_assignment): Assert that bitpos is positive.
6920         (store_field): Likewise
6921         (expand_expr_real_1): Make sure that bitpos is positive.
6922         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6923         integer overflow.
6925 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6927         * Makefile.in (CPP_ID_DATA_H): Delete.
6928         (CPP_INTERNAL_H): Don't add it.
6929         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6930         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6932 2018-08-20  Richard Biener  <rguenther@suse.de>
6934         PR tree-optimization/78655
6935         * tree-vrp.c (extract_range_from_binary_expr_1): Make
6936         pointer + offset nonnull if either operand is nonnull work.
6938 2018-08-20  Tom de Vries  <tdevries@suse.de>
6940         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6941         unless the referenced die describes the added property using
6942         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
6943         Otherwise, add a DW_AT_location to the referenced die.
6945 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
6947         PR target/86994
6948         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6949         register_operand when calling ix86_set_reg_reg_cost.
6950         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6951         Set *total to 0 for operands that satisfy x86_64_immediate_operand
6952         predicate and to 1 otherwise.
6954 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
6956         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6957         emit a diagnostic that it is not supported and reset the option.
6958         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6959         supported and consume the option.  (ASM_FINAL_SPEC): New.
6961 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6963         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6964         a sentence.
6966 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
6968         C-SKY port: Documentation
6970         * doc/extend.texi (C-SKY Function Attributes): New section.
6971         * doc/invoke.texi (Option Summary): Add C-SKY options.
6972         (C-SKY Options): New section.
6973         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6975 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6976             Huibin Wang  <huibin_wang@c-sky.com>
6977             Sandra Loosemore  <sandra@codesourcery.com>
6978             Chung-Lin Tang  <cltang@codesourcery.com>
6980         C-SKY port: Backend implementation
6982         * config/csky/*: New.
6983         * common/config/csky/*: New.
6985 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6986             Huibin Wang  <huibin_wang@c-sky.com>
6987             Sandra Loosemore  <sandra@codesourcery.com>
6988             Chung-Lin Tang  <cltang@codesourcery.com>
6989             Andrew Jenner  <andrew@codesourcery.com>
6991         C-SKY port: Configury
6993         * config.gcc (csky-*-*): New.
6994         * configure.ac: Add csky to targets for dwarf2 debug_line support.
6995         * configure: Regenerated.
6997 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
6999         * dump-context.h: Include "dumpfile.h".
7000         (dump_context::dump_printf_va): Convert final param from va_list
7001         to va_list *.  Convert from ATTRIBUTE_PRINTF to
7002         ATTRIBUTE_GCC_DUMP_PRINTF.
7003         (dump_context::dump_printf_loc_va): Likewise.
7004         * dumpfile.c: Include "stringpool.h".
7005         (make_item_for_dump_printf_va): Delete.
7006         (make_item_for_dump_printf): Delete.
7007         (class dump_pretty_printer): New class.
7008         (dump_pretty_printer::dump_pretty_printer): New ctor.
7009         (dump_pretty_printer::emit_items): New member function.
7010         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
7011         function.
7012         (dump_pretty_printer::emit_item): New member function.
7013         (dump_pretty_printer::stash_item): New member function.
7014         (dump_pretty_printer::format_decoder_cb): New member function.
7015         (dump_pretty_printer::decode_format): New member function.
7016         (dump_context::dump_printf_va): Reimplement in terms of
7017         dump_pretty_printer.
7018         (dump_context::dump_printf_loc_va): Convert final param from va_list
7019         to va_list *.
7020         (dump_context::begin_scope): Reimplement call to
7021         make_item_for_dump_printf.
7022         (dump_printf): Update for change to dump_printf_va.
7023         (dump_printf_loc): Likewise.
7024         (selftest::test_capture_of_dump_calls): Convert "stmt" from
7025         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
7026         with %T, %E, and %G.
7027         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
7028         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
7029         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
7030         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
7031         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
7032         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
7033         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
7034         within a dump_printf_loc call to "%wu".
7035         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
7036         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
7037         missing space after "=".
7038         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
7039         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
7040         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
7041         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
7042         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
7043         duplicate "vectorized" from message.
7045 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
7047         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
7048         polyNxK_t element's TYPE_STRING_FLAG.
7050 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
7052         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
7053         (they were unnamed before).  Fix comments.
7055 2018-08-17  Nathan Sidwell  <nathan@acm.org>
7057         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
7059 2018-08-17  Richard Biener  <rguenther@suse.de>
7061         PR tree-optimization/86841
7062         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
7064 2018-08-17  Martin Liska  <mliska@suse.cz>
7066         * common.opt: Remove Warn, Init and Report for options with
7067         Ignore/Deprecated flag. Warning is done automatically for
7068         Deprecated flags.
7069         * config/i386/i386.opt: Likewise.
7070         * config/ia64/ia64.opt: Likewise.
7071         * config/rs6000/rs6000.opt: Likewise.
7072         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
7073         Remove usage of flag_check_pointer_bounds.
7074         * lto-wrapper.c (merge_and_complain): Do not handle
7075         OPT_fcheck_pointer_bounds.
7076         (append_compiler_options): Likewise.
7077         * opt-functions.awk: Do not handle Deprecated.
7078         * optc-gen.awk: Check that Var, Report and Init are not
7079         used for an option with Ignore/Deprecated flag.
7080         * opts-common.c (decode_cmdline_option): Do not report
7081         CL_ERR_DEPRECATED.
7082         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
7083         options.
7084         * opts.h (struct cl_option): Remove cl_deprecated flag.
7085         (CL_ERR_DEPRECATED): Remove error enum value.
7087 2018-08-17  Richard Biener  <rguenther@suse.de>
7089         PR middle-end/86505
7090         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
7091         across a va-arg-pack using call adjust its return value accordingly.
7093 2018-08-16  Martin Sebor  <msebor@redhat.com>
7095         PR tree-optimization/86853
7096         * gimple-ssa-sprintf.c (struct format_result): Rename member.
7097         (struct fmtresult): Add member and initialize it in ctors.
7098         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
7099         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
7100         (format_directive): Set POSUNDER4K when MAYFAIL is set.
7101         (parse_directive): Handle %C same as %c.
7102         (sprintf_dom_walker::compute_format_length): Adjust.
7103         (is_call_safe): Adjust.
7105 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7107         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
7108         for determining how to count the elements.
7109         * builtins.h (c_strlen): Adjust prototype.
7110         * expr.c (string_constant): Add new parameter mem_size.
7111         Set *mem_size appropriately.
7112         * expr.h (string_constant): Adjust protoype.
7113         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
7114         * gimple-fold.h (get_range_strlen): Adjust prototype.
7115         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
7116         (format_string): Call get_string_length with eltsize.
7118 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
7120         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
7121         to emit the span, rather than setting it as the prefix.
7123 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
7125         * diagnostic-show-locus.c (layout::start_annotation_line): Add
7126         "margin_char" parameter, defaulting to space.  Use it in place
7127         of pp_space for the initial part of the margin.
7128         (layout::print_leading_fixits): Use '+' when filling the margin
7129         of line-insertion fix-it hints.
7131 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7133         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
7134         Delete.
7136 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7138         * config/rs6000/altivec.md: Don't set length attribute to the default
7139         value.
7140         * config/rs6000/darwin.md: Ditto.
7141         * config/rs6000/dfp.md: Ditto.
7142         * config/rs6000/htm.md: Ditto.
7143         * config/rs6000/rs6000.md: Ditto.
7144         * config/rs6000/sync.md: Ditto.
7145         * config/rs6000/vsx.md: Ditto.
7147 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7149         * config/rs6000/altivec.md: Don't set length attribute to the default
7150         value, for branch instructions.
7151         * config/rs6000/darwin.md: Ditto.
7152         * config/rs6000/rs6000.md: Ditto.
7154 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
7156         * config/rs6000/rs6000.md (length): Always define as const_int 4.
7157         (unnamed conditional branch define_insn): Set length to 4 or 8
7158         depending on offset.
7159         (<bd>_<mode>): Similar, for alternative 0.
7160         (<bd>tf_<mode>): Ditto.
7162 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7164         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
7166 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
7168         * doc/rtl.texi: Replace old RTX class names with new names.
7171 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
7173         * expmed.h (canonicalize_comparison): New declaration.
7174         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
7175         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
7176         * optabs.c (prepare_cmp_insn): Likewise.
7177         * rtl.h (unsigned_condition_p): New function which checks if a
7178         comparison operator is unsigned.
7180 2018-08-16  Nathan Sidwell  <nathan@acm.org>
7182         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
7183         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
7185 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7187         PR target/84711
7188         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
7189         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
7190         (mov<mov>): ..this and enable unconditionally.
7192 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
7194         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
7196 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
7198         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
7199         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
7200         "Common" with "Target".
7202 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
7204         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
7205         * doc/invoke.texi (mmitigate-rop): Remove.
7206         * config/i386/i386.c: Do not include "regrename.h".
7207         (ix86_rop_should_change_byte_p, reg_encoded_number)
7208         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
7209         Remove.
7210         (ix86_reorg): Remove call to ix86_mitigate_rop.
7211         * config/i386/i386.md (attr "modrm_class"): Remove.
7212         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
7213         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
7214         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
7216 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7218         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
7219         allow folding of mergeh() and mergel() for the float and double types.
7220         (fold_mergehl_helper): Rework to handle building a permute tree
7221         for float vectors.
7223 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
7225         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
7226         for TARGET_SSE.
7228 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
7230         * common.opt (fdiagnostics-show-labels): New option.
7231         * diagnostic-show-locus.c (class layout_range): Add field
7232         "m_label".
7233         (class layout): Add field "m_show_labels_p".
7234         (layout_range::layout_range): Add param "label" and use it to
7235         initialize m_label.
7236         (make_range): Pass in NULL for new "label" param of layout_range's
7237         ctor.
7238         (layout::layout): Initialize m_show_labels_p.
7239         (layout::maybe_add_location_range): Pass in loc_range->m_label
7240         when constructing layout_range instances.
7241         (struct line_label): New struct.
7242         (layout::print_any_labels): New member function.
7243         (layout::print_line): Call it if label-printing is enabled.
7244         (selftest::test_one_liner_labels): New test.
7245         (selftest::test_diagnostic_show_locus_one_liner): Call it.
7246         * diagnostic.c (diagnostic_initialize): Initialize
7247         context->show_labels_p.
7248         * diagnostic.h (struct diagnostic_context): Add field
7249         "show_labels_p".
7250         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7251         -fno-diagnostics-show-labels.
7252         * dwarf2out.c (gen_producer_string): Add
7253         OPT_fdiagnostics_show_labels to the ignored options.
7254         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
7255         param.
7256         (gcc_rich_location::maybe_add_expr): Likewise.
7257         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
7258         label" param, defaulting to NULL.
7259         (gcc_rich_location::add_expr): Add "label" param.
7260         (gcc_rich_location::maybe_add_expr): Likewise.
7261         (class text_range_label): New class.
7262         (class range_label_for_type_mismatch): New class.
7263         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
7264         of format_warning_va.
7265         (fmtwarn_n): Likewise for new params of format_warning_n_va.
7266         * lto-wrapper.c (merge_and_complain): Add
7267         OPT_fdiagnostics_show_labels to the "pick one setting" options.
7268         (append_compiler_options): Likewise to the dropped options.
7269         (append_diag_options): Likewise to the passed-on options.
7270         * opts.c (common_handle_option): Handle the new option.
7271         * selftest-diagnostic.c
7272         (test_diagnostic_context::test_diagnostic_context): Enable
7273         show_labels_p.
7274         * substring-locations.c: Include "gcc-rich-location.h".
7275         (format_warning_n_va): Add "fmt_label" and "param_label" params
7276         and use them as appropriate.
7277         (format_warning_va): Add "fmt_label" and "param_label" params,
7278         passing them on to format_warning_n_va.
7279         (format_warning_at_substring): Likewise.
7280         (format_warning_at_substring_n): Likewise.
7281         * substring-locations.h (format_warning_va): Add "fmt_label" and
7282         "param_label" params.
7283         (format_warning_n_va): Likewise.
7284         (format_warning_at_substring): Likewise.
7285         (format_warning_at_substring_n): Likewise.
7286         * toplev.c (general_init): Initialize global_dc->show_labels_p.
7288 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
7290         PR testsuite/86519
7291         * builtins.c (expand_builtin_memcmp): Do not expand the call
7292         when overflow is detected.
7294 2018-08-15  Martin Sebor  <msebor@redhat.com>
7296         PR tree-optimization/71625
7297         * config/aarch64/aarch64-builtins.c
7298         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
7300 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
7302         * config/s390/s390.c (s390_reorg): Remove loop.
7304 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7306         * config/darwin.c
7307          (darwin_function_switched_text_sections): Delete.
7308         * gcc/config/darwin.h
7309          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
7311 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
7313         PR target/81685
7314         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
7315         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
7316         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
7318 2018-08-15  Martin Liska  <mliska@suse.cz>
7320         PR tree-optimization/86925
7321         * predict.c (expr_expected_value_1): When taking
7322         later predictor, assign also probability.
7323         Use fold_build2_initializer_loc in order to fold
7324         the expression in -frounding-math.
7326 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
7328         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
7329         patterns.
7330         (expand_vec_perm_1): Try the new method.
7332 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7334         PR target/86547
7335         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
7336         Check whether lra_live_max_point is 0 before dividing.
7338 2018-08-14  Martin Sebor  <msebor@redhat.com>
7340         PR tree-optimization/86650
7341         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
7342         (vrp_prop::check_mem_ref): Same.
7344 2018-08-13  Liu Hao <lh_mouse@126.com>
7346         * pretty-print.c (eat_esc_sequence): Swap the foreground and
7347         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
7348         and clear it thereafter, as it only works for DBCS.
7350 2018-08-13  Liu Hao <lh_mouse@126.com>
7352         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
7353         handle returned by _get_osf_handle().
7355 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7357         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
7358         for folding vec_perm.
7360 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
7362         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
7363         Add support for gimple-folding of vec_pack() and vec_unpack()
7364         intrinsics.
7366 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
7368         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
7369         vec_xst variants to the list.
7370         (rs6000_gimple_fold_builtin): Add support for folding unaligned
7371         vector loads and stores.
7373 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
7375         * config.gcc (rs6000-ibm-aix4.x): Delete.
7376         (rs6000-ibm-aix5.1): Delete.
7377         (rs6000-ibm-aix5.2): Delete.
7378         (rs6000-ibm-aix5.3): Delete.
7379         * config/rs6000/aix43.h: Delete.
7380         * config/rs6000/aix51.h: Delete.
7381         * config/rs6000/aix52.h: Delete.
7382         * config/rs6000/t-aix43: Delete.
7384 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
7386         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7387         New function.
7388         (s390_decompose_address): Factor out constant pool ref
7389         decomposition.
7391 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7393         * config/nds32/nds32-predicates.c
7394         (nds32_can_use_bclr_p): Change return type as bool.
7395         (nds32_can_use_bset_p): Ditto.
7396         (nds32_can_use_btgl_p): Ditto.
7397         (nds32_can_use_bitci_p): Ditto.
7398         * config/nds32/nds32-protos.h
7399         (nds32_can_use_bclr_p): Change declaration.
7400         (nds32_can_use_bset_p): Ditto.
7401         (nds32_can_use_btgl_p): Ditto.
7402         (nds32_can_use_bitci_p): Ditto.
7404 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7406         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
7407         Support -msched-prolog-epilog option.
7408         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
7410 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7412         * common/config/nds32/nds32-common.c
7413         (nds32_option_optimization_table): Enalbe -malways-align.
7415 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
7417         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
7418         extra_headers.
7419         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
7420         OPT_misr_secure_ case.
7421         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
7422         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
7423         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
7424         secure attribute.
7425         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
7426         (nds32_isr_info): New field security_level.
7427         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
7428         * config/nds32/nds32.md (return_internal): Consider critical attribute.
7429         * config/nds32/nds32.opt (misr-secure): New option.
7430         * config/nds32/nds32_init.inc: New file.
7431         * config/nds32/nds32_isr.h: New file.
7433 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
7435         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
7436         Update comment for atomic instructions.
7437         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
7438         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
7439         Remove.
7440         (atomic_loaddi): Revise fence expansion to only emit fence prior to
7441         load for __ATOMIC_SEQ_CST model.
7442         (atomic_loaddi_1): Remove float register target.
7443         (atomic_storedi): Handle CONST_INT values.
7444         (atomic_storedi_1): Remove float register source.  Add special case
7445         for zero value.
7446         (memory_barrier): New expander and insn.
7448 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
7450         PR tree-optimization/86835
7451         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
7452         new_stmt after def_gsi, make sure to insert new_square_stmt after
7453         that stmt, not 2 stmts before it.
7455 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
7457         PR target/82418
7458         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
7459         instead of SWI48.
7461 2018-08-10  Martin Liska  <mliska@suse.cz>
7463         PR target/83610
7464         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
7465         function type.
7466         * builtins.c (expand_builtin_expect_with_probability):
7467         New function.
7468         (expand_builtin_expect_with_probability): New function.
7469         (build_builtin_expect_predicate): Add new argumnet probability
7470         for BUILT_IN_EXPECT_WITH_PROBABILITY.
7471         (fold_builtin_expect):
7472         (fold_builtin_2):
7473         (fold_builtin_3):
7474         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
7475         * builtins.h (fold_builtin_expect): Set new argument.
7476         * doc/extend.texi: Document __builtin_expect_with_probability.
7477         * doc/invoke.texi: Likewise.
7478         * gimple-fold.c (gimple_fold_call): Pass new argument.
7479         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
7480         also BUILT_IN_EXPECT_WITH_PROBABILITY.
7481         * predict.c (get_predictor_value): New function.
7482         (expr_expected_value): Add new argument probability. Assume
7483         that predictor and probability are always non-null.
7484         (expr_expected_value_1): Likewise.  For __builtin_expect and
7485         __builtin_expect_with_probability set probability.  Handle
7486         combination in binary expressions.
7487         (tree_predict_by_opcode): Simplify code by simply calling
7488         get_predictor_value.
7489         (pass_strip_predict_hints::execute): Add handling of
7490         BUILT_IN_EXPECT_WITH_PROBABILITY.
7491         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
7492         new predictor.
7493         * tree.h (DECL_BUILT_IN_P): New function.
7495 2018-08-10  Martin Liska  <mliska@suse.cz>
7497         PR tree-optimization/85799
7498         * passes.def: Add argument for pass_strip_predict_hints.
7499         * predict.c (class pass_strip_predict_hints): Add new argument
7500         early_p.
7501         (strip_predictor_early): New function.
7502         (pass_strip_predict_hints::execute): Call the function to
7503         strip predictors.
7504         (strip_predict_hints): New function.
7505         * predict.def: Fix comment.
7507 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
7509         * Makefile.in: Clarify which tm.texi to copy over to assert the
7510         right to grant a GFDL license for all.
7512 2018-08-09  Jeff Law  <law@redhat.com>
7514         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
7515         unused variable.
7517 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
7519         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
7520         prototype.
7522 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7524         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
7525         reductions for variable-length vectors.
7527 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
7529         PR other/84889
7530         * common.opt (fdiagnostics-show-line-numbers): New option.
7531         * diagnostic-show-locus.c (class layout): Add fields
7532         "m_show_line_numbers_p" and "m_linenum_width";
7533         (num_digits): New function.
7534         (test_num_digits): New function.
7535         (layout::layout): Initialize new fields.  Update m_x_offset
7536         logic to handle any left margin.
7537         (layout::print_source_line): Print line number when requested.
7538         (layout::start_annotation_line): New member function.
7539         (layout::print_annotation_line): Call it.
7540         (layout::print_leading_fixits): Likewise.
7541         (layout::print_trailing_fixits): Likewise.  Update calls to
7542         move_to_column for new parameter.
7543         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
7544         it to potentially call start_annotation_line.
7545         (layout::show_ruler): Call start_annotation_line.
7546         (selftest::test_line_numbers_multiline_range): New selftest.
7547         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
7548         and selftest::test_line_numbers_multiline_range.
7549         * diagnostic.c (diagnostic_initialize): Initialize
7550         show_line_numbers_p.
7551         * diagnostic.h (struct diagnostic_context): Add field
7552         "show_line_numbers_p".
7553         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
7554         -fno-diagnostics-show-line-numbers.
7555         * dwarf2out.c (gen_producer_string): Add
7556         OPT_fdiagnostics_show_line_numbers to the ignored options.
7557         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
7558         one setting" options.
7559         (append_compiler_options): Likewise to the dropped options.
7560         (append_diag_options): Likewise to the passed-on options.
7561         * opts.c (common_handle_option): Handle the new option.
7562         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
7564 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7566         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
7567         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
7568         third argument of type "const signed char" to descriptions of
7569         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
7570         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
7571         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
7572         __builtin_bcdsub_ov functions.
7574 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7576         PR tree-optimization/86858
7577         * tree-vect-loop.c (vect_is_simple_reduction): Restore
7578         flow_bb_inside_loop_p calls.
7580 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7582         PR tree-optimization/86871
7583         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
7584         instead of gimple_assign_lhs.
7586 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
7588         PR target/86887
7589         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
7590         register constraint to operand 0.
7591         (add<mode>3_carryinC): Likewise.
7592         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
7594 2018-08-09  Martin Liska  <mliska@suse.cz>
7596         PR c/86895
7597         * common.opt: Remove extra line.
7599 2018-08-09  Martin Liska  <mliska@suse.cz>
7601         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
7602         at the end of a line, make first letter capital and end up
7603         a sentence with a dot.
7604         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
7605         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
7606         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
7607         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
7608         (PARAM_MAX_ISL_OPERATIONS): Likewise.
7609         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
7610         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
7611         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
7612         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
7613         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
7614         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
7615         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
7616         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7617         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7618         (PARAM_TREE_REASSOC_WIDTH): Likewise.
7619         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7620         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7621         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7623 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
7625         PR target/84332
7626         * config/s390/s390.c (s390_option_override_internal): Reduce the
7627         stack-clash-protection-probe-interval param if it would be too big
7628         for z900.
7630 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
7632         PR target/46179
7633         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7634         * config/m68k/m68k.c (handle_move_double): Don't call
7635         m68k_final_prescan_insn.
7636         (m68k_adjust_decorated_operand): Renamed from
7637         m68k_final_prescan_insn, remove first and third operand and
7638         simplify.
7639         (print_operand): Call it.
7640         (print_operand_address): Call it.
7642 2018-08-08  Nathan Sidwell  <nathan@acm.org>
7644         * diagnostic.c (diagnostic_report_current_module): Use
7645         linemap_included_from & linemap_included_from_linemap.
7647 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
7649         * config/aarch64/aarch64-cores.def: Add phecda core.
7650         * config/aarch64/aarch64-tune.md: Regenerate.
7651         * doc/invoke.texi: Add phecda core.
7653 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7655         PR target/85295
7656         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7657         definitions.
7658         * config/s390/s390.md ("movti"): Add more alternatives for
7659         constant to GPR copies.
7661 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7663         * config/s390/s390.c: Fix whitespace damage throughout the file.
7664         * config/s390/s390.h: Likewise.
7665         * config/s390/tpf.h: Likewise.
7667 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7669         * config/s390/s390.c (s390_loadrelative_operand_p):
7670         Remove TARGET_CPU_ZARCH usages.
7671         (s390_rtx_costs): Likewise.
7672         (s390_legitimate_constant_p): Likewise.
7673         (s390_cannot_force_const_mem): Likewise.
7674         (legitimate_reload_constant_p): Likewise.
7675         (s390_preferred_reload_class): Likewise.
7676         (legitimize_pic_address): Likewise.
7677         (legitimize_tls_address): Likewise.
7678         (s390_split_branches): Removed.
7679         (s390_add_execute): Removed.
7680         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7681         (s390_mainpool_start): Likewise.
7682         (s390_mainpool_finish): Likewise.
7683         (s390_mainpool_cancel): Removed.
7684         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7685         (s390_chunkify_cancel): Likewise.
7686         (s390_return_addr_rtx): Likewise.
7687         (s390_register_info): Remove split_branches_pending_p uages.
7688         (s390_optimize_register_info): Likewise.
7689         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7690         split_branches_pending_p usages.
7691         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7692         (s390_load_got): Likewise.
7693         (s390_expand_split_stack_prologue): Likewise.
7694         (output_asm_nops): Likewise.
7695         (s390_function_profiler): Likewise.
7696         (s390_emit_call): Likewise.
7697         (s390_conditional_register_usage): Likewise.
7698         (s390_optimize_prologue): Likewise.
7699         (s390_reorg): Remove TARGET_CPU_ZARCH and
7700         split_branches_pending_p usages.
7701         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7702         usages.
7703         (s390_output_indirect_thunk_function): Likewise.
7704         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7705         (TARGET_CPU_ZARCH_P): Removed.
7706         (struct machine_function): Remove split_branches_pending_p.
7707         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7709 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7711         * common/config/s390/s390-common.c (processor_flags_table):
7712         Remove flags.
7713         * config.gcc: Remove with_arch/with_tune support.
7714         * config/s390/2064.md: Remove cpu attribute comparisons.
7715         * config/s390/driver-native.c (s390_host_detect_local_cpu):
7716         Remove MTN.
7717         * config/s390/linux.h (ASM_SPEC):
7718         Remove -march support.
7719         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7720         Use a table to get an arch level.
7721         * config/s390/s390-opts.h (enum processor_type):
7722         Remove enum values.
7723         * config/s390/s390.c
7724         (processor_table): Remove entries, add arch_level values.
7725         (s390_issue_rate): Remove cases.
7726         (s390_option_override): Adjust
7727         s390_option_override_internal() call.
7728         (s390_option_override_internal): Remove deprecation warning.
7729         (s390_valid_target_attribute_tree): Adjust
7730         s390_option_override_internal() call.
7731         * config/s390/s390.h (struct s390_processor):
7732         Share with s390-c.c, add arch_level field.
7733         * config/s390/s390.md:
7734         Remove occurrences in cpu attribute.
7735         * config/s390/s390.opt: Remove -march/-mtune support.
7736         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7737         * doc/invoke.texi: Remove deprecation warning.
7739 2018-08-08  Luis Machado  <luis.machado@linaro.org>
7741         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7742         global.
7743         (qdf24xx_tunings): Set vector cost structure to
7744         qdf24xx_vector_cost.
7746         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7747         <register_sextend>: Set to 3.
7749 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
7751         PR target/86838
7752         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7753         * config/aarch64/aarch64-simd.md
7754         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7755         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7756         (aarch64_frecpx<mode>): ...this new pattern.
7757         * config/aarch64/aarch64-simd-builtins.def: Remove comment
7758         about aarch64_frecp<FRECP:frecp_suffix><mode>.
7760 2018-08-07  Martin Liska  <mliska@suse.cz>
7762         PR middle-end/83023
7763         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7764         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7765         * predict.def (PRED_MALLOC_NONNULL): New predictor.
7766         * doc/extend.texi: Document that malloc attribute adds
7767         hit to compiler.
7769 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
7771         PR target/86785
7772         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7773         Define to speculation_safe_value_not_needed.
7775 2018-08-06  Jeff Law  <law@redhat.com>
7777         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7778         the vr_values instance to cprop_into_stmt.
7779         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7780         (cprop_operand): Also query EVRP to determine if OP is a constant.
7782 2018-08-06  Nathan Sidwell  <nathan@acm.org>
7784         * diagnostic.c (diagnostic_report_current_module): Reroll
7785         included-at loop.  Translate text.
7787 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
7789         * function-tests.c (selftest::test_expansion_to_rtl): Call
7790         free_after_compilation.
7792 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7794         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7796 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
7798         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7799         loops with memory block operations from getting unrolled.
7801 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
7803         PR target/86807
7804         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7805         Define to speculation_safe_value_not_needed.
7807 2018-08-06  Jeff Law  <law@redhat.com>
7809         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7810         assert.
7812 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7814         PR target/86662
7815         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7816         with all enabled __intN types.
7818         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7820 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7822         * alias.c (record_set): Check for clobber high.
7823         * cfgexpand.c (expand_gimple_stmt): Likewise.
7824         * combine-stack-adj.c (single_set_for_csa): Likewise.
7825         * combine.c (find_single_use_1): Likewise.
7826         (set_nonzero_bits_and_sign_copies): Likewise.
7827         (get_combine_src_dest): Likewise.
7828         (is_parallel_of_n_reg_sets): Likewise.
7829         (try_combine): Likewise.
7830         (record_dead_and_set_regs_1): Likewise.
7831         (reg_dead_at_p_1): Likewise.
7832         (reg_dead_at_p): Likewise.
7833         * dce.c (deletable_insn_p): Likewise.
7834         (mark_nonreg_stores_1): Likewise.
7835         (mark_nonreg_stores_2): Likewise.
7836         * df-scan.c (df_find_hard_reg_defs): Likewise.
7837         (df_uses_record): Likewise.
7838         (df_get_call_refs): Likewise.
7839         * dwarf2out.c (mem_loc_descriptor): Likewise.
7840         * haifa-sched.c (haifa_classify_rtx): Likewise.
7841         * ira-build.c (create_insn_allocnos): Likewise.
7842         * ira-costs.c (scan_one_insn): Likewise.
7843         * ira.c (equiv_init_movable_p): Likewise.
7844         (rtx_moveable_p): Likewise.
7845         (interesting_dest_for_shprep): Likewise.
7846         * jump.c (mark_jump_label_1): Likewise.
7847         * postreload-gcse.c (record_opr_changes): Likewise.
7848         * postreload.c (reload_cse_simplify): Likewise.
7849         (struct reg_use): Add source expr.
7850         (reload_combine): Check for clobber high.
7851         (reload_combine_note_use): Likewise.
7852         (reload_cse_move2add): Likewise.
7853         (move2add_note_store): Likewise.
7854         * print-rtl.c (print_pattern): Likewise.
7855         * recog.c (decode_asm_operands): Likewise.
7856         (store_data_bypass_p): Likewise.
7857         (if_test_bypass_p): Likewise.
7858         * regcprop.c (kill_clobbered_value): Likewise.
7859         (kill_set_value): Likewise.
7860         * reginfo.c (reg_scan_mark_refs): Likewise.
7861         * reload1.c (maybe_fix_stack_asms): Likewise.
7862         (eliminate_regs_1): Likewise.
7863         (elimination_effects): Likewise.
7864         (mark_not_eliminable): Likewise.
7865         (scan_paradoxical_subregs): Likewise.
7866         (forget_old_reloads_1): Likewise.
7867         * reorg.c (find_end_label): Likewise.
7868         (try_merge_delay_insns): Likewise.
7869         (redundant_insn): Likewise.
7870         (own_thread_p): Likewise.
7871         (fill_simple_delay_slots): Likewise.
7872         (fill_slots_from_thread): Likewise.
7873         (dbr_schedule): Likewise.
7874         * resource.c (update_live_status): Likewise.
7875         (mark_referenced_resources): Likewise.
7876         (mark_set_resources): Likewise.
7877         * rtl.c (copy_rtx): Likewise.
7878         * rtlanal.c (reg_referenced_p): Likewise.
7879         (single_set_2): Likewise.
7880         (noop_move_p): Likewise.
7881         (note_stores): Likewise.
7882         * sched-deps.c (sched_analyze_reg): Likewise.
7883         (sched_analyze_insn): Likewise.
7885 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7887         * cse.c (invalidate_reg): New function extracted from...
7888         (invalidate): ...here.
7889         (canonicalize_insn): Check for clobber high.
7890         (invalidate_from_clobbers): invalidate clobber highs.
7891         (invalidate_from_sets_and_clobbers): Likewise.
7892         (count_reg_usage): Check for clobber high.
7893         (insn_live_p): Likewise.
7894         * cselib.c (cselib_expand_value_rtx_1):Likewise.
7895         (cselib_invalidate_regno): Check for clobber in setter.
7896         (cselib_invalidate_rtx): Pass through setter.
7897         (cselib_invalidate_rtx_note_stores):
7898         (cselib_process_insn): Check for clobber high.
7899         * cselib.h (cselib_invalidate_rtx): Add operand.
7901 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7903         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7904         (mark_not_eliminable): Likewise.
7905         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7906         * lra-lives.c (process_bb_lives): Check for clobber high.
7907         * lra.c (new_insn_reg): Remember clobber highs.
7908         (collect_non_operand_hard_regs): Check for clobber high.
7909         (lra_set_insn_recog_data): Likewise.
7910         (add_regs_to_insn_regno_info): Likewise.
7911         (lra_update_insn_regno_info): Likewise.
7913 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7915         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7916         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7918 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7920         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7921         (copy_insn_1): Likewise.
7922         (gen_hard_reg_clobber_high): New gen function.
7923         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7924         * genemit.c (gen_exp): Likewise.
7925         (gen_emit_seq): Pass through info.
7926         (gen_insn): Check for CLOBBER_HIGH.
7927         (gen_expand): Pass through info.
7928         (gen_split): Likewise.
7929         (output_add_clobbers): Likewise.
7930         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7931         (remove_clobbers): Likewise.
7932         * rtl.h (gen_hard_reg_clobber_high): New declaration.
7934 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7936         * doc/rtl.texi (clobber_high): Add.
7937         (parallel): Add in clobber high
7938         * rtl.c (rtl_check_failed_code3): Add function.
7939         * rtl.def (CLOBBER_HIGH): Add expression.
7940         * rtl.h (RTL_CHECKC3): Add macro.
7941         (rtl_check_failed_code3): Add declaration.
7942         (XC3EXP): Add macro.
7944 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
7946         PR target/86386
7947         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7948         cfun->machine->max_used_stack_alignment if needed.
7950 2018-08-04  Martin Sebor  <msebor@redhat.com>
7952         PR tree-optimization/86571
7953         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7954         NaN output to 4.
7956 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
7958         PR target/86799
7959         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7960         Define.
7962 2018-08-03  Jeff Law  <law@redhat.com>
7964         PR target/86795
7965         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7966         Define to speculation_safe_value_not_needed.
7968 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
7970         * doc/gcov.texi (-x): Remove duplicate "to".
7971         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7972         (-Wif-not-aligned): Remove duplicate "is".
7973         (-flto): Remove duplicate "the".
7974         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7975         duplicate "v5.00.b".
7976         (MSP430 Options): Remove duplicate "and" from the description
7977         of "-mgprel-sec=regexp".
7978         (x86 Options): Remove duplicate copies of "vmldLog102" and
7979         vmlsLog104 from description of "-mveclibabi=type".
7981 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
7983         * internal-fn.h (first_commutative_argument): Declare.
7984         * internal-fn.c (first_commutative_argument): New function.
7985         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7986         restrictions for pattern statements.  Use first_commutative_argument
7987         to look for commutative operands in calls to internal functions.
7989 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
7991         * Makefile.in (wide-int-range.o): New.
7992         * tree-vrp.c: Move all the wide_int_* functions to...
7993         * wide-int-range.cc: ...here.
7994         * tree-vrp.h: Move all the wide_int_* prototypes to...
7995         * wide-int-range.h: ...here.
7997 2018-08-03  Tom de Vries  <tdevries@suse.de>
7999         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
8000         UI_NONE.
8001         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
8002         * except.c (output_function_exception_table): Do early exit if
8003         targetm_common.except_unwind_info (&global_options) == UI_NONE.
8005 2018-08-03  Martin Liska  <mliska@suse.cz>
8007         * predict.c (dump_prediction): Change to 2 digits
8008         in fraction part.
8010 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
8012         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
8013         neon_dup_q to...
8014         (falkor_am_1_gtov_gtov): ... a new insn reservation.
8016 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
8018         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
8019         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
8020         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
8021         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
8022         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
8024 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
8026         * diagnostic-show-locus.c (diagnostic_show_locus): Use
8027         pp_take_prefix when saving the existing prefix.
8028         * diagnostic.c (diagnostic_append_note): Likewise.
8029         * langhooks.c (lhd_print_error_function): Likewise.
8030         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
8031         param's type.  Free the existing prefix.
8032         (pp_take_prefix): New function.
8033         (pretty_printer::pretty_printer): Drop the prefix parameter.
8034         Rename the length parameter to match the comment.
8035         (pretty_printer::~pretty_printer): Free the prefix.
8036         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
8037         parameter.
8038         (struct pretty_printer): Drop the "const" from "prefix" field's
8039         type and clarify memory management.
8040         (pp_set_prefix): Drop the "const" from the 2nd param.
8041         (pp_take_prefix): New decl.
8043 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
8045         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
8046         for word_mode_ok here instead of passing as argument.
8047         (expand_block_compare): Change select_block_compare_mode() call.
8048         (expand_strncmp_gpr_sequence): New function.
8049         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
8051 2018-08-02  Jeff Law  <law@redhat.com>
8053         PR target/86790
8054         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8055         Define to speculation_safe_value_not_needed.
8057         PR target/86784
8058         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8059         Define to speculation_safe_value_not_needed.
8061 2018-08-02  Tom de Vries  <tdevries@suse.de>
8063         PR target/86660
8064         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
8065         function.  Return UI_TARGET unconditionally.
8066         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
8067         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
8069 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8071         * genemit.c (print_overload_test): Fix typo.
8073 2018-08-02  Richard Biener  <rguenther@suse.de>
8075         PR tree-optimization/86816
8076         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
8077         which checks for value availability before querying it.
8078         (gvn_uses_equal): Use it.
8079         (same_succ_hash): Likewise.
8080         (gimple_equal_p): Likewise.
8082 2018-08-02  Nick Clifton  <nickc@redhat.com>
8084         PR target/86813
8085         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8086         Define to speculation_safe_value_not_needed.
8088         PR target/86810
8089         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8090         Define to speculation_safe_value_not_needed.
8092         PR target/86810
8093         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8094         Define to speculation_safe_value_not_needed.
8096         PR target/86803
8097         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8098         Define to speculation_safe_value_not_needed.
8100         PR target/86797
8101         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8102         Define to speculation_safe_value_not_needed.
8104         PR target/86791
8105         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8106         Define to speculation_safe_value_not_needed.
8108         PR target/86789
8109         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8110         Define to speculation_safe_value_not_needed.
8112         PR target/86787
8113         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8114         Define to speculation_safe_value_not_needed.
8116         PR target/86782
8117         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8118         speculation_safe_value_not_needed.
8120         PR target/86781
8121         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
8122         to speculation_safe_value_not_needed.
8124 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8126         * doc/md.texi: Expand the documentation of instruction names
8127         to mention port-local uses.  Document '@' in pattern names.
8128         * read-md.h (overloaded_instance, overloaded_name): New structs.
8129         (mapping): Declare.
8130         (md_reader::handle_overloaded_name): New member function.
8131         (md_reader::get_overloads): Likewise.
8132         (md_reader::m_first_overload): New member variable.
8133         (md_reader::m_next_overload_ptr): Likewise.
8134         (md_reader::m_overloads_htab): Likewise.
8135         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
8136         m_next_overload_ptr and m_overloads_htab.
8137         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
8138         (get_mode_token, get_code_token, get_int_token): New functions.
8139         (map_attr_string): Add an optional argument that passes back
8140         the associated iterator.
8141         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
8142         (md_reader::handle_overloaded_name, add_overload_instance): New
8143         functions.
8144         (apply_iterators): Handle '@' names.  Report an error if '@'
8145         is used without iterators.
8146         (initialize_iterators): Initialize the new iterator_group fields.
8147         * genopinit.c (handle_overloaded_code_for)
8148         (handle_overloaded_gen): New functions.
8149         (main): Use them to print declarations of maybe_code_for_* and
8150         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
8151         * genemit.c (print_overload_arguments, print_overload_test)
8152         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
8153         (main): Use it to print definitions of maybe_code_for_* and
8154         maybe_gen_* functions.
8155         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
8156         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
8157         instead of explicit mode checks.
8158         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
8159         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
8160         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
8161         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
8162         (aarch64_expand_compare_and_swap): Likewise
8163         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
8164         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
8165         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
8166         (aarch64_constant_pool_reload_icode): Delete.
8167         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
8168         instead of aarch64_constant_pool_reload_icode.  Use
8169         code_for_aarch64_reload_mov instead of explicit mode checks.
8170         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
8171         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
8172         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
8173         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
8174         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
8175         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
8176         (aarch64_atomic_load_op_code): Delete.
8177         (aarch64_emit_atomic_load_op): Likewise.
8178         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
8179         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
8180         instead of aarch64_emit_atomic_load_op.
8181         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
8182         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
8183         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
8184         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
8185         character before the pattern name.
8186         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8187         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
8188         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
8189         (aarch64_frecps<mode>): Likewise.
8190         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
8191         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
8192         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
8193         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
8194         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
8196 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8198         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
8199         Allow HFmode constants if TARGET_FP_F16INST.
8201 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
8203         PR target/86014
8204         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
8205         No longer check last store for clobber of address register.
8207 2018-08-02  Martin Liska  <mliska@suse.cz>
8209         PR gcov-profile/86817
8210         * gcov.c (process_all_functions): New function.
8211         (main): Call it.
8212         (process_file): Move functions processing to
8213         process_all_functions.
8215 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
8217         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
8218         "const" to the "gimple *" and "rtx_insn *" parameters.
8219         * dumpfile.h (dump_user_location_t::dump_user_location_t):
8220         Likewise.
8221         (dump_location_t::dump_location_t): Likewise.
8223 2018-08-01  Martin Sebor  <msebor@redhat.com>
8225         PR tree-optimization/86650
8226         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
8227         rather than a "gcall *".  Directly pass the data of interest
8228         to percent_K_format, rather than building a temporary CALL_EXPR
8229         to hold it.
8230         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
8231         (gimple_fold_builtin_strncat): Adjust.
8232         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
8233         gcall* argument with gimple*.
8234         * gimple-ssa-warn-restrict.c (check_call): Same.
8235         (wrestrict_dom_walker::before_dom_children): Same.
8236         (builtin_access::builtin_access): Same.
8237         (check_bounds_or_overlap): Same
8238         (maybe_diag_overlap): Same.
8239         (maybe_diag_offset_bounds): Same.
8240         * tree-diagnostic.c (default_tree_printer): Move usage of
8241         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
8242         to this callsite.
8243         * tree-pretty-print.c (percent_K_format): Add argument.
8244         * tree-pretty-print.h: Add argument.
8245         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
8246         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
8247         (maybe_diag_stxncpy_trunc): Same.
8248         (handle_builtin_stxncpy): Same.
8249         (handle_builtin_strcat): Same.
8251 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8253         * match.pd: Optimise pointer range checks.
8255 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8257         PR tree-optimization/86758
8258         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
8259         to remove pattern statements.
8261 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8263         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
8264         result of dfs_enumerate_from when constructing stmt_vec_infos,
8265         instead of additionally calling get_loop_body.
8267 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8269         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
8270         parameter.
8271         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
8272         When creating an iv, assert that the step is not known to be zero.
8273         (vect_setup_realignment): Update call accordingly.
8274         * tree-vect-stmts.c (vectorizable_store): Likewise.
8275         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
8277 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8279         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
8280         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
8281         (vectorizable_reduction): Likewise.
8282         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8283         (vect_detect_hybrid_slp_stmts): Likewise.
8284         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8286 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
8288         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
8289         (wide_int_set_zero_nonzero_bits): ...this.
8290         (zero_nonzero_bits_from_vr): Rename to...
8291         (vrp_set_zero_nonzero_bits): ...this.
8292         (extract_range_from_multiplicative_op_1): Abstract wide int
8293         code...
8294         (wide_int_range_multiplicative_op): ...here.
8295         (extract_range_from_binary_expr_1): Extract wide int binary
8296         operations into their own functions.
8297         (wide_int_range_lshift): New.
8298         (wide_int_range_can_optimize_bit_op): New.
8299         (wide_int_range_shift_undefined_p): New.
8300         (wide_int_range_bit_xor): New.
8301         (wide_int_range_bit_ior): New.
8302         (wide_int_range_bit_and): New.
8303         (wide_int_range_trunc_mod): New.
8304         (extract_range_into_wide_ints): New.
8305         (vrp_shift_undefined_p): New.
8306         (extract_range_from_multiplicative_op): New.
8307         (vrp_can_optimize_bit_op): New.
8308         * tree-vrp.h (value_range::dump): New.
8309         (wide_int_range_multiplicative_op): New.
8310         (wide_int_range_lshift):New.
8311         (wide_int_range_shift_undefined_p): New.
8312         (wide_int_range_bit_xor): New.
8313         (wide_int_range_bit_ior): New.
8314         (wide_int_range_bit_and): New.
8315         (wide_int_range_trunc_mod): New.
8316         (zero_nonzero_bits_from_bounds): Rename to...
8317         (wide_int_set_zero_nonzero_bits): ...this.
8318         (zero_nonzero_bits_from_vr): Rename to...
8319         (vrp_set_zero_nonzero_bits): ...this.
8320         (range_easy_mask_min_max): Rename to...
8321         (wide_int_range_can_optimize_bit_op): this.
8322         * vr-values.c (simplify_bit_ops_using_ranges): Rename
8323         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
8325 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8327         * tree-vectorizer.h (vect_orig_stmt): New function.
8328         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
8329         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
8330         (vect_create_epilog_for_reduction): Likewise.
8331         (vectorizable_live_operation): Likewise.
8332         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8333         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
8334         * tree-vect-stmts.c (vectorizable_call): Likewise.
8335         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
8337 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8339         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
8340         argument.
8341         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
8342         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
8343         (vect_transform_loop): Likewise.
8344         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8346 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8348         * tree-vectorizer.h (vect_schedule_slp): Return void.
8349         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
8350         (vect_schedule_slp): Likewise.
8352 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8354         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
8355         argument.
8356         (vect_transform_loop): Update calls accordingly.  Schedule SLP
8357         instances before the main loop, if any exist.
8359 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
8361         PR tree-optimization/86749
8362         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
8363         If the lhs is used in a COND_EXPR, check that it is being used
8364         as the "then" or "else" value.
8366 2018-08-01  Tom de Vries  <tdevries@suse.de>
8368         PR target/86800
8369         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8370         speculation_safe_value_not_needed.
8372 2018-08-01  Richard Biener  <rguenther@suse.de>
8374         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
8375         as base and offset.
8377 2018-08-01  Martin Liska  <mliska@suse.cz>
8379         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
8380         format how successful transformation is dumped.
8381         (gimple_mod_pow2_value_transform): Likewise.
8382         (gimple_mod_subtract_transform): Likewise.
8383         (gimple_stringops_transform): Likewise.
8385 2018-08-01  Martin Liska  <mliska@suse.cz>
8387         PR value-prof/35543
8388         * value-prof.c (interesting_stringop_to_profile_p):
8389         Simplify the code and add BUILT_IN_MEMMOVE.
8390         (gimple_stringops_transform): Likewise.
8392 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
8394         * config/aarch64/aarch64-simd.md
8395         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
8396         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
8397         use GPI iterator instead of SI mode.
8399 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8401         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
8402         rs6000_speculation_barrier.
8403         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
8404         new barrier pattern name.
8406 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
8408         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
8409         (speculation_barrier): New insn.
8411 2018-08-01  Richard Biener  <rguenther@suse.de>
8413         PR bootstrap/86724
8414         * graphite.h: Include isl/id.h and isl/space.h to allow build
8415         with ISL 0.20.
8417 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
8419         PR target/86651
8420         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
8421         mode for COFF targets.
8422         * defaults.h (TARGET_COFF): Define.
8423         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
8424         TARGET_COFF): Define.
8425         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
8426         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
8427         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
8429 2018-07-31  Alexandre Oliva <oliva@adacore.com>
8431         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
8432         * gimple-streamer-out.c (output_bb): Save it.
8433         * lto-streamer-in.c (input_struct_function_base): Restore
8434         instance discriminator if available.  Create map on demand.
8435         * lto-streamer-out.c (output_struct_function_base): Save it if
8436         available.
8437         * final.c (decl_to_instance_map): Document LTO strategy.
8439 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
8440             Olivier Hainque  <hainque@adacore.com>
8442         * debug.h (decl_to_instance_map_t): New type.
8443         (decl_to_instance_map): Declare.
8444         (maybe_create_decl_to_instance_map): New inline function.
8445         * final.c (bb_discriminator, last_bb_discriminator): New statics,
8446         to track basic block discriminators.
8447         (final_start_function_1): Initialize them.
8448         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
8449         bb_discriminator.
8450         (decl_to_instance_map): New variable.
8451         (map_decl_to_instance, maybe_set_discriminator): New functions.
8452         (notice_source_line): Set discriminator.
8454 2018-07-31  Ian Lance Taylor  <iant@golang.org>
8456         * targhooks.c (default_have_speculation_safe_value): Add
8457         ATTRIBUTE_UNUSED.
8459 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8461         * dump-context.h: Include "pretty-print.h".
8462         (dump_context::refresh_dumps_are_enabled): New decl.
8463         (dump_context::emit_item): New decl.
8464         (class dump_context): Add fields "m_test_pp" and
8465         "m_test_pp_flags".
8466         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8467         (temp_dump_context::get_dumped_text): New decl.
8468         (class temp_dump_context): Add field "m_pp".
8469         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
8470         (dump_context::refresh_dumps_are_enabled): ...and add a test for
8471         m_test_pp.
8472         (set_dump_file): Update for above change.
8473         (set_alt_dump_file): Likewise.
8474         (dump_loc): New overload, taking a pretty_printer *.
8475         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
8476         to any test pretty-printer.
8477         (make_item_for_dump_gimple_stmt): New function, adapted from
8478         optinfo::add_gimple_stmt.
8479         (dump_context::dump_gimple_stmt): Call it, and use the result,
8480         eliminating the direct usage of dump_file and alt_dump_file in
8481         favor of indirectly using them via emit_item.
8482         (make_item_for_dump_gimple_expr): New function, adapted from
8483         optinfo::add_gimple_expr.
8484         (dump_context::dump_gimple_expr): Call it, and use the result,
8485         eliminating the direct usage of dump_file and alt_dump_file in
8486         favor of indirectly using them via emit_item.
8487         (make_item_for_dump_generic_expr): New function, adapted from
8488         optinfo::add_tree.
8489         (dump_context::dump_generic_expr): Call it, and use the result,
8490         eliminating the direct usage of dump_file and alt_dump_file in
8491         favor of indirectly using them via emit_item.
8492         (make_item_for_dump_printf_va): New function, adapted from
8493         optinfo::add_printf_va.
8494         (make_item_for_dump_printf): New function.
8495         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
8496         and use the result, eliminating the direct usage of dump_file and
8497         alt_dump_file in favor of indirectly using them via emit_item.
8498         (make_item_for_dump_dec): New function.
8499         (dump_context::dump_dec): Call it, and use the result,
8500         eliminating the direct usage of dump_file and alt_dump_file in
8501         favor of indirectly using them via emit_item.
8502         (make_item_for_dump_symtab_node): New function, adapted from
8503         optinfo::add_symtab_node.
8504         (dump_context::dump_symtab_node): Call it, and use the result,
8505         eliminating the direct usage of dump_file and alt_dump_file in
8506         favor of indirectly using them via emit_item.
8507         (dump_context::begin_scope): Reimplement, avoiding direct usage
8508         of dump_file and alt_dump_file in favor of indirectly using them
8509         via emit_item.
8510         (dump_context::emit_item): New member function.
8511         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
8512         Set up test pretty-printer on the underlying context.  Call
8513         refresh_dumps_are_enabled.
8514         (temp_dump_context::~temp_dump_context): Call
8515         refresh_dumps_are_enabled.
8516         (temp_dump_context::get_dumped_text): New member function.
8517         (selftest::verify_dumped_text): New function.
8518         (ASSERT_DUMPED_TEXT_EQ): New macro.
8519         (selftest::test_capture_of_dump_calls): Run all tests twice, with
8520         and then without optinfo enabled.  Add uses of
8521         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
8522         * dumpfile.h: Update comment for the dump_* API.
8523         * optinfo-emit-json.cc
8524         (selftest::test_building_json_from_dump_calls): Update for new
8525         param for temp_dump_context ctor.
8526         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
8527         and "m_owned" field.
8528         (optinfo_item::~optinfo_item): Likewise.
8529         (optinfo::add_item): New member function.
8530         (optinfo::emit): Update comment.
8531         (optinfo::add_string): Delete.
8532         (optinfo::add_printf): Delete.
8533         (optinfo::add_printf_va): Delete.
8534         (optinfo::add_gimple_stmt): Delete.
8535         (optinfo::add_gimple_expr): Delete.
8536         (optinfo::add_tree): Delete.
8537         (optinfo::add_symtab_node): Delete.
8538         (optinfo::add_dec): Delete.
8539         * optinfo.h (class dump_context): New forward decl.
8540         (optinfo::add_item): New decl.
8541         (optinfo::add_string): Delete.
8542         (optinfo::add_printf): Delete.
8543         (optinfo::add_printf_va): Delete.
8544         (optinfo::add_gimple_stmt): Delete.
8545         (optinfo::add_gimple_expr): Delete.
8546         (optinfo::add_tree): Delete.
8547         (optinfo::add_symtab_node): Delete.
8548         (optinfo::add_dec): Delete.
8549         (optinfo::add_poly_int): Delete.
8550         (optinfo_item::optinfo_item): Remove "owned" param.
8551         (class optinfo_item): Remove field "m_owned".
8553 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
8555         PR middle-end/86705
8556         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
8557         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
8558         requested variable alignment.
8559         (expand_one_ssa_partition): Likewise.
8560         (expand_one_var): Likewise.
8562 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8564         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8565         to speculation_safe_value_not_needed.
8567 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8569         * targhooks.h (speculation_safe_value_not_needed): New prototype.
8570         * targhooks.c (speculation_safe_value_not_needed): New function.
8571         * target.def (have_speculation_safe_value): Update documentation.
8572         * doc/tm.texi: Regenerated.
8574 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8576         * config/aarch64/iterators.md (ALLI_TI): New iterator.
8577         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
8578         expand.
8579         (despeculate_copy<ALLI:mode>_insn): New insn.
8580         (despeculate_copyti_insn): New insn.
8581         (despeculate_simple<ALLI:mode>): New insn
8582         (despeculate_simpleti): New insn.
8583         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
8584         function.
8585         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
8586         aarch64_speculation_safe_value.
8587         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
8589 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8591         * config/aarch64/aarch64-speculation.cc: New file.
8592         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
8593         before pass_reorder_blocks.
8594         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
8595         prototype.
8596         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
8597         X14 and X15 when tracking speculation.
8598         * config/aarch64/aarch64.md (register name constants): Add
8599         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
8600         (unspec): Add UNSPEC_SPECULATION_TRACKER.
8601         (speculation_barrier): New insn attribute.
8602         (cmp<mode>): Allow SP in comparisons.
8603         (speculation_tracker): New insn.
8604         (speculation_barrier): Add speculation_barrier attribute.
8605         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
8606         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
8607         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
8609 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8611         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
8612         aarch64_track_speculation is true.
8613         (tb<optab><mode>1): Likewise.
8614         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
8615         generate CB[N]Z when tracking speculation.
8616         (aarch64_split_compare_and_swap): Likewise.
8617         (aarch64_split_atomic_op): Likewise.
8619 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8621         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8623 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8625         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8626         (speculation_barrier): New insn.
8628 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8630         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8631         * config/arm/arm.md (speculation_barrier): New expand.
8632         (speculation_barrier_insn): New pattern.
8634 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8636         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8637         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8638         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8639         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8640         list.
8641         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8642         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8643         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8644         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8645         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8646         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8647         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8648         * builtins.c (expand_speculation_safe_value): New function.
8649         (expand_builtin): Call it.
8650         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8651         * doc/extend.texi: Document __builtin_speculation_safe_value.
8652         * doc/md.texi: Document "speculation_barrier" pattern.
8653         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8654         TARGET_HAVE_SPECULATION_SAFE_VALUE.
8655         * doc/tm.texi: Regenerated.
8656         * target.def (have_speculation_safe_value, speculation_safe_value): New
8657         hooks.
8658         * targhooks.c (default_have_speculation_safe_value): New function.
8659         (default_speculation_safe_value): New function.
8660         * targhooks.h (default_have_speculation_safe_value): Add prototype.
8661         (default_speculation_safe_value): Add prototype.
8663 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8665         * dump-context.h (dump_context::dump_loc): New decl.
8666         * dumpfile.c (dump_context::dump_loc): New member function.
8667         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8668         and dump_gimple_stmt.
8669         (dump_context::dump_gimple_expr_loc): Likewise, using
8670         dump_gimple_expr.
8671         (dump_context::dump_generic_expr_loc): Likewise, using
8672         dump_generic_expr.
8673         (dump_context::dump_printf_loc_va): Likewise, using
8674         dump_printf_va.
8675         (dump_context::begin_scope): Explicitly using the global function
8676         "dump_loc", rather than the member function.
8678 2018-07-31  Martin Sebor  <msebor@redhat.com>
8680         PR tree-optimization/86741
8681         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8683 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
8685         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8686         depend on whether prefetch instructions will be emitted or not.
8687         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8688         will be emitted or not.
8689         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8690         (TARGET_SETMEM_PFD): New macros.
8692 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8694         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8695         (NULL_STMT_VEC_INFO): Delete.
8696         (stmt_vec_info::operator*): Likewise.
8697         (stmt_vec_info::operator gimple *): Likewise.
8698         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8699         of NULL_STMT_VEC_INFO.
8700         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8701         (vect_reassociating_reduction_p): Likewise.
8702         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8703         (vectorizable_store): Likewise.
8704         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8705         (vec_info::free_stmt_vec_infos): Likewise.
8707 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8709         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8710         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8712 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8714         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8715         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8716         (vec_info::free_stmt_vec_info): New private member functions.
8717         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8718         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8719         * tree-parloops.c (gather_scalar_reductions): Remove calls to
8720         set_stmt_vec_info_vec and free_stmt_vec_infos.
8721         * tree-vect-loop.c (_loop_vec_info): Remove call to
8722         set_stmt_vec_info_vec.
8723         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8724         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8725         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8726         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8727         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
8728         assignments in {vec_info::,}new_stmt_vec_info that are redundant
8729         with the clearing in the xcalloc.
8730         (stmt_vec_info_vec): Delete.
8731         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8732         (vectorize_loops): Likewise.
8733         (vec_info::~vec_info): Remove argument from call to
8734         free_stmt_vec_infos.
8735         (vec_info::add_stmt): Remove vinfo argument from call to
8736         new_stmt_vec_info.
8738 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8740         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8741         rather than a gimple stmt.
8742         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
8743         information for pattern statements when passed the original
8744         statement; instead wait to be passed the pattern statement itself.
8745         Don't call set_vinfo_for_stmt here.
8746         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8747         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8748         stmt_vec_infos here.
8749         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8750         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8751         stmt_vec_infos entry.
8753 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8755         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8756         * tree-vectorizer.c (vec_info::replace_stmt): New function.
8757         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8758         * tree-vect-stmts.c (vectorizable_call): Likewise.
8759         (vectorizable_simd_clone_call): Likewise.
8761 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8763         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8764         * tree-vectorizer.c (vec_info::remove_stmt): New function.
8765         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8766         * tree-vect-loop.c (vect_transform_loop): Likewise.
8767         * tree-vect-slp.c (vect_schedule_slp): Likewise.
8768         * tree-vect-stmts.c (vect_remove_stores): Likewise.
8770 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8772         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8773         (vect_dr_stmt): Delete.
8774         * tree-vectorizer.c (vec_info::lookup_dr): New function.
8775         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8776         of DR_VECT_AUX.
8777         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8778         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8779         (vect_verify_datarefs_alignment, vect_peeling_supportable)
8780         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8781         (vect_analyze_data_refs): Likewise.
8782         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
8783         argument.
8784         (vect_find_same_alignment_drs): Likewise.
8785         (vect_slp_analyze_node_dependences): Update calls accordingly.
8786         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8787         instead of DR_VECT_AUX.
8788         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8789         of a vector data references.  Use vec_info::lookup_dr instead of
8790         DR_VECT_AUX.
8791         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8792         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8793         instead of DR_VECT_AUX.
8795 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8797         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8798         dr_vec_info.
8799         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8800         accordingly.
8801         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8802         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8803         (vect_gen_prolog_loop_niters): Likewise.
8805 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8807         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8808         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8809         (vect_known_alignment_in_bytes, vect_dr_behavior)
8810         (vect_get_scalar_dr_size): Take references as dr_vec_infos
8811         instead of data_references.  Update calls to other routines for
8812         which the same change has been made.
8813         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8814         dr_vec_infos instead of stmt_vec_infos.
8815         (vect_analyze_data_ref_dependence): Update call accordingly.
8816         (vect_slp_analyze_data_ref_dependence)
8817         (vect_record_base_alignments): Use DR_VECT_AUX.
8818         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8819         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8820         (vector_alignment_reachable_p, vect_get_data_access_cost)
8821         (vect_peeling_supportable, vect_analyze_group_access_1)
8822         (vect_analyze_group_access, vect_analyze_data_ref_access)
8823         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8824         (vect_compile_time_alias, vect_small_gap_p)
8825         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8826         (vect_supportable_dr_alignment): Take references as dr_vec_infos
8827         instead of data_references.  Update calls to other routines for
8828         which the same change has been made.
8829         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8830         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8831         (vect_slp_analyze_and_verify_node_alignment)
8832         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8833         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8834         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
8835         above changes.
8836         (_vect_peel_info::dr): Replace with...
8837         (_vect_peel_info::dr_info): ...this new field.
8838         (vect_peeling_hash_get_most_frequent)
8839         (vect_peeling_hash_choose_best_peeling): Update accordingly.
8840         (vect_peeling_hash_get_lowest_cost):
8841         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
8842         routines for which the same change has been made.
8843         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
8844         data_reference.
8845         * tree-vect-loop-manip.c (get_misalign_in_elems)
8846         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
8847         above changes.
8848         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8849         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8850         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8851         (get_group_load_store_type, get_negative_load_store_type)
8852         (vect_get_data_ptr_increment, vectorizable_store)
8853         (vectorizable_load): Likewise.
8854         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8855         Update calls to other routines for which the same change has been made.
8857 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8859         * tree-vectorizer.h (vec_info::move_dr): New member function.
8860         (dataref_aux): Rename to...
8861         (dr_vec_info): ...this and add "dr" and "stmt" fields.
8862         (_stmt_vec_info::dr_aux): Update accordingly.
8863         (_stmt_vec_info::data_ref_info): Delete.
8864         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8865         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8866         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8867         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8868         of data_ref.
8869         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
8870         (STMT_VINFO_DR_INFO): New macro.
8871         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8872         (set_dr_misalignment): Update after rename of dataref_aux.
8873         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
8874         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8875         initialization of STMT_VINFO_DATA_REF.
8876         * tree-vectorizer.c (vec_info::move_dr): New function.
8877         * tree-vect-patterns.c (vect_recog_bool_pattern)
8878         (vect_recog_mask_conversion_pattern)
8879         (vect_recog_gather_scatter_pattern): Use it.
8880         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8881         the "dr" and "stmt" fields of dr_vec_info instead of
8882         STMT_VINFO_DATA_REF.
8884 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8886         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8887         (is_pattern_stmt_p): Use it.
8888         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8889         on pattern statements.
8891 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8893         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8894         original stmt as a stmt_vec_info rather than a gimple stmt.
8895         (vect_pattern_recog_1): Take the statement directly as a
8896         stmt_vec_info, rather than via a gimple_stmt_iterator.
8897         Update call to vect_mark_pattern_stmts.
8898         (vect_pattern_recog): Update calls accordingly.
8900 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8902         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8903         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8904         a vect_def_type for the first argument.
8905         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8906         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
8907         operand if it isn't defined by a vectorized statement.
8908         (vect_build_gather_load_calls): Remove the mask_dt argument and
8909         update calls to vect_get_vec_def_for_stmt_copy.
8910         (vectorizable_bswap): Likewise the dt argument.
8911         (vectorizable_call): Update calls to vectorizable_bswap and
8912         vect_get_vec_def_for_stmt_copy.
8913         (vectorizable_simd_clone_call, vectorizable_assignment)
8914         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8915         (vectorizable_comparison): Update calls to
8916         vect_get_vec_def_for_stmt_copy.
8917         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
8918         vect_is_simple_use.
8919         (vect_get_loop_based_defs): Remove dt argument and update call
8920         to vect_get_vec_def_for_stmt_copy.
8921         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8922         and vect_get_vec_def_for_stmt_copy.
8923         (vectorizable_load): Update calls to vect_build_gather_load_calls
8924         and vect_get_vec_def_for_stmt_copy.
8925         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8926         (vectorizable_reduction, vectorizable_live_operation): Update calls
8927         to vect_get_vec_def_for_stmt_copy.
8929 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8931         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8932         and gimple stmt arguments with a stmt_vec_info.
8933         (vect_record_base_alignments): Update calls accordingly.
8934         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8935         and gimple stmt arguments with a stmt_vec_info.
8936         (vect_build_slp_tree_1): Remove vinfo argument and update call
8937         to vect_record_max_nunits.
8938         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8939         and vect_record_max_nunits.
8941 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8943         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8944         file and take a stmt_vec_info instead of a gimple stmt.
8945         (supportable_widening_operation, vect_finish_replace_stmt)
8946         (vect_finish_stmt_generation, vect_get_store_rhs)
8947         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8948         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8949         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8950         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8951         (vect_create_data_ref_ptr, bump_vector_ptr)
8952         (vect_permute_store_chain, vect_setup_realignment)
8953         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8954         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8955         (vectorizable_reduction, vectorizable_induction)
8956         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8957         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8958         than gimple stmts as arguments.
8959         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8960         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8961         (can_group_stmts_p, vect_check_gather_scatter)
8962         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8963         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8964         (vect_permute_load_chain, vect_shift_permute_load_chain)
8965         (vect_transform_grouped_load)
8966         (vect_record_grouped_load_vectors): Likewise.
8967         * tree-vect-loop.c (vect_fixup_reduc_chain)
8968         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8969         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8970         (vectorizable_reduction, vectorizable_induction)
8971         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8972         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8973         (vect_get_load_store_mask): Likewise.
8974         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8975         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8976         * tree-vect-stmts.c (vect_mark_relevant)
8977         (is_simple_and_all_uses_invariant)
8978         (exist_non_indexing_operands_for_use_p, process_use)
8979         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8980         (vect_get_vec_def_for_operand, vect_get_vec_defs)
8981         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8982         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8983         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8984         (get_negative_load_store_type, get_load_store_type)
8985         (vect_check_load_store_mask, vect_check_store_rhs)
8986         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8987         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8988         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8989         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8990         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8991         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8992         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8993         (supportable_widening_operation): Likewise.
8995 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8997         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8998         a stmt_vec_info instead of a gcall.
8999         (vect_check_gather_scatter): Update call accordingly.
9000         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
9001         of a gphi.
9002         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
9003         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
9004         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
9005         instead of a gimple stmt.
9006         (vect_transform_loop): Update calls accordingly.
9007         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
9008         stmt_vec_infos instead of gimple stmts.
9009         (vect_analyze_slp_instance): Update use accordingly.
9010         * tree-vect-stmts.c (read_vector_array, write_vector_array)
9011         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
9012         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
9013         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
9014         (vect_gen_widened_results_half, vect_get_loop_based_defs)
9015         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
9016         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
9017         down to subroutines.
9019 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9021         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
9022         of the worklist from a vector of gimple stmts to a vector of
9023         stmt_vec_infos.
9024         * tree-vect-stmts.c (vect_mark_relevant, process_use)
9025         (vect_mark_stmts_to_be_vectorized): Likewise
9027 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9029         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
9030         statement before passing it to vect_analyze_stmt.
9031         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
9032         the chain of phi vector definitions.  Track the exit phi via its
9033         stmt_vec_info.
9034         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
9035         STMT_VINFO_REDUC_DEF.
9036         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
9037         stmt_vec_infos to handle the statement chains.
9038         (vect_get_slp_defs): Record the first statement in the node
9039         using a stmt_vec_info.
9040         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
9041         statements here and pass their stmt_vec_info down to subroutines.
9042         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
9043         down to vect_finish_stmt_generation.
9044         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
9045         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
9046         stmt_vec_infos to subroutines.
9047         (vect_remove_stores): Use stmt_vec_infos to handle the statement
9048         chains.
9050 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9052         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
9053         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
9054         (vect_permute_store_chain, vect_setup_realignment)
9055         (vect_permute_load_chain, vect_shift_permute_load_chain)
9056         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
9057         stmts internally, and when passing values to other vectorizer routines.
9058         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
9059         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
9060         (vect_analyze_loop_operations, get_initial_def_for_reduction)
9061         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
9062         (vectorizable_reduction, vectorizable_induction)
9063         (vectorizable_live_operation, vect_transform_loop_stmt)
9064         (vect_transform_loop): Likewise.
9065         * tree-vect-patterns.c (vect_reassociating_reduction_p)
9066         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
9067         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
9068         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9069         (vect_slp_analyze_node_operations_1): Likewise.
9070         * tree-vect-stmts.c (vect_mark_relevant, process_use)
9071         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
9072         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
9073         (vect_finish_stmt_generation_1, get_group_load_store_type)
9074         (get_load_store_type, vect_build_gather_load_calls)
9075         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9076         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9077         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9078         (vectorizable_store, vectorizable_load, vectorizable_condition)
9079         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
9080         (supportable_widening_operation): Likewise.
9081         (vect_get_vector_types_for_stmt): Likewise.
9082         * tree-vectorizer.h (vect_dr_behavior): Likewise.
9084 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9086         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9087         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
9088         (vect_permute_store_chain, vect_permute_load_chain)
9089         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
9090         repeated stmt_vec_info lookups.
9091         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
9092         (vect_update_ivs_after_vectorizer): Likewise.
9093         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
9094         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9095         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
9096         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9097         (vect_bb_slp_scalar_cost): Likewise.
9098         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
9100 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9102         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
9103         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
9104         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
9105         to get gassigns and gcalls, rather than operating on generc gimple
9106         stmts.
9107         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
9108         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
9109         (vectorizable_load, vect_analyze_stmt): Likewise.
9110         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
9112 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9114         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
9115         return stmt_vec_infos rather than gimple stmts.  Do not accept
9116         null arguments.
9117         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
9118         of a gimple stmt.
9119         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9120         Update use of get_later_stmt.
9121         (vect_get_constant_vectors): Update call accordingly.
9122         (vect_schedule_slp_instance): Likewise
9123         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
9124         (vect_slp_analyze_instance_dependence): Likewise.
9125         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
9127 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9129         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
9130         (stmt_info_for_cost::stmt_info): ...this new field.
9131         (add_stmt_costs): Update accordingly.
9132         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9133         (vect_get_known_peeling_cost): Likewise.
9134         (vect_estimate_min_profitable_iters): Likewise.
9135         * tree-vect-stmts.c (record_stmt_cost): Likewise.
9137 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9139         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
9140         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9141         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
9142         accordingly.
9143         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
9145 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9147         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
9148         a gimple stmt to a stmt_vec_info.
9149         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
9151 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9153         * tree-vectorizer.h (vec_info::grouped_stores): Change from
9154         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9155         (_loop_vec_info::reduction_chains): Likewise.
9156         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
9157         accordingly.
9158         * tree-vect-slp.c (vect_analyze_slp): Likewise.
9160 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9162         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
9163         a gimple stmt to a stmt_vec_info.
9164         (_stmt_vec_info::next_element): Likewise.
9165         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
9166         (vect_slp_analyze_and_verify_node_alignment)
9167         (vect_analyze_group_access_1, vect_analyze_group_access)
9168         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
9169         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
9170         (vect_supportable_dr_alignment): Update accordingly.
9171         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
9172         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
9173         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
9174         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9175         * tree-vect-slp.c (vect_build_slp_tree_1)
9176         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
9177         (vect_split_slp_store_group, vect_analyze_slp_instance)
9178         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
9179         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
9180         (get_group_load_store_type, get_load_store_type)
9181         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
9182         (vect_transform_stmt, vect_remove_stores): Likewise.
9184 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9186         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
9187         than a gimple stmt.
9188         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9189         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
9190         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
9191         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
9192         (vector_alignment_reachable_p, vect_get_data_access_cost)
9193         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
9194         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
9195         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
9196         (vect_analyze_group_access_1, vect_analyze_group_access)
9197         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
9198         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
9199         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
9200         result of vect_dr_stmt and use the stmt_vec_info instead of
9201         the associated gimple stmt.
9202         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9203         (vect_gen_prolog_loop_niters): Likewise.
9204         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9206 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9208         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
9209         to a vec<stmt_vec_info>.
9210         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
9211         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
9212         vec<stmt_vec_info>.
9213         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
9214         to a vec<stmt_vec_info>.
9215         (bst_traits::value_type, bst_traits::value_type): Likewise.
9216         (bst_traits::hash): Update accordingly.
9217         (vect_get_and_check_slp_defs): Change the stmts parameter from
9218         a vec<gimple *> to a vec<stmt_vec_info>.
9219         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
9220         (vect_build_slp_tree): Likewise.
9221         (vect_build_slp_tree_2): Likewise.  Update uses of
9222         SLP_TREE_SCALAR_STMTS.
9223         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
9224         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
9225         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
9226         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
9227         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
9228         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
9229         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
9230         (vect_get_constant_vectors, vect_get_slp_defs)
9231         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
9232         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
9233         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
9234         instead of gimple stmts.
9235         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
9236         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
9237         (vect_slp_analyze_instance_dependence): Update uses of
9238         SLP_TREE_SCALAR_STMTS.
9239         (vect_slp_analyze_and_verify_node_alignment): Likewise.
9240         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
9241         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
9242         (get_initial_defs_for_reduction): Likewise.
9243         (vect_create_epilog_for_reduction): Likewise.
9244         (vectorize_fold_left_reduction): Likewise.
9245         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
9246         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
9247         (can_vectorize_live_stmts): Likewise.
9249 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9251         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
9252         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
9253         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
9254         than gimple stmts.
9255         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
9256         of a gimple stmt.
9257         (gather_scalar_reductions): Update after above interface changes.
9258         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
9259         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
9260         than gimple stmts.
9261         (vect_force_simple_reduction): Likewise.
9262         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
9263         LOOP_VINFO_REDUCTIONS.
9264         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9266 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9268         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
9269         a gimple stmt to a stmt_vec_info.
9270         * tree-vect-loop.c (vect_active_double_reduction_p)
9271         (vect_force_simple_reduction, vectorizable_reduction): Update
9272         accordingly.
9274 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9276         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
9277         vec<gimple *> to a vec<stmt_vec_info>.
9278         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
9279         the reduction_phis argument from a vec<gimple *> to a
9280         vec<stmt_vec_info>.
9281         (vectorizable_reduction): Likewise the phis local variable that
9282         is passed to vect_create_epilog_for_reduction.  Update for new type
9283         of SLP_TREE_VEC_STMTS.
9284         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
9285         (vectorizable_live_operation): Likewise.
9286         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
9287         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
9289 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9291         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
9292         a gimple stmt to a stmt_vec_info.
9293         (vectorizable_condition, vectorizable_live_operation)
9294         (vectorizable_reduction, vectorizable_induction): Pass back the
9295         vectorized statement as a stmt_vec_info.
9296         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
9297         use of STMT_VINFO_VEC_STMT.
9298         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
9299         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
9300         as stmt_vec_infos rather than gimple stmts.
9301         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
9302         to a stmt_vec_info.
9303         (vectorizable_live_operation): Likewise.
9304         (vectorizable_reduction, vectorizable_induction): Likewise,
9305         updating use of STMT_VINFO_VEC_STMT.
9306         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
9307         of STMT_VINFO_VEC_STMT.
9308         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
9309         (vectorizable_simd_clone_call, vectorizable_conversion)
9310         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9311         (vectorizable_store, vectorizable_load, vectorizable_condition)
9312         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
9313         from a gimple stmt to a stmt_vec_info.
9314         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
9315         pointer to a stmt_vec_info to the vectorizable_* routines.
9317 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9319         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
9320         a gimple stmt to a stmt_vec_info.
9321         (is_pattern_stmt_p): Update accordingly.
9322         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
9323         (vect_record_grouped_load_vectors): Likewise.
9324         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
9325         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
9326         (vect_model_reduction_cost): Likewise.
9327         (vect_create_epilog_for_reduction): Likewise.
9328         (vectorizable_reduction, vectorizable_induction): Likewise.
9329         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9330         Return the stmt_vec_info for the pattern statement.
9331         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
9332         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
9333         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9334         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
9335         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
9336         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
9337         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
9338         (free_stmt_vec_info, vect_is_simple_use): Likewise.
9340 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9342         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
9343         (vect_finish_stmt_generation): Likewise.
9344         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9345         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
9346         (vect_build_gather_load_calls): Use the return value of the above
9347         functions instead of a separate call to vinfo_for_stmt.  Use narrow
9348         scopes for the input gimple stmt and wider scopes for the associated
9349         stmt_vec_info.  Use vec_info::lookup_def when setting these
9350         stmt_vec_infos from an SSA_NAME definition.
9351         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9352         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9353         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9354         (vectorizable_store, vectorizable_load, vectorizable_condition)
9355         (vectorizable_comparison): Likewise.
9356         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
9357         (vectorizable_reduction): Likewise.
9359 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9361         * tree-vectorizer.h (vect_is_simple_use): Add an optional
9362         stmt_vec_info * parameter before the optional gimple **.
9363         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9364         (process_use, vect_get_vec_def_for_operand_1): Update callers.
9365         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
9366         * tree-vect-loop.c (vectorizable_reduction): Likewise.
9367         (vectorizable_live_operation): Likewise.
9368         * tree-vect-patterns.c (type_conversion_p): Likewise.
9369         (vect_look_through_possible_promotion): Likewise.
9370         (vect_recog_rotate_pattern): Likewise.
9371         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9373 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9375         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
9376         a typedef to a wrapper class.
9377         (NULL_STMT_VEC_INFO): New macro.
9378         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
9379         (stmt_vec_info::operator*): New function.
9380         (stmt_vec_info::operator gimple *): Likewise.
9381         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
9382         (add_stmt_costs): Likewise.
9383         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
9384         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
9385         (vect_get_known_peeling_cost): Likewise.
9386         (vect_estimate_min_profitable_iters): Likewise.
9387         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9388         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
9389         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9390         (vectorizable_store, free_stmt_vec_infos): Likewise.
9391         (new_stmt_vec_info): Change return type of xcalloc to
9392         _stmt_vec_info *.
9394 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9396         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
9397         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
9398         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
9399         a single_imm_use-based sequence.
9400         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
9402 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9404         * tree-vectorizer.h (vec_info::lookup_def): Declare.
9405         * tree-vectorizer.c (vec_info::lookup_def): New function.
9406         * tree-vect-patterns.c (vect_get_internal_def): Use it.
9407         (vect_widened_op_tree): Likewise.
9408         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9409         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
9410         (vectorizable_reduction): Likewise.
9411         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
9412         of a gimple *.
9413         (vect_is_slp_reduction): Update calls accordingly.  Use
9414         vec_info::lookup_def.
9415         (vect_is_simple_reduction): Likewise
9416         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
9418 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9420         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
9421         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
9422         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
9423         of vinfo_for_stmt.
9424         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
9425         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
9426         (vect_update_vf_for_slp, vect_analyze_loop_operations)
9427         (vect_is_slp_reduction, vectorizable_induction)
9428         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
9429         * tree-vect-patterns.c (vect_init_pattern_stmt):
9430         (vect_determine_min_output_precision_1, vect_determine_precisions)
9431         (vect_pattern_recog): Likewise.
9432         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
9433         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
9434         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
9435         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
9436         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
9437         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
9438         info field from a loop to a loop_vec_info.
9440 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9442         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
9443         (vec_info::add_stmt): Declare.
9444         * tree-vectorizer.c (vec_info::add_stmt): New function.
9445         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
9446         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
9447         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9448         (vectorizable_induction): Likewise.
9449         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
9450         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
9451         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
9452         (vectorizable_load): Likewise.
9453         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9454         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
9455         (vect_recog_gather_scatter_pattern): Likewise.
9456         (append_pattern_def_seq): Likewise.  Remove a check that is
9457         performed by add_stmt itself.
9459 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9461         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
9462         which make_ssa_name was called with new_stmt before new_stmt
9463         had been created.
9465 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9467         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
9468         split out from...
9469         (vect_is_slp_reduction): ...here...
9470         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
9471         that are already known to be false.
9473 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9475         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
9476         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
9477         STMT_VINFO_NUM_SLP_USES when it's true.
9478         (vect_free_slp_instance): Add a final_p parameter and pass it to
9479         vect_free_slp_tree.
9480         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
9481         (vect_analyze_slp_instance): Likewise.
9482         (vect_slp_analyze_operations): Likewise.
9483         (vect_slp_analyze_bb_1): Likewise.
9484         * tree-vectorizer.c (vec_info): Likewise.
9485         * tree-vect-loop.c (vect_transform_loop): Likewise.
9487 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9489         * tree-vect-loop.c (vectorizable_reduction): Assert that the
9490         function is not called for second and subsequent members of
9491         a reduction group.
9493 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9495         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
9496         cases for nested loops from here to ...
9497         (vect_create_epilog_for_reduction): ...here.  Only call
9498         vect_is_simple_use for inner-loop reductions.
9500 2018-07-31  Martin Liska  <mliska@suse.cz>
9502         PR gcov-profile/85338
9503         PR gcov-profile/85350
9504         PR gcov-profile/85372
9505         * profile.c (struct location_triplet): New.
9506         (struct location_triplet_hash): Likewise.
9507         (output_location): Do not output a BB that
9508         is already recorded for a line.
9509         (branch_prob): Use streamed_locations.
9511 2018-07-31  Martin Liska  <mliska@suse.cz>
9513         PR gcov-profile/85370
9514         * coverage.c (coverage_begin_function): Do not mark target
9515         clones as artificial functions.
9517 2018-07-31  Martin Liska  <mliska@suse.cz>
9519         PR gcov-profile/83813
9520         PR gcov-profile/84758
9521         PR gcov-profile/85217
9522         PR gcov-profile/85332
9523         * profile.c (branch_prob): Do not record GOTO expressions
9524         for GIMPLE statements which locations are already streamed.
9526 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9528         * gcc.c (handle_spec_function): Accept a soft_matched_part
9529         argument, as do_spec_1.  Pass it down to ...
9530         (eval_spec_function): Accept a soft_matched_part argument,
9531         and pass it down to ...
9532         (do_spec_2): Accept a soft_matched_part argument, and pass
9533         it down to do_spec_1.
9534         (do_spec_1): Pass soft_matched_part to handle_spec_function.
9535         (handle_braces): Update call to handle_spec_function.
9536         (driver::set_up_specs): Update calls to do_spec_2.
9537         (compare_debug_dump_opt_spec_function): Likewise.
9538         (compare_debug_self_opt_spec_function): Likewise.
9540 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9542         * common.opt (nolibc): New option.
9543         * doc/invoke.texi (Link Options): Document it.
9544         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
9545         * config/alpha/linux.h: Likewise.
9546         * config/arc/elf.h: Likewise.
9547         * config/arm/uclinux-elf.h: Likewise.
9548         * config/arm/unknown-elf.h: Likewise.
9549         * config/avr/avrlibc.h: Likewise.
9550         * config/bfin/bfin.h: Likewise.
9551         * config/bfin/linux.h: Likewise.
9552         * config/bfin/uclinux.h: Likewise.
9553         * config/darwin.h: Likewise.
9554         * config/darwin10.h: Likewise.
9555         * config/darwin12.h: Likewise.
9556         * config/gnu-user.h: Likewise.
9557         * config/lm32/uclinux-elf.h: Likewise.
9558         * config/pa/pa-hpux11.h: Likewise.
9559         * config/pa/pa64-hpux.h: Likewise.
9560         * config/sparc/sparc.h: Likewise.
9562 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9564         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
9565         undefined variables.
9567 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9569         PR target/86640
9570         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
9571         instead of GEN_INT.
9573 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9575         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9576         terminated string literal.
9578 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9580         PR rtl-optimization/85160
9581         * combine.c (is_just_move): New function.
9582         (try_combine): Allow combining two instructions into two if neither of
9583         the original instructions was a move.
9585 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
9587         PR target/86673
9588         * doc/extend.texi (Global Register Variables): Discourage use of type
9589         qualifiers.
9590         (Local Register Variables): Likewise.
9592 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
9594         PR tree-optimization/86506
9595         * hwint.h (ceil_log2): Resync with hwint.c implementation.
9597 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
9599         PR target/86547
9600         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
9601         hard_regno, make sure no insn between `from` and `to` clobbers it.
9603 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
9604             Tom de Vries  <tdevries@suse.de>
9606         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
9607         (PTX_DEFAULT_RUNTIME_DIM): ... this.
9608         (nvptx_goacc_validate_dims): Set default worker and gang dims to
9609         PTX_DEFAULT_RUNTIME_DIM.
9610         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
9612 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
9614         * config/pa/pa.c (pa_output_addr_vec): Align address table.
9615         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
9616         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9618 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
9620         * config/rs6000/constraints.md (wG constraint): Delete, no longer
9621         used.
9622         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9623         predicate to reflect toc fusion has been deleted.
9624         (toc_fusion_mem_raw): Delete, no longer used.
9625         (toc_fusion_mem_wrapped): Likewise.
9626         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9627         fusion mask bit.
9628         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9629         Delete, no longer used.
9630         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9631         meant to be used for toc fusion.
9632         (rs6000_debug_print_mode): Delete toc fusion debugging.
9633         (rs6000_debug_reg_global): Likewise.
9634         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9635         fusion and secondary reload support that were never used.
9636         (rs6000_option_override_internal): Delete TOC fusion, that was only
9637         partially defined, and it did not work unless you also used the
9638         -mcmodel= switch.
9639         (rs6000_legitimate_address_p): Delete TOC fusion support.
9640         (rs6000_opt_masks): Likewise.
9641         (fusion_wrap_memory_address): Delete function, no longer used.
9642         (fusion_split_address); Delete TOC fusion support.
9643         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9644         longer used with toc fusion being deleted.
9645         (TARGET_TOC_FUSION_FP): Likewise.
9646         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9647         UNSPEC.
9648         (toc fusion spliter): Delete TOC fusion support.
9649         (toc_fusionload_<mode>): Likewise.
9650         (toc_fusionload_di): Likewise.
9651         (fusion_gpr_load_<mode>): Delete generator function, this insn no
9652         longer needs to be named.  Rename predicate to delete TOC fusion.
9653         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9654         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9655         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9656         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9657         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9659 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9661         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9662         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
9663         __int128 in built-in function prototypes.
9664         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9665         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9667 2018-07-27  Martin Sebor  <msebor@redhat.com>
9669         PR tree-optimization/86696
9670         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9671         types, including enums.
9672         (handle_char_store): Be prepared for the above function to fail.
9674 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
9676         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9677         when optimization level is lower than 2 or optimize for size.
9679 2018-07-26  Martin Sebor  <msebor@redhat.com>
9681         PR tree-optimization/86043
9682         PR tree-optimization/86042
9683         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9684         (get_string_cst_length): Rename...
9685         (get_min_string_length): ...to this.  Add argument.
9686         (handle_char_store): Extend to handle multi-character stores by
9687         MEM_REF.
9688         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
9689         * tree.h (initializer_zerop): Add argument.
9691 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
9693         PR middle-end/86660
9694         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9695         declare target to variables if they have always,{to,from,tofrom} map
9696         kinds.
9698 2018-07-26  Martin Liska  <mliska@suse.cz>
9700         PR lto/86548
9701         * lto-wrapper.c: Add linker_output as prefix
9702         for ltrans_output_file.
9704 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
9706         PR rtl-optimization/85805
9707         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9708         value for hard registers if that was written in the same mode.
9710 2018-07-26  Martin Liska  <mliska@suse.cz>
9712         PR gcov-profile/86536
9713         * gcov.c (format_gcov): Use printf format %.*f directly
9714         and do not handle special values.
9716 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9718         * common/config/arc/arc-common.c (arc_option_optimization_table):
9719         Update default optimizations for size.
9721 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9723         * config/arc/arc.md (movsf_insn): Add short instruction selection.
9724         * config/arc/constraints.md (CfZ): New constraint.
9725         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9726         (subsf3_fpu): Likewise.
9727         (cmpsf_fpu): Likewise.
9728         (cmpsf_fpu_uneq): Likewise.
9730 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9732         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9733         accesses as well.
9734         (arc_is_uncached_mem_p): uncached applies to both the variable and
9735         the pointer.
9737 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9739         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9740         register names.
9742 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
9744         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9745         field "m_scopes" from vec to auto_vec.
9747 2018-07-25  Martin Liska  <mliska@suse.cz>
9749         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9750         return type.
9752 2018-07-25  Richard Biener  <rguenther@suse.de>
9754         PR debug/86654
9755         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9756         special wrt context_die late.
9757         (gen_subprogram_die): Re-use DIEs in local scope.
9759 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
9761         PR tree-optimization/86644
9762         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
9764 2018-07-25  Martin Liska  <mliska@suse.cz>
9766         PR middle-end/86645
9767         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9768         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9770 2018-07-25  Martin Liska  <mliska@suse.cz>
9772         PR sanitizer/79635
9773         * params.def: Explain ASan abbreviation and provide
9774         a documentation link.
9776 2018-07-24  Martin Sebor  <msebor@redhat.com>
9778         PR tree-optimization/86622
9779         PR tree-optimization/86532
9780         * builtins.h (string_length): Declare.
9781         * builtins.c (c_strlen): Correct handling of non-constant offsets.
9782         (check_access): Be prepared for non-constant length ranges.
9783         (string_length): Make extern.
9784         * expr.c (string_constant): Only handle the minor non-constant
9785         array index.  Use string_constant to compute the length of
9786         a generic string constant.
9788 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
9790         PR tree-optimization/86618
9791         * tree-vect-stmts.c (vectorizable_call): Don't take the address
9792         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9794 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
9796         PR tree-optimization/86636
9797         * json.cc (json::object::set): Fix comment.  Add assertions.
9798         (json::array::append): Move here from json.h.  Add comment and an
9799         assertion.
9800         (json::string::string): Likewise.
9801         * json.h (json::array::append): Move to json.cc.
9802         (json::string::string): Likewise.
9803         * optinfo-emit-json.cc
9804         (optrecord_json_writer::impl_location_to_json): Assert that we
9805         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9806         wrapper around it.  Expand the location once, rather than three
9807         times.
9808         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9809         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9810         wrappers.
9811         (optrecord_json_writer::optinfo_to_json): Likewise, in four
9812         places.  Fix some overlong lines.
9814 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
9816         * config/aarch64/aarch64-simd.md
9817         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9818         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9819         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9820         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9821         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9822         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9823         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9824         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9825         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9827 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
9829         PR middle-end/86627
9830         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9831         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
9832         and abs_d == d, do the power of two handling if profitable.
9834 2018-07-24  Richard Biener  <rguenther@suse.de>
9836         * match.pd: Add BIT_FIELD_REF canonicalizations.
9838 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9840         PR c/86617
9841         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9843 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9845         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9846         terminated STRING_CST object.
9848 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9850         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9852 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9854         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9855         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9856         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9857         the elements into a register.
9858         (rs6000_split_v4si_init_di_reg): Delete.
9859         (rs6000_split_v4si_init): Delete.
9860         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9861         (vsx_init_v4si): Rewrite as a define_expand.
9863 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9865         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9866         zero_extend argument from memory): New.
9868 2018-07-22  Martin Sebor  <msebor@redhat.com>
9870         PR bootstrap/86621
9871         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9872         diagnosing calls with unknown arguments unless -Walloca-larger-than
9873         is restricted to less than PTRDIFF_MAX bytes.
9875 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
9877         * doc/gcov.texi (Invoking Gcov): Editorial changes.
9879 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9881         * pretty-print.c (text_info::set_location): Remove redundant
9882         "line_table" parameter from call to rich_location::set_range.
9884 2018-07-20  Martin Sebor  <msebor@redhat.com>
9886         PR middle-end/82063
9887         * builtins.c (expand_builtin_alloca): Adjust.
9888         * calls.c (alloc_max_size): Simplify.
9889         * cgraphunit.c (cgraph_node::expand): Adjust.
9890         * common.opt (larger_than_size, warn_frame_larger_than): Remove
9891         variables.
9892         (frame_larger_than_size): Same.
9893         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9894         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9895         Initialize.
9896         * doc/invoke.texi (GCC Command Options): Document option arguments.
9897         Explain byte-size arguments and suffixes.
9898         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9899         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9900         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9901         * doc/options.texi (UInteger): Expand.
9902         (Host_Wide_Int, ByteSize): Document new properties.
9903         * final.c (final_start_function_1): Include sizes in an error message.
9904         * function.c (frame_offset_overflow): Same.
9905         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9906         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9907         Diagnose unbounded alloca calls only for limits of less than
9908         PTRDIFF_MAX.
9909         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
9910         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
9911         for alloca(0).
9912         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
9913         only for limits of less than PTRDIFF_MAX.
9914         * langhooks-def.h (lhd_handle_option): Change function argument
9915         to HOST_WIDE_INT.
9916         * langhooks.c (lhd_handle_option): Same.
9917         * langhooks.h (handle_option): Same.
9918         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9919         ByteSize flags.
9920         (var_type, var_type_struct): Same.
9921         (var_set): Handle ByteSize flag.
9922         * optc-gen.awk: Add comments to output to ease debugging.  Make
9923         use of HOST_WIDE_INT where appropriate.
9924         * opts-gen-save.awk:  Use %lx to format unsigned long.
9925         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9926         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9927         arguments.  Parse bytes-size suffixes.
9928         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9929         (enum_value_to_arg): Same.
9930         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
9931         (handle_option): Adjust.
9932         (generate_option): Change function argument to HOST_WIDE_INT.
9933         (cmdline_handle_error): Adjust.
9934         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9935         (set_option): Change function argument to HOST_WIDE_INT.
9936         (option_enabled): Handle cl_host_wide_int.
9937         (get_option_state): Handle CLVC_SIZE.
9938         (control_warning_option): Same.
9939         * opts.c (common_handle_option): Change function argument to
9940         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
9941         OPT_Wvla_larger_than_.
9942         * opts.h (enum cl_var_type): Add an enumerator.
9943         * stor-layout.c (layout_decl): Print a more meaningful warning.
9944         * toplev.c (output_stack_usage): Adjust.
9946 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
9948         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9949         call to inline_expand_builtin_string_cmp.
9950         (expand_builtin_strcmp): Likewise.
9951         (expand_builtin_strncmp): Likewise.
9952         (inline_string_cmp): Delete the last parameter, change char_type_node
9953         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9954         two operands.
9955         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9956         the inlining expansion on target where the type of the call has same or
9957         narrower precision than unsigned char.
9959 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9961         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9962         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9963         * common.opt (fsave-optimization-record): New option.
9964         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9965         * doc/invoke.texi (-fsave-optimization-record): New option.
9966         * dumpfile.c: Include "optinfo-emit-json.h".
9967         (struct kv_pair): Move to coretypes.h.
9968         (optgroup_options): Make non-static.
9969         (dump_context::end_scope): Call
9970         optimization_records_maybe_pop_dump_scope.
9971         * dumpfile.h (optgroup_options): New decl.
9972         * json.cc: New file.
9973         * json.h: New file.
9974         * optinfo-emit-json.cc: New file.
9975         * optinfo-emit-json.h: New file.
9976         * optinfo.cc: Include "optinfo-emit-json.h".
9977         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9978         (optinfo_enabled_p): Check optimization_records_enabled_p.
9979         (optinfo_wants_inlining_info_p): Likewise.
9980         * optinfo.h: Update comment.
9981         * profile-count.c (profile_quality_as_string): New function.
9982         * profile-count.h (profile_quality_as_string): New decl.
9983         (profile_count::quality): New accessor.
9984         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9985         and optinfo_emit_json_cc_tests.
9986         * selftest.h (selftest::json_cc_tests): New decl.
9987         (selftest::optinfo_emit_json_cc_tests): New decl.
9988         * toplev.c: Include "optinfo-emit-json.h".
9989         (compile_file): Call optimization_records_finish.
9990         (do_compile): Call optimization_records_start.
9991         * tree-ssa-live.c: Include optinfo.h.
9992         (remove_unused_scope_block_p): Retain inlining information if
9993         optinfo_wants_inlining_info_p returns true.
9995 2018-07-20  Richard Biener  <rguenther@suse.de>
9997         PR debug/86585
9998         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9999         to cover -flto-partition=none.
10001 2018-07-20  Martin Liska  <mliska@suse.cz>
10003         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
10004         (get_decl_source_range): Remove unused function.
10006 2018-07-20  Richard Biener  <rguenther@suse.de>
10008         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
10009         (struct vn_phi_s): Likewise.
10010         (struct vn_reference_s): Likewise.
10011         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
10012         for searching the slot of an entry known to be in the hash itself.
10013         (vn_phi_hasher::equal): Likewise.
10014         (vn_reference_hasher::equal): Likewise.
10015         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
10016         globals.
10017         (optimistic_info, current_info): Remove, keeping only valid_info.
10018         (vn_reference_lookup_1): Remove fallback lookup.
10019         (vn_reference_lookup_2): Likewise.
10020         (vn_nary_op_lookup_1): Likewise.
10021         (vn_phi_lookup): Likewise.
10022         (vn_nary_build_or_lookup_1): Make sure to not chain the built
10023         hash element.
10024         (vn_reference_insert): Adjust, chain the inserted hash element
10025         at last_inserted_ref.
10026         (vn_reference_insert_pieces): Likewise.
10027         (visit_reference_op_call): Likewise.
10028         (vn_nary_op_insert_into): Chain the inserted hash element at
10029         last_inserted_nary.
10030         (vn_nary_op_insert_pieces): Adjust.
10031         (vn_nary_op_insert): Likewise.
10032         (vn_nary_op_insert_stmt): Likewise.
10033         (vn_phi_insert): Adjust, chain the inserted hash element at
10034         last_inserted_phi.
10035         (process_scc): Remove clearing and copying the optimistic
10036         table.  Instead remove elements inserted during an optimistic
10037         iteration from the single table we maintain.
10038         (init_scc_vn): Adjust.
10039         (free_scc_vn): Likewise.
10040         (sccvn_dom_walker::record_cond): Likewise.
10041         (sccvn_dom_walker::after_dom_children): Likewise.
10043 2018-07-19  Martin Sebor  <msebor@redhat.com>
10045         PR tree-optimization/84047
10046         PR tree-optimization/83776
10047         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
10048         (check_array_bounds): Call it.
10050 2018-07-19  Martin Sebor  <msebor@redhat.com>
10052         * align.h (align_flags): Use member initialization.
10054 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
10056         * Makefile.in (OBJS): Add optinfo.o.
10057         * coretypes.h (class symtab_node): New forward decl.
10058         (struct cgraph_node): New forward decl.
10059         (class varpool_node): New forward decl.
10060         * dump-context.h: New file.
10061         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
10062         "tree-pass.h".
10063         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
10064         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
10065         (set_alt_dump_file): Likewise.
10066         (dump_context::~dump_context): New dtor.
10067         (dump_gimple_stmt): Move implementation to...
10068         (dump_context::dump_gimple_stmt): ...this new member function.
10069         Add the stmt to any pending optinfo, creating one if need be.
10070         (dump_gimple_stmt_loc): Move implementation to...
10071         (dump_context::dump_gimple_stmt_loc): ...this new member function.
10072         Start a new optinfo and add the stmt to it.
10073         (dump_gimple_expr): Move implementation to...
10074         (dump_context::dump_gimple_expr): ...this new member function.
10075         Add the stmt to any pending optinfo, creating one if need be.
10076         (dump_gimple_expr_loc): Move implementation to...
10077         (dump_context::dump_gimple_expr_loc): ...this new member function.
10078         Start a new optinfo and add the stmt to it.
10079         (dump_generic_expr): Move implementation to...
10080         (dump_context::dump_generic_expr): ...this new member function.
10081         Add the tree to any pending optinfo, creating one if need be.
10082         (dump_generic_expr_loc): Move implementation to...
10083         (dump_context::dump_generic_expr_loc): ...this new member
10084         function.  Add the tree to any pending optinfo, creating one if
10085         need be.
10086         (dump_printf): Move implementation to...
10087         (dump_context::dump_printf_va): ...this new member function.  Add
10088         the text to any pending optinfo, creating one if need be.
10089         (dump_printf_loc): Move implementation to...
10090         (dump_context::dump_printf_loc_va): ...this new member function.
10091         Start a new optinfo and add the stmt to it.
10092         (dump_dec): Move implementation to...
10093         (dump_context::dump_dec): ...this new member function.  Add the
10094         value to any pending optinfo, creating one if need be.
10095         (dump_context::dump_symtab_node): New member function.
10096         (dump_context::get_scope_depth): New member function.
10097         (dump_context::begin_scope): New member function.
10098         (dump_context::end_scope): New member function.
10099         (dump_context::ensure_pending_optinfo): New member function.
10100         (dump_context::begin_next_optinfo): New member function.
10101         (dump_context::end_any_optinfo): New member function.
10102         (dump_context::s_current): New global.
10103         (dump_context::s_default): New global.
10104         (dump_scope_depth): Delete global.
10105         (dumpfile_ensure_any_optinfo_are_flushed): New function.
10106         (dump_symtab_node): New function.
10107         (get_dump_scope_depth): Reimplement in terms of dump_context.
10108         (dump_begin_scope): Likewise.
10109         (dump_end_scope): Likewise.
10110         (selftest::temp_dump_context::temp_dump_context): New ctor.
10111         (selftest::temp_dump_context::~temp_dump_context): New dtor.
10112         (selftest::verify_item): New function.
10113         (ASSERT_IS_TEXT): New macro.
10114         (ASSERT_IS_TREE): New macro.
10115         (ASSERT_IS_GIMPLE): New macro.
10116         (selftest::test_capture_of_dump_calls): New test.
10117         (selftest::dumpfile_c_tests): Call it.
10118         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
10119         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
10120         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
10121         descriptive comment.
10122         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
10123         (dump_node, dump_bb): Move these unrelated decls.
10124         (class dump_manager): Add leading comment.
10125         * optinfo.cc: New file.
10126         * optinfo.h: New file.
10128 2018-07-19  Michael Collison  <michael.collison@arm.com>
10129             Richard Henderson <rth@redhat.com>
10131         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
10132         (subti): Handle op1 zero.
10133         (subvti4, usub4ti4): New.
10134         (*sub<GPI>3_compare1_imm): New.
10135         (sub<GPI>3_carryinCV): New.
10136         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
10137         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
10139 2018-07-19  Michael Collison  <michael.collison@arm.com>
10140             Richard Henderson <rth@redhat.com>
10142         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
10143         (addti3): Create simpler code if low part is already known to be 0.
10144         (addvti4, uaddvti4): New.
10145         (*add<GPI>3_compareC_cconly_imm): New.
10146         (*add<GPI>3_compareC_cconly): New.
10147         (*add<GPI>3_compareC_imm): New.
10148         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
10149         handle constants within this pattern..
10150         (*add<GPI>3_compareV_cconly_imm): New.
10151         (*add<GPI>3_compareV_cconly): New.
10152         (*add<GPI>3_compareV_imm): New.
10153         (add<GPI>3_compareV): New.
10154         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
10155         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
10156         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
10157         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
10158         with 'comparison' operator.
10159         (*add<GPI>3_compareV_cconly_imm): Ditto.
10160         (*add<GPI>3_compareV_cconly): Ditto.
10161         (*add<GPI>3_compareV_imm): Ditto.
10162         (add<GPI>3_compareV): Ditto.
10163         (add<mode>3_carryinC): Ditto.
10164         (*add<mode>3_carryinC_zero): Ditto.
10165         (*add<mode>3_carryinC): Ditto.
10166         (add<mode>3_carryinV): Ditto.
10167         (*add<mode>3_carryinV_zero): Ditto.
10168         (*add<mode>3_carryinV): Ditto.
10170 2018-07-19  Michael Collison  <michael.collison@arm.com>
10171             Richard Henderson <rth@redhat.com>
10173         * config/aarch64/aarch64-modes.def (CC_V): New.
10174         * config/aarch64/aarch64-protos.h
10175         (aarch64_addti_scratch_regs): Declare
10176         (aarch64_subvti_scratch_regs): Declare.
10177         (aarch64_expand_subvti): Declare.
10178         (aarch64_gen_unlikely_cbranch): Declare
10179         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
10180         for signed overflow using CC_Vmode.
10181         (aarch64_get_condition_code_1): Handle CC_Vmode.
10182         (aarch64_gen_unlikely_cbranch): New function.
10183         (aarch64_addti_scratch_regs): New function.
10184         (aarch64_subvti_scratch_regs): New function.
10185         (aarch64_expand_subvti): New function.
10187 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10189         * config/aarch64/aarch64-option-extensions.def: New entry for profile
10190         extension.
10191         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
10192         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
10193         extension.
10195 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10197         PR target/83009
10198         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10199         address check not strict.
10201 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10203         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
10204         Umq with Umn.
10205         (store_pair_lanes<mode>): Likewise.
10206         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
10207         enum value 'ADDR_QUERY_LDP_STP_N'.
10208         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
10209         (aarch64_print_address_internal): Add declaration.
10210         (aarch64_print_ldpstp_address): Remove.
10211         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
10212         (aarch64_print_operand): Change printing of 'y'.
10213         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
10214         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
10215         'true' rather than '1'.
10216         * config/aarch64/constraints.md (Uml): Likewise.
10217         (Uml): Rename to Umn.
10218         (Umq): Remove.
10220 2018-07-19  Richard Biener  <rguenther@suse.de>
10222         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
10223         a trailing array.
10224         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
10225         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
10226         (vn_reference_hasher): Likewise.
10227         (struct vn_tables_s): Remove obstack and alloc-pool members.
10228         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
10229         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
10230         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
10231         (vn_reference_insert_pieces): Likewise.
10232         (alloc_vn_nary_op_noinit): Adjust.
10233         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
10234         (vn_phi_eq): Adjust.
10235         (shared_lookup_phiargs): Remove.
10236         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
10237         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
10238         (visit_reference_op_call): Likewise.
10239         (copy_nary, copy_phi, copy_reference): Remove.
10240         (process_scc): Rewind the obstack when iterating.  Do not
10241         copy the elements to valid_info but just move them from one
10242         hashtable to the other.
10243         (allocate_vn_table): Adjust.
10244         (free_vn_table): Likewise.
10245         (init_scc_vn): Likewise.
10246         (free_scc_vn): Likewise.
10248 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
10250         PR target/86560
10251         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
10252         indirect_return as function type attribute.
10253         (ix86_attribute_table): Change indirect_return to function
10254         type attribute.
10255         * doc/extend.texi: Update indirect_return attribute.
10257 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
10259         * wide-int.h (widest2_int): New.
10260         * gimple-fold.c (arith_overflowed_p): Use it.
10261         * tree.h (widest2_int_cst): New.
10262         * tree-vrp.c (wide_int_binop_overflow): Rename from
10263         vrp_int_const_binop.
10264         Rewrite to work on trees.
10265         (extract_range_from_multiplicative_op_1): Abstract code to...
10266         (wide_int_range_min_max): ...here.
10267         (wide_int_range_cross_product): ...and here.
10268         (extract_range_from_binary_expr_1): Abstract overflow code to...
10269         (wide_int_range_mult_wrapping): ...here.
10270         * tree-vrp.h (wide_int_range_cross_product): New.
10271         (wide_int_range_mult_wrapping): New.
10273 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
10274             Julia Koval  <julia.koval@intel.com>
10276         * config/i386/x86-tune-costs.h (skylake_memcpy,
10277         skylake_memset): Replace rep_prefix with unrolling for size 512.
10279 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
10281         PR middle-end/86544
10282         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
10283         comparision with EQ_EXPR in last stmt.
10285 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10287         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
10288         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
10289         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
10290         previously known as "PowerPC AltiVec Built-in Functions".  Move
10291         some material to new subsubsections "PowerPC AltiVec Built-in
10292         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
10293         ISA 2.07".
10294         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
10295         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
10296         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
10297         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
10299 2018-07-18  Richard Biener  <rguenther@suse.de>
10301         PR tree-optimization/86557
10302         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
10303         EXACT_DIV_EXPR.
10305 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
10307         * config/s390/s390.c (s390_function_profiler): Generate CFI.
10309 2018-07-17  Jeff Law  <law@redhat.com>
10311         * config/arm/arm.c (get_label_padding): Update for recent
10312         changes to label_to_alignment.
10314         PR tree-optimization/86010
10315         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
10317         * config/mips/mips.c (vr4130_align_insns): Update for recent
10318         changes to label_to_alignment.
10320         * config/frv/frv.c (frv_label_align): Update for recent changes
10321         to label_to_alignment.
10323         * config/nios2/nios2.c (nios2_label_align): Update for recent
10324         changes which dropped ALIGN_LABELS_LOG.
10326 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
10328         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
10329         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
10331 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
10333         * config/arc/arc.c (arc_label_align): Use align_labels instead of
10334         deprecated align_labels_log.
10336 2018-07-17  Richard Biener  <rguenther@suse.de>
10338         PR lto/86456
10339         * dwarf2out.c (init_sections_and_labels): Always generate
10340         a debug_line_str_section for early LTO debug.
10341         (dwarf2out_finish): Reset debug_line_str_hash output early.
10342         Bump counter for extra dwarf5 .debug_loc labels to not conflict
10343         with fat LTO part.
10344         (dwarf2out_early_finish): Output debug_line_str.
10346 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10348         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
10349         index register on z196 or later.
10351 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
10353         * config/s390/s390.c (s390_default_align): Set default function
10354         alignment to 16.
10355         (s390_override_options_after_change): Call s390_default align.
10356         (s390_option_override_internal): Call s390_default_align.
10357         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
10359 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10361         PR middle-end/86542
10362         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
10363         to _looptemp_ clauses, other than the first two.
10365 2018-07-17  Martin Liska  <mliska@suse.cz>
10367         * opts.c: Do not enable OPT_falign_* for -Os.
10369 2018-07-17  Martin Liska  <mliska@suse.cz>
10371         * align.h (MAX_CODE_ALIGN): New.
10372         (MAX_CODE_ALIGN_VALUE): New.
10373         * common/config/i386/i386-common.c (ix86_handle_option):
10374         (MAX_CODE_ALIGN): Moved to align.h.
10375         * final.c (MAX_CODE_ALIGN): Likewise.
10376         * opts.c (parse_and_check_align_values):
10377         (MAX_CODE_ALIGN): Likewise.
10378         (MAX_CODE_ALIGN_VALUE): Likewise.
10380 2018-07-17  Martin Liska  <mliska@suse.cz>
10382         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
10383         in order to fulfil coding style.
10384         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
10385         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10386         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10387         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
10388         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
10389         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
10390         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
10391         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
10392         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10394 2018-07-17  Martin Liska  <mliska@suse.cz>
10396         * align.h: New file.
10397         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
10398         directly.
10399         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
10400         align_flags of label_to_alignment.
10401         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
10402         align_flags class.
10403         * config/m68k/m68k.c: Do not use removed align_labels_value and
10404         align_loops_value.
10405         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
10406         (LOOP_ALIGN): Likewise.
10407         (LABEL_ALIGN): Likewise.
10408         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
10409         Remove not used macro.
10410         (rs6000_loop_align): Change return type to align_flags.
10411         (rs6000_loop_align_max_skip): Remove.
10412         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
10413         Change return type to align_flags.
10414         * config/rs6000/rs6000.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/rx/rx.h (JUMP_ALIGN): Wrap integer values
10419         * config/rx/rx-protos.h (rx_align_for_label): Make it
10420         static function.
10421         * config/rx/rx.c (rx_align_for_label): Change return type
10422         to align_flags.
10423         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
10424         macro definitions.
10425         into align_flags class.
10426         (LABEL_ALIGN): Likewise.
10427         (LOOP_ALIGN): Likewise.
10428         * config/s390/s390.c (s390_label_align): Use align_flags
10429         class member.
10430         (s390_asm_output_function_label): Likewise.
10431         * config/sh/sh.c (sh_override_options_after_change):
10432         Use align_flags class directly without macros.
10433         (find_barrier): Likewise.
10434         (barrier_align): Likewise.
10435         (sh_loop_align): Likewise.
10436         * config/spu/spu.c (spu_option_override):
10437         Use align_flags_tuple::get_value instead of removed macros.
10438         (spu_sched_init): Likewise.
10439         * config/spu/spu.h (GTY): Likewise.
10440         * config/visium/visium.c (visium_option_override):
10441         Set "8" as default secondary alignment.
10442         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
10443         in order to guarantee secondary alignment of 8.
10444         * coretypes.h: Include align.h header file.
10445         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
10446         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
10447         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
10448         * doc/tm.texi.in: Likewise.
10449         * final.c (struct label_alignment): Remove not used structure.
10450         (LABEL_ALIGN): Change type to align_flags.
10451         (LOOP_ALIGN): Likewise.
10452         (JUMP_ALIGN): Likewise.
10453         (default_loop_align_max_skip): Remove.
10454         (default_label_align_max_skip): Likewise.
10455         (default_jump_align_max_skip): Likewise.
10456         (default_label_align_after_barrier_max_skip):
10457         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
10458         (LABEL_TO_MAX_SKIP): Remove.
10459         (label_to_alignment): Return align_flags type instead of integer.
10460         (label_to_max_skip): Remove.
10461         (align_fuzz): Use align_flags type.
10462         (compute_alignments): Use align_flags type and use align_flags::max
10463         to combine multiple alignments.
10464         (grow_label_align): Grow vec instead of C array.
10465         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
10466         (shorten_branches):  Use align_flags type and use align_flags::max
10467         to combine multiple alignments.
10468         (final_scan_insn_1): Remove usage of secondary alignment that comes
10469         from label alignment, but instead use proper secondary alignment
10470         which is computed in grow_label_align.
10471         * flags.h (struct align_flags_tuple): Move to align.h.
10472         (struct align_flags): Likewise.
10473         (state_align_loops): Rename to align_loops.
10474         (state_align_jumps): Rename to align_jumps.
10475         (state_align_labels): Rename to align_labels.
10476         (state_align_functions): Rename to align_functions.
10477         (align_loops_log): Remove.
10478         (align_jumps_log): Remove.
10479         (align_labels_log): Remove.
10480         (align_functions_log): Remove.
10481         (align_loops_max_skip): Remove.
10482         (align_jumps_max_skip): Remove.
10483         (align_labels_max_skip): Remove.
10484         (align_functions_max_skip): Remove.
10485         (align_loops_value): Remove.
10486         (align_jumps_value): Remove.
10487         (align_labels_value): Remove.
10488         (align_functions_value): Remove.
10489         * output.h (label_to_alignment): Change return type to align_flags.
10490         (label_to_max_skip): Remove.
10491         * target.def: Remove loop_align_max_skip, label_align_max_skip,
10492         jump_align_max_skip macros.
10493         * targhooks.h (default_loop_align_max_skip): Remove.
10494         (default_label_align_max_skip): Likewise.
10495         (default_jump_align_max_skip): Likewise.
10496         (default_label_align_after_barrier_max_skip): Remove.
10497         * toplev.c (read_log_maxskip): Use ::normalize function.
10498         (parse_N_M): Remove not used argument and also call ::normalize.
10499         (parse_alignment_opts): Do not pass unused arguments.
10500         * varasm.c (assemble_start_function): Use directly align_functions
10501         instead of removed macros.
10502         * system.h: Do not poison removed macros.
10504 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
10506         PR middle-end/86539
10507         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
10508         and cond temporaries don't have reference type if iterator has
10509         pointer type.  For init use &for_pre_body instead of pre_p if
10510         for_pre_body is non-empty.
10512 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10514         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
10515         double-double modes to SFmode directly directly.
10516         (trunc<mode>sf2_fprs): Delete.
10518 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10520         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
10521         for conversions between IFmode and the decimal floating point modes.
10522         (init_float128_ieee): Use the correct names for conversions between
10523         KFmode and the decimal floating point modes.
10525 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
10527         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
10528         for the conversions between TDmode and IFmode.
10529         (init_float128_ieee): Use more correct names for the conversions
10530         between TDmode and KFmode.
10532 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
10534         PR tree-optimization/86526
10535         * builtins.c (expand_builtin_memcmp): Formatting fixes.
10536         (inline_expand_builtin_string_cmp): Likewise.
10537         (inline_string_cmp): Likewise.  Use c_readstr instead of
10538         builtin_memcpy_read_str.  Add unit_mode temporary.
10540 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10542         PR middle-end/86528
10543         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
10544         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
10546 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10548         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
10549         Alphabetize prototypes of built-in functions, separating out
10550         built-in functions that are listed in this section but should be
10551         described elsewhere.
10553 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
10555         PR target/86511
10556         * expmed.c (emit_store_flag): Do not emit setcc followed by a
10557         conditional move when trapping comparison was split to a
10558         non-trapping one (and vice versa).
10560 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10562         * config/s390/s390.c (s390_function_profiler): Generate nops
10563         instead of profiler call sequences.
10564         * config/s390/s390.opt: Add the new option.
10566 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10568         * config/s390/s390.c (s390_function_profiler): Generate
10569         __mcount_loc section.
10570         * config/s390/s390.opt: Add the new option.
10572 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10574         * common.opt: Add the new warning.
10575         * config/s390/s390.c (s390_function_profiler): Emit "brasl
10576         %r0,__fentry__" when -mfentry is specified.
10577         (s390_option_override_internal): Disallow -mfentry for 31-bit
10578         CPUs.
10579         * config/s390/s390.opt: Add the new option.
10581 2018-07-16  Richard Biener  <rguenther@suse.de>
10583         PR lto/86523
10584         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
10585         for function-local FUNCTION_DECL and RESULT_DECL immediately.
10587 2018-07-16  Martin Liska  <mliska@suse.cz>
10589         PR ipa/86529
10590         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
10591         to ::get_create.
10593 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10595         * config/arc/arcHS.md: Update ARCHS scheduling rules.
10597 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10599         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
10600         for ARCHS4x.
10601         * config/arc/arc-cpus.def (hs4x): New cpu.
10602         (hs4xd): Likewise.
10603         * config/arc/arc-tables.opt: Regenerate.
10604         * config/arc/arc.c (arc_sched_issue_rate): New function.
10605         (TARGET_SCHED_ISSUE_RATE): Define.
10606         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
10607         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
10608         fpu_cvt.
10609         (attr tune): Add ARCHS4x tune values.
10610         (attr tune_dspmpy): Define.
10611         (*tst): Correct instruction type.
10612         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
10613         * config/arc/arcHS4x.md: New file.
10614         * config/arc/fpu.md: Update instruction type attributes.
10615         * config/arc/t-multilib: Regenerate.
10617 2018-07-16  Tom de Vries  <tdevries@suse.de>
10619         PR debug/86455
10620         * var-tracking.c (vt_initialize): Fix pre_dec handling.
10622 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
10624         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10625         early clobber.
10627 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
10629         PR tree-optimization/86514
10630         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10631         conversion to a boolean type from a type with greater precision.
10633 2018-07-16  Tom de Vries  <tdevries@suse.de>
10635         * var-tracking.c (vt_initialize): Print adjusted insn slim if
10636         dump_flags request TDF_SLIM.
10638 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
10640         * fold-const.c (int_const_binop_1): Abstract...
10641         (wide_int_binop): ...wide int code here.
10642         (poly_int_binop): ...poly int code here.
10643         Abstract the rest of int_const_binop_1 into int_const_binop.
10644         * fold-const.h (wide_int_binop): New.
10645         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10646         Remove useless PLUS/MINUS_EXPR case.
10647         (zero_nonzero_bits_from_vr): Move wide int code...
10648         (zero_nonzero_bits_from_bounds): ...here.
10649         (extract_range_from_binary_expr_1): Move mask optimization code...
10650         (range_easy_mask_min_max): ...here.
10651         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10652         (range_easy_mask_min_max): New.
10654 2018-07-15  Jeff Law  <law@redhat.com>
10656         PR target/85993
10657         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10658         block.
10660 2018-07-14  Jim Wilson  <jimw@sifive.com>
10662         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10664 2018-07-14  Paul Koning  <ni1d@arrl.net>
10666         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10668 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
10670         * lto-streamer-out.c (copy_function_or_variable): Dump info about
10671         copying section.
10673 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
10674             Steve Munroe  <munroesj52@gmail.com>
10676         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10677         (_mm_andnot_si128): Likewise.
10678         (_mm_or_si128): Likewise.
10679         (_mm_xor_si128): Likewise.
10681 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
10683         PR middle-end/78809
10684         * builtins.c (expand_builtin_memcmp): Inline the calls first
10685         when result_eq is false.
10686         (expand_builtin_strcmp): Inline the calls first.
10687         (expand_builtin_strncmp): Likewise.
10688         (inline_string_cmp): New routine. Expand a string compare
10689         call by using a sequence of char comparison.
10690         (inline_expand_builtin_string_cmp): New routine. Inline expansion
10691         a call to str(n)cmp/memcmp.
10692         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10693         New option.
10694         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10696 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
10698         * config/arm/driver-arm.c: Include arm-native.h.
10699         (host_detect_local_cpu): Use auto-generated data tables.
10700         (vendors, arm_cpu_table): Delete.  Move part information to ...
10701         * config/arm/arm-cpus.in: ... here.
10702         * config/arm/parsecpu.awk (gen_native): New function.
10703         (vendor, part): New CPU fields.
10704         (END): Add support for building the native CPU detection tables.
10705         * config/arm/t-arm (arm-native.h): Add build rule.
10706         (driver-arm.o): Add dependency on arm-native.h.
10708 2018-07-13  Richard Biener  <rguenther@suse.de>
10710         PR middle-end/85974
10711         * match.pd (addr1 - addr2): Allow either of the operand to
10712         have a conversion.
10714 2018-07-13  Tom de Vries  <tdevries@suse.de>
10716         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10717         in remap_ssa_name.
10719 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
10721         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10722         arrays instead of numbered variables.
10724 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
10726         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10727         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10729 2018-07-13  Richard Biener  <rguenther@suse.de>
10731         PR debug/86452
10732         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10733         instead of get_context_die.
10735 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
10736             Richard Biener  <rguenther@suse.de>
10738         PR middle-end/86489
10739         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10740         that the loop latch destination where phi is defined.
10742 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
10744         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10745         (riscv_expand_epilogue): Add assertion to check interrupt mode.
10746         (riscv_set_current_function): Extract getting interrupt type to new
10747         function.
10748         (riscv_get_interrupt_type): New function.
10749         (riscv_merge_decl_attributes): New function, checking interrupt type is
10750         same.
10751         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10753 2018-07-12  Paul Koning  <ni1d@arrl.net>
10755         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10756         directive.
10758 2018-07-12  Paul Koning  <ni1d@arrl.net>
10760         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10761         zero reference, add doloop_end instead.
10762         * doc/md.texi (decrement_and_branch_until_zero): Remove.
10763         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
10764         detail for doloop_end.
10766 2018-07-12  Martin Sebor  <msebor@redhat.com>
10768         PR c/86453
10769         * attribs.c (decl_attributes): Reject conflicting attributes before
10770         calling attribute handlers.
10772 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10774         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10775          parameter.
10776         (gcc::dump_manager::get_dump_file_name): likewise.
10777         (dump_begin): Likewise.
10778         * dumpfile.h (dump_begin): Update prototype.
10779         (gcc::dump_manager::get_dump_file_name,
10780         gcc::dump_manager::get_dump_file_name): Update prototype.
10782 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10784         * internal-fn.h (vectorizable_internal_fn_p): New function.
10785         * tree-vect-slp.c (compatible_calls_p): Likewise.
10786         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
10787         to internal functions.
10788         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10790 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10792         * fold-const.h (inverse_conditions_p): Declare.
10793         * fold-const.c (inverse_conditions_p): New function.
10794         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
10795         that test the inverse condition of a conditional internal function.
10796         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10797         * internal-fn.c (internal_fn_mask_index): Handle conditional
10798         internal functions.
10799         (vectorized_internal_fn_supported_p): New function.
10800         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10801         (any_pred_load_store): Replace with...
10802         (need_to_predicate): ...this new variable.
10803         (redundant_ssa_names): New variable.
10804         (ifcvt_can_use_mask_load_store): Move initial checks to...
10805         (ifcvt_can_predicate): ...this new function.  Handle tree codes
10806         for which a conditional internal function exists.
10807         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10808         instead of ifcvt_can_use_mask_load_store.  Update after variable
10809         name change.
10810         (predicate_load_or_store): New function, split out from
10811         predicate_mem_writes.
10812         (check_redundant_cond_expr): New function.
10813         (value_available_p): Likewise.
10814         (predicate_rhs_code): Likewise.
10815         (predicate_mem_writes): Rename to...
10816         (predicate_statements): ...this.  Use predicate_load_or_store
10817         and predicate_rhs_code.
10818         (combine_blocks, tree_if_conversion): Update after above name changes.
10819         (ifcvt_local_dce): Handle redundant_ssa_names.
10820         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10821         general conditional functions.
10822         * tree-vect-stmts.c (vectorizable_call): Likewise.
10824 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10825             Alan Hayward  <alan.hayward@arm.com>
10826             David Sherwood  <david.sherwood@arm.com>
10828         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10829         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10830         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10831         plus and minus and convert them into IFN_COND_FMA-based sequences.
10832         (convert_mult_to_fma): Handle conditional plus and minus.
10834 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10836         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10837         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10838         (cond_fnms_optab): New optabs.
10839         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10840         internal functions.
10841         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10842         * internal-fn.h (get_conditional_internal_fn): Declare.
10843         (get_unconditional_internal_fn): Likewise.
10844         * internal-fn.c (cond_ternary_direct): New macro.
10845         (expand_cond_ternary_optab_fn): Likewise.
10846         (direct_cond_ternary_optab_supported_p): Likewise.
10847         (FOR_EACH_COND_FN_PAIR): Likewise.
10848         (get_conditional_internal_fn): New function.
10849         (get_unconditional_internal_fn): Likewise.
10850         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10851         (gimple_match_op::gimple_match_op): Add a new overload for 5
10852         operands.
10853         (gimple_match_op::set_op): Likewise.
10854         (gimple_resimplify5): Declare.
10855         * genmatch.c (decision_tree::gen): Generate simplifications for
10856         5 operands.
10857         * gimple-match-head.c (gimple_simplify): Define an overload for
10858         5 operands.  Handle calls with 5 arguments in the top-level overload.
10859         (convert_conditional_op): Handle conversions from unconditional
10860         internal functions to conditional ones.
10861         (gimple_resimplify5): New function.
10862         (build_call_internal): Pass a fifth operand.
10863         (maybe_push_res_to_seq): Likewise.
10864         (try_conditional_simplification): Try converting conditional
10865         internal functions to unconditional internal functions.
10866         Handle 3-operand unconditional forms.
10867         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10868         Define ternary equivalents of the current rules for binary conditional
10869         internal functions.
10870         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10871         ternary operations.
10872         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10873         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10874         (optab): Handle them.
10875         (SVE_COND_FP_TERNARY): New int iterator.
10876         (sve_fmla_op, sve_fmad_op): New int attributes.
10877         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10878         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10879         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10881 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10883         * target.def (preferred_else_value): New target hook.
10884         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10885         * doc/tm.texi: Regenerate.
10886         * targhooks.h (default_preferred_else_value): Declare.
10887         * targhooks.c (default_preferred_else_value): New function.
10888         * internal-fn.h (conditional_internal_fn_code): Declare.
10889         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10890         (get_conditional_internal_fn): Use it.
10891         (conditional_internal_fn_code): New function.
10892         * gimple-match.h (gimple_match_cond): New struct.
10893         (gimple_match_op): Add a cond member function.
10894         (gimple_match_op::gimple_match_op): Update all forms to take a
10895         gimple_match_cond.
10896         * genmatch.c (expr::gen_transform): Use the same condition as res_op
10897         for the suboperation, but don't specify a particular else_value.
10898         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10899         (visit_nary_op, visit_reference_op_load): Pass
10900         gimple_match_cond::UNCOND to the gimple_match_op constructor.
10901         * gimple-match-head.c: Include tree-eh.h
10902         (convert_conditional_op): New function.
10903         (maybe_resimplify_conditional_op): Likewise.
10904         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10905         (gimple_resimplify2): Likewise.
10906         (gimple_resimplify3): Likewise.
10907         (gimple_resimplify4): Likewise.
10908         (maybe_push_res_to_seq): Return null for conditional operations.
10909         (try_conditional_simplification): New function.
10910         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
10911         constructor.
10912         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10913         IFN_COND_* call.
10914         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10915         function.
10916         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10918 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10920         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10921         DECL_FCONTEXT
10922         (hash_tree): Do not hash DECL_FCONTEXT
10923         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10924         Do not stream DECL_FCONTEXT.
10925         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10926         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10928 2018-07-12  Richard Biener  <rguenther@suse.de>
10930         PR debug/86462
10931         * dwarf2out.c (gen_block_die): Only output blocks when they have
10932         at least one !DECL_IGNORED_P variable.
10934 2018-07-12  Richard Biener  <rguenther@suse.de>
10936         PR target/84829
10937         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10938         Remove -mieee-fp handling.
10940 2018-07-12  Richard Biener  <rguenther@suse.de>
10942         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10943         left-over from last patch.
10945 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
10947         PR tree-optimization/86492
10948         * gimple-ssa-store-merging.c
10949         (imm_store_chain_info::coalesce_immediate_stores): Call
10950         check_no_overlap even for the merge_overlapping case.  Formatting fix.
10952 2018-07-12  Richard Biener  <rguenther@suse.de>
10954         PR middle-end/86479
10955         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10956         move possibly trapping operations into the conditional.
10958 2018-07-12  Richard Biener  <rguenther@suse.de>
10960         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10961         (vn_lookup_simplify_result): Remove recursion limit applied
10962         here.
10963         (vn_nary_build_or_lookup_1): Adjust.
10964         (try_to_simplify): Likewise.
10965         * gimple-match-head.c (gimple_resimplify1): Instead apply one
10966         here.
10967         (gimple_resimplify2): Likewise.
10968         (gimple_resimplify3): Likewise.
10969         (gimple_resimplify4): Likewise.
10971 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
10973         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10974         Use __mmask64 type instead of __mmask8 for __M argument.
10975         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10976         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10977         __U argument.
10978         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10979         __mmask16 for __M argument.
10980         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10981         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10982         to __mmask16 instead of __mmask8.
10983         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10984         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10985         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10986         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10987         instead of __mmask16 for __U argument.
10988         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10989         __mmask16 instead of __mmask8 for __U argument.
10990         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10991         __U argument.
10992         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10993         __mmask16.
10994         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10995         argument.
10996         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10997         __U argument.
10998         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10999         __mmask16.
11000         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
11001         of __mmask16.
11002         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
11003         __U argument.
11004         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
11005         __U argument.
11006         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
11007         __U argument.
11008         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
11009         __U argument.
11010         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
11011         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
11012         return type as well as __M argument type and all casts from __mmask8
11013         to __mmask32.
11014         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
11015         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
11016         return type as well as __M argument type and all casts from __mmask8
11017         to __mmask16.
11018         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
11019         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
11020         return type as well as __M argument type and all casts from __mmask8
11021         to __mmask32.
11022         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
11023         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
11024         return type as well as __M argument type and all casts from __mmask8
11025         to __mmask16.
11026         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
11027         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
11028         __mmask16.
11030 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
11032         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
11033         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
11034         for __U argument.
11036 2018-07-11  Paul Koning  <ni1d@arrl.net>
11038         * doc/md.texi (define_subst): Document how multiple occurrences of
11039         the same argument in the replacement pattern are handled.
11041 2018-07-11  Paul Koning  <ni1d@arrl.net>
11043         * doc/extend.texi (Common Variable Attributes): Move "mode" into
11044         alphabetical order.
11045         (Common Type Attributes): Add "mode" attribute.
11047 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
11049         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
11050         stream DECL_ORIGINAL_TYPE.
11051         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
11052         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
11053         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
11054         Do not walk original type.
11055         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
11056         external decls.
11057         (write_ts_decl_non_common_tree_pointers): Do not stream
11058         DECL_ORIGINAL_TYPE
11059         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
11060         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
11062 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
11064         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
11065         thread twice from the same starting edge.
11067 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
11069         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
11070         * gimple.c (gimple_call_nonnull_result_p): ...here...
11071         (gimple_call_nonnull_arg): ...and here.
11072         * gimple.h (gimple_call_nonnull_result_p): New.
11073         (gimple_call_nonnull_arg): New.
11075 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
11077         * config/arm/arm-cpus.in: Move information from fpu field of each
11078         cpu definition to the isa field.
11079         * config/arm/parsecpu.awk (fpu): Delete match rule.
11080         (gen_comm_data): Don't add bits from the CPU's FPU entry.
11082 2018-07-11  Richard Biener  <rguenther@suse.de>
11084         PR debug/86457
11085         * dwarf2out.c (init_sections_and_labels): Use
11086         output_asm_line_debug_info consistently.
11087         (dwarf2out_early_finish): Likewise.
11088         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
11089         type units.
11091 2018-07-11  Richard Biener  <rguenther@suse.de>
11093         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
11094         Rework father_bb setting in a way to avoid propagating constants
11095         multiple times on a loop body.
11097 2018-07-10  Mark Wielaard  <mark@klomp.org>
11099         PR debug/86459
11100         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
11102 2018-07-10  Richard Biener  <rguenther@suse.de>
11104         * hash-map.h (hash_map::iterator::operator*): Return
11105         references to key and value.
11107 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
11109         PR c++/86443
11110         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
11111         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
11112         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
11113         (gimplify_omp_for): For composite loops, move outer
11114         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
11115         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
11116         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
11117         TREE_LIST for both the original class iterator and the "last" helper
11118         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
11119         loop, remember has_decl_expr from outer composite loops for the
11120         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
11122 2018-07-09  Martin Sebor  <msebor@redhat.com>
11124         PR middle-end/77357
11125         PR middle-end/86428
11126         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
11127         accessing implicitly initialized array elements.
11128         * expr.c (string_constant): Handle string initializers of
11129         character arrays within aggregates.
11130         * gimple-fold.c (fold_array_ctor_reference): Add argument.
11131         Store element offset.  As a special case, handle zero size.
11132         (fold_nonarray_ctor_reference): Same.
11133         (fold_ctor_reference): Add argument.  Store subobject offset.
11134         * gimple-fold.h (fold_ctor_reference): Add argument.
11136 2018-07-09  Paul Koning  <ni1d@arrl.net>
11138         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
11139         (pdp11_insn_cost): New function.
11140         (pdp11_md_asm_adjust): New function.
11141         (TARGET_INVALID_WITHIN_DOLOOP): Define.
11142         (pdp11_rtx_costs): Update to match machine better.
11143         (output_addr_const_pdp11): Correct format mismatch warnings.
11144         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
11145         * config/pdp11/pdp11.md: General change to add base_cost and/or
11146         length attributes for use by new pdp11_insn_cost function.
11147         (MIN_BRANCH): Correct definition.
11148         (MIN_SOB): Ditto.
11149         (doloop_end): Use standard pattern name for looping pattern.
11150         (doloop_end_nocc): New.
11151         (movsf): Add another constraint alternative.
11152         (zero_extendqihi2): Add constraint alternatives for not in place
11153         extend.
11154         (zero_extendhisi2): Remove.
11155         (shift patterns): Add CC handling variants.
11156         (bswaphi2): New.
11157         (bswapsi2): New.
11158         (rothi3): New.
11159         (define_peephole2): New peephole to recognize mov that sets CC for
11160         subsequent test.
11162 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11164         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
11165         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
11166         wi::add.
11168 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
11170         PR c/86420
11171         * real.c (real_nextafter): Return true if result is denormal.
11173 2018-07-09  Martin Liska  <mliska@suse.cz>
11175         * common.opt: Add back wrongly removed attribute.
11177 2018-07-09  Richard Biener  <rguenther@suse.de>
11179         PR debug/86413
11180         * dwarf2out.c (gen_block_die): For an early generated DIE
11181         always output high/low PC attributes.
11183 2018-07-09  Tom de Vries  <tdevries@suse.de>
11185         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
11186         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
11187         onto VAR_DECL with abstract origin.
11189 2018-07-07  Jim Wilson  <jimw@sifive.com>
11191         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
11193 2018-07-07  Tom de Vries  <tdevries@suse.de>
11195         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
11197 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
11199         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
11200         overflow_type.
11201         (combine_bound): Use wide-int overflow calculation instead of
11202         rolling our own.
11203         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
11204         overflow_type.
11205         * fold-const.c (int_const_binop_2): Same.
11206         (extract_muldiv_1): Same.
11207         (fold_div_compare): Same.
11208         (fold_abs_const): Same.
11209         * match.pd: Same.
11210         * poly-int.h (add): Same.
11211         (sub): Same.
11212         (neg): Same.
11213         (mul): Same.
11214         * predict.c (predict_iv_comparison): Same.
11215         * profile-count.c (slow_safe_scale_64bit): Same.
11216         * simplify-rtx.c (simplify_const_binary_operation): Same.
11217         * tree-chrec.c (tree_fold_binomial): Same.
11218         * tree-data-ref.c (split_constant_offset_1): Same.
11219         * tree-if-conv.c (idx_within_array_bound): Same.
11220         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
11221         * tree-ssa-phiopt.c (minmax_replacement): Same.
11222         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
11223         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
11224         * vr-values.c (vr_values::adjust_range_with_scev): Same.
11225         * wide-int.cc (wi::add_large): Same.
11226         (wi::mul_internal): Same.
11227         (wi::sub_large): Same.
11228         (wi::divmod_internal): Same.
11229         * wide-int.h: Change overflow type to overflow_type for neg, add,
11230         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
11231         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
11232         mul_internal, divmod_internal.
11233         (overflow_type): New enum.
11234         (accumulate_overflow): New.
11236 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
11238         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
11239         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
11241 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11243         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
11244         argument is checked for zero before entering loop, avoid checking again.
11246 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
11248         * gimplify.h (generic_expr_could_trap_p): Set as global function.
11249         * gimplify.h (generic_expr_could_trap_p): Likwise.
11250         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
11252 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
11254         PR tree-optimization/86401
11255         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
11256         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
11257         (fold_bit_and_mask): ... here.  New helper function for match.pd.
11258         * fold-const.h (fold_bit_and_mask): Declare.
11259         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
11261 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
11263         PR target/86324
11264         * target.def (translate_mode_attribute): New hook.
11265         * targhooks.h (default_translate_mode_attribute): Declare.
11266         * targhooks.c (default_translate_mode_attribute): New function.
11267         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
11268         * doc/tm.texi: Regenerate.
11269         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
11270         (rs6000_translate_mode_attribute): New function.
11272 2018-07-06  Paul Koning  <ni1d@arrl.net>
11274         * doc/md.texi (define_split): Document DONE and FAIL.
11275         (define_peephole2): Ditto.
11277 2018-07-05  Jeff Law  <law@redhat.com>
11279         PR tree-optimization/86010
11280         * tree-ssa-dse.c (compute_trims): More aggressively trim at
11281         both the head and tail of mem* and str* calls.
11283 2018-07-05  Jim Wilson  <jimw@sifive.com>
11285         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
11287 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
11289         * config/aarch64/aarch64-simd.md: correct flags text for
11290         MIN_EXPR replacement.
11292 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
11294         * configure: Regenerated.
11296 2018-07-05  Carl Love  <cel@us.ibm.com>
11298         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
11299         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
11300         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
11301         VSX_BUILTIN_DOUBLEL_V4SF.
11303 2018-07-05  Martin Sebor  <msebor@redhat.com>
11305         PR c++/86400
11306         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
11307         than its domain to compute its the upper bound of a char array.
11309 2018-07-05  Nathan Sidwell  <nathan@acm.org>
11311         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
11312         * doc/cpp.texi: Update comment.
11313         * doc/tm.texi: Rebuilt.
11314         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
11315         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
11316         * doc/extend.texi (Backwards Compatibility): Clarify it is system
11317         headers affected by extern "C".
11318         * system.h: Poison NO_IMPLICIT_EXTERN_C.
11319         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
11320         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
11321         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
11322         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
11323         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
11324         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
11325         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
11326         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
11327         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
11328         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
11329         config/sparc/sp64-elf.h, config/spu/spu.h,
11330         config/stormy16/stormy16.h, config/v850/v850.h,
11331         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
11332         define NO_IMPLICIT_EXTERN_C.
11333         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
11335 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11337         PR target/84711
11338         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11339         instead of GET_MODE_SIZE when comparing Units.
11341 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11343         PR target/84711
11344         * rtlanal.c (set_noop_p): Constrain on mode change,
11345         include hard-reg-set.h
11347 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
11349         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
11351 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
11353         Revert
11354         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11356         PR sanitizer/84250
11357         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11358         libasan.
11359         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11361 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
11363         PR sanitizer/84250
11364         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
11365         libasan.
11366         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
11368 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
11370         PR middle-end/86380
11371         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
11373 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
11375         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
11376         neg_*_op* variables.
11378 2018-07-04  Martin Liska  <mliska@suse.cz>
11380         * tree-switch-conversion.c: Define
11381         max_ratio_for_speed and max_ratio_for_size constants.
11383 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11384             Martin Liska  <mliska@suse.cz>
11386         PR middle-end/66240
11387         PR target/45996
11388         PR c/84100
11389         * common.opt: Rename align options with 'str_' prefix.
11390         * common/config/i386/i386-common.c (set_malign_value): New
11391         function.
11392         (ix86_handle_option): Use it to set -falign-* options/
11393         * config/aarch64/aarch64-protos.h (struct tune_params): Change
11394         type from int to string.
11395         * config/aarch64/aarch64.c: Update default values from int
11396         to string.
11397         * config/alpha/alpha.c (alpha_override_options_after_change):
11398         Likewise.
11399         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
11400         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11401         max skip conditionally.
11402         * config/i386/freebsd.h (SUBALIGN_LOG): New.
11403         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11404         max skip conditionally.
11405         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11406         max skip conditionally.
11407         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
11408         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11409         max skip conditionally.
11410         * config/i386/i386.c (struct ptt): Change type from int to
11411         string.
11412         (ix86_default_align): Set default values.
11413         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
11414         max skip conditionally.
11415         * config/i386/iamcu.h (SUBALIGN_LOG): New.
11416         (ASM_OUTPUT_MAX_SKIP_ALIGN):
11417         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
11418         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11419         max skip conditionally.
11420         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
11421         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
11422         * config/i386/x86-64.h (SUBALIGN_LOG): New.
11423         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
11424         max skip conditionally.
11425         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11426         * config/ia64/ia64.c (ia64_option_override): Set default values
11427         for alignment options.
11428         * config/m68k/m68k.c: Handle new str_align_* options.
11429         * config/mips/mips.c (mips_set_compression_mode): Change
11430         type of constants.
11431         (mips_option_override): Set default values for options.
11432         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
11433         Likewise.
11434         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11435         Likewise.
11436         * config/rx/rx.c (rx_option_override): Likewise.
11437         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
11438         (LABEL_ALIGN): Use align_labels_log.
11439         (LOOP_ALIGN): Use align_loops_align.
11440         * config/s390/s390.c (s390_asm_output_function_label): Use new
11441         macros.
11442         * config/sh/sh.c (sh_override_options_after_change):
11443         Change type of constants.
11444         * config/spu/spu.c (spu_sched_init): Likewise.
11445         * config/sparc/sparc.c (sparc_option_override): Set default
11446         values for options.
11447         * config/visium/visium.c (visium_option_override): Likewise.
11448         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
11449         emit p2align format with last argument if it's not needed.
11450         * doc/invoke.texi: Document extended format of -falign-*.
11451         * final.c: Use align_labels alignment.
11452         * flags.h (struct target_flag_state): Change type to use
11453         align_flags.
11454         (struct align_flags_tuple): New.
11455         (struct align_flags): Likewise.
11456         (align_loops_log): Redefine macro to use new types.
11457         (align_loops_max_skip): Redefine macro to use new types.
11458         (align_jumps_log): Redefine macro to use new types.
11459         (align_jumps_max_skip): Redefine macro to use new types.
11460         (align_labels_log): Redefine macro to use new types.
11461         (align_labels_max_skip): Redefine macro to use new types.
11462         (align_functions_log): Redefine macro to use new types.
11463         (align_loops): Redefine macro to use new types.
11464         (align_jumps): Redefine macro to use new types.
11465         (align_labels): Redefine macro to use new types.
11466         (align_functions): Redefine macro to use new types.
11467         (align_functions_max_skip): Redefine macro to use new types.
11468         (align_loops_value): New macro.
11469         (align_jumps_value): New macro.
11470         (align_labels_value): New macro.
11471         (align_functions_value): New macro.
11472         * function.c (invoke_set_current_function_hook): Propagate
11473         alignment values from flags to global variables default in
11474         topleev.h.
11475         * ipa-icf.c (sem_function::equals_wpa): Use
11476         cl_optimization_option_eq instead of memcmp.
11477         * lto-streamer.h (cl_optimization_stream_out): Support streaming
11478         of string types.
11479         (cl_optimization_stream_in): Likewise.
11480         * optc-save-gen.awk: Support strings in cl_optimization.
11481         * opth-gen.awk: Likewise.
11482         * opts.c (finish_options): Remove error checking of invalid
11483         value ranges.
11484         (MAX_CODE_ALIGN): Remove.
11485         (MAX_CODE_ALIGN_VALUE): Likewise.
11486         (parse_and_check_align_values): New function.
11487         (check_alignment_argument): Likewise.
11488         (common_handle_option): Use check_alignment_argument.
11489         * opts.h (parse_and_check_align_values): Declare.
11490         * toplev.c (init_alignments): Remove.
11491         (read_log_maxskip): New.
11492         (parse_N_M): Likewise.
11493         (parse_alignment_opts): Likewise.
11494         (backend_init_target): Remove usage of init_alignments.
11495         * toplev.h (parse_alignment_opts): Declare.
11496         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
11497         argument.
11498         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
11499         * tree.c (cl_option_hasher::equal): New.
11500         * varasm.c: Use new global macros.
11502 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
11504         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
11505         Use a simpler align directive also if MAXSKIP = ALIGN-1.
11506         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11507         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11508         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11509         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11510         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
11511         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
11512         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
11513         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11514         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11515         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11516         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11518 2018-07-04  Martin Liska  <mliska@suse.cz>
11519             Jonathan Wakely  <jwakely@redhat.com>
11521         * coverage.c: Use correct type.
11522         * doc/invoke.texi: Language correction.
11524 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
11526         PR target/85620
11527         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
11528         ENDBRANCH for non-tail call which may return via indirect branch.
11529         * doc/extend.texi: Document indirect_return attribute.
11531 2018-07-03  Martin Sebor  <msebor@redhat.com>
11533         PR tree-optimization/86274
11534         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
11535         precondition.
11536         (format_floating): Correct handling of infinities and NaNs.
11538 2018-07-03  Martin Sebor  <msebor@redhat.com>
11540         * print-tree.c (print_real_cst): New function.
11541         (print_node_brief): Call it.
11542         (print_node): Ditto.
11544 2018-07-03  Jeff Law  <law@redhat.com>
11546         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
11547         into a single pattern.
11549         * config/h8300/h8300.md (ors code_iterator): New.
11550         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
11551         a single pattern and single splitter.
11552         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
11553         (iorqi3_1, xorqi3_1): Likewise.
11554         (iorqi3, xorqi3 expanders): Similarly.
11556         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
11557         (movmd_internal) into a single pattern using the P mode iterator.
11558         (movmd splitters): Similarly.
11559         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
11560         (movsd splitters): Similarly.
11562         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
11563         ADDB, ADDW and ADDL into a single ADD attribute which selects the
11564         right table based on the size of the operand.
11565         * config/h8300/h8300.md (length_table): Corresponding changes. All
11566         references to "addb", "addw" and "addl" changed to "add".
11567         (btst patterns): Merge two variants into a single pattern.
11568         (tstqi, tsthi): Likewise.
11569         (addhi3_incdec, addsi3_incdec): Likewise.
11570         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
11571         (mulhi3, mulsi3): Likewise.
11572         (udivhi3, udivsi3): Likewise.
11573         (divhi3, divsi3): Likewise.
11574         (andorqi3, andorhi3, andorsi3): Likewise.
11576 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
11578         PR target/85694
11579         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
11580         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
11582 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11584         PR tree-optimization/85694
11585         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
11586         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
11587         UNSPEC_URHADD.
11588         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
11589         (<u>avg<mode>3_ceil): New patterns.
11591 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
11593         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
11594         scan-tree-dump directive.
11595         * gcc.dg/vect/slp-perm-2.c: Likewise.
11596         * gcc.dg/vect/slp-perm-3.c: Likewise.
11597         * gcc.dg/vect/slp-perm-5.c: Likewise.
11598         * gcc.dg/vect/slp-perm-6.c: Likewise.
11599         * gcc.dg/vect/slp-perm-7.c: Likewise.
11600         * gcc.dg/vect/slp-perm-8.c: Likewise.
11602 2018-07-03  Marek Polacek  <polacek@redhat.com>
11604         PR middle-end/86202
11605         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
11607 2018-07-03  Richard Biener  <rguenther@suse.de>
11609         PR ipa/86389
11610         * tree-ssa-structalias.c (find_func_clobbers): Properly
11611         handle indirect calls.
11613 2018-07-03  Jeff Law  <law@redhat.com>
11615         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
11616         (shifts): New code iterator.
11617         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11618         expander.  Fix HImode handling on H8/SX.
11619         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11620         (subqi3, subhi3, subsi3 expanders): Likewise.
11621         (andqi3, andhi3, andsi3 expanders): Likewise.
11622         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11623         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11624         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11625         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11626         (zero_extendqihi2, zero_extendqisi2): Likewise.
11627         (extendqihi2, extendqisi2): Likewise.
11628         (rotlqi3, rotlhi3, rotlsi3): Likewise.
11629         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11630         (rotlqi3_1, rotlhi3_1): Likewise.
11631         (logicalhi3_sn, logicalsi3_sn): Likewise.
11632         (logicalhi3, logicalsi3): Likewise.
11634 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11636         * tree-vect-patterns.c (vect_recog_rotate_pattern)
11637         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11638         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11639         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11640         type to append_pattern_def_seq instead of creating a stmt_vec_info
11641         directly.
11642         (build_mask_conversion): Likewise.  Remove vinfo argument.
11643         (vect_add_conversion_to_patterm): Likewise, renaming to...
11644         (vect_add_conversion_to_pattern): ...this.
11645         (vect_recog_mask_conversion_pattern): Update call to
11646         build_mask_conversion.  Pass the vector type to
11647         append_pattern_def_seq here too.
11648         (vect_recog_gather_scatter_pattern): Update call to
11649         vect_add_conversion_to_pattern.
11651 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11653         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11654         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11655         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11656         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11657         STMT_VINFO_PATTERN_DEF_SEQ to null here.
11658         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11659         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11660         append_pattern_def_seq instead of new_pattern_def_seq.
11661         (vect_recog_divmod_pattern): Do both of the above.
11662         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11663         is null.
11665 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11667         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11668         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11669         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11670         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11671         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11672         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11673         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11674         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11675         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11676         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11677         parameter with a single stmt_vec_info.
11678         (vect_recog_func_ptr): Likewise.
11679         (vect_recog_gather_scatter_pattern): Likewise, folding in...
11680         (vect_try_gather_scatter_pattern): ...this.
11681         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11682         the stmt_vec_info of the statement to be matched.  Don't clear
11683         STMT_VINFO_RELATED_STMT.
11684         (vect_pattern_recog): Update call accordingly.
11686 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11688         PR tree-optimization/85694
11689         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11690         (uavgM3_ceil): Document new optabs.
11691         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11692         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11693         functions.
11694         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11695         (savg_ceil_optab): New optabs.
11696         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11697         (vect_vect_recog_func_ptrs): Add it.
11698         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11699         constant directly from the associated lhs.
11701 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11703         * tree-vect-patterns.c (vect_split_statement): New function.
11704         (vect_convert_input): Use it to try to split an existing cast.
11706 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11708         * poly-int.h (print_hex): New function.
11709         * dumpfile.h (dump_dec, dump_hex): Declare.
11710         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11711         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11712         min_input_precision, operation_precision and operation_sign.
11713         * tree-vect-patterns.c (vect_get_range_info): New function.
11714         (vect_same_loop_or_bb_p, vect_single_imm_use)
11715         (vect_operation_fits_smaller_type): Delete.
11716         (vect_look_through_possible_promotion): Add an optional
11717         single_use_p parameter.
11718         (vect_recog_over_widening_pattern): Rewrite to use new
11719         stmt_vec_info infomration.  Handle one operation at a time.
11720         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11721         (vect_truncatable_operation_p, vect_set_operation_type)
11722         (vect_set_min_input_precision): New functions.
11723         (vect_determine_min_output_precision_1): Likewise.
11724         (vect_determine_min_output_precision): Likewise.
11725         (vect_determine_precisions_from_range): Likewise.
11726         (vect_determine_precisions_from_users): Likewise.
11727         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11728         (vect_vect_recog_func_ptrs): Put over_widening first.
11729         Add cast_forwprop.
11730         (vect_pattern_recog): Call vect_determine_precisions.
11732 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11734         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11735         statements that have been replaced by further pattern statements.
11736         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11738 2018-07-03  Richard Biener  <rguenther@suse.de>
11740         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11741         always set *dt.  Dump vectype in vectype overload.
11742         * dumpfile.h (dump_gimple_expr): New function.
11743         (dump_gimple_expr_loc): Likewise.
11744         * dumpfile.c (dump_gimple_expr): New function.
11745         (dump_gimple_expr_loc): Likewise.
11747 2018-07-02  Jeff Law  <law@redhat.com>
11749         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11750         the H8/300, H8/300H and H8/S variants into a single pattern.
11751         (movhi_h8300, movqi_h8300hs): Similarly.
11752         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11753         (QHI mode iterator): New.
11755         * config/h8300/h8300.md: Remove trailing whitespace.
11757 2018-07-02  Jim Wilson  <jimw@sifive.com>
11759         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11760         instead of emit_insn for interrupt returns.
11761         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11762         (riscv_sret, riscv_uret): Likewise.
11764 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11766         * pretty-print.c (selftest::test_pp_format): Move save and restore
11767         of quotes to class auto_fix_quotes, and add an instance.
11768         * selftest.c: Include "intl.h".
11769         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11770         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11771         * selftest.h (selftest::auto_fix_quotes): New class.
11773 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
11775         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11776         (aarch64_sve_prepare_conditional_op): Remove.
11777         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11778         Allow aarch64_simd_reg_or_zero as select operand; remove
11779         the aarch64_sve_prepare_conditional_op call.
11780         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11781         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11782         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11783         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11784         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11785         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11786         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11787         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11788         and a splitters to match all of the *_any patterns.
11789         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11791         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11792         (SVE_COND_FP_BINARY_REV): Remove.
11793         (sve_int_op_rev, sve_fp_op_rev): New.
11794         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11795         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11796         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11797         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11798         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11799         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11800         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11801         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11802         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11804         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11805         Remove match_dup 1 from the inner unspec.
11806         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11808         * config/aarch64/aarch64.md (movprfx): New attr.
11809         (length): Default movprfx to 8.
11810         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11811         (*madd<SVE_I>, *msub<SVE_I): Likewise.
11812         (*<su>mul<SVE_I>3_highpart): Likewise.
11813         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11814         (*v<ASHIFT><SVE_I>3): Likewise.
11815         (*<su><MAXMIN><SVE_I>3): Likewise.
11816         (*<su><MAXMIN><SVE_F>3): Likewise.
11817         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11818         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11819         (*div<SVE_F>4): Likewise.
11821 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
11823         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11824         in dump string.
11826 2018-07-02  Richard Biener  <rguenther@suse.de>
11828         PR tree-optimization/86363
11829         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11830         memset argument refers to a non-variable address.
11832 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11834         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11835         {PLUS,MINUS}_EXPR code to...
11836         (adjust_symbolic_bound): ...here,
11837         (combine_bound): ...here,
11838         (set_value_range_with_overflow): ...and here.
11840 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11842         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11843         code...
11844         (extract_range_from_abs_expr): ...here.
11846 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
11848         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11849         -fno-omit-frame-pointer when not optimizing.
11851 2018-07-02  Martin Liska  <mliska@suse.cz>
11853         PR ipa/86279
11854         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11855         (propagate_nothrow): Likewise.
11857 2018-07-02  Martin Liska  <mliska@suse.cz>
11859         PR ipa/86323
11860         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11862 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11864         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11865         function in r262149, changing "loc" param from source_location to
11866         const dump_location_t &.
11867         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11868         declaration, as above.
11870 2018-07-01  Paul Koning  <ni1d@arrl.net>
11872         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11873         -munit-asm, -mgnu-asm, -mdec-asm.
11874         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11875         (pdp11_output_labelref): New.
11876         (pdp11_output_def): New.
11877         (pdp11_output_addr_vec_elt): New.
11878         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
11879         %# and %@ format codes.
11880         (pdp11_option_override): New.
11881         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11882         (pdp11_output_ident): New.
11883         (pdp11_asm_named_section): New.
11884         (pdp11_asm_init_sections): New.
11885         (pdp11_file_start): New.
11886         (pdp11_file_end): New.
11887         (output_ascii): Use .ascii/.asciz for -mdec-asm.
11888         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
11889         %o, like %c but octal.
11890         (pdp11_option_override): New.
11891         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11892         -mdec-asm.
11893         (DATA_SECTION_ASM_OP): Ditto.
11894         (READONLY_DATA_SECTION_ASM_OP): New.
11895         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11896         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11897         (ASM_OUTPUT_LABELREF): Ditto.
11898         (ASM_OUTPUT_DEF): Ditto.
11899         (ASM_OUTPUT_EXTERNAL): New.
11900         (ASM_OUTPUT_SOURCE_FILENAME): New.
11901         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11902         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11903         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
11904         %# and %@ format codes.
11905         * config/pdp11/pdp11.opt (mgnu-asm): New.
11906         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11907         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11908         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11910 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
11912         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11913         dereferencing path[] beyond its length.
11914         (debug_path): New.
11915         (debug_all_paths): New.
11916         (rewire_first_differing_edge): New.
11917         (adjust_paths_after_duplication): New.
11918         (duplicate_thread_path): Call adjust_paths_after_duplication.
11919         Add new argument.
11920         (thread_through_all_blocks): Add new argument to
11921         duplicate_thread_path.
11923 2018-06-30  Jim Wilson  <jimw@sifive.com>
11925         * config/riscv/predicates.md (p2m1_shift_operand): New.
11926         (high_mask_shift_operand): New.
11927         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11928         pattern using p2m1_shift_operand.
11929         (lshsi3_zero_extend_3+2): New combiner pattern using
11930         high_mask_shift_operand.
11932 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11934         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11935         split out from...
11936         (vect_recog_rotate_pattern): ...here.
11937         (vect_convert_input): Try to insert casts of invariants in the
11938         preheader.
11939         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11940         preheader to be empty.
11942 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11944         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11945         vector type.  If given, install it in the new statement's
11946         STMT_VINFO_VECTYPE.
11947         (vect_element_precision): New function.
11948         (vect_unpromoted_value): New struct.
11949         (vect_unpromoted_value::vect_unpromoted_value): New function.
11950         (vect_unpromoted_value::set_op): Likewise.
11951         (vect_look_through_possible_promotion): Likewise.
11952         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11953         (vect_widened_op_tree, vect_convert_input): Likewise.
11954         (vect_convert_inputs, vect_convert_output): Likewise.
11955         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11956         to handle the optional cast of the multiplication result and
11957         vect_widened_op_tree to detect the widened multiplication itself.
11958         Do not require the input and output of promotion casts to have
11959         the same sign, but base the signedness of the operation on the
11960         input rather than the result.  If the pattern includes two
11961         promotions, check that those promotions have the same sign.
11962         Do not restrict the MULT_EXPR handling to a double-width result;
11963         handle quadruple-width results and wider.  Use vect_convert_inputs
11964         to convert the inputs to the common type.
11965         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
11966         to handle the optional cast of the ABS result.  Also allow a sign
11967         change or a sign extension between the ABS and MINUS.
11968         Use vect_widened_op_tree to detect the widened subtraction and use
11969         vect_convert_inputs to convert the inputs to the common type.
11970         (vect_handle_widen_op_by_const): Delete.
11971         (vect_recog_widen_op_pattern): New function.
11972         (vect_recog_widen_mult_pattern): Use it.
11973         (vect_recog_widen_shift_pattern): Likewise.
11974         (vect_recog_widen_sum_pattern): Use
11975         vect_look_through_possible_promotion to handle the promoted
11976         PLUS_EXPR operand.
11978 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11980         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11981         the containing gimple_seq *.
11982         * gimple-iterator.h (gsi_for_stmt): Declare it.
11983         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11984         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11985         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11986         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11987         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11988         checks.
11989         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11990         split out from...
11991         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
11992         statement being replaced is part of an existing pattern
11993         definition sequence, inserting the new pattern statements before
11994         the original one.
11995         (vect_pattern_recog_1): Don't return a bool.  If the statement
11996         is already part of a pattern, instead apply pattern matching
11997         to the pattern definition statements.  Don't clear the
11998         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11999         (vect_pattern_recog): Don't break after the first match;
12000         continue processing the pattern definition statements instead.
12001         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
12003 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12005         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12006         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12007         (vect_recog_widen_sum_pattern): Use it.
12009 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12011         * tree-vect-loop.c (vectorizable_reduction): Assert that the
12012         phi is not a pattern statement and has not been replaced by
12013         a pattern statement.
12014         * tree-vect-patterns.c (type_conversion_p): Don't check
12015         STMT_VINFO_IN_PATTERN_P.
12016         (vect_recog_vector_vector_shift_pattern): Likewise.
12017         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
12018         the pattern statement rather than the original statement; check
12019         directly for a WIDEN_MULT_EXPR here.
12020         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
12021         vect_is_simple_use to return the pattern statement rather
12022         than the original statement; use is_pattern_stmt_p to check
12023         for such a pattern statement.
12024         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
12025         to return the pattern statement rather than the original statement;
12026         don't do the same transformation here.
12027         (vect_is_simple_use): If the defining statement has been replaced
12028         by a pattern statement, return the pattern statement instead.
12029         Remove the corresponding (local) transformation from the vectype
12030         overload.
12032 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12034         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
12035         end and default to null.
12036         * tree-vect-loop.c (vect_create_epilog_for_reduction)
12037         (vectorizable_reduction): Update calls accordingly, dropping the
12038         gimple ** argument if the passed-back statement isn't needed.
12039         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
12040         (vect_recog_rotate_pattern): Likewise.
12041         (vect_recog_mask_conversion_pattern): Likewise.
12042         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
12043         (vect_mask_constant_operand_p): Likewise.
12044         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
12045         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
12046         (get_group_load_store_type, get_load_store_type): Likewise.
12047         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
12048         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
12049         (vectorizable_conversion, vectorizable_assignment): Likewise.
12050         (vectorizable_shift, vectorizable_operation): Likewise.
12051         (vectorizable_store, vect_is_simple_cond): Likewise.
12052         (vectorizable_condition, vectorizable_comparison): Likewise.
12053         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
12054         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
12055         and move it to the end.  Cope with null def_stmt_outs.
12057 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12059         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
12061 2018-06-29  Jeff Law  <law@redhat.com>
12063         * config/v850/v850.c (v850_legitimate_address_p): Handle large
12064         displacements for TARGET_V850E2V3 and newer.
12065         (TARGET_LRA_P): Remove.  Defaults to LRA now.
12066         * config/v850/v850.md (sign23byte_load): Remove.
12067         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
12068         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
12070 2018-06-29  Martin Liska  <mliska@suse.cz>
12072         PR lto/85759
12073         * coverage.c (coverage_init): Mangle full path name.
12074         * doc/invoke.texi: Document the change.
12075         * gcov-io.c (mangle_path): New.
12076         * gcov-io.h (mangle_path): Likewise.
12077         * gcov.c (mangle_name): Use mangle_path for path mangling.
12079 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12081         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
12082         if starting source register is not even.
12084 2018-06-29  Martin Liska  <mliska@suse.cz>
12086         PR tree-optimization/86263
12087         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
12088         Make edge redirection.
12090 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
12092         * dumpfile.c (dump_loc): Add indentation based on scope depth.
12093         (dump_scope_depth): New variable.
12094         (get_dump_scope_depth): New function.
12095         (dump_begin_scope): New function.
12096         (dump_end_scope): New function.
12097         * dumpfile.h (get_dump_scope_depth): New declaration.
12098         (dump_begin_scope): New declaration.
12099         (dump_end_scope): New declaration.
12100         (class auto_dump_scope): New class.
12101         (AUTO_DUMP_SCOPE): New macro.
12102         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
12103         AUTO_DUMP_SCOPE.
12105 2018-06-29  Richard Biener  <rguenther@suse.de>
12107         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
12108         compute_all_dependences succeeds.
12109         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
12110         exceed --param loop-max-datarefs-for-datadeps.
12112 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
12114         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
12116 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
12118         PR target/86348
12119         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
12120         alternative 0 in preferred_for_speed attribute.
12122 2018-06-28  Paul Koning  <ni1d@arrl.net>
12124         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
12125         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
12126         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
12127         * config/pdp11/pdp11.md: Correct "length" attribute calculation
12128         for shift insn patterns.
12130 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
12132         * cgraph.c (cgraph_node::get_body): Replace assignments to
12133         "dump_file" with calls to set_dump_file.
12134         * dumpfile.c (alt_dump_file): Make static, and group with...
12135         (alt_flags): ...this definition.
12136         (dumps_are_enabled): New variable.
12137         (refresh_dumps_are_enabled): New function.
12138         (set_dump_file): New function.
12139         (set_alt_dump_file): New function.
12140         (gcc::dump_manager::dump_start): Replace assignments to
12141         "dump_file" and "alt_dump_file" with calls to set_dump_file and
12142         set_alt_dump_file.
12143         (gcc::dump_manager::dump_finish): Likewise.
12144         * dumpfile.h (alt_dump_file): Delete decl.
12145         (dumps_are_enabled): New variable decl.
12146         (set_dump_file): New function decl.
12147         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
12148         global.
12149         * tree-nested.c (lower_nested_functions): Replace assignments to
12150         "dump_file" with calls to set_dump_file.
12152 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
12154         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
12155         goto_locus of each outgoing edge of each basic block.
12157 2018-06-28  Richard Biener  <rguenther@suse.de>
12159         * dwarf2out.c (decl_scope_table): Remove.
12160         (push_decl_scope): Likewise.
12161         (pop_decl_scope): Likewise.
12162         (gen_type_die_for_member): Do not call push/pop_decl_scope.
12163         (gen_struct_or_union_type_die): Likewise.
12164         (gen_tagged_type_die): Likewise.
12165         (dwarf2out_init): Do not initialize decl_scope_table.
12166         (dwarf2out_c_finalize): Do not free it.
12168 2018-06-28  Richard Biener  <rguenther@suse.de>
12170         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
12171         deciding whether to not re-use a DIE.
12173 2018-06-28  Richard Biener  <rguenther@suse.de>
12175         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
12176         DW_AT_abstract_origin attribute.
12178 2018-06-28  Martin Liska  <mliska@suse.cz>
12180         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12181         Use newly introduced constants.
12182         * tree-switch-conversion.h (struct jump_table_cluster):
12183         Define max_ratio_for_size and max_ratio_for_speed.
12185 2018-06-28  Martin Liska  <mliska@suse.cz>
12187         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12188         Add new checking assert to catch invalid state.
12189         (jump_table_cluster::can_be_handled): Handle single case
12190         clusters.
12191         (jump_table_cluster::is_beneficial): Bail out for such case.
12192         (bit_test_cluster::find_bit_tests):
12193         Add new checking assert to catch invalid state.
12194         (bit_test_cluster::can_be_handled): Handle single case
12195         clusters.
12196         (bit_test_cluster::is_beneficial): Bail out for such case.
12197         (switch_decision_tree::analyze_switch_statement):
12198         Fix comment.
12200 2018-06-28  Martin Liska  <mliska@suse.cz>
12202         * common.opt: Introduce -completion option.
12203         * gcc.c (driver_handle_option): Handle it.
12204         (driver::main): Print completions if completion
12205         is set.
12206         * opt-suggestions.c (option_proposer::get_completions):
12207         New function.
12208         (option_proposer::suggest_completion): Likewise.
12209         (option_proposer::find_param_completions): Likewise.
12210         (verify_autocompletions): Likewise.
12211         (test_completion_valid_options): Likewise.
12212         (test_completion_valid_params): Likewise.
12213         (in_completion_p): Likewise.
12214         (empty_completion_p): Likewise.
12215         (test_completion_partial_match): Likewise.
12216         (test_completion_garbage): Likewise.
12217         (opt_proposer_c_tests): Likewise.
12218         * opt-suggestions.h: Declare new functions.
12219         * opts.c (common_handle_option): Handle OPT__completion_.
12220         * selftest-run-tests.c (selftest::run_tests): Add
12221         opt_proposer_c_tests.
12222         * selftest.c (assert_str_startswith): New.
12223         * selftest.h (assert_str_startswith): Likewise.
12224         (opt_proposer_c_tests): New.
12225         (ASSERT_STR_STARTSWITH): Likewise.
12227 2018-06-28  Martin Liska  <mliska@suse.cz>
12229         * Makefile.in: Add opt-suggestions.o.
12230         * gcc-main.c: Include opt-suggestions.h.
12231         * gcc.c (driver::driver): Likewise.
12232         (driver::~driver): Remove m_option_suggestions.
12233         (driver::build_option_suggestions): Moved to option_proposer.
12234         (driver::suggest_option): Likewise.
12235         (driver::handle_unrecognized_options): Use option_proposer.
12236         * gcc.h (class driver): Add new memver m_option_proposer.
12237         * opt-suggestions.c: New file.
12238         * opt-suggestions.h: New file.
12240 2018-06-28  Martin Liska  <mliska@suse.cz>
12242         * vec.h (class auto_string_vec): New (moved from auto_argvec).
12243         (auto_string_vec::~auto_string_vec): Likewise.
12245 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
12247         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
12248         prevent_decl_creation_for_types fields up and add reset_location field.
12249         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
12250         statement if id->reset_location is true.
12251         (copy_edges_for_bb): Do not set goto_locus on the new edges if
12252         id->reset_location is true.
12253         (copy_phis_for_bb): Force input_location on the arguments if
12254         id->reset_location is true.
12255         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
12256         is set on the function to be inlined.
12258 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
12260         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
12262 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
12264         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
12265         registers for Pmode.
12266         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
12267         hard registers for the clobbered pseudo.
12269 2018-06-27  Paul Koning  <ni1d@arrl.net>
12271         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12272         mutually exclusive options.
12273         * config/pdp11/constraints.md (h): New constraint.
12274         (O): Update definition to match shift code generation.
12275         (D): New constraint.
12276         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
12277         (CCFP): Remove.
12278         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
12279         function.
12280         (output_jump): Change arguments.
12281         (pdp11_fixed_cc_regs): New function.
12282         (pdp11_cc_mode): Ditto.
12283         (pdp11_expand_shift): Ditto.
12284         (pdp11_assemble_shift): Ditto.
12285         (pdp11_small_shift): Ditto.
12286         (pdp11_branch_cost): Remove.
12287         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
12288         from output.
12289         (pdp11_register_move_cost): Update for CC registers.
12290         (pdp11_rtx_costs): Add case for LSHIFTRT.
12291         (pdp11_output_jump): Add CCNZ mode conditional branches.
12292         (notice_update_cc_on_set): Remove.
12293         (pdp11_cc_mode): New function.
12294         (simple_memory_operand): Correct pre/post decrement case.
12295         (no_side_effect_operand): New function.
12296         (pdp11_regno_reg_class): Add CC_REGS class.
12297         (pdp11_fixed_cc_regs): New function.
12298         (pdp11_small_shift): New function.
12299         (pdp11_expand_shift): New function to expand shift insns.
12300         (pdp11_assemble_shift): New function to output shifts.
12301         (pdp11_branch_cost): Remove.
12302         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
12303         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
12304         (WCHAR_TYPE): Ditto.
12305         (PTRDIFF_TYPE): Ditto.
12306         (ADJUST_INSN_LENGTH): New macro.
12307         (FIXED_REGISTERS): Add CC registers.
12308         (CALL_USED_REGISTERS): Ditto.
12309         (reg_class): Ditto.
12310         (REG_CLASS_NAMES): Ditto.
12311         (REG_CLASS_CONTENTS): Ditto.
12312         (SELECT_CC_MODE): Use new function.
12313         (TARGET_FLAGS_REGNUM): New macro.
12314         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
12315         (cc0_reg_rtx): Remove.
12316         (CC_STATUS_MDEP): Remove.
12317         (CC_STATUS_MDEFP_INIT): Remove.
12318         (CC_IN_FPU): Remove.
12319         (NOTICE_UPDATE_CC): Remove.
12320         (REGISTER_NAMES): Add CC registers.
12321         (BRANCH_COST): Change to constant 1.
12322         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
12323         handling.
12324         * config/pdp11/pdp11.opt (mbcopy): Remove.
12325         (mbcopy-builtin): Remove.
12326         (mbranch-cheap): Remove.
12327         (mbranch-expensive): Remove.
12328         * config/pdp11/predicates.md (expand_shift_operand): Update to
12329         match shift code generation.
12330         (ccnz_operator): New predicate.
12331         * doc/invoke.texi (PDP-11 Options): Remove deleted options
12332         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
12333         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
12334         exclusive options.
12335         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
12336         description of O constraint.
12338 2018-06-27  Jeff Law  <law@redhat.com>
12339             Austin Law  <austinklaw@gmail.com>
12341         * config/v850/v850.md (addsi3_set_flags): New pattern.
12342         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
12343         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
12344         (zero_extendhisi2_v850_set_flags): Likewise.
12345         (zero_extendqisi2_v850_set_flags): Likewise.
12346         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
12347         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
12348         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
12350         * config/v850/v850-protos.h (notice_update_cc): Remove.
12351         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
12352         (v850_print_operand): Handle 'D' and "d".
12353         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
12354         Add handling of arithmetic/logical operations compared against zero.
12355         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
12356         Do not look at v850_compare_op, instead get mode from last argument.
12357         (v850_gen_compare): Remove
12358         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
12359         after reload for prologue insns.
12360         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
12361         patterns.
12362         (construct_save_jarl): Likewise.
12363         (TARGET_FLAGS_REGNUM): Define.
12364         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
12365         (NOTICE_UPDATE_CC): Remove.
12366         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
12367         than cc0.  Conditionalize on reload_completed.
12368         (cmpsi_insn, setfcc_insn): Likewise.
12369         (tst1 splitter): Turn into define_and_split which sets the flags
12370         after reload.
12371         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
12372         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
12373         (cstoresf4, cstoredf4): Clobber the flags.
12374         (cmpsi, cmpsf, cmpdf): Remove expanders.
12375         (setf_insn): Remove pattern.
12376         (addsi3): Turn into define_and_split which clobbers the flags after
12377         reload and a suitable pattern (addsi3_clobber_flags) for use after
12378         reload.
12379         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
12380         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
12381         (ashrsi3, ashrsi3_v850e2): Likewise.
12382         (bins): Clobber the flags.
12383         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
12384         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
12385         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
12386         (call_value_internal_short, call_value_internal_long): Likewise.
12387         (callt_save_interrupt, callt_return_interrupt): Likewise.
12388         (save_interrupt, return_interrupt): Likewise.
12389         (callt_save_all_interrupt, save_all_interrupt): Likewise.
12390         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
12391         (restore_all_interrupt, _restore_all_interrupt): Likewise.
12392         (All FP comparisons): Only allow after reload has completed.
12393         (trfsr): Likewise.
12394         (divh, divhu): Tweak output template.
12395         (branch_z_normal, branch_z_invert): Remove
12396         (branch_nz_normal, branch_nz_invert): Likewise.
12397         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
12399         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
12400         * config/v850/v850.c (notice_update_cc): Remove.
12401         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
12402         (CC_NO_CARRY): Likewise.
12403         (NOTICE_UPDATE_CC): Define to nothing.
12404         * config/v850/v850.md: Remove block comment on cc0 handling
12405         Remove "cc" attribute from all patterns.  Remove cc_status handling
12406         from all patterns.  Minor formatting fixes.
12408 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12410         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
12411         (cortex-a76.cortex-a55): Likewise.
12412         * config/aarch64/aarch64-tune.md: Regenerate.
12413         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
12414         cortex-a76.cortex-a55.
12416 2018-06-27  Jeff Law  <law@redhat.com>
12418         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
12419         (MULTILIB_DIRNAMES): Similarly.
12421 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
12423         * gimple.h (gimple_return_retbnd): Delete.
12424         (gimple_return_set_retbnd): Likewise.
12425         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
12426         gimple_return_set_retbnd.
12427         * gimple-pretty-print.c (dump_gimple_return): Remove call to
12428         gimple_return_retbnd and adjust.
12429         * tree-inline.h (struct copy_body_data): Remove retbnd field.
12430         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
12431         Explicitly return NULL in a couple more cases.  Move assertion
12432         on debug statements and remove unreachable code.
12433         (reset_debug_binding): Do not test id->retbnd.
12434         (expand_call_inline): Do not set it.
12436 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
12438         * configure.ac: Add --disable-gcov option.
12439         * configure: Regenerate.
12440         * Makefile.in: Honour @enable_gcov@.
12441         * doc/install.texi: Document --disable-gcov.
12443 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12445         * config/arm/arm-cpus.in (cortex-a76): New entry.
12446         (cortex-a76.cortex-a55): Likewise.
12447         * config/arm/arm-tables.opt: Regenerate.
12448         * config/arm/arm-tune.md: Likewise.
12449         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
12450         * doc/invoke.texi (ARM Options): Document cortex-a76 and
12451         cortex-a76.cortex-a55.
12453 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
12455         PR target/85769
12456         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
12458 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
12460         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
12461         comment.
12462         (EPILOGUE_USES): Likewise.
12464 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
12466         * tree-inline.c (remap_location): New function extracted from...
12467         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
12468         (copy_phis_for_bb): ...here.  Call remap_location.
12469         (copy_cfg_body): Adjust call to copy_edges_for_bb.
12471 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
12473         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
12474         unaligned vsx for 16B memset.
12476 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
12478         PR target/86285
12479         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
12480         ieee128_float_type_node to long_double_type_node unless
12481         TARGET_LONG_DOUBLE_128 is set.
12483 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
12485         * cfgloop.c (get_loop_location): Convert return type from
12486         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
12487         by implicit construction from rtx_insn *, and using
12488         dump_user_location_t::from_function_decl for the fallback case.
12489         * cfgloop.h (get_loop_location): Convert return type from
12490         location_t to dump_user_location_t.
12491         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
12492         dump_printf_loc to pass in a dump_location_t rather than a
12493         location_t, via the gimple stmt.
12494         * coverage.c (get_coverage_counts): Update calls to
12495         dump_printf_loc to pass in dump_location_t rather than a
12496         location_t.
12497         * doc/optinfo.texi (Dump types): Convert example of
12498         dump_printf_loc from taking "locus" to taking "insn".  Update
12499         description of the "_loc" calls to cover dump_location_t.
12500         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
12501         "selftest.h".
12502         (dump_user_location_t::dump_user_location_t): New constructors,
12503         from gimple *stmt and rtx_insn *.
12504         (dump_user_location_t::from_function_decl): New function.
12505         (dump_loc): Make static.
12506         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
12507         const dump_location_t &.
12508         (dump_generic_expr_loc): Delete.
12509         (dump_printf_loc): Convert param "loc" from location_t to
12510         const dump_location_t &.
12511         (selftest::test_impl_location): New function.
12512         (selftest::dumpfile_c_tests): New function.
12513         * dumpfile.h: Include "profile-count.h".
12514         (class dump_user_location_t): New class.
12515         (struct dump_impl_location_t): New struct.
12516         (class dump_location_t): New class.
12517         (dump_printf_loc): Convert 2nd param from source_location to
12518         const dump_location_t &.
12519         (dump_generic_expr_loc): Delete.
12520         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
12521         const dump_location_t &.
12522         * gimple-fold.c (fold_gimple_assign): Update call to
12523         dump_printf_loc to pass in a dump_location_t rather than a
12524         location_t, via the gimple stmt.
12525         (gimple_fold_call): Likewise.
12526         * gimple-loop-interchange.cc
12527         (loop_cand::analyze_iloop_reduction_var): Update for change to
12528         check_reduction_path.
12529         (tree_loop_interchange::interchange): Update for change to
12530         find_loop_location.
12531         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
12532         change in return-type of find_loop_location.
12533         (graphite_regenerate_ast_isl): Likewise.
12534         * graphite-optimize-isl.c (optimize_isl): Likewise.
12535         * graphite.c (graphite_transform_loops): Likewise.
12536         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
12537         pass in a dump_location_t rather than a location_t, via the
12538         gimple stmt.
12539         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12540         * ipa.c (walk_polymorphic_call_targets): Likewise.
12541         * loop-unroll.c (report_unroll): Convert "locus" param from
12542         location_t to dump_location_t.
12543         (decide_unrolling): Update for change to get_loop_location's
12544         return type.
12545         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
12546         location_t to dump_user_location_t.
12547         (grid_find_single_omp_among_assignments_1): Updates calls to
12548         dump_printf_loc to pass in a dump_location_t rather than a
12549         location_t, via the gimple stmt.
12550         (grid_parallel_clauses_gridifiable): Convert "tloc" from
12551         location_t to dump_location_t.  Updates calls to dump_printf_loc
12552         to pass in a dump_location_t rather than a location_t, via the
12553         gimple stmt.
12554         (grid_inner_loop_gridifiable_p): Likewise.
12555         (grid_dist_follows_simple_pattern): Likewise.
12556         (grid_gfor_follows_tiling_pattern): Likewise.
12557         (grid_target_follows_gridifiable_pattern): Likewise.
12558         (grid_attempt_target_gridification): Convert initialization
12559         of local "grid" from memset to zero-initialization; FIXME: does
12560         this require C++11?  Update call to dump_printf_loc to pass in a
12561         optinfo_location rather than a location_t, via the gimple stmt.
12562         * profile.c (read_profile_edge_counts): Updates call to
12563         dump_printf_loc to pass in a dump_location_t rather than a
12564         location_t
12565         (compute_branch_probabilities): Likewise.
12566         * selftest-run-tests.c (selftest::run_tests): Call
12567         dumpfile_c_tests.
12568         * selftest.h (dumpfile_c_tests): New decl.
12569         * tree-loop-distribution.c (pass_loop_distribution::execute):
12570         Update for change in return type of find_loop_location.
12571         * tree-parloops.c (parallelize_loops): Likewise.
12572         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
12573         "locus" from location_t to dump_user_location_t.
12574         (canonicalize_loop_induction_variables): Likewise.
12575         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
12576         for change in return type of find_loop_location.
12577         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
12578         to dump_printf_loc to pass in a dump_location_t rather than a
12579         location_t, via the stmt.
12580         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
12581         Likewise.
12582         * tree-vect-loop-manip.c (find_loop_location): Convert return
12583         type from source_location to dump_user_location_t.
12584         (vect_do_peeling): Update for above change.
12585         (vect_loop_versioning): Update for change in type of
12586         vect_location.
12587         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
12588         from location_t to dump_user_location_t.
12589         (vect_estimate_min_profitable_iters): Update for change in type
12590         of vect_location.
12591         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
12592         location_t to dump_location_t.
12593         (vect_slp_bb): Update for change in type of vect_location.
12594         * tree-vectorizer.c (vect_location): Convert from source_location
12595         to dump_user_location_t.
12596         (try_vectorize_loop_1): Update for change in vect_location's type.
12597         (vectorize_loops): Likewise.
12598         (increase_alignment): Likewise.
12599         * tree-vectorizer.h (vect_location): Convert from source_location
12600         to dump_user_location_t.
12601         (find_loop_location): Convert return type from source_location to
12602         dump_user_location_t.
12603         (check_reduction_path): Convert 1st param from location_t to
12604         dump_user_location_t.
12605         * value-prof.c (check_counter): Update call to dump_printf_loc to
12606         pass in a dump_user_location_t rather than a location_t; update
12607         call to error_at for change in type of "locus".
12608         (check_ic_target): Update call to dump_printf_loc to
12609         pass in a dump_user_location_t rather than a location_t, via the
12610         call_stmt.
12612 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12614         * config/s390/s390.h (enum processor_flags): Do not use
12615         default tune parameter when -march was specified.
12617 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
12619         PR target/86314
12620         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12621         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12622         operands.
12624 2018-06-26  Richard Biener  <rguenther@suse.de>
12626         PR tree-optimization/86287
12627         PR bootstrap/86316
12628         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12629         (vect_analyze_loop): Initialize n_stmts.
12631 2018-06-26  Richard Biener  <rguenther@suse.de>
12633         PR middle-end/86271
12634         * fold-const.c (fold_convertible_p): Pointer extension
12635         isn't valid.
12637 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
12639         PR debug/86064
12640         * dwarf2out.c (loc_list_has_views): Adjust comments.
12641         (dw_loc_list): Split single cross-partition range with
12642         nonzero locview.
12644 2018-06-25  Jeff Law  <law@redhat.com>
12646         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12647         on -mbig-switch by default.
12649         * config/v850/predicates.md (const_float_1_operand): Fix match_code
12650         test.
12651         (const_float_0_operand): Remove unused predicate.
12652         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12653         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12654         (recipsf2): New expander.  Original pattern now called
12655         (recipsf2_insn).
12656         (recipdf2, recipdf2_insn): Similarly.
12657         (rsqrtsf2, rsqrtsf2_insn): Similarly
12658         (rsqrtdf2, rsqrtdf2_insn): Similarly
12660 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
12662         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12663         Simplify logic for FreeBSD (twice).
12665 2018-06-25  Martin Sebor  <msebor@redhat.com>
12667         PR tree-optimization/86204
12668         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12669         a strnlen result if it's less than the length of the string.
12671 2018-06-25  Martin Sebor  <msebor@redhat.com>
12673         PR tree-optimization/85700
12674         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12675         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12676         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12678 2018-06-25  Martin Sebor  <msebor@redhat.com>
12680         * doc/extend.texi (Zero-length arrays): Update and clarify.
12682 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
12684         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12685         added IEEE/IBM long double multilib support on PowerPC little
12686         endian Linux systems.
12687         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12688         (MULTILIB_DEFAULTS): Likewise.
12689         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12690         Likewise.
12691         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12692         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12693         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12695 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
12697         PR middle-end/86311
12698         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12699         (REORDER_45): Likewise.
12701 2018-06-25  Jeff Law  <law@redhat.com>
12703         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12704         dividend to 32 bits.  Adjust length.
12705         (udivmodhi4): Cleanup output template.  Fix length.
12707 2018-06-25  Carl Love  <cel@us.ibm.com>
12709         * config/rs6000/vsx.md: Change word selector to prefered location.
12711 2018-06-25  Richard Biener  <rguenther@suse.de>
12713         PR tree-optimization/86304
12714         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12715         epilogue-if-converted loops as well.
12717 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
12719         * lto-section-out.c (lto_begin_section): Do not print section
12720         name for noaddr and unnumbered dumps.
12722 2018-06-25  Richard Biener  <rguenther@suse.de>
12724         * tree-vectorizer.h (struct vec_info_shared): New structure
12725         with parts split out from struct vec_info and loop_nest from
12726         struct _loop_vec_info.
12727         (struct vec_info): Adjust accordingly.
12728         (struct _loop_vec_info): Likewise.
12729         (LOOP_VINFO_LOOP_NEST): Adjust.
12730         (LOOP_VINFO_DATAREFS): Likewise.
12731         (LOOP_VINFO_DDRS): Likewise.
12732         (struct _bb_vec_info): Likewise.
12733         (BB_VINFO_DATAREFS): Likewise.
12734         (BB_VINFO_DDRS): Likewise.
12735         (struct _stmt_vec_info): Add dr_aux member.
12736         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12737         (DR_MISALIGNMENT_UNINITIALIZED): New.
12738         (set_dr_misalignment): Adjust.
12739         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12740         (vect_analyze_loop): Adjust prototype.
12741         (vect_analyze_loop_form): Likewise.
12742         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12743         Compute dependences lazily.
12744         (vect_record_base_alignments): Use shared datarefs/ddrs.
12745         (vect_verify_datarefs_alignment): Likewise.
12746         (vect_analyze_data_refs_alignment): Likewise.
12747         (vect_analyze_data_ref_accesses): Likewise.
12748         (vect_analyze_data_refs): Likewise.
12749         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12750         constructor parameter for shared part.
12751         (vect_analyze_loop_form): Pass in shared part and adjust.
12752         (vect_analyze_loop_2): Pass in storage for the number of
12753         stmts.  Move loop nest finding to the caller.  Compute
12754         datarefs lazily.
12755         (vect_analyze_loop): Pass in shared part.
12756         (vect_transform_loop): Verify shared datarefs are unchanged.
12757         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12758         constructor parameter for shared part.
12759         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12760         (vect_slp_bb): Verify shared datarefs are unchanged before
12761         transform.
12762         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12763         change.
12764         (new_stmt_vec_info): Initialize DR_AUX misalignment to
12765         DR_MISALIGNMENT_UNINITIALIZED.
12766         * tree-vectorizer.c (vec_info::vec_info): Add constructor
12767         parameter for shared part.
12768         (vec_info::~vec_info): Adjust.
12769         (vec_info_shared::vec_info_shared): New.
12770         (vec_info_shared::~vec_info_shared): Likewise.
12771         (vec_info_shared::save_datarefs): Likewise.
12772         (vec_info_shared::check_datarefs): Likewise.
12773         (try_vectorize_loop_1): Construct shared part live for analyses
12774         of a single loop for multiple vector sizes.
12775         * tree-parloops.c (gather_scalar_reductions): Adjust.
12777 2018-06-25  Richard Biener  <rguenther@suse.de>
12779         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12780         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12781         (vect_analyze_data_refs): Remove similar code from here and
12782         simplify accordingly.
12784 2018-06-25  Richard Biener  <rguenther@suse.de>
12786         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12787         for reverse storage order accesses rather than asserting
12788         they cannot happen here.
12790 2018-06-25  Tom de Vries  <tdevries@suse.de>
12792         PR debug/86257
12793         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12794         Use data16 instead of .byte for insn prefix.
12796 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
12798         PR C++/86082
12799         * parser.c (make_char_string_pack): Pass this literal chars
12800         through cpp_interpret_string.
12801         (cp_parser_userdef_numeric_literal): Check the result of
12802         make_char_string_pack.
12804 2018-06-24  Maya Rashish  <coypu@sdf.org>
12806         * ginclude/stddef.h: Simplify conditions around avoiding
12807         re-definition of __size_t.
12809 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12811         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12812         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12814 2018-06-22  Maya Rashish  <coypu@sdf.org>
12816         * doc/invoke.texi (mno-fancy-math-387): Update for changes
12817         made to OpenBSD and NetBSD through the years.
12819 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12821         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12822         behavior of vec_pack (vector double, vector double) to match
12823         behavior of vec_float2 (vector double, vector double).
12825 2018-06-22  Olivier Hainque  <hainque@adacore.com>
12827         * gimplify.c (gimplify_function_tree): Prevent creation
12828         of a trampoline for the address of the current function
12829         passed to entry/exit instrumentation hooks.
12831 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
12833         PR target/86222
12834         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12835         correctly.
12837 2018-06-22  Martin Liska  <mliska@suse.cz>
12839         PR tree-optimization/86263
12840         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12841         Bail out if is_enabled is false.
12842         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12843         New declaration.
12844         (jump_table_cluster::is_enabled): New function.
12846 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12848         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12849         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12850         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12851         (lto_input_ts_binfo_tree_pointers): Likewise.
12852         * tree-streamer-out.c (streamer_write_tree_bitfields,
12853         write_ts_binfo_tree_pointers): Likewise.
12854         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12856 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12858         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12860 2018-06-22  Martin Liska  <mliska@suse.cz>
12862         * symbol-summary.h (get): Make it pure and inline move
12863         functionality from ::get function.
12864         (get): Remove and inline into ::get and ::get_create.
12865         (get_create): Move code from ::get function.
12867 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12869         PR target/85994
12870         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12871         -x assembler-with-cpp.
12873 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12875         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12876         _FILE_OFFSET_BITS=64 for C++.
12878 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
12880         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12881         conversion insn that shows up when pr85657-3.c is compiled using
12882         IEEE 128-bit long double.
12883         (neg<mode>2_internal): Use the correct mode to check whether the
12884         mode is IBM extended.
12885         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12886         multiply and divide external functions from being created more
12887         than once.
12889 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12891         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12892         functions.
12893         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
12894         the edge can be forwarded.
12895         (cfg_layout_merge_blocks): Likewise.
12897 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12899         * except.c (finish_eh_generation): Commit edge insertions only after
12900         the EH edges have been redirected from post-landing to landing pads.
12902 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12904         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12905         create_tmp_var_for to create the FRAME decl.
12906         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12908 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12910         * tree-inline.c (copy_edges_for_bb): Minor tweak.
12911         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12912         debug statement when resetting its value.
12913         (expand_call_inline): Copy the locus of the call onto the assignment
12914         of the return value, if any.  Use local variable in more cases.
12916 2018-06-21  Martin Liska  <mliska@suse.cz>
12918         * ipa-pure-const.c (propagate_nothrow): Use
12919         funct_state_summaries->get.
12920         (dump_malloc_lattice): Likewise.
12921         (propagate_malloc): Likewise.
12923 2018-06-21  Richard Biener  <rguenther@suse.de>
12925         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12926         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12927         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12928         comment.
12929         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12930         BLOCK_ABSTRACT_ORIGIN unconditionally.
12932 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
12934         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12935         deleting it.
12936         * ipa-reference.c (ipa_reference_c_finalize): Delete
12937         ipa_ref_opt_sum_summaries and set it to NULL.
12939 2018-06-21  Tom de Vries  <tdevries@suse.de>
12941         PR tree-optimization/85859
12942         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12943         test with comment from bb_no_side_effects_p.
12945 2018-06-21  Richard Biener  <rguenther@suse.de>
12947         PR tree-optimization/86232
12948         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12949         max for constant niter.
12951 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12953         * config/aarch64/aarch64-simd.md
12954         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12956 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12958         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12959         Make opernads of the unspec commutative.
12961 2018-06-21  Richard Biener  <rguenther@suse.de>
12963         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12964         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12965         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12966         (vect_analyze_data_ref_dependence): Re-order checks to deal with
12967         NULL DR_STEP.
12968         (vect_record_base_alignments): Do not record base alignment
12969         for gathers or scatters.
12970         (vect_compute_data_ref_alignment): Drop return value that is always
12971         true.  Bail out early for gathers or scatters.
12972         (vect_enhance_data_refs_alignment): Bail out early for gathers
12973         or scatters.
12974         (vect_find_same_alignment_drs): Likewise.
12975         (vect_analyze_data_refs_alignment): Remove dead code.
12976         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12977         (vect_analyze_data_refs): For possible gathers or scatters do
12978         not create an alternate DR, just check their possible validity
12979         and mark them.  Adjust DECL_NONALIASED handling to not rely
12980         on DR_BASE_ADDRESS.
12981         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12982         update inits of gathers or scatters.
12983         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12984         Also copy gather/scatter flag to pattern vinfo.
12986 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12988         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12989         behavior of vec_packsu (vector unsigned long long, vector unsigned
12990         long long) to match behavior of vec_packs with same signature.
12992 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
12993             Thomas Schwinge <thomas@codesourcery.com>
12994             Cesar Philippidis  <cesar@codesourcery.com>
12996         * gimplify.c (gimplify_scan_omp_clauses): Add support for
12997         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12998         (gimplify_adjust_omp_clauses): Likewise.
12999         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
13000         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
13001         (gimplify_omp_target_update): Update handling of acc update and
13002         enter/exit data.
13003         * omp-low.c (install_var_field): Remove unused parameter
13004         base_pointers_restrict.
13005         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
13006         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
13007         FINALIZE}
13008         (omp_target_base_pointers_restrict_p): Delete.
13009         (scan_omp_target): Update call to scan_sharing_clauses.
13010         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
13011         FINALIZE}.
13012         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
13013         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13014         (convert_local_omp_clauses): Likewise.
13015         * tree-pretty-print.c (dump_omp_clause): Likewise.
13016         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
13017         FINALIZE}.
13018         (omp_clause_code_name): Likewise.
13020 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
13022         PR debug/86194
13023         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
13024         be narrowed.
13026         PR tree-optimization/86231
13027         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
13028         anti-range don't overwrite *vr0min before using it to compute *vr0max.
13030 2018-06-20  Tom de Vries  <tdevries@suse.de>
13032         PR tree-optimization/86097
13033         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
13034         iv type if signedness of iv type is not the same as that of *nit.
13036 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
13038         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
13039         EDGE_EH edges, verify they are all EDGE_EH.
13041 2018-06-20  Maya Rashish  <coypu@sdf.org>
13043         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
13045 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13047         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
13048         * config/aarch64/aarch64.c (xgene1_tunings): Add
13049         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
13050         (aarch64_mode_valid_for_sched_fusion_p):
13051         Allow 16-byte modes.
13052         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
13053         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
13054         128-bit modes.
13055         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
13056         New pattern.
13057         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
13058         * config/aarch64/iterators.md (VQ2): New mode iterator.
13060 2018-06-20  Martin Liska  <mliska@suse.cz>
13062         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13063         Change default ratio from 10 to 8.
13065 2018-06-20  Martin Liska  <mliska@suse.cz>
13067         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13068         New.
13069         (bit_test_cluster::find_bit_tests): Likewise.
13070         (switch_decision_tree::analyze_switch_statement): Find clusters.
13071         * tree-switch-conversion.h (struct jump_table_cluster): Document
13072         hierarchy.
13074 2018-06-20  Martin Liska  <mliska@suse.cz>
13076         * tree-switch-conversion.c (switch_conversion::collect):
13077         Record m_uniq property.
13078         (switch_conversion::expand): Bail out for special conditions.
13079         (group_cluster::~group_cluster): New.
13080         (group_cluster::group_cluster): Likewise.
13081         (group_cluster::dump): Likewise.
13082         (jump_table_cluster::emit): New.
13083         (switch_decision_tree::fix_phi_operands_for_edges): New.
13084         (struct case_node): Remove struct.
13085         (jump_table_cluster::can_be_handled): New.
13086         (case_values_threshold): Moved to header.
13087         (reset_out_edges_aux): Likewise.
13088         (jump_table_cluster::is_beneficial): New.
13089         (bit_test_cluster::can_be_handled): Likewise.
13090         (add_case_node): Remove.
13091         (bit_test_cluster::is_beneficial): New.
13092         (case_bit_test::cmp): New.
13093         (bit_test_cluster::emit): New.
13094         (expand_switch_as_decision_tree_p): Remove.
13095         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
13096         (fix_phi_operands_for_edge): Likewise.
13097         (switch_decision_tree::analyze_switch_statement): New.
13098         (compute_cases_per_edge): Move ...
13099         (switch_decision_tree::compute_cases_per_edge): ... here.
13100         (try_switch_expansion): Likewise.
13101         (switch_decision_tree::try_switch_expansion): Likewise.
13102         (record_phi_operand_mapping): Likewise.
13103         (switch_decision_tree::record_phi_operand_mapping): Likewise.
13104         (emit_case_decision_tree): Likewise.
13105         (switch_decision_tree::emit): Likewise.
13106         (balance_case_nodes): Likewise.
13107         (switch_decision_tree::balance_case_nodes): Likewise.
13108         (dump_case_nodes): Likewise.
13109         (switch_decision_tree::dump_case_nodes): Likewise.
13110         (emit_jump): Likewise.
13111         (switch_decision_tree::emit_jump): Likewise.
13112         (emit_cmp_and_jump_insns): Likewise.
13113         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
13114         (emit_case_nodes): Likewise.
13115         (switch_decision_tree::emit_case_nodes): Likewise.
13116         (conditional_probability): Remove.
13117         * tree-switch-conversion.h (enum cluster_type): New.
13118         (PRINT_CASE): New.
13119         (struct cluster): Likewise.
13120         (cluster::cluster): Likewise.
13121         (struct simple_cluster): Likewise.
13122         (simple_cluster::simple_cluster): Likewise.
13123         (struct group_cluster): Likewise.
13124         (struct jump_table_cluster): Likewise.
13125         (struct bit_test_cluster): Likewise.
13126         (struct min_cluster_item): Likewise.
13127         (struct case_tree_node): Likewise.
13128         (case_tree_node::case_tree_node): Likewise.
13129         (jump_table_cluster::case_values_threshold): Likewise.
13130         (struct case_bit_test): Likewise.
13131         (struct switch_decision_tree): Likewise.
13132         (struct switch_conversion): Likewise.
13133         (switch_decision_tree::reset_out_edges_aux): Likewise.
13135 2018-06-20  Martin Liska  <mliska@suse.cz>
13137         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
13138         (hoist_edge_and_branch_if_true): Likewise.
13139         (expand_switch_using_bit_tests_p): Likewise.
13140         (struct case_bit_test): Likewise.
13141         (case_bit_test_cmp): Likewise.
13142         (emit_case_bit_tests): Likewise.
13143         (switch_conversion::switch_conversion): New class.
13144         (struct switch_conv_info): Remove old struct.
13145         (collect_switch_conv_info): More to ...
13146         (switch_conversion::collect): ... this.
13147         (check_range): Likewise.
13148         (switch_conversion::check_range): Likewise.
13149         (check_all_empty_except_final): Likewise.
13150         (switch_conversion::check_all_empty_except_final): Likewise.
13151         (check_final_bb): Likewise.
13152         (switch_conversion::check_final_bb): Likewise.
13153         (create_temp_arrays): Likewise.
13154         (switch_conversion::create_temp_arrays): Likewise.
13155         (free_temp_arrays): Likewise.
13156         (gather_default_values): Likewise.
13157         (switch_conversion::gather_default_values): Likewise.
13158         (build_constructors): Likewise.
13159         (switch_conversion::build_constructors): Likewise.
13160         (constructor_contains_same_values_p): Likewise.
13161         (switch_conversion::contains_same_values_p): Likewise.
13162         (array_value_type): Likewise.
13163         (switch_conversion::array_value_type): Likewise.
13164         (build_one_array): Likewise.
13165         (switch_conversion::build_one_array): Likewise.
13166         (build_arrays): Likewise.
13167         (switch_conversion::build_arrays): Likewise.
13168         (gen_def_assigns): Likewise.
13169         (switch_conversion::gen_def_assigns): Likewise.
13170         (prune_bbs): Likewise.
13171         (switch_conversion::prune_bbs): Likewise.
13172         (fix_phi_nodes): Likewise.
13173         (switch_conversion::fix_phi_nodes): Likewise.
13174         (gen_inbound_check): Likewise.
13175         (switch_conversion::gen_inbound_check): Likewise.
13176         (process_switch): Use the newly created class.
13177         (switch_conversion::expand): New.
13178         (switch_conversion::~switch_conversion): New.
13179         * tree-switch-conversion.h: New file.
13181 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13183         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
13184         tree-vect-patterns.c.
13185         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
13186         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
13187         (vect_recog_sad_pattern): Likewise.
13188         (vect_recog_widen_sum_pattern): Likewise.
13189         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
13190         (vect_recog_widen_shift_pattern): Remove the type_in argument.
13191         (vect_recog_rotate_pattern): Likewise.
13192         (vect_recog_mult_pattern): Likewise.
13193         (vect_recog_vector_vector_shift_pattern): Likewise.
13194         (vect_recog_divmod_pattern): Likewise.
13195         (vect_recog_mixed_size_cond_pattern): Likewise.
13196         (vect_recog_bool_pattern): Likewise.
13197         (vect_recog_mask_conversion_pattern): Likewise.
13198         (vect_try_gather_scatter_pattern): Likewise.
13199         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
13200         (vect_recog_over_widening_pattern): Likewise.
13201         (vect_recog_gather_scatter_pattern): Likewise.
13202         (vect_recog_func_ptr): Move from tree-vectorizer.h
13203         (vect_vect_recog_func_ptrs): Move further down the file.
13204         (vect_recog_func): Likewise.  Remove the third argument.
13205         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
13206         (vect_pattern_recog_1): Expect the pattern function to do any
13207         necessary target tests.  Also expect it to provide a vector type.
13208         Remove the type_in handling.
13210 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13212         * tree-vect-patterns.c (vect_pattern_detected): New function.
13213         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
13214         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
13215         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
13216         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
13217         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13218         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13219         (vect_recog_mask_conversion_pattern)
13220         (vect_try_gather_scatter_pattern): Likewise.
13222 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13224         * tree-vect-patterns.c (vect_get_internal_def): New function.
13225         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13226         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
13227         (search_type_for_mask_1): Use it.
13229 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13231         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13232         redundant WIDEN_SUM_EXPR handling.
13233         (vect_recog_sad_pattern): Likewise.
13235 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13237         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
13238         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
13239         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
13240         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
13241         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
13243 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13245         * tree-vect-stmts.c (vectorizable_call): Make sure that we
13246         use the stmt_vec_info of the original bb statement for the
13247         new zero assignment, even if the call is part of a pattern.
13249 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13251         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
13252         that the sequence is attached to the original statement rather
13253         than the pattern statement.
13254         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
13255         PATTERN_DEF_SEQ from the original statement rather than
13256         the main pattern statement.
13257         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
13258         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
13259         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
13261 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
13263         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
13264         definition statements before the early exit for statements that aren't
13265         live or relevant.
13266         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
13267         split out from...
13268         (vect_transform_loop): ...here.  Process pattern definition
13269         statements without first checking whether the main pattern
13270         statement is live or relevant.
13272 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13274         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
13275         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
13277 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
13279         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
13280         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
13281         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
13282         (expand_block_compare): Change select_block_compare_mode call.
13283         (expand_strncmp_align_check): Use new functions, fix comment.
13284         (emit_final_str_compare_gpr): New function.
13285         (expand_strn_compare): Refactor and clean up code.
13286         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
13288 2018-06-19  Tony Reix  <tony.reix@atos.com>
13289             Damien Bergamini  <damien.bergamini@atos.com>
13290             David Edelsohn  <dje.gcc@gmail.com>
13292         * collect2.c (static_obj): New variable.
13293         (static_libs): New variable.
13294         (is_in_list): Uncomment declaration.
13295         (main): Track AIX libraries linked statically.
13296         (is_in_list): Uncomment definition.
13297         (scan_prog_file): Don't add AIX shared libraries initializer
13298         to constructor list if linking statically.
13300 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
13302         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
13303         constant.
13304         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
13306 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
13308         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
13309         blocks.
13311 2018-06-19  Martin Liska  <mliska@suse.cz>
13313         * config/i386/i386.c (ix86_can_inline_p): Do not use
13314         ipa_fn_summaries::get_create.
13315         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
13316         get.
13317         (devirtualization_time_bonus): Likewise.
13318         (ipcp_propagate_stage): Likewise.
13319         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13320         (edge_set_predicate): Likewise.
13321         (evaluate_conditions_for_known_args): Likewise.
13322         (evaluate_properties_for_edge): Likewise.
13323         (ipa_call_summary::reset): Tranform to ...
13324         (ipa_call_summary::~ipa_call_summary): ... this.
13325         (ipa_fn_summary::reset): Transform to ...
13326         (ipa_fn_summary::~ipa_fn_summary): ... this.
13327         (ipa_fn_summary_t::remove): Rename to ...
13328         (ipa_fn_summary_t::remove_callees): ... this.
13329         (ipa_fn_summary_t::duplicate): Use placement new
13330         instead of memory copy.
13331         (ipa_call_summary_t::duplicate): Likewise.
13332         (ipa_call_summary_t::remove): Remove.
13333         (dump_ipa_call_summary): Change get_create to get.
13334         (ipa_dump_fn_summary): Dump only when summary exists.
13335         (analyze_function_body): Use symbol_summary::get instead
13336         of get_create.
13337         (compute_fn_summary): Likewise.
13338         (estimate_edge_devirt_benefit): Likewise.
13339         (estimate_edge_size_and_time): Likewise.
13340         (inline_update_callee_summaries): Likewise.
13341         (remap_edge_change_prob): Likewise.
13342         (remap_edge_summaries): Likewise.
13343         (ipa_merge_fn_summary_after_inlining): Likewise.
13344         (write_ipa_call_summary): Likewise.
13345         (ipa_fn_summary_write): Likewise.
13346         (ipa_free_fn_summary): Likewise.
13347         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
13348         (struct ipa_call_summary): Likewise.
13349         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
13350         of get_create.
13351         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
13352         (estimate_size_after_inlining): Likewise.
13353         (estimate_growth): Likewise.
13354         (growth_likely_positive): Likewise.
13355         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13356         (inline_call): Likewise.
13357         * ipa-inline.c (caller_growth_limits): Likewise.
13358         (can_inline_edge_p): Likewise.
13359         (can_inline_edge_by_limits_p): Likewise.
13360         (compute_uninlined_call_time): Likewise.
13361         (compute_inlined_call_time): Likewise.
13362         (want_inline_small_function_p): Likewise.
13363         (edge_badness): Likewise.
13364         (update_caller_keys): Likewise.
13365         (update_callee_keys): Likewise.
13366         (inline_small_functions): Likewise.
13367         (inline_to_all_callers_1): Likewise.
13368         (dump_overall_stats): Likewise.
13369         (early_inline_small_functions): Likewise.
13370         (early_inliner): Likewise.
13371         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13372         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13373         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13374         * ipa-split.c (execute_split_functions): Likewise.
13375         * symbol-summary.h: Likewise.
13376         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13378 2018-06-19  Richard Biener  <rguenther@suse.de>
13380         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
13381         (vectorize_loops): ... here.  Fix dbgcnt handling.
13382         (try_vectorize_loop): Wrap try_vectorize_loop_1.
13384 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
13386         PR target/86197
13387         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
13388         ieee128 argument takes up only one (vector) register, not two (floating
13389         point) registers.
13391 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
13393         * gimplify.c (gimplify_init_constructor): Really never clear for an
13394         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
13396 2018-06-19  Richard Biener  <rguenther@suse.de>
13398         PR tree-optimization/86179
13399         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
13400         after failed recognition.
13402 2018-06-18  Martin Sebor  <msebor@redhat.com>
13404         PR middle-end/85602
13405         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
13406         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
13407         Handle integer subtraction.
13408         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
13409         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
13411 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13413         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
13414         param from rtx to rtx_insn *.
13415         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
13416         param.
13417         (frv_ifcvt_modify_insn): Likwise.
13418         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
13419         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
13420         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
13421         as_a <rtx_insn *> cast to local "unprotected_region" once
13422         it's been established that it's not NULL or pc_rtx.
13423         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
13424         param "sethi" from rtx to rtx_insn *.
13425         (nds32_group_float_insns): Likewise for param "insn".
13426         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
13427         param.
13428         (vax_output_int_subtract): Likewise.
13429         * config/vax/vax.c (vax_output_int_add): Likewise for param
13430         "insn".
13431         (vax_output_int_subtract): Likewise.
13432         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
13433         (emit_pattern_after): Likewise for param "after".
13434         (emit_insn_after): Likewise.
13435         (emit_jump_insn_after): Likewise.
13436         (emit_call_insn_after): Likewise.
13437         (emit_debug_insn_after): Likewise.
13438         (emit_pattern_before): Likewise for param "before".
13439         (emit_insn_before): Likewise.
13440         (emit_jump_insn_before): Likewise.
13441         * final.c (get_insn_template): Likewise for param "insn", removing
13442         a cast.
13443         * output.h (get_insn_template): Likewise for 2nd param.
13444         * rtl.h (emit_insn_before): Likewise.
13445         (emit_jump_insn_before): Likewise.
13446         (emit_debug_insn_before_noloc): Likewise.
13447         (emit_insn_after): Likewise.
13448         (emit_jump_insn_after): Likewise.
13449         (emit_call_insn_after): Likewise.
13450         (emit_debug_insn_after): Likewise.
13451         (set_insn_deleted): Likewise for param.
13453 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
13455         PR target/85358
13456         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
13457         floating point modes, so that IFmode is numerically greater than
13458         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
13459         to declare the ordering.  This prevents IFmode from being
13460         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
13461         machine.  Include rs6000-modes.h to share the fractional values
13462         between genmodes* and the rest of the compiler.
13463         (IFmode): Likewise.
13464         (KFmode): Likewise.
13465         (TFmode): Likewise.
13466         * config/rs6000/rs6000-modes.h: New file.
13467         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
13468         meaning of rs6000_long_double_size so that 126..128 selects an
13469         appropriate 128-bit floating point type.
13470         (rs6000_option_override_internal): Likewise.
13471         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
13472         (TARGET_LONG_DOUBLE_128): Change the meaning of
13473         rs6000_long_double_size so that 126..128 selects an appropriate
13474         128-bit floating point type.
13475         (LONG_DOUBLE_TYPE_SIZE): Update comment.
13476         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
13477         source and destination to match the standard usage.
13478         (truncifkf2): Likewise.
13479         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
13480         ISA 2.07 to use an explicit clobber, instead of passing in a
13481         temporary.
13482         (copysign<mode>3_soft): Likewise.
13484 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
13486         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13487         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
13488         (vect_slp_analyze_instance_dependence): Likewise.
13489         (vect_enhance_data_refs_alignment): Likewise.
13490         (vect_analyze_data_refs_alignment): Likewise.
13491         (vect_slp_analyze_and_verify_instance_alignment
13492         (vect_analyze_data_ref_accesses): Likewise.
13493         (vect_prune_runtime_alias_test_list): Likewise.
13494         (vect_analyze_data_refs): Likewise.
13495         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
13496         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13497         (vect_analyze_scalar_cycles_1): Likewise.
13498         (vect_get_loop_niters): Likewise.
13499         (vect_analyze_loop_form_1): Likewise.
13500         (vect_update_vf_for_slp): Likewise.
13501         (vect_analyze_loop_operations): Likewise.
13502         (vect_analyze_loop): Likewise.
13503         (vectorizable_induction): Likewise.
13504         (vect_transform_loop): Likewise.
13505         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
13506         * tree-vect-slp.c (vect_analyze_slp): Likewise.
13507         (vect_make_slp_decision): Likewise.
13508         (vect_detect_hybrid_slp): Likewise.
13509         (vect_slp_analyze_operations): Likewise.
13510         (vect_slp_bb): Likewise.
13511         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
13512         (vectorizable_bswap): Likewise.
13513         (vectorizable_call): Likewise.
13514         (vectorizable_simd_clone_call): Likewise.
13515         (vectorizable_conversion): Likewise.
13516         (vectorizable_assignment): Likewise.
13517         (vectorizable_shift): Likewise.
13518         (vectorizable_operation): Likewise.
13519         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
13521 2018-06-18  Martin Sebor  <msebor@redhat.com>
13523         PR tree-optimization/81384
13524         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
13525         * builtins.c (expand_builtin_strnlen): New function.
13526         (expand_builtin): Call it.
13527         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
13528         * builtins.def (BUILT_IN_STRNLEN): New.
13529         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
13530         Warn for bounds in excess of maximum object size.
13531         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
13532         single-value ranges.  Handle strnlen.
13533         (handle_builtin_strlen): Handle strnlen.
13534         (strlen_check_and_optimize_stmt): Same.
13535         * doc/extend.texi (Other Builtins): Document strnlen.
13537 2018-06-18  Maya Rashish  <coypu@sdf.org>
13539         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
13540         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
13541         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
13543         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
13544         here to ...
13545         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
13547 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13549         * tree.c (escaped_string::escape): Replace cast to char * by
13550         const_cast<char *> (unescaped).
13552 2018-06-18  Nick Clifton  <nickc@redhat.com>
13554         PR 84195
13555         * tree.c (escaped_string): New class.  Converts an unescaped
13556         string into its escaped equivalent.
13557         (warn_deprecated_use): Use the new class to convert the
13558         deprecation message, if present.
13559         (test_escaped_strings): New self test.
13560         (test_c_tests): Add test_escaped_strings.
13561         * doc/extend.texi (deprecated): Add a note that the
13562         deprecation message is affected by the -fmessage-length
13563         option, and that control characters will be escaped.
13564         (#pragma GCC error): Document this pragma.
13565         (#pragma GCC warning): Likewise.
13566         * doc/invoke.texi (-fmessage-length): Document this option's
13567         effect on the #warning and #error preprocessor directives and
13568         the deprecated attribute.
13570 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
13572         * tree.c (decl_value_expr_lookup): Revert latest change.
13573         (decl_value_expr_insert): Likewise.
13575 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
13577         * gimplify.c (nonlocal_vlas): Delete.
13578         (nonlocal_vla_vars): Likewise.
13579         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
13580         referenced VLAs.
13581         (gimplify_body): Do not create and destroy nonlocal_vlas.
13582         * tree-nested.c: Include diagnostic.h.
13583         (use_pointer_in_frame): Tweak.
13584         (lookup_field_for_decl): Add assertion and declare the transformation.
13585         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
13586         internal error when the reference is in a wrong context.  Do not
13587         create a debug decl by default.
13588         (note_nonlocal_block_vlas): Delete.
13589         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
13590         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
13591         create a debug decl by default.
13592         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
13593         call is in a wrong context.
13594         (fixup_vla_decls): New function.
13595         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
13596         debug variables were created.
13597         * tree.c (decl_value_expr_lookup): Add checking assertion.
13598         (decl_value_expr_insert): Likewise.
13600 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13602         PR middle-end/82479
13603         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
13604         * tree-scalar-evolution.c (interpret_expr): Likewise.
13605         (expression_expensive_p): Likewise.
13606         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
13607         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
13608         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
13609         (ssa_defined_by_minus_one_stmt_p): New.
13611 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13613         PR middle-end/64946
13614         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
13615         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
13616         * dojump.c (do_jump): Likewise.
13617         * expr.c (expand_expr_real_2): Check operand type's sign.
13618         * fold-const.c (const_unop): Handle ABSU_EXPR.
13619         (fold_abs_const): Likewise.
13620         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13621         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13622         (strip_sign_op_1): Likesise.
13623         * match.pd: Add new pattern to generate ABSU_EXPR.
13624         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13625         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13626         * tree-eh.c (operation_could_trap_helper_p): Likewise.
13627         * tree-inline.c (estimate_operator_cost): Likewise.
13628         * tree-pretty-print.c (dump_generic_node): Likewise.
13629         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13630         * tree.def (ABSU_EXPR): New.
13632 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
13634         PR middle-end/86095
13635         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13636         documented as preserved for backward compatibility only.
13637         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13639         PR rtl-optimization/86108
13640         * bb-reorder.c (create_forwarder_block): Renamed to ...
13641         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
13642         jump from new landing pad to the second part.
13643         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13644         Adjust callers.
13646 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
13648         PR middle-end/85878
13649         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13650         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13651         Only call store_expr for halves if the mode is the same.
13653         PR middle-end/86123
13654         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13655         Fix up comment formatting.
13657 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13659         * typed-splay-tree.h (typed_splay_tree::remove): New function.
13660         (typed_splay_tree::closure,
13661         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13662         (typed_splay_tree::typed_splay_tree,
13663         typed_splay_tree::operator =): Declared private.
13664         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13665         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13666         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13667         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13668         typed_splay_tree::splay_tree_splay,
13669         typed_splay_tree::splay_tree_foreach_helper,
13670         typed_splay_tree::splay_tree_insert,
13671         typed_splay_tree::splay_tree_remove,
13672         typed_splay_tree::splay_tree_lookup,
13673         typed_splay_tree::splay_tree_predecessor,
13674         typed_splay_tree::splay_tree_successor,
13675         typed_splay_tree::splay_tree_min,
13676         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13677         (typed_splay_tree::root, typed_splay_tree::comp,
13678         typed_splay_tree::delete_key,
13679         typed_splay_tree::delete_value): New data members.
13680         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13681         typed_splay_tree::remove.
13683 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
13685         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13686         -mginv and -mno-ginv to the assembler.
13687         * config/mips/mips.opt (-mcrc): New option.
13688         (-mginv): Likewise.
13689         * doc/invoke.text (-mcrc): Document.
13690         (-mginv): Likewise.
13692 2018-06-15  Nick Clifton  <nickc@redhat.com>
13694         PR 84195
13695         * tree.c (escaped_string): New class.  Converts an unescaped
13696         string into its escaped equivalent.
13697         (warn_deprecated_use): Use the new class to convert the
13698         deprecation message, if present.
13699         (test_escaped_strings): New self test.
13700         (test_c_tests): Add test_escaped_strings.
13701         * doc/extend.texi (deprecated): Add a note that the
13702         deprecation message is affected by the -fmessage-length
13703         option, and that control characters will be escaped.
13704         (#pragma GCC error): Document this pragma.
13705         (#pragma GCC warning): Likewise.
13706         * doc/invoke.texi (-fmessage-length): Document this option's
13707         effect on the #warning and #error preprocessor directives and
13708         the deprecated attribute.
13710 2018-06-15  Richard Biener  <rguenther@suse.de>
13712         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13713         here, also noting vector size used.
13714         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
13715         size used in MSG_OPTIMIZED_LOCATIONS dump.
13716         (pass_slp_vectorize::execute): Adjust.
13718 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
13720         PR target/85968
13721         * config/arc/arc.c (arc_return_address_register): Fix
13722         if-condition.
13724 2018-06-15  Richard Biener  <rguenther@suse.de>
13726         PR middle-end/86159
13727         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13728         leave useless conversion stripping to force_gimple_operand_gsi.
13729         (gimplify_build2): Likewise.
13730         (gimplify_build1): Likewise.
13732 2018-06-15  Richard Biener  <rguenther@suse.de>
13734         PR middle-end/86076
13735         * tree-cfg.c (move_stmt_op): unshare invariant addresses
13736         before adjusting their block.
13738 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13740         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13741         multilibs for *-*-rtems*.
13742         * config/riscv/t-rtems: New file.
13744 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
13746         PR middle-end/86122
13747         * match.pd ((A +- CST1) +- CST2): Punt if last resort
13748         unsigned_type_for returns NULL.
13750         PR target/85945
13751         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13752         subregs of multi-word pseudos unless the float mode has word size.
13754 2018-06-14  Richard Biener  <rguenther@suse.de>
13756         PR middle-end/86139
13757         * tree-vect-generic.c (build_word_mode_vector_type): Remove
13758         duplicate and harmful type_hash_canon.
13759         * tree.c (type_hash_canon): Assert we didn't find ourselves.
13761 2018-06-14  Richard Biener  <rguenther@suse.de>
13763         PR ipa/86124
13764         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13765         NULL cgraph_node.
13767 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13769         * config/rtems.h (STDINT_LONG32): Define.
13771 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
13772             Prachi Godbole  <prachi.godbole@imgtec.com>
13774         * config/mips/mips-cpus.def: Define P6600.
13775         * config/mips/mips-tables.opt: Regenerate.
13776         * config/mips/mips.c (mips_ucbranch_type): New enum.
13777         (mips_rtx_cost_data): Add support for P6600.
13778         (mips_issue_rate): Likewise.
13779         (mips_multipass_dfa_lookahead): Likewise.
13780         (mips_avoid_hazard): Likewise.
13781         (mips_reorg_process_insns): Likewise.
13782         (mips_classify_branch_p6600): New function.
13783         * config/mips/mips.h (TUNE_P6600): New define.
13784         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13785         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13786         * config/mips/mips.md: Include p6600.md.
13787         (processor): Add p6600.
13788         * config/mips/p6600.md: New file.
13789         * doc/invoke.texi: Add p6600 to supported architectures.
13791 2018-06-13  Martin Sebor  <msebor@redhat.com>
13793         PR tree-optimization/86114
13794         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13795         of integer types.
13796         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13798 2018-06-13  Richard Biener  <rguenther@suse.de>
13800         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13801         Properly set vector type of the intermediate stmt.
13802         * tree-vect-stmts.c (vectorizable_operation): The destination
13803         var always has vectype_out type.
13805 2018-06-13  Jeff Law  <law@redhat.com>
13807         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13808         integer 0 for argument to print_rtl_with_bb.
13809         (rl78_reorg): Likewise.
13811 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
13813         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13814         from rtx to rtx_insn *.
13815         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13816         "label".
13817         (add_sched_insns_for_speculation): Likewise for local "target",
13818         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13819         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13820         from rtx_insn ** to rtx_code_label **.
13821         (reorg_emit_nops): Likewise.
13822         (c6x_reorg): Likewise for local "call_labels".
13823         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13824         rtx to rtx_insn *.
13825         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13826         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13827         the loops over LABEL_REFS.
13828         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13829         braf_label.
13830         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13831         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13832         (split_branches): Strengthen local "olabel" from rtx to
13833         rtx_insn *, adding a safe_as_a cast.
13834         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13835         to "rtx_insn *".
13836         (add_insn_after): Likewise for first two params.
13837         (add_insn_before): Likewise.
13838         (remove_insn): Likewise for param.
13839         (emit_pattern_before_noloc): Likewise for second and third params.
13840         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13841         (emit_call_insn_before_noloc): Likewise.
13842         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13843         to "rtx_insn *".
13844         (emit_barrier_before): Likewise.
13845         (emit_label_before): Strengthen "label" param from "rtx" to
13846         "rtx_code_label *".  Strengthen "before" param from "rtx" to
13847         "rtx_insn *".
13848         (emit_insn_after_1): Strengthen "after" param from "rtx" to
13849         "rtx_insn *".
13850         (emit_pattern_after_noloc): Likewise.
13851         (emit_insn_after_noloc): Likewise.
13852         (emit_jump_insn_after_noloc): Likewise.
13853         (emit_call_insn_after_noloc): Likewise.
13854         (emit_debug_insn_after_noloc): Likewise.
13855         (emit_barrier_after): Likewise.
13856         (emit_label_after): Likewise for both params.
13857         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
13858         "loc" param from "int" to "location_t".
13859         (emit_insn_after_setloc): Likewise.
13860         (emit_jump_insn_after_setloc): Likewise.
13861         (emit_call_insn_after_setloc): Likewise.
13862         (emit_debug_insn_after_setloc): Likewise.
13863         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
13864         "loc" param from "int" to "location_t".
13865         (emit_pattern_before): Convert NULL_RTX to NULL.
13866         (emit_insn_before_setloc): Convert "loc" param from "int" to
13867         "location_t".
13868         (emit_jump_insn_before_setloc): Likewise.
13869         (emit_call_insn_before_setloc): Likewise.
13870         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13871         rtx_insn *.  Convert "loc" param from "int" to "location_t".
13872         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13873         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13874         Convert 3rd param from "int" to "location_t".
13875         (emit_barrier_before, emit_barrier_after, next_real_insn):
13876         Strengthen param from rtx to rtx_insn *.
13877         (emit_label_before): Strengthen 1st param from "rtx" to
13878         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
13879         "rtx_insn *".
13880         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13881         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13882         Strengthen 2nd param from "rtx" to "rtx_insn *".
13883         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13884         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13885         Likewise. Convert 3rd param from "int" to "location_t".
13886         (emit_label_after): Strengthen 1st param from "rtx" to
13887         "rtx_code_label *".
13888         (next_real_insn, remove_insn): Strengthen param from "rtx" to
13889         "rtx_insn *".
13890         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13891         from "rtx" to "rtx_insn *".
13893 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
13895         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13896         bodies streamed in with -Q.
13897         * dumpfile.c (dump_files): Add lto-stream-out dump file.
13898         * dumpfile.h (tree_dump_index): Add lto_stream_out.
13899         * gimple-streamer-out.c: Include gimple-pretty-print.h
13900         (output_bb): Dump stmts streamed.
13901         * lto-section-out.c: Include print-tree.h
13902         (lto_begin_section): Dump sections created.
13903         (lto_output_decl_index): Dump decl encoded.
13904         * lto-streamer-out.c: Include print-tree.h
13905         (create_output_block): Dump output block created.
13906         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13907         (output_function): Dump function output.
13908         (output_constructor): Dump constructor streamed.
13909         (write_global_stream): Output indexes encoded.
13910         (produce_asm_for_decls): Dump streams encoded.
13911         * lto-streamer.c (streamer_dump_file): New global var.
13912         * lto-streamer.h (streamer_dump_file): Declare.
13913         * passes.c (ipa_write_summaries): Initialize streamer dump.
13914         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13915         in.
13917 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
13919         PR target/86048
13920         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13921         offsets for register save directives.  Emit a second batch of save
13922         directives, if need be, when the function accesses prior frames.
13924 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13926         * config/arc/fpu.md (fmasf4): Force operand to register.
13927         (fnmasf4): Likewise.
13929 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13931         * config/arc/arc-protos.h (arc_pad_return): Remove.
13932         * config/arc/arc.c (machine_function): Remove force_short_suffix
13933         and size_reason.
13934         (arc_print_operand): Adjust printing of '&'.
13935         (arc_verify_short): Remove conditional printing of short suffix.
13936         (arc_final_prescan_insn): Remove reference to size_reason.
13937         (pad_return): New function.
13938         (arc_reorg): Call pad_return.
13939         (arc_pad_return): Remove.
13940         (arc_init_machine_status): Remove reference to force_short_suffix.
13941         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13942         (attr length): When attribute iscompact is true force to 2
13943         regardless; in the case of maybe check if we want to force the
13944         instruction to have 4 bytes length.
13945         (nopv): Change it to generate 4 byte long nop as well.
13946         (blockage): New pattern.
13947         (simple_return): Remove call to arc_pad_return.
13948         (p_return_i): Likewise.
13950 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13952         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13954 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13956         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13957         ARC cores.
13959 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13961         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13962         for ARC700 and ARCv2.
13964 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
13966         PR target/86076
13967         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13968         operands[2] instead of operands[1].
13971 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13973         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13974         case, check whether the outer register overlaps an unallocatable
13975         register, not just whether it fits the required class.
13977 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13979         * poly-int.h (can_div_trunc_p): Add new overload in which all values
13980         are poly_ints.
13981         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13982         (memrefs_conflict_p): Likewise.
13983         (init_alias_analysis): Likewise.
13984         * cfgexpand.c (expand_debug_expr): Likewise.
13985         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13986         * cse.c (fold_rtx): Likewise.
13987         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13988         * expr.c (emit_block_move_hints): Likewise.
13989         (clear_storage_hints, push_block, emit_push_insn): Likewise.
13990         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13991         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13992         (emit_group_store): Likewise.
13993         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
13994         to read the PRE/POST_MODIFY increment.
13995         * calls.c (store_one_arg): Use strip_offset.
13996         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13997         poly_int_rtx_p.
13998         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13999         by a VEC_SELECT.
14000         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
14001         (simplify_binary_operation_1): Extend CONST_INT handling to
14002         poly_int_rtx_p.
14003         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
14004         than a HOST_WIDE_INT.
14005         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
14006         poly_int64.
14007         (adjust_mems, add_stores): Update accodingly.
14008         (vt_canonicalize_addr): Track polynomial offsets.
14009         (emit_note_insn_var_location): Likewise.
14010         (vt_add_function_parameter): Likewise.
14011         (vt_initialize): Likewise.
14013 2018-06-12  Jeff Law  <law@redhat.com>
14015         * config.gcc (alpha*-*-freebsd*): Remove.
14016         * config/alpha/freebsd.h: Remove.
14018 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
14020         PR other/69968
14021         * spellcheck-tree.c (levenshtein_distance): Rename to...
14022         (get_edit_distance): ...this, and update for underlying renaming.
14023         * spellcheck-tree.h (levenshtein_distance): Rename to...
14024         (get_edit_distance): ...this.
14025         * spellcheck.c (levenshtein_distance): Rename to...
14026         (get_edit_distance): ...this.  Convert from Levenshtein distance
14027         to Damerau-Levenshtein distance by supporting transpositions of
14028         adjacent characters.  Rename "v1" to "v_next" and "v0" to
14029         "v_one_ago".
14030         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
14031         (selftest::test_edit_distance_unit_test_oneway): ...this, and
14032         update for underlying renaming.
14033         (selftest::levenshtein_distance_unit_test): Rename to...
14034         (selftest::test_get_edit_distance_unit): ...this, and update for
14035         underlying renaming.
14036         (selftest::test_find_closest_string): Add example from PR 69968
14037         where transposition helps
14038         (selftest::test_metric_conditions): Update for renaming.
14039         (selftest::test_metric_conditions): Likewise.
14040         (selftest::spellcheck_c_tests): Likewise.
14041         * spellcheck.h (levenshtein_distance): Rename both overloads to...
14042         (get_edit_distance): ...this.
14043         (best_match::consider): Update for renaming.
14045 2018-06-12  Martin Sebor  <msebor@redhat.com>
14047         PR tree-optimization/85259
14048         * builtins.c (compute_objsize): Handle constant offsets.
14049         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
14050         true iff a warning has been issued.
14051         * gimple.h (gimple_nonartificial_location): New function.
14052         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
14053         gimple_nonartificial_location and handle -Wno-system-headers.
14054         (handle_builtin_stxncpy): Same.
14056 2018-06-12  Martin Sebor  <msebor@redhat.com>
14058         PR c/85931
14059         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
14061 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
14063         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
14064         BUILTIN_VEC_XST entries for pointer to double and long long.
14066 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
14068         PR target/85990
14069         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14070         Update comments.
14071         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14072         Likewise.
14074 2018-06-12  Martin Liska  <mliska@suse.cz>
14076         * doc/options.texi: Document IntegerRange.
14078 2018-06-12  Martin Liska  <mliska@suse.cz>
14080         * config/i386/i386.opt: Make MPX-related options as Deprecated.
14081         * opt-functions.awk: Handle Deprecated flag.
14082         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
14083         and report error.
14084         (read_cmdline_option): Report warning for a deprecated option.
14085         * opts.h (struct cl_option): Add new field cl_deprecated.
14086         (CL_ERR_DEPRECATED): New.
14088 2018-06-12  Martin Liska  <mliska@suse.cz>
14090         * doc/options.texi: Document Deprecated option flag.
14092 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14094         * config/arc/arc-arch.h (arc_extras): New enum.
14095         (arc_cpu_t):Add field extra.
14096         (arc_cpu_types): Consider the extras.
14097         * config/arc/arc-cpus.def: Add extras info.
14098         * config/arc/arc-opts.h (processor_type): Consider extra field.
14099         * config/arc/arc.c (arc_override_options): Handle extra field.
14101 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14103         * config/arc/arc-arch.h: Update ARC_OPTX macro.
14104         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
14105         field.
14106         * config/arc/arc.c (arc_init): Update pic warning.
14107         (irq_range): Update irq range parsing warnings.
14108         (arc_override_options): Update various warning messages.
14109         (arc_handle_aux_attribute): Likewise.
14111 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
14113         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
14115 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14117         * doc/sourcebuild.texi: Document usage of line number 0 in verify
14118         compiler messages directives.
14120 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
14122         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
14123         * config/mips/mips-tables.opt: Regenerate.
14124         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
14125         mips64r6.
14126         * doc/invoke.texi: Document -march=i6500.
14128 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
14130         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
14131         (i6400_gpmul): Add cpu_unit.
14132         (i6400_gpdiv): Likewise.
14133         (i6400_msa_add_d): Update reservations.
14134         (i6400_msa_int_add) Likewise.
14135         (i6400_msa_short_logic3) Likewise.
14136         (i6400_msa_short_logic2) Likewise.
14137         (i6400_msa_short_logic) Likewise.
14138         (i6400_msa_move) Likewise.
14139         (i6400_msa_cmp) Likewise.
14140         (i6400_msa_short_float2) Likewise.
14141         (i6400_msa_div_d) Likewise.
14142         (i6400_msa_long_logic1) Likewise.
14143         (i6400_msa_long_logic2) Likewise.
14144         (i6400_msa_mult) Likewise.
14145         (i6400_msa_long_float2) Likewise.
14146         (i6400_msa_long_float4) Likewise.
14147         (i6400_msa_long_float5) Likewise.
14148         (i6400_msa_long_float8) Likewise.
14149         (i6400_fpu_fadd): Include frint type.
14150         (i6400_fpu_store): New define_insn_reservation.
14151         (i6400_fpu_load): Likewise.
14152         (i6400_fpu_move): Likewise.
14153         (i6400_fpu_fcmp): Likewise.
14154         (i6400_fpu_fmadd): Likewise.
14155         (i6400_int_mult): Include imul3nc type and update reservation.
14156         (i6400_int_div): Include idiv3 type and update reservation.
14157         (i6400_int_load): Update to check type not move_type.
14158         (i6400_int_store): Likewise.
14159         (i6400_int_prefetch): Set zero latency.
14161 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
14163         * gcc.c: Document new %@{...} sequence.
14164         (LINK_COMMAND_SPEC): Use it for the -L switches.
14165         (cpp_unique_options): Use it for the -I switches.
14166         (at_file_argbuf): New global variable.
14167         (in_at_file): Likewise.
14168         (alloc_args): Create at_file_argbuf.
14169         (clear_args): Truncate at_file_argbuf.
14170         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
14171         (open_at_file): New function.
14172         (close_at_file): Likewise.
14173         (create_at_file): Delete.
14174         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
14175         <'o'>: Likewise.
14176         <'@'>: New case.
14177         (validate_switches_from_spec): Deal with %@{...} sequence.
14178         (validate_switches): Likewise.
14179         (driver::finalize): Call clear_args.
14181 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
14183         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
14185 2018-06-11  Martin Sebor  <msebor@redhat.com>
14187         * doc/invoke.texi (-Wall): List -Wc++17-compat.
14188         (Wno-class-memaccess): Add @opindex.
14189         (Wno-templates, Wno-multiple-inheritance): Same.
14190         (Wno-virtual-inheritance, Wno-namespaces): Same.
14191         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
14192         (Wno-format-overflow, Wno-format-truncation): Same.
14193         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
14194         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
14195         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
14196         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
14197         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
14198         (Wno-misspelled-isr): Same.
14200 2018-06-11  Martin Sebor  <msebor@redhat.com>
14202         * PR tree-optimization/86083
14203         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
14205 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
14207         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
14209 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
14211         PR target/85755
14212         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
14213         on the correct operand.
14214         (*movdi_internal64): Ditto.
14216 2018-06-11  Martin Liska  <mliska@suse.cz>
14218         PR tree-optimization/86089
14219         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
14221 2018-06-11  Julia Koval  <julia.koval@intel.com>
14223         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
14224         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
14225         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
14227 2018-06-11  Olivier Hainque  <hainque@adacore.com>
14229         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
14230         for Ada with strict dwarf2.
14232 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
14234         PR target/85755
14235         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
14236         addresses.
14238 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
14240         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
14242 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
14244         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
14245         TARGET_ELF.
14247 2018-06-08  Martin Liska  <mliska@suse.cz>
14249         * tree-cfg.h (debug_function): Fix argument type to match
14250         implementation.
14252 2018-06-08  Martin Liska  <mliska@suse.cz>
14254         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
14255         Remove usage of MPX-related (and removed) fields.
14256         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
14258 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
14260         * cfg.c (debug): Use TDF_NONE rather than 0.
14261         * cfghooks.c (debug): Likewise.
14262         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
14263         (struct dump_option_value_info): Convert to...
14264         (struct kv_pair): ...this template type.
14265         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
14266         rather than 0.
14267         (optinfo_verbosity_options): Likewise.
14268         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
14269         OPTGROUP_NONE.
14270         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
14271         than int for "optgroup_flags" param.
14272         (dump_generic_expr_loc): Use dump_flags_t rather than int for
14273         "dump_kind" param.
14274         (dump_dec): Likewise.
14275         (dump_finish): Use TDF_NONE rather than 0.
14276         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
14277         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
14278         than 0.  Update for change to option_ptr.
14279         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
14280         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
14281         0.  Update for changes to optinfo_verbosity_options and
14282         optgroup_options.
14283         (opt_info_switch_p): Convert optgroup_flags from int to
14284         optgroup_flags_t.
14285         (dump_basic_block): Use dump_flags_t rather than int
14286         for "dump_kind" param.
14287         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
14288         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
14289         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
14290         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
14291         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
14292         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
14293         TDF_NONE): Convert from macros to...
14294         (enum dump_flag): ...this new enum.
14295         (dump_flags_t): Update to use enum.
14296         (operator|, operator&, operator~, operator|=, operator&=):
14297         Implement for dump_flags_t.
14298         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
14299         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
14300         Convert from macros to...
14301         (enum optgroup_flag): ...this new enum.
14302         (optgroup_flags_t): New typedef.
14303         (operator|, operator|=): Implement for optgroup_flags_t.
14304         (struct dump_file_info): Convert field "alt_flags" to
14305         dump_flags_t.  Convert field "optgroup_flags" to
14306         optgroup_flags_t.
14307         (dump_basic_block): Use dump_flags_t rather than int for param.
14308         (dump_generic_expr_loc): Likewise.
14309         (dump_dec): Likewise.
14310         (dump_register): Convert param "optgroup_flags" to
14311         optgroup_flags_t.
14312         (opt_info_enable_passes): Likewise.
14313         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
14314         than 0.
14315         * gimple-pretty-print.c (debug): Likewise.
14316         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
14317         (merged_store_group::apply_stores): Likewise.
14318         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
14319         * gimple.c (verify_gimple_pp): Likewise.
14320         * graphite-poly.c (print_pbb_body): Likewise.
14321         * passes.c (pass_manager::register_one_dump_file): Convert
14322         local "optgroup_flags" to optgroup_flags_t.
14323         * print-tree.c (print_node): Use TDF_NONE rather than 0.
14324         (debug): Likewise.
14325         (debug_body): Likewise.
14326         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
14327         to optgroup_flags_t.
14328         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
14329         than 0.
14330         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
14331         (convert_mult_to_fma): Likewise.
14332         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
14333         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
14334         * tree-vect-data-refs.c (dump_lower_bound): Convert param
14335         "dump_kind" to dump_flags_t.
14337 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14339         * config/rs6000/rs6000.c (min, max): Delete.
14341 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
14343         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
14344         -mabi=spe and -mabi=no-spe.
14346 2018-06-08  Martin Liska  <mliska@suse.cz>
14348         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
14349         where we expect an existing summary.
14351 2018-06-08  Martin Liska  <mliska@suse.cz>
14353         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
14354         * ipa-inline.h (estimate_edge_growth): Likewise.
14356 2018-06-08  Martin Liska  <mliska@suse.cz>
14358         * cgraph.c (function_version_hasher::hash): Use
14359         cgraph_node::get_uid ().
14360         (function_version_hasher::equal):
14361         * cgraph.h (cgraph_node::get_uid): New method.
14362         * ipa-inline.c (update_caller_keys): Use
14363         cgraph_node::get_uid ().
14364         (update_callee_keys): Likewise.
14365         * ipa-utils.c (searchc): Likewise.
14366         (ipa_reduced_postorder): Likewise.
14367         * lto-cgraph.c (input_node): Likewise.
14368         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
14369         * symbol-summary.h (symtab_insertion): Likewise.
14370         (symtab_removal): Likewise.
14371         (symtab_duplication): Likewise.
14372         * tree-pretty-print.c (dump_function_header): Likewise.
14373         * tree-sra.c (convert_callers_for_node): Likewise.
14375 2018-06-08  Martin Liska  <mliska@suse.cz>
14377         * cgraph.c (symbol_table::create_edge): Always assign a new
14378         unique number.
14379         (symbol_table::free_edge): Do not recycle numbers.
14380         * cgraph.h (cgraph_edge::get): New method.
14381         * symbol-summary.h (symtab_removal): Use it.
14382         (symtab_duplication): Likewise.
14383         (call_summary::hashable_uid): Remove.
14385 2018-06-08  Martin Liska  <mliska@suse.cz>
14387         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
14388         (initialize_growth_caches): Remove.
14389         (free_growth_caches): Likewise.
14390         (do_estimate_edge_time): Use edge_growth_cache.
14391         (do_estimate_edge_size): Likewise.
14392         (do_estimate_edge_hints): Likewise.
14393         * ipa-inline.c (reset_edge_caches): Likewise.
14394         (recursive_inlining): Likewise.
14395         (inline_small_functions): Likewise.
14396         * ipa-inline.h (initialize_growth_caches): Remove.
14397         (estimate_edge_size): Likewise.
14398         (estimate_edge_time): Likewise.
14399         (estimate_edge_hints): Likewise.
14400         (reset_edge_growth_cache): Likewise.
14401         * symbol-summary.h (call_summary::remove): New method.
14403 2018-06-08  Martin Liska  <mliska@suse.cz>
14405         * ipa-cp.c (class edge_clone_summary): New summary.
14406         (grow_edge_clone_vectors): Remove.
14407         (ipcp_edge_duplication_hook): Remove.
14408         (class edge_clone_summary_t): New call_summary class.
14409         (ipcp_edge_removal_hook): Remove.
14410         (edge_clone_summary_t::duplicate): New function.
14411         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
14412         (create_specialized_node): Likewise.
14413         (ipcp_driver): Initialize edge_clone_summaries and do not
14414         register hooks.
14416 2018-06-08  Martin Liska  <mliska@suse.cz>
14418         * symbol-summary.h (get): New function.
14419         (call_summary::m_initialize_when_cloning): New class member.
14421 2018-06-08  Martin Liska  <mliska@suse.cz>
14423         * cgraph.c (cgraph_node::remove): Do not recycle uid.
14424         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
14425         (symbol_table::allocate_cgraph_symbol): Do not set uid.
14426         * passes.c (uid_hash_t): Record removed_nodes by their uids.
14427         (remove_cgraph_node_from_order): Use the removed_nodes set.
14428         (do_per_function_toporder): Likwise.
14429         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
14430         instead of summary_uid.
14431         (symtab_removal): Likewise.
14432         (symtab_duplication): Likewise.
14434 2018-06-08  Martin Liska  <mliska@suse.cz>
14436         * ipa-cp.c (ipcp_store_bits_results): Use
14437         ipcp_transformation_sum.
14438         (ipcp_store_vr_results): Likewise.
14439         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
14440         to ...
14441         (ipcp_transformation_initialize): ... this.
14442         (ipa_set_node_agg_value_chain):
14443         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
14444         (write_ipcp_transformation_info): Likewise.
14445         (read_ipcp_transformation_info): Likewise.
14446         (ipcp_update_bits): Likewise.
14447         (ipcp_update_vr): Likewise.
14448         (ipcp_transform_function): Likewise.
14449         * ipa-prop.h: Rename ipcp_transformation_summary to
14450         ipcp_transformation.
14451         (class ipcp_transformation_t): New function summary.
14452         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
14453         (ipa_get_agg_replacements_for_node): Likewise.
14455 2018-06-08  Martin Liska  <mliska@suse.cz>
14457         * ipa-pure-const.c (struct funct_state_d): Do it class instead
14458         of struct.
14459         (class funct_state_summary_t): New function_summary class.
14460         (has_function_state): Remove.
14461         (get_function_state): Likewise.
14462         (set_function_state): Likewise.
14463         (add_new_function): Likewise.
14464         (funct_state_summary_t::insert): New function.
14465         (duplicate_node_data): Remove.
14466         (remove_node_data): Remove.
14467         (funct_state_summary_t::duplicate): New function.
14468         (register_hooks): Create new funct_state_summaries.
14469         (pure_const_generate_summary): Use it.
14470         (pure_const_write_summary): Likewise.
14471         (pure_const_read_summary): Likewise.
14472         (propagate_pure_const): Likewise.
14473         (propagate_nothrow): Likewise.
14474         (dump_malloc_lattice): Likewise.
14475         (propagate_malloc): Likewise.
14476         (execute): Do not register hooks, just remove summary
14477         instead.
14478         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
14479         constructor.
14481 2018-06-08  Martin Liska  <mliska@suse.cz>
14483         * ipa-reference.c (remove_node_data): Remove.
14484         (duplicate_node_data): Likewise.
14485         (class ipa_ref_var_info_summary_t): New class.
14486         (class ipa_ref_opt_summary_t): Likewise.
14487         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
14488         (get_reference_optimization_summary): Use
14489         ipa_ref_opt_sum_summaries.
14490         (set_reference_vars_info): Remove.
14491         (set_reference_optimization_summary): Likewise.
14492         (ipa_init): Create summaries.
14493         (init_function_info): Use function summary.
14494         (ipa_ref_opt_summary_t::duplicate): New function.
14495         (ipa_ref_opt_summary_t::remove): New function.
14496         (get_read_write_all_from_node): Fix GNU coding style.
14497         (propagate): Use function summary.
14498         (write_node_summary_p): Fix GNU coding style.
14499         (stream_out_bitmap): Likewise.
14500         (ipa_reference_read_optimization_summary): Use function summary.
14501         (ipa_reference_c_finalize): Do not release hooks.
14503 2018-06-08  Martin Liska  <mliska@suse.cz>
14505         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
14506         (analyze_function_body): Extract multiple calls of get_create.
14507         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14508         * ipa-inline.c (recursive_inlining): Use ::get method.
14509         * ipa-inline.h (estimate_edge_growth): Likewise.
14511 2018-06-08  Martin Liska  <mliska@suse.cz>
14513         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
14514         HSA_INVALID.
14515         (hsa_function_summary::hsa_function_summary): Use the new enum
14516         value.
14517         (hsa_gpu_implementation_p): Use hsa_summaries::get.
14518         * hsa-gen.c (hsa_get_host_function): Likewise.
14519         (get_brig_function_name): Likewise.
14520         * ipa-hsa.c (process_hsa_functions): Likewise.
14521         (ipa_hsa_write_summary): Likewise.
14522         * symbol-summary.h (symtab_duplication): Use ::get function/
14523         (get): New function.
14525 2018-06-08  Martin Liska  <mliska@suse.cz>
14527         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
14528         of get.
14529         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
14530         (hsa_register_kernel): Likewise.
14531         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
14532         * hsa-gen.c (hsa_get_host_function): Likewise.
14533         (get_brig_function_name): Likewise.
14534         (generate_hsa): Likewise.
14535         (pass_gen_hsail::execute): Likewise.
14536         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
14537         (devirtualization_time_bonus): Likewise.
14538         (ipcp_propagate_stage): Likewise.
14539         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14540         (edge_set_predicate): Likewise.
14541         (evaluate_conditions_for_known_args): Likewise.
14542         (evaluate_properties_for_edge): Likewise.
14543         (ipa_fn_summary::reset): Likewise.
14544         (ipa_fn_summary_t::duplicate): Likewise.
14545         (dump_ipa_call_summary): Likewise.
14546         (ipa_dump_fn_summary): Likewise.
14547         (analyze_function_body): Likewise.
14548         (compute_fn_summary): Likewise.
14549         (estimate_edge_devirt_benefit): Likewise.
14550         (estimate_edge_size_and_time): Likewise.
14551         (estimate_calls_size_and_time): Likewise.
14552         (estimate_node_size_and_time): Likewise.
14553         (inline_update_callee_summaries): Likewise.
14554         (remap_edge_change_prob): Likewise.
14555         (remap_edge_summaries): Likewise.
14556         (ipa_merge_fn_summary_after_inlining): Likewise.
14557         (ipa_update_overall_fn_summary): Likewise.
14558         (read_ipa_call_summary): Likewise.
14559         (inline_read_section): Likewise.
14560         (write_ipa_call_summary): Likewise.
14561         (ipa_fn_summary_write): Likewise.
14562         (ipa_free_fn_summary): Likewise.
14563         * ipa-hsa.c (process_hsa_functions): Likewise.
14564         (ipa_hsa_write_summary): Likewise.
14565         (ipa_hsa_read_section): Likewise.
14566         * ipa-icf.c (sem_function::merge): Likewise.
14567         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14568         (do_estimate_edge_time): Likewise.
14569         (estimate_size_after_inlining): Likewise.
14570         (estimate_growth): Likewise.
14571         (growth_likely_positive): Likewise.
14572         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14573         (inline_call): Likewise.
14574         * ipa-inline.c (caller_growth_limits): Likewise.
14575         (can_inline_edge_p): Likewise.
14576         (can_inline_edge_by_limits_p): Likewise.
14577         (compute_uninlined_call_time): Likewise.
14578         (compute_inlined_call_time): Likewise.
14579         (want_inline_small_function_p): Likewise.
14580         (edge_badness): Likewise.
14581         (update_caller_keys): Likewise.
14582         (update_callee_keys): Likewise.
14583         (recursive_inlining): Likewise.
14584         (inline_small_functions): Likewise.
14585         (inline_to_all_callers_1): Likewise.
14586         (dump_overall_stats): Likewise.
14587         (early_inline_small_functions): Likewise.
14588         (early_inliner): Likewise.
14589         * ipa-inline.h (estimate_edge_growth): Likewise.
14590         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14591         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14592         * ipa-prop.h (IPA_NODE_REF): Likewise.
14593         (IPA_EDGE_REF): Likewise.
14594         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14595         (propagate_malloc): Likewise.
14596         * ipa-split.c (execute_split_functions): Likewise.
14597         * symbol-summary.h: Rename get to get_create.
14598         (get): Likewise.
14599         (get_create): Likewise.
14600         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14602 2018-06-08  Martin Liska  <mliska@suse.cz>
14604         * symbol-summary.h (release): Move definition out of class
14605         declaration.
14606         (symtab_removal): Likewise.
14607         (symtab_duplication): Likewise.
14609 2018-06-08  Martin Liska  <mliska@suse.cz>
14611         * symbol-summary.h (function_summary): Move constructor
14612         implementation out of class declaration.
14613         (release): Likewise.
14614         (symtab_insertion): Likewise.
14615         (symtab_removal): Likewise.
14616         (symtab_duplication): Likewise.
14617         (get): Likewise.
14619 2018-06-08  Martin Liska  <mliska@suse.cz>
14621         * Makefile.in: Remove support for MPX (macros, related functions,
14622         fields in cgraph_node, ...).
14623         * builtin-types.def (BT_BND): Likewise.
14624         (BT_FN_BND_CONST_PTR): Likewise.
14625         (BT_FN_CONST_PTR_BND): Likewise.
14626         (BT_FN_VOID_PTR_BND): Likewise.
14627         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14628         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14629         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14630         (expand_builtin_mempcpy_with_bounds): Likewise.
14631         (expand_builtin_memset_with_bounds): Likewise.
14632         (expand_builtin_memset_args): Likewise.
14633         (std_expand_builtin_va_start): Likewise.
14634         (expand_builtin): Likewise.
14635         (expand_builtin_with_bounds): Likewise.
14636         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14637         (DEF_LIB_BUILTIN_CHKP): Likewise.
14638         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14639         (DEF_CHKP_BUILTIN): Likewise.
14640         (BUILT_IN_MEMCPY): Likewise.
14641         (BUILT_IN_MEMMOVE): Likewise.
14642         (BUILT_IN_MEMPCPY): Likewise.
14643         (BUILT_IN_MEMSET): Likewise.
14644         (BUILT_IN_STPCPY): Likewise.
14645         (BUILT_IN_STRCAT): Likewise.
14646         (BUILT_IN_STRCHR): Likewise.
14647         (BUILT_IN_STRCPY): Likewise.
14648         (BUILT_IN_STRLEN): Likewise.
14649         (BUILT_IN_MEMCPY_CHK): Likewise.
14650         (BUILT_IN_MEMMOVE_CHK): Likewise.
14651         (BUILT_IN_MEMPCPY_CHK): Likewise.
14652         (BUILT_IN_MEMSET_CHK): Likewise.
14653         (BUILT_IN_STPCPY_CHK): Likewise.
14654         (BUILT_IN_STRCAT_CHK): Likewise.
14655         (BUILT_IN_STRCPY_CHK): Likewise.
14656         * calls.c (store_bounds): Likewise.
14657         (emit_call_1): Likewise.
14658         (special_function_p): Likewise.
14659         (maybe_warn_nonstring_arg): Likewise.
14660         (initialize_argument_information): Likewise.
14661         (finalize_must_preallocate): Likewise.
14662         (compute_argument_addresses): Likewise.
14663         (expand_call): Likewise.
14664         * cfgexpand.c (expand_call_stmt): Likewise.
14665         (expand_return): Likewise.
14666         (expand_gimple_stmt_1): Likewise.
14667         (pass_expand::execute): Likewise.
14668         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14669         (cgraph_node::remove): Likewise.
14670         (cgraph_node::dump): Likewise.
14671         (cgraph_node::verify_node): Likewise.
14672         * cgraph.h (chkp_function_instrumented_p): Likewise.
14673         (symtab_node::get_alias_target): Likewise.
14674         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14675         (cgraph_local_p): Likewise.
14676         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14677         (cgraph_edge::rebuild_references): Likewise.
14678         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14679         (walk_polymorphic_call_targets): Likewise.
14680         (cgraph_node::expand_thunk): Likewise.
14681         (symbol_table::output_weakrefs): Likewise.
14682         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14683         (ix86_handle_option): Likewise.
14684         * config/i386/constraints.md: Likewise.
14685         * config/i386/i386-builtin-types.def (BND): Likewise.
14686         (VOID): Likewise.
14687         (PVOID): Likewise.
14688         (ULONG): Likewise.
14689         * config/i386/i386-builtin.def (BDESC_END): Likewise.
14690         (BDESC_FIRST): Likewise.
14691         (BDESC): Likewise.
14692         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14693         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14694         * config/i386/i386.c (enum reg_class): Likewise.
14695         (ix86_target_string): Likewise.
14696         (ix86_option_override_internal): Likewise.
14697         (ix86_conditional_register_usage): Likewise.
14698         (ix86_valid_target_attribute_inner_p): Likewise.
14699         (ix86_set_indirect_branch_type): Likewise.
14700         (ix86_set_current_function): Likewise.
14701         (ix86_function_arg_regno_p): Likewise.
14702         (init_cumulative_args): Likewise.
14703         (ix86_function_arg_advance): Likewise.
14704         (ix86_function_arg): Likewise.
14705         (ix86_pass_by_reference): Likewise.
14706         (ix86_function_value_regno_p): Likewise.
14707         (ix86_function_value_1): Likewise.
14708         (ix86_function_value_bounds): Likewise.
14709         (ix86_return_in_memory): Likewise.
14710         (ix86_setup_incoming_vararg_bounds): Likewise.
14711         (ix86_va_start): Likewise.
14712         (indirect_thunk_need_prefix): Likewise.
14713         (print_reg): Likewise.
14714         (ix86_print_operand): Likewise.
14715         (ix86_expand_call): Likewise.
14716         (ix86_output_function_return): Likewise.
14717         (reg_encoded_number): Likewise.
14718         (BDESC_VERIFYS): Likewise.
14719         (ix86_init_mpx_builtins): Likewise.
14720         (ix86_init_builtins): Likewise.
14721         (ix86_emit_cmove): Likewise.
14722         (ix86_emit_move_max): Likewise.
14723         (ix86_expand_builtin): Likewise.
14724         (ix86_builtin_mpx_function): Likewise.
14725         (ix86_get_arg_address_for_bt): Likewise.
14726         (ix86_load_bounds): Likewise.
14727         (ix86_store_bounds): Likewise.
14728         (ix86_load_returned_bounds): Likewise.
14729         (ix86_store_returned_bounds): Likewise.
14730         (ix86_class_likely_spilled_p): Likewise.
14731         (ix86_hard_regno_mode_ok): Likewise.
14732         (x86_order_regs_for_local_alloc): Likewise.
14733         (ix86_mitigate_rop): Likewise.
14734         (ix86_bnd_prefixed_insn_p): Likewise.
14735         (ix86_mpx_bound_mode): Likewise.
14736         (ix86_make_bounds_constant): Likewise.
14737         (ix86_initialize_bounds): Likewise.
14738         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14739         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14740         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14741         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14742         (TARGET_CHKP_BOUND_MODE): Likewise.
14743         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14744         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14745         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14746         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14747         * config/i386/i386.h (TARGET_MPX): Likewise.
14748         (TARGET_MPX_P): Likewise.
14749         (VALID_BND_REG_MODE): Likewise.
14750         (FIRST_BND_REG): Likewise.
14751         (LAST_BND_REG): Likewise.
14752         (enum reg_class): Likewise.
14753         (BND_REG_P): Likewise.
14754         (BND_REGNO_P): Likewise.
14755         (BNDmode): Likewise.
14756         (ADJUST_INSN_LENGTH): Likewise.
14757         * config/i386/i386.md: Likewise.
14758         * config/i386/i386.opt: Likewise.
14759         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14760         (defined): Likewise.
14761         (LINK_MPX): Likewise.
14762         (MPX_SPEC): Likewise.
14763         (LIBMPX_SPEC): Likewise.
14764         (LIBMPXWRAPPERS_SPEC): Likewise.
14765         (CHKP_SPEC): Likewise.
14766         * config/i386/predicates.md: Likewise.
14767         * dbxout.c (dbxout_type): Likewise.
14768         * doc/extend.texi: Likewise.
14769         * doc/invoke.texi: Likewise.
14770         * doc/md.texi: Likewise.
14771         * doc/tm.texi: Likewise.
14772         * doc/tm.texi.in: Likewise.
14773         * dwarf2out.c (is_base_type): Likewise.
14774         (gen_formal_types_die): Likewise.
14775         (gen_subprogram_die): Likewise.
14776         (gen_type_die_with_usage): Likewise.
14777         (gen_decl_die): Likewise.
14778         (dwarf2out_late_global_decl): Likewise.
14779         * expr.c (expand_assignment): Likewise.
14780         (emit_storent_insn): Likewise.
14781         (store_expr_with_bounds): Likewise.
14782         (store_expr): Likewise.
14783         (expand_expr_real_1): Likewise.
14784         * expr.h (store_expr_with_bounds): Likewise.
14785         * function.c (use_register_for_decl): Likewise.
14786         (struct bounds_parm_data): Likewise.
14787         (assign_parms_augmented_arg_list): Likewise.
14788         (assign_parm_find_entry_rtl): Likewise.
14789         (assign_parm_is_stack_parm): Likewise.
14790         (assign_parm_load_bounds): Likewise.
14791         (assign_bounds): Likewise.
14792         (assign_parms): Likewise.
14793         (expand_function_start): Likewise.
14794         * gcc.c (CHKP_SPEC): Likewise.
14795         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14796         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14797         (wrestrict_dom_walker::check_call): Likewise.
14798         * gimple.c (gimple_build_call_from_tree): Likewise.
14799         * gimple.h (enum gf_mask): Likewise.
14800         (gimple_call_with_bounds_p): Likewise.
14801         (gimple_call_set_with_bounds): Likewise.
14802         * gimplify.c (gimplify_init_constructor): Likewise.
14803         * ipa-cp.c (initialize_node_lattices): Likewise.
14804         (propagate_constants_across_call): Likewise.
14805         (find_more_scalar_values_for_callers_subset): Likewise.
14806         * ipa-hsa.c (process_hsa_functions): Likewise.
14807         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14808         * ipa-icf.c (sem_function::merge): Likewise.
14809         * ipa-inline.c (early_inliner): Likewise.
14810         * ipa-pure-const.c (warn_function_noreturn): Likewise.
14811         (warn_function_cold): Likewise.
14812         (propagate_pure_const): Likewise.
14813         * ipa-ref.h (enum GTY): Likewise.
14814         * ipa-split.c (find_retbnd): Likewise.
14815         (consider_split): Likewise.
14816         (split_function): Likewise.
14817         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14818         * ipa.c (walk_polymorphic_call_targets): Likewise.
14819         (symbol_table::remove_unreachable_nodes): Likewise.
14820         (process_references): Likewise.
14821         (cgraph_build_static_cdtor_1): Likewise.
14822         * lto-cgraph.c (lto_output_node): Likewise.
14823         (output_refs): Likewise.
14824         (compute_ltrans_boundary): Likewise.
14825         (input_overwrite_node): Likewise.
14826         (input_node): Likewise.
14827         (input_cgraph_1): Likewise.
14828         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14829         * passes.c (pass_manager::execute_early_local_passes): Likewise.
14830         (class pass_chkp_instrumentation_passes): Likewise.
14831         (make_pass_chkp_instrumentation_passes): Likewise.
14832         * passes.def: Likewise.
14833         * rtl.h (struct GTY): Likewise.
14834         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14835         * stor-layout.c (layout_type): Likewise.
14836         * symtab.c: Likewise.
14837         * target.def: Likewise.
14838         * targhooks.c (default_chkp_bound_type): Likewise.
14839         (default_chkp_bound_mode): Likewise.
14840         (default_builtin_chkp_function): Likewise.
14841         (default_chkp_function_value_bounds): Likewise.
14842         (default_chkp_make_bounds_constant): Likewise.
14843         (default_chkp_initialize_bounds): Likewise.
14844         * targhooks.h (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         * toplev.c (compile_file): Likewise.
14851         (process_options): Likewise.
14852         * tree-core.h (DEF_BUILTIN): Likewise.
14853         (DEF_BUILTIN_CHKP): Likewise.
14854         * tree-inline.c (declare_return_variable): Likewise.
14855         (remap_gimple_stmt): Likewise.
14856         (copy_bb): Likewise.
14857         (initialize_inlined_parameters): Likewise.
14858         (expand_call_inline): Likewise.
14859         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14860         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14861         (make_pass_ipa_chkp_produce_thunks): Likewise.
14862         (make_pass_chkp): Likewise.
14863         (make_pass_chkp_opt): Likewise.
14864         (make_pass_chkp_instrumentation_passes): Likewise.
14865         * tree-pretty-print.c (dump_generic_node): Likewise.
14866         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14867         * tree-ssa-dce.c (propagate_necessity): Likewise.
14868         (eliminate_unnecessary_stmts): Likewise.
14869         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14870         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14871         * tree-ssa-sccvn.h: Likewise.
14872         * tree-ssa-strlen.c (get_string_length): Likewise.
14873         (valid_builtin_call): Likewise.
14874         (adjust_last_stmt): Likewise.
14875         (handle_builtin_strchr): Likewise.
14876         (handle_builtin_strcpy): Likewise.
14877         (handle_builtin_stxncpy): Likewise.
14878         (handle_builtin_memcpy): Likewise.
14879         (handle_builtin_strcat): Likewise.
14880         (strlen_check_and_optimize_stmt): Likewise.
14881         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14882         * tree-streamer-in.c: Likewise.
14883         * tree-streamer.c (record_common_node): Likewise.
14884         * tree.c (tree_code_size): Likewise.
14885         (wide_int_to_tree_1): Likewise.
14886         (type_contains_placeholder_1): Likewise.
14887         (build_common_tree_nodes): Likewise.
14888         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14889         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14890         (POINTER_BOUNDS_P): Likewise.
14891         (BOUNDED_TYPE_P): Likewise.
14892         (BOUNDED_P): Likewise.
14893         (CALL_WITH_BOUNDS_P): Likewise.
14894         (pointer_bounds_type_node): Likewise.
14895         * value-prof.c (gimple_ic): Likewise.
14896         * var-tracking.c (vt_add_function_parameters): Likewise.
14897         * varasm.c (make_decl_rtl): Likewise.
14898         (assemble_start_function): Likewise.
14899         (output_constant): Likewise.
14900         (maybe_assemble_visibility): Likewise.
14901         * varpool.c (ctor_for_folding): Likewise.
14902         * chkp-builtins.def: Remove.
14903         * ipa-chkp.c: Remove.
14904         * ipa-chkp.h: Remove.
14905         * rtl-chkp.c: Remove.
14906         * rtl-chkp.h: Remove.
14907         * tree-chkp-opt.c: Remove.
14908         * tree-chkp.c: Remove.
14909         * tree-chkp.h: Remove.
14911 2018-06-07  Carl Love  <cel@us.ibm.com>
14913         * config/rs6000/vsx.md (vextract_fp_from_shorth,
14914         vextract_fp_from_shortl): Add BE support.
14916 2018-06-07  Paul Koning  <ni1d@arrl.net>
14918         * compare-elim.c (try_merge_compare): Don't merge compare if
14919         address contains a side effect.
14920         (try_eliminate_compare): Likewise.
14922 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
14924         * config.gcc: Support "tremont".
14925         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14926         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14927         PROCESSOR_TREMONT.
14928         * config/i386/i386.c (m_TREMONT): Define.
14929         (processor_target_table): Add "tremont".
14930         (PTA_TREMONT): Define.
14931         (ix86_lea_outperforms): Add TARGET_TREMONT.
14932         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14933         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14934         and M_INTEL_GOLDMONT_PLUS.
14935         (fold_builtin_cpu): Add "tremont".
14936         (ix86_add_stmt_cost): Add TARGET_TREMONT.
14937         (ix86_option_override_internal): Add "tremont".
14938         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14939         (processor_type): Add PROCESSOR_TREMONT.
14940         * config/i386/x86-tune.def: Add m_TREMONT.
14941         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14943 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14945         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14946         symbol defined for msp430i* devices to be lower case.
14948 2018-06-07  Richard Biener  <rguenther@suse.de>
14950         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14951         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
14952         Properly wrap signed arithmetic if overflow wraps.
14954 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
14956         PR tree-optimization/69615
14957         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14958         of a cast from a same precision integral SSA_NAME in a bb dominated
14959         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
14960         cast to utype if rhs2 has already a compatible type.
14962 2018-06-07  Richard Biener  <rguenther@suse.de>
14964         PR tree-optimization/85935
14965         * graphite-scop-detection.c (find_params_in_bb): Analyze
14966         condition operands with respect to the correct loop.  Assert
14967         the analysis doesn't fail.
14969 2018-06-04  Carl Love  <cel@us.ibm.com>
14971         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14972         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
14973         as it is slightly cheaper.
14974         (first_match_or_eos_index_<mode>):
14975         Calculate index using natural element order.
14976         (first_match_index_<mode>):
14977         Calculate index using natural element order.
14978         (first_match_or_eos_index_<mode>):
14979         Calculate index using natural order.
14980         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14981         for BE and LE modes.
14982         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14983         P9V_BUILTIN_VCLZLSBB_V16QI.
14984         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14985         specific.
14987 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14989         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14990         indentation and line wrap for many prototypes.  Add missing
14991         @smallexample directives around block of prototypes for vec_xl and
14992         vec_xst.
14994 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
14996         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14997         track if we pass or return IEEE 128-bit floating point.
14998         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14999         C++ mangling that is compatible with GCC 8.1.
15000         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
15001         (init_cumulative_args): Note if we pass or return IEEE 128-bit
15002         floating point types.
15003         (rs6000_function_arg_advance_1): Likewise.
15004         (rs6000_mangle_type): Optionally generate mangled names that match
15005         what GCC 8.1 generated for IEEE 128-bit floating point types.
15006         (rs6000_globalize_decl_name): If we have an external function that
15007         passes or returns IEEE 128-bit types, generate a weak reference
15008         from the mangled name used in GCC 8.1 to the current mangled
15009         name.
15010         (rs6000_init_builtins): Make __ibm128 use the long double type if
15011         long double is IBM extended double.  Make __float128 use the long
15012         double type if long double is IEEE 128-bit.
15014         PR target/85657
15015         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
15016         macro for __ibm128 built-in functions.
15017         (PACK_IF): Add __ibm128 pack/unpack functions.
15018         (UNPACK_IF): Likewise.
15019         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
15020         enable long double built-in functions if long double is IEEE
15021         128-bit floating point.
15022         (rs6000_invalid_builtin): Update long double built-in function
15023         error message.
15024         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
15025         functions, adjust the built-in function to use the long double
15026         built-in function if __ibm128 and long double are the same type.
15027         * doc/extend.texi (PowerPC builtins): Update documention for
15028         __builtin_{,un}pack_longdouble.  Add documentation for
15029         __builtin_{,un}pack_ibm128.
15031 2018-06-06  Jim Wilson  <jimw@sifive.com>
15033         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
15034         (struct machine_function): New field interrupt_mode.
15035         (riscv_handle_type_attribute): New function.  Add forward declaration.
15036         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
15037         (riscv_expand_epilogue): Check interrupt_mode field.
15038         (riscv_set_current_function): Check interrupt attribute args and
15039         set interrupt_mode field.
15040         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
15041         (riscv_sret, riscv_uret): New.
15042         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
15043         new arguments to interrupt attribute.
15045 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
15047         PR target/63177
15048         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
15049         Don't handle -mcpu=power8 if -mpower9-vector is also used.
15051 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15053         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
15054         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
15055         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15056         several redundant entries.
15058 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
15060         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
15061         type from "rtx" to "rtx_insn *".
15062         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
15063         for local "call_insn", removing cast.
15064         (ix86_expand_call): Likewise, introducing a "call_insn" local.
15066 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
15068         PR tree-optimization/86066
15069         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
15070         for BIT_INSERT_EXPR stores.
15072 2018-06-06  Richard Biener  <rguenther@suse.de>
15074         PR tree-optimization/86062
15075         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15076         component refs ontop
15077         of to be offsetted base.
15079 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15081         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
15082         to be static and remove check on interrupt attribute name.
15084 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15086         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15087         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
15089 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
15091         PR target/79924
15092         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
15093         second argument.
15094         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
15095         Remove second argument, change how error is called.
15096         (aarch64_layout_arg): Remove second argument from
15097         aarch64_err_no_fpadvsimd call.
15098         (aarch64_init_cumulative_args): Ditto.
15099         (aarch64_gimplify_va_arg_expr): Ditto.
15100         * config/aarch64/aarch64.md (mov<mode>): Ditto.
15102 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
15104         * config/i386/i386.md (simple_return_indirect_internal): New expander.
15105         (*simple_return_indirect_internal<mode>): Rename from
15106         simple_return_indirect_internal.  Use W mode iterator.
15107         (rstorssp): New expander.
15108         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
15109         (clrssbsy): New expander.
15110         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
15112 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15114         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
15115         __typeof__.
15116         (cmse_check_pointed_object): Likewise.
15118 2018-06-05  Martin Liska  <mliska@suse.cz>
15120         PR gcov-profile/47618
15121         * doc/invoke.texi: Document how -fprofile-dir format
15122         is extended.
15124 2018-06-05  Richard Biener  <rguenther@suse.de>
15126         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
15127         removal pretend DOM info isn't available so we do not update
15128         it and only remove edges, not dominated blocks.  Actually free
15129         DOM info in case we removed something.  Remove unreachable blocks.
15130         (mfb_keep_latches): Work with either DOM info or marked backedges.
15131         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
15132         first.  Mark backedges if DOM info isn't available.
15133         (Re-)compute DOM info after cleanup_control_flow_pre.
15135 2018-06-05  Richard Biener  <rguenther@suse.de>
15137         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
15138         (locus_discrim_hasher::hash): Adjust.
15139         (locus_discrim_hasher::equal): Likewise.
15140         (next_discriminator_for_locus): Work on line directly.
15141         (same_line_p): Pass in expanded locus1 as well.
15142         (assign_discriminators): Avoid redundant location expansions.
15144 2018-06-05  Richard Biener  <rguenther@suse.de>
15146         PR tree-optimization/86046
15147         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
15148         if required after clearing TREE_ADDRESSABLE.
15150 2018-06-05  Richard Biener  <rguenther@suse.de>
15152         PR tree-optimization/86047
15153         * tree-ssa-loop.c (for_each_index): Glob handling of all
15154         decls and constants and really handle all of them.
15156 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15158         PR target/81497
15159         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
15160         qualifier_void_pointer and qualifier_const_void_pointer.
15161         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
15162         (arm_init_builtins): Handle the above.
15163         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
15164         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
15165         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
15166         void intrinsics.
15168 2018-06-05  Martin Liska  <mliska@suse.cz>
15170         * auto-profile.c (read_autofdo_file): Do not use
15171         gcov_ctr_summary struct.
15172         (afdo_callsite_hot_enough_for_early_inline): Likewise.
15173         * coverage.c (struct counts_entry): Likewise.
15174         (read_counts_file): Read just single summary entry.
15175         (get_coverage_counts): Use gcov_summary struct.
15176         * coverage.h (get_coverage_counts): Likewise.
15177         * gcov-dump.c (dump_working_sets): Likewise.
15178         (tag_summary): Dump just single summary.
15179         * gcov-io.c (gcov_write_summary): Write just histogram
15180         summary.
15181         (gcov_read_summary): Read just single summary.
15182         (compute_working_sets): Use gcov_summary struct.
15183         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
15184         of GCOV_COUNTERS_SUMMABLE.
15185         (GCOV_COUNTERS_SUMMABLE): Remove.
15186         (GCOV_FIRST_VALUE_COUNTER): Replace with
15187         GCOV_COUNTER_V_INTERVAL.
15188         (struct gcov_ctr_summary): Remove.
15189         (struct gcov_summary): Directly use fields of former
15190         gcov_ctr_summary.
15191         (compute_working_sets): Use gcov_summary struct.
15192         * gcov.c (read_count_file): Do not use ctrs fields.
15193         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
15194         struct.
15195         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
15196         struct.
15197         * profile.c: Likewise.
15198         * profile.h: Likewise.
15200 2018-06-05  Martin Liska  <mliska@suse.cz>
15202         PR gcov-profile/84846
15203         * gcov.c (output_lines): Print working directory only
15204         in intermediate format.
15206 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
15208         * config/s390/s390-builtin-types.def: Add void function type.
15209         * config/s390/s390-builtins.def: Use the function type for the
15210         tbeginc builtin.
15212 2018-06-04  Jim Wilson  <jimw@sifive.com>
15214         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
15215         to int.
15216         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
15217         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
15218         handle EH_RETURN_DATA_REGNO registers properly.
15219         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
15220         (riscv_expand_epilogue): Update comment.  Change argument name and
15221         type.  Update code to use new name and type.  Pass new args to
15222         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
15223         EXCEPTION_RETURN.
15224         * config/riscv/riscv.md (NORMAL_RETURN): New.
15225         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
15226         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
15227         (eh_return): Call gen_eh_return_internal and emit barrier.
15228         (eh_return_internal): Call riscv_expand_epilogue.
15230 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
15232         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
15233         bit_insertion field and declare can_be_merged_into method.
15234         (merged_store_group::can_be_merged_into): New method.
15235         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
15236         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
15237         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
15239 2018-06-04  Richard Biener  <rguenther@suse.de>
15241         PR tree-optimization/85955
15242         * builtins.c (fold_builtin_sincos): Convert pointers to
15243         destination to appropriate type before dereferencing.
15245 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
15247         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
15249 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
15251         * expr.c (expand_expr_real_1): Force the operand into memory if
15252         its TYPE_MODE is BLKmode and if there is no integer mode for
15253         the number of bits being extracted.
15255 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15257         PR target/85832
15258         PR target/86036
15259         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15260         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
15262 2018-06-04  Richard Biener  <rguenther@suse.de>
15264         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
15265         (cleanup_tree_cfg_noloop): ... single caller.  Do
15266         start_recording_case_labels later.
15268 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
15270         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
15271         to _IMMINTRIN_H_INCLUDED.
15272         * config/i386/pconfigintrin.h: Ditto.
15273         * config/i386/waitpkgintrin.h: Ditto.
15274         * config/i386/immintrin.h: Add includes for sgxintrin.h,
15275         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
15276         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
15277         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
15278         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
15279         waitpkgintrin.h and cldemoteintrin.h.
15281 2018-06-04  Richard Biener  <rguenther@suse.de>
15283         PR tree-optimization/86038
15284         * tracer.c (find_best_successor): Check probability for
15285         being initialized, bail out if not.
15287 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
15289         PR target/86003
15290         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
15291         of bits to ignore when comparing architectures.
15293 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
15295         PR tree-optimization/69615
15296         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
15297         maximum or minimum of the type, try to merge it also as if
15298         range1 is + [-, x - 1] or + [x + 1, -].
15300         PR c++/86025
15301         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
15303 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
15305         PR tree-optimization/86034
15306         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
15307         the unsigned bitfield type in a bit insertion sequence if it does not
15308         have a larger precision than the bitfield size.
15309         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
15311 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
15313         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
15315 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15317         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
15318         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
15319         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
15320         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
15322 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
15324         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
15325         Disable -fdelete-null-pointer-checks for ELF toolchain.
15327 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15328             Kito Cheng  <kito.cheng@gmail.com>
15330         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
15331         (nds32le-*-*, nds32be-*-*): Integrate checking process.
15332         (nds32*-*-*): Add glibc and uclibc conditions.
15333         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
15334         (TARGET_EXCEPT_UNWIND_INFO): Define.
15335         * config/nds32/elf.h: New file.
15336         * config/nds32/linux.h: New file.
15337         * config/nds32/nds32-elf.opt: New file.
15338         * config/nds32/nds32-linux.opt: New file.
15339         * config/nds32/nds32-fp-as-gp.c
15340         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
15341         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
15342         TARGET_LINUX_ABI.
15343         (nds32_asm_file_end): Ditto.
15344         (nds32_print_operand): Ditto.
15345         (nds32_insert_attributes): Ditto.
15346         (nds32_init_libfuncs): New function.
15347         (TARGET_HAVE_TLS): Define.
15348         (TARGET_INIT_LIBFUNCS): Define.
15349         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
15350         spec content.
15351         (TARGET_ELF): Apply different mcmodel setting.
15352         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
15353         been migrated into elf.h and linux.h files.
15354         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
15355         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
15356         (mcmodel): The content has been migrated into nds32-elf.opt and
15357         nds32-linux.opt files.
15358         * config/nds32/t-elf: New file.
15359         * config/nds32/t-linux: New file.
15361 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
15362             Shiva Chen  <shiva0217@gmail.com>
15364         * config/nds32/constants.md (unspec_volatile_element): Add
15365         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
15366         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
15367         optimization.
15368         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
15369         (make_pass_nds32_fp_as_gp): Declare.
15370         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
15371         optmization pass.
15372         (nds32_asm_function_end_prologue): Remove unused asm output.
15373         (nds32_asm_function_begin_epilogue): Remove unused asm output.
15374         (nds32_asm_file_start): Output necessary fp_as_gp information.
15375         (nds32_option_override): Adjust register usage.
15376         (nds32_expand_prologue): Consider fp_as_gp situation.
15377         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
15378         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
15379         (epilogue): Ditto.
15380         (return): Ditto.
15381         (simple_return): Ditto.
15382         (omit_fp_begin): Output special directive for fp_as_gp.
15383         (omit_fp_end): Output special directive for fp_as_gp.
15384         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
15385         mforbid-fp-as-gp): New options.
15387 2018-06-01  Mark Wielaard  <mark@klomp.org>
15389         * dwarf2out.c (dwarf2out_finish): Remove generation of
15390         DW_AT_loclists_base.
15392 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
15394         * gimple-ssa-store-merging.c: Include gimple-fold.h.
15395         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
15396         (struct merged_store_group): Add bit_insertion field.
15397         (dump_char_array): Use standard hexadecimal format.
15398         (merged_store_group::merged_store_group): Set bit_insertion to false.
15399         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
15400         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
15401         also print the mask in the dump file.
15402         (pass_store_merging::gate): Minor tweak.
15403         (imm_store_chain_info::coalesce_immediate): Fix wrong association
15404         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
15405         stores with INTEGER_CST stores.
15406         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
15407         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
15408         and use it throughout.  Generate bit insertion sequences if need be.
15409         (pass_store_merging::process_store): Remove redundant condition.
15410         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
15412 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
15414         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
15415         the 128-bit floating point types.  Fix function comment.
15417 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15419         * config/aarch64/aarch64-simd.md
15420         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
15421         mnemonics.
15422         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
15423         mnemonics.
15425 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15427         PR tree-optimization/85989
15428         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
15429         variable.
15430         (backprop::intersect_uses): Check it when deciding whether this
15431         is a backedge reference.
15432         (backprop::process_block): Add each phi to m_visited_phis
15433         after visiting it, then clear it at the end.
15435 2018-06-01  Richard Biener  <rguenther@suse.de>
15437         * tree-vectorizer.h (vect_dr_stmt): New function.
15438         (vect_get_load_cost): Adjust.
15439         (vect_get_store_cost): Likewise.
15440         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
15441         Use vect_dr_stmt instead of DR_SMTT.
15442         (vect_record_base_alignments): Likewise.
15443         (vect_calculate_target_alignment): Likewise.
15444         (vect_compute_data_ref_alignment): Likewise and make static.
15445         (vect_update_misalignment_for_peel): Likewise.
15446         (vect_verify_datarefs_alignment): Likewise.
15447         (vector_alignment_reachable_p): Likewise.
15448         (vect_get_data_access_cost): Likewise.  Pass down
15449         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
15450         (vect_get_peeling_costs_all_drs): Likewise.
15451         (vect_peeling_hash_get_lowest_cost): Likewise.
15452         (vect_enhance_data_refs_alignment): Likewise.
15453         (vect_find_same_alignment_drs): Likewise.
15454         (vect_analyze_data_refs_alignment): Likewise.
15455         (vect_analyze_group_access_1): Likewise.
15456         (vect_analyze_group_access): Likewise.
15457         (vect_analyze_data_ref_access): Likewise.
15458         (vect_analyze_data_ref_accesses): Likewise.
15459         (vect_vfa_segment_size): Likewise.
15460         (vect_small_gap_p): Likewise.
15461         (vectorizable_with_step_bound_p): Likewise.
15462         (vect_prune_runtime_alias_test_list): Likewise.
15463         (vect_analyze_data_refs): Likewise.
15464         (vect_supportable_dr_alignment): Likewise.
15465         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
15466         (vect_gen_prolog_loop_niters): Likewise.
15467         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
15468         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
15469         modify DR_STMT.
15470         (vect_recog_mask_conversion_pattern): Likewise.
15471         (vect_try_gather_scatter_pattern): Likewise.
15472         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
15473         to vect_get_store_cost.
15474         (vect_get_store_cost): Get stmt_info instead of DR.
15475         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
15476         (vect_get_load_cost): Get stmt_info instead of DR.
15478 2018-06-01  Richard Biener  <rguenther@suse.de>
15480         PR middle-end/86017
15481         * gimple-fold.c (var_decl_component_p): Also allow offsetted
15482         vars wrapped in MEM_REFs.
15484 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
15486         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15487         Fix subreg tests so that we only return a choice between
15488         GENERAL_REGS and FP_REGS if the original classes included both.
15490 2018-06-01  Richard Biener  <rguenther@suse.de>
15492         PR ipa/85960
15493         * tree-ssa-structalias.c (get_function_part_constraint):
15494         Handle NULL fi->decl.
15495         (find_func_aliases_for_call): Properly handle indirect
15496         fi from direct call.
15497         (find_func_clobbers): Likewise.
15498         (ipa_pta_execute): Likewise.
15499         (create_variable_info_for): For functions that are ifunc_resolver
15500         resolve to a varinfo that contains the result of the resolver call.
15501         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
15502         aliases.
15504 2018-05-31  Michael Collison  <michael.collison@arm.com>
15506         * config/aarch64/aarch64.md:
15507         (*fix_to_zero_extenddfdi2): New pattern.
15508         * gcc.target/aarch64/fix_extend1.c: New testcase.
15510 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
15512         PR middle-end/78809
15513         PR middle-end/83026
15514         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
15515         and BUILT_IN_STRNCMP_EQ.
15516         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
15517         BUILT_IN_STRNCMP_EQ.
15518         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
15519         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15520         (gimple_fold_builtin): Likewise.
15521         * tree-ssa-strlen.c (compute_string_length): New function.
15522         (determine_min_obsize): New function.
15523         (handle_builtin_string_cmp): New function to handle calls to
15524         string compare functions.
15525         (strlen_optimize_stmt): Add handling to builtin string compare
15526         calls.
15527         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
15528         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
15529         * tree.c (build_common_builtin_nodes): Add new defines of
15530         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
15532 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
15534         PR target/85984
15535         * bb-reorder.c (pass_partition_blocks::gate): Return false for
15536         functions with naked attribute.
15538 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
15540         * config/i386/sse.md (avx_vec_concat<mode>):
15541         Substitute concat_tg_mode mode attribute with xtg_mode.
15542         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
15543         (concat_tg_mode): Remove mode attribute.
15545 2018-05-31  Martin Sebor  <msebor@redhat.com>
15547         PR c/82063
15548         * calls.c (alloc_max_size): Correct a logic error/typo.
15549         Treat excessive arguments as infinite.  Warn for invalid arguments.
15550         * doc/invoke.texi (-Walloc-size-larger-than): Update.
15552 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
15554         PR target/85829
15555         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
15556         and movx for Haswell.
15558 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
15559             Cesar Philippidis  <cesar@codesourcery.com>
15561         PR middle-end/85879
15562         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
15563         when emitting error on private/firstprivate reductions.
15564         * omp-low.c (lower_omp_target): Avoid reference-type processing
15565         on pointers for firstprivate clause.
15567 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15569         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15570         (st1x2): Likewise.
15571         (st1x3): Likewise.
15572         * config/aarch64/aarch64-simd.md
15573         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
15574         (aarch64_ld1_x3_<mode>): Likewise
15575         (aarch64_st1x2<VALLDIF:mode>): Likewise
15576         (aarch64_st1_x2_<mode>): Likewise
15577         (aarch64_st1x3<VALLDIF:mode>): Likewise
15578         (aarch64_st1_x3_<mode>): Likewise
15579         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
15580         (vld1_s8_x3): Likewise.
15581         (vld1_u16_x3): Likewise.
15582         (vld1_s16_x3): Likewise.
15583         (vld1_u32_x3): Likewise.
15584         (vld1_s32_x3): Likewise.
15585         (vld1_u64_x3): Likewise.
15586         (vld1_s64_x3): Likewise.
15587         (vld1_f16_x3): Likewise.
15588         (vld1_f32_x3): Likewise.
15589         (vld1_f64_x3): Likewise.
15590         (vld1_p8_x3): Likewise.
15591         (vld1_p16_x3): Likewise.
15592         (vld1_p64_x3): Likewise.
15593         (vld1q_u8_x3): Likewise.
15594         (vld1q_s8_x3): Likewise.
15595         (vld1q_u16_x3): Likewise.
15596         (vld1q_s16_x3): Likewise.
15597         (vld1q_u32_x3): Likewise.
15598         (vld1q_s32_x3): Likewise.
15599         (vld1q_u64_x3): Likewise.
15600         (vld1q_s64_x3): Likewise.
15601         (vld1q_f16_x3): Likewise.
15602         (vld1q_f32_x3): Likewise.
15603         (vld1q_f64_x3): Likewise.
15604         (vld1q_p8_x3): Likewise.
15605         (vld1q_p16_x3): Likewise.
15606         (vld1q_p64_x3): Likewise.
15607         (vst1_s64_x2): Likewise.
15608         (vst1_u64_x2): Likewise.
15609         (vst1_f64_x2): Likewise.
15610         (vst1_s8_x2): Likewise.
15611         (vst1_p8_x2): Likewise.
15612         (vst1_s16_x2): Likewise.
15613         (vst1_p16_x2): Likewise.
15614         (vst1_s32_x2): Likewise.
15615         (vst1_u8_x2): Likewise.
15616         (vst1_u16_x2): Likewise.
15617         (vst1_u32_x2): Likewise.
15618         (vst1_f16_x2): Likewise.
15619         (vst1_f32_x2): Likewise.
15620         (vst1_p64_x2): Likewise.
15621         (vst1q_s8_x2): Likewise.
15622         (vst1q_p8_x2): Likewise.
15623         (vst1q_s16_x2): Likewise.
15624         (vst1q_p16_x2): Likewise.
15625         (vst1q_s32_x2): Likewise.
15626         (vst1q_s64_x2): Likewise.
15627         (vst1q_u8_x2): Likewise.
15628         (vst1q_u16_x2): Likewise.
15629         (vst1q_u32_x2): Likewise.
15630         (vst1q_u64_x2): Likewise.
15631         (vst1q_f16_x2): Likewise.
15632         (vst1q_f32_x2): Likewise.
15633         (vst1q_f64_x2): Likewise.
15634         (vst1q_p64_x2): Likewise.
15635         (vst1_s64_x3): Likewise.
15636         (vst1_u64_x3): Likewise.
15637         (vst1_f64_x3): Likewise.
15638         (vst1_s8_x3): Likewise.
15639         (vst1_p8_x3): Likewise.
15640         (vst1_s16_x3): Likewise.
15641         (vst1_p16_x3): Likewise.
15642         (vst1_s32_x3): Likewise.
15643         (vst1_u8_x3): Likewise.
15644         (vst1_u16_x3): Likewise.
15645         (vst1_u32_x3): Likewise.
15646         (vst1_f16_x3): Likewise.
15647         (vst1_f32_x3): Likewise.
15648         (vst1_p64_x3): Likewise.
15649         (vst1q_s8_x3): Likewise.
15650         (vst1q_p8_x3): Likewise.
15651         (vst1q_s16_x3): Likewise.
15652         (vst1q_p16_x3): Likewise.
15653         (vst1q_s32_x3): Likewise.
15654         (vst1q_s64_x3): Likewise.
15655         (vst1q_u8_x3): Likewise.
15656         (vst1q_u16_x3): Likewise.
15657         (vst1q_u32_x3): Likewise.
15658         (vst1q_u64_x3): Likewise.
15659         (vst1q_f16_x3): Likewise.
15660         (vst1q_f32_x3): Likewise.
15661         (vst1q_f64_x3): Likewise.
15662         (vst1q_p64_x3): Likewise.
15664 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15666         * config/msp430/msp430.c (msp430_output_labelref): Prepend
15667         user_label_prefix to name.
15669         * tree-core.h: Update comment about the format of NAME string
15670         passed to handler in attribute_spec.
15672         * config/msp430/msp430.md: Remove erroneous subreg expression from
15673         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15674         zero_extend{q,h}isi2.
15676 2018-05-30  Borislav Petkov  <bp@suse.de>
15678         * doc/extend.texi: Document some architecture specific
15679         constraints and sort entries.
15681 2018-05-30  Martin Sebor  <msebor@redhat.com>
15683         PR middle-end/85369
15684         * builtins.c (expand_builtin_stpcpy_1): New function.
15685         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15686         only if the former succeeds.
15688 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15690         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15691         in saphira.
15693 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15695         * doc/invoke.texi (-flinker-output): Document
15697 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15699         * passes.c (ipa_write_summaries): Only modify statements if body
15700         is in memory.
15701         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15702         incrementally linking.
15703         (ipa_passes): Likewise.
15704         * lto-cgraph.c (lto_output_node): When incrementally linking do not
15705         pass down resolution info.
15706         * common.opt (flag_incremental_link): Update info.
15707         * gcc.c (plugin specs): Turn flinker-output=* to
15708         -plugin-opt=-linker-output-known
15709         * toplev.c (compile_file): Also cut compilation when doing incremental
15710         link.
15711         * flag-types. (enum lto_partition_model): Add
15712         LTO_LINKER_OUTPUT_NOLTOREL.
15713         (invoke.texi): Add -flinker-output docs.
15714         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15715         link same way as WPA; do not stream in dead initializers.
15717         * dwarf2out.c (dwarf2out_die_ref_for_decl,
15718         darf2out_register_external_decl): Support incremental link.
15720 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15722         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15724 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15726         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15727         it down to simple_object_copy_lto_debug_sections.
15728         (run_gcc): Determine incremental LTO link time and configure
15729         lto1 into non-wpa mode, disable renaming of debug sections.
15731 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15733         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15734         descriptions of various incorrectly documented functions.
15736 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15738         Revert:
15739         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15740         address check not strict.
15742 2018-05-30  Richard Biener  <rguenther@suse.de>
15744         PR tree-optimization/85964
15745         * tracer.c (better_p): Drop initialized count check, we only
15746         call the function with initialized counts now.
15747         (find_best_successor): Do find a best edge if one
15748         has uninitialized count.
15749         (find_best_predecessor): Likewise.  Do BB frequency check only
15750         if count is initialized.
15752 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
15754         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15755         (aarch64_ldrstr_offset_compare): New.
15756         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15757         load/store orderings.
15758         (aarch64_gen_adjusted_ldpstp): Likewise.
15760 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
15762         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15763         Check for subset of GENERAL_REGS and FP_REGS.
15764         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15765         r=w alternative.
15767 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
15769         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15770         and wi::to_poly_offset.  Add the current offset and then check
15771         whether the sum fits, rather than using an unchecked addition of
15772         a checked term.  Check for a shwi rather than a uhwi.
15773         * expr.c (get_bit_range): Use tree_to_poly_uint64.
15774         (store_constructor): Use poly_int_tree_p.
15775         (expand_expr_real_1): Likewise.
15776         * function.c (assign_temp): Likewise.
15777         * fold-const.c (const_binop): Use poly_int_tree_p and
15778         wi::to_poly_offset.
15779         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
15780         division.
15781         * ipa-icf-gimple.c (func_checker::compare_operand): Use
15782         to_poly_offset for MEM offsets.
15783         * ipa-icf.c (sem_variable::equals): Likewise.
15784         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15785         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15786         wi::to_poly_offset for BIT_FIELD_REF offsets.
15787         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15788         wi::to_poly_offset.
15789         * var-tracking.c (emit_note_insn_var_location): Use
15790         tree_to_poly_uint64.
15792 2018-05-29  Jim Wilson  <jimw@sifive.com>
15794         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15796 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
15798         PR target/85950
15799         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15800         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15801         sequence.
15802         (sse4_1_round<mode>2): Use nonimmediate_operand
15803         for operand 1 predicate.
15805 2018-05-29  Martin Sebor  <msebor@redhat.com>
15806             Richard Biener  <rguenther@suse.de>
15808         PR testsuite/85888
15809         * calls.c (get_size_range): Call determine_value_range instead
15810         of get_value_range..
15811         * tree-vrp.h (determine_value_range): Declared new function.
15812         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15814 2018-05-29  Richard Biener  <rguenther@suse.de>
15816         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15817         sure to use non-pattern stmts for get_earlier_stmt arguments.
15818         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15819         called on pattern stmts.
15820         (get_later_stmt): Likewise.
15822 2018-05-29  Martin Liska  <mliska@suse.cz>
15824         PR gcov-profile/85759
15825         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15826         env variables.
15828 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
15830         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15831         VEC_UNPACK_*_EXPR.
15832         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15833         VEC_PACK_*_EXPR.
15835         PR target/85918
15836         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15837         VEC_PACK_FLOAT_EXPR): New tree codes.
15838         * tree-pretty-print.c (op_code_prio): Handle
15839         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15840         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15841         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15842         * tree-inline.c (estimate_operator_cost): Likewise.
15843         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15844         * fold-const.c (const_binop): Likewise.
15845         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15846         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15847         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15848         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15849         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15850         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15851         * expr.c (expand_expr_real_2): Likewise.
15852         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15853         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15854         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15855         optabs.
15856         * optabs.c (expand_widen_pattern_expr): For
15857         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15858         sign from result type rather than operand's type.
15859         (expand_binop_directly): For vec_packu_float_optab and
15860         vec_packs_float_optab allow result type to be different from operand's
15861         type.
15862         * optabs-tree.c (optab_for_tree_code): Handle
15863         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15864         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
15865         * tree-vect-generic.c (expand_vector_operations_1):  Handle
15866         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15867         VEC_PACK_FLOAT_EXPR.
15868         * tree-vect-stmts.c (supportable_widening_operation): Handle
15869         FIX_TRUNC_EXPR.
15870         (supportable_narrowing_operation): Handle FLOAT_EXPR.
15871         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15872         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15873         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
15874         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15875         mode attributes.
15876         (vec_pack<floatprefix>_float_<mode>): New expander.
15877         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15878         attributes.
15879         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15880         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15881         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15882         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15883         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15884         Document.
15885         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15886         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15887         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15888         VEC_PACK_FLOAT_EXPR): Document.
15890 2018-05-29  Richard Biener  <rguenther@suse.de>
15892         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15893         member.
15894         (stmt_vec_info_vec): Make pointer.
15895         (init_stmt_vec_info_vec): Remove.
15896         (free_stmt_vec_info_vec): Likewise.
15897         (set_stmt_vec_info_vec): New function.
15898         (free_stmt_vec_infos): Likewise.
15899         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15900         (set_vinfo_for_stmt): Likewise.
15901         (get_earlier_stmt): Likewise.
15902         (get_later_stmt): Likewise.
15903         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15904         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15905         (vec_info::~vec_info): Free stmt_vec_infos.
15906         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15907         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15908         (pass_slp_vectorize::execute): Likewise.
15909         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15910         (free_stmt_vec_info_vec): Likewise.
15911         (set_stmt_vec_info_vec): New function.
15912         (free_stmt_vec_infos): Likewise.
15913         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15914         the global stmt_vec_info_vec.
15915         * tree-parloops.c (gather_scalar_reductions): Use
15916         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15917         vector.
15919 2018-05-29  Richard Biener  <rguenther@suse.de>
15921         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15923 2018-05-29  Martin Liska  <mliska@suse.cz>
15924             David Malcolm  <dmalcolm@redhat.com>
15926         * vec.c (test_reverse): New.
15927         (vec_c_tests): Add new test.
15928         * vec.h (vl_ptr>::reverse): New function.
15930 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
15932         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15934         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15935         and later.
15937 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15939         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15941 2018-05-28  Richard Biener  <rguenther@suse.de>
15943         PR tree-optimization/85933
15944         * tree-vect-data-refs.c (vect_record_base_alignments): Only
15945         look at stmts marked as vectorizable.
15947 2018-05-28  Richard Biener  <rguenther@suse.de>
15949         PR tree-optimization/85934
15950         * tree-vect-generic.c (expand_vector_operations_1): Hoist
15951         vector boolean check before scalar optimization.
15953 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
15955         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15956         for armv5te.
15958 2018-05-28  Mark Wielaard  <mark@klomp.org>
15960         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15961         if it is an expression containing a minus sign.
15963 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
15965         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15967 2018-05-27  Paul Koning  <ni1d@arrl.net>
15969         * config/pdp11/pdp11.md (truncsihi2): Remove.
15971 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
15972             Chung-Ju Wu  <jasonwucj@gmail.com>
15974         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15975         implementation.
15976         (unaligned_store_dw): Ditto.
15977         * config/nds32/nds32-memory-manipulation.c
15978         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15979         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15980         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15981         (emit_setmem_word_loop): Rename to ...
15982         (emit_setmem_doubleword_loop): ... this.
15983         (nds32_gen_dup_4_byte_to_word_value): New function.
15984         (nds32_gen_dup_8_byte_to_double_word_value): New function.
15985         (nds32_expand_setmem_loop): Refine implementation.
15986         (nds32_expand_setmem_loop_v3m): Ditto.
15987         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15988         pattern.
15990 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15992         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15994 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15996         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15997         (nds32_init_machine_status): Initialize machine->attr_naked_p and
15998         machine->attr_no_prologue_p.
15999         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
16000         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
16001         (nds32_expand_epilogue): Consider attr_naked_p.
16002         (nds32_expand_epilogue_v3pop): Likewise.
16003         (nds32_can_use_return_insn): Likewise.
16004         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
16005         attr_no_prologue_p fields.
16006         * config/nds32/nds32.opt (mret-in-naked-func): New option.
16008 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
16010         PR target/85918
16011         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
16012         attributes.
16013         * config/i386/sse.md
16014         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
16015         Rename to ...
16016         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
16017         ... this.
16018         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
16019         Rename to ...
16020         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
16021         ... this.
16022         (*<floatsuffix>floatv2div2sf2): Rename to ...
16023         (*float<floatunssuffix>v2div2sf2): ... this.
16024         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
16025         (float<floatunssuffix>v2div2sf2_mask): ... this.
16026         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
16027         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
16028         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
16029         to ...
16030         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
16031         ... this.
16032         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16033         Rename to ...
16034         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16035         ... this.
16036         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16037         Rename to ...
16038         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16039         ... this.
16040         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
16041         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
16042         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
16043         gen_ufix_truncv8dfv8si2.
16044         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
16045         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
16046         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
16047         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
16048         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
16049         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
16050         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
16051         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
16053 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
16055         PR target/85900
16056         PR target/85345
16057         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
16059 2018-05-25  Jim Wilson  <jimw@sifive.com>
16061         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
16062         * config/riscv/riscv.c (struct machine_function): Add
16063         interrupt_handler_p and attribute_checked_p fields.
16064         (riscv_attribute_table): Add interrupt.
16065         (riscv_interrupt_type_p): New.
16066         (riscv_save_reg_p): Save extra regs for interrupt handler.
16067         (riscv_use_save_libcall): Return false  for interrupt handler.
16068         (riscv_first_stack_step): Add forward declaration.
16069         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
16070         for interrupt handler with large frame.  Use it for saved reg list.
16071         (riscv_expand_prologue): Move flag_stack_usage_info support to
16072         eliminate duplication.
16073         (riscv_expand_epilogue): Generate mret for interrupt handler.
16074         (riscv_epilogue_uses): New.
16075         (riscv_can_use_return_insn): Return false for interrupt handler.
16076         (riscv_function_ok_for_sibcall): Likewise.
16077         (riscv_set_current_function): Add interrupt handler support.
16078         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
16079         * config/riscv/riscv.md (UNSPECV_MRET): New.
16080         (GP_REGNUM): New.
16081         (riscv_frflags, riscv_fsflags): Use tab after opcode.
16082         (riscv_mret): New.
16083         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
16085 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
16087         PR tree-optimization/85712
16088         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
16089         this candidate has already been replaced in-situ by a copy.
16091 2018-05-25  Jason Merrill  <jason@redhat.com>
16093         PR c++/80485 - inline function non-zero address.
16094         * symtab.c (nonzero_address): Check DECL_COMDAT.
16096 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
16098         PR target/83628
16099         * config/alpha/alpha.md (ashlsi3): New insn pattern.
16100         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
16101         extension of SImode operation.  Use const123_operand predicate.
16102         (*saddsi_1): Remove.
16103         (*saddl_se_1): Ditto.
16104         (*ssubsi_1): Ditto.
16105         (*ssubl_se_1): Ditto.
16106         * config/alpha/predicates.md (const123_operand): New predicate.
16107         * config/alpha/constraints.md (P): Use IN_RANGE.
16109 2018-05-25  Richard Biener  <rguenther@suse.de>
16111         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
16112         defaulted to true.
16113         (ref_maybe_used_by_stmt_p): Likewise.
16114         (stmt_may_clobber_ref_p): Likewise.
16115         (stmt_may_clobber_ref_p_1): Likewise.
16116         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
16117         and pass it along.
16118         (ref_maybe_used_by_stmt_p): Likewise.
16119         (stmt_may_clobber_ref_p): Likewise.
16120         (stmt_may_clobber_ref_p_1): Likewise.
16121         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
16122         the alias oracle to disambiguate DRs with stmts DR analysis
16123         couldn't handle.
16124         (vect_analyze_data_refs): Do not give up on not analyzable
16125         DRs for BB vectorization.  Remove code truncating the dataref
16126         vector.
16128 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
16130         PR target/85832
16131         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16132         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
16133         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
16135 2018-05-25  Richard Biener  <rguenther@suse.de>
16137         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
16138         function, combining stmt data ref gathering and fatal analysis
16139         parts.
16140         (vect_analyze_data_refs): Remove now redudnant code and simplify.
16141         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
16142         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
16143         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
16144         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
16146 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16148         PR tree-optimization/85720
16149         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
16150         SCC if all partitions are builtins.
16151         (version_loop_by_alias_check): New parameter.  Generate cancelable
16152         runtime alias check if all partitions are builtins.
16153         (distribute_loop): Update call to above function.
16155 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16157         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
16158         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
16159         (parm_default_def_partition_arg): Ditto.
16160         (set_parm_default_def_partition): Ditto.
16161         (get_parm_default_def_partitions): Ditto and make it static.
16162         (get_undefined_value_partitions): Ditto and make it static.
16163         (remove_ssa_form): Refactor call to init_var_map here.
16164         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
16165         computation for loop region.
16166         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
16167         (register_default_def): Delete.
16168         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
16169         (parm_default_def_partition_arg): Ditto.
16170         (set_parm_default_def_partition): Ditto.
16171         (get_parm_default_def_partitions): Ditto and make it static.
16172         (get_undefined_value_partitions): Ditto and make it static.
16173         (coalesce_with_default, coalesce_with_default): Update comment.
16174         (create_coalesce_list_for_region): New func factored out from
16175         create_outofssa_var_map.
16176         (populate_coalesce_list_for_outofssa): New func factored out from
16177         create_outofssa_var_map and coalesce_ssa_name.
16178         (create_outofssa_var_map): Delete.
16179         (coalesce_ssa_name): Refactor to support live range computation.
16180         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
16181         (get_parm_default_def_partitions): Delete.
16182         (get_undefined_value_partitions): Ditto.
16183         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
16184         computation for loop region.
16185         (new_tree_live_info, loe_visit_block): Ditto.
16186         (live_worklist, set_var_live_on_entry): Ditto.
16187         (calculate_live_on_exit, verify_live_on_entry): Ditto.
16188         * tree-ssa-live.h (struct _var_map): New fields.
16189         (init_var_map): Change decl.
16190         (region_contains_p): New.
16192 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
16194         * tree-ssa-live.h (live_merge_and_clear): Delete.
16196 2018-05-25  Richard Biener  <rguenther@suse.de>
16198         PR c++/85912
16199         * tree-dump.c (dequeue_and_dump): Remove access to removed
16200         operand 2 of a SWITCH_EXPR.
16202 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16204         * doc/sourcebuild.texi (vect_double_cond_arith): Include
16205         multiplication and division.
16206         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
16207         (cond_udiv@var{m}, cond_umod@var{m}): Document.
16208         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
16209         (cond_udiv_optab, cond_umod_optab): New optabs.
16210         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
16211         (IFN_COND_RDIV): New internal functions.
16212         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
16213         TRUNC_MOD_EXPR and RDIV_EXPR.
16214         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
16215         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
16216         New unspecs.
16217         (SVE_INT_BINARY): Include mult.
16218         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
16219         (optab, sve_int_op): Handle mult.
16220         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
16221         UNSPEC_COND_DIV.
16222         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
16223         for SVE_INT_BINARY_SD.
16225 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16227         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
16228         (optab, sve_int_op): Handle div and udiv.
16229         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
16230         for SVE_INT_BINARY_SD.
16231         (*<optab><mode>3): New insn for the same.
16233 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16235         * tree-vect-patterns.c: Include predict.h.
16236         (vect_recog_divmod_pattern): Restrict check for division support
16237         to when optimizing for size.
16239 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16241         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
16242         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
16243         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
16244         (gimple_match_op::set_op): Likewise.
16245         (gimple_resimplify4): Declare.
16246         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
16247         (expr::gen_transform): Likewise.
16248         (decision_tree::gen): Generate a simplification routine for 4 operands.
16249         * gimple-match-head.c (gimple_simplify): Add an overload for
16250         4 operands.  In the top-level function, handle up to 4 call
16251         arguments and call gimple_resimplify4.
16252         (gimple_resimplify4): New function.
16253         (build_call_internal): Pass a fourth operand.
16254         (maybe_push_to_seq): Likewise.
16255         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
16256         Fold VEC_COND_EXPRs of an operation and a default value into
16257         an IFN_COND_* function if possible.
16258         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
16259         New unspecs.
16260         (SVE_COND_FP_BINARY): Include them.
16261         (optab, sve_fp_op): Handle them.
16262         (SVE_INT_BINARY_REV): New code iterator.
16263         (SVE_COND_FP_BINARY_REV): New int iterator.
16264         (commutative): New int attribute.
16265         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
16266         Declare.
16267         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
16268         function.
16269         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
16270         (*cond_<optab><mode>): New patterns for reversed operands.
16272 2018-05-25  Richard Biener  <rguenther@suse.de>
16274         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
16275         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
16276         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
16277         (STMT_VINFO_GROUPED_ACCESS): Adjust.
16278         * tree-vect-data-refs.c (everywhere): Adjust users.
16279         * tree-vect-loop.c (everywhere): Likewise.
16280         * tree-vect-slp.c (everywhere): Likewise.
16281         * tree-vect-stmts.c (everywhere): Likewise.
16282         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
16284 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16286         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
16287         Rename to...
16288         (gcc_cv_as_section_exclude): ... this.
16289         Try Solaris as #exclude syntax.
16290         * configure: Regenerate.
16291         * config.in: Regenerate.
16292         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
16293         SECTION_EXCLUDE.
16294         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
16295         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
16297         * varasm.c (default_elf_asm_named_section): Don't check if
16298         HAVE_GAS_SECTION_EXCLUDE is defined.
16300 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16302         * doc/md.texi: Update the documentation of the cond_* optabs
16303         to mention the new final operand.  Fix GET_MODE_NUNITS call.
16304         Describe the scalar case too.
16305         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
16306         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
16307         instead of 2.
16308         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
16309         (get_conditional_internal_fn): Update comment.
16310         * tree-vect-loop.c (vectorizable_reduction): Pass the original
16311         accumulator value as a final argument to conditional functions.
16312         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
16313         a define_expand and add an "else" operand.  Assert for now that
16314         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
16315         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
16316         (*cond_<optab><mode>): New patterns.
16317         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
16318         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16319         (UNSPEC_COND_EOR): Delete.
16320         (optab): Remove associated mappings.
16321         (SVE_INT_BINARY): New code iterator.
16322         (sve_int_op): Remove int attribute and add "minus" to the code
16323         attribute.
16324         (SVE_COND_INT_OP): Delete.
16325         (SVE_COND_FP_OP): Rename to...
16326         (SVE_COND_FP_BINARY): ...this.
16328 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
16330         * optabs.c (can_reuse_operands_p): New function.
16331         (maybe_legitimize_operands): Try to reuse the results for
16332         earlier operands.
16334 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16336         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
16337         Add {q} suffix to insn mnemonic.
16339 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16341         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
16342         (msp430_warn_func_return): New.
16344 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
16346         * fold-const.c (tree_nonzero_bits): New function.
16347         * fold-const.h (tree_nonzero_bits): Likewise.
16348         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
16349         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
16351 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
16353         PR target/85900
16354         PR target/85345
16355         * varasm.c (assemble_alias): Check ifunc_resolver only on
16356         FUNCTION_DECL.
16358 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
16360         PR target/85903
16361         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
16362         when memory input operand is handled.
16364 2018-05-24  Luis Machado  <luis.machado@linaro.org>
16366         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
16367         global.
16368         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
16370 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16372         * match.pd: Delay FMA folds until after vectorization.
16374 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16376         PR target/83009
16377         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16378         address check not strict.
16380 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
16382         * gimple-match.h (gimple_match_op): New class.
16383         (mprts_hook): Replace parameters with a gimple_match_op *.
16384         (maybe_build_generic_op): Likewise.
16385         (gimple_simplified_result_is_gimple_val): Replace parameters with
16386         a const gimple_match_op *.
16387         (gimple_simplify): Replace code_helper * and tree * parameters with
16388         a gimple_match_op * parameter.
16389         (gimple_resimplify1): Replace code_helper *, tree and tree *
16390         parameters with a gimple_match_op * parameter.
16391         (gimple_resimplify2): Likewise.
16392         (gimple_resimplify3): Likewise.
16393         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
16394         parameters with a gimple_match_op * parameter.
16395         * gimple-match-head.c (gimple_simplify): Change prototypes of
16396         auto-generated functions to take a gimple_match_op * instead of
16397         separate code_helper * and tree * parameters.  Make the same
16398         change in the top-level overload and update calls to the
16399         gimple_resimplify routines.  Update calls to the auto-generated
16400         functions and to maybe_push_res_to_seq in the publicly-facing
16401         operation-specific gimple_simplify overloads.
16402         (gimple_match_op::MAX_NUM_OPS): Define.
16403         (gimple_resimplify1): Replace rcode and ops with a single res_op
16404         parameter.  Update call to gimple_simplify.
16405         (gimple_resimplify2): Likewise.
16406         (gimple_resimplify3): Likewise.
16407         (mprts_hook): Replace parameters with a gimple_match_op *.
16408         (maybe_build_generic_op): Likewise.
16409         (build_call_internal): Replace type, nargs and ops with
16410         a gimple_match_op *.
16411         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
16412         with a single gimple_match_op *.  Update calls to mprts_hook,
16413         build_call_internal and gimple_simplified_result_is_gimple_val.
16414         Factor out code that is common to the tree_code and combined_fn cases.
16415         * genmatch.c (expr::gen_transform): Replace tem_code and
16416         tem_ops with a gimple_match_op called tem_op.  Update calls
16417         to the gimple_resimplify functions and maybe_push_res_to_seq.
16418         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
16419         res_ops.  Update call to the gimple_resimplify functions.
16420         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
16421         (decision_tree::gen): Make the functions take a gimple_match_op *
16422         called res_op instead of separate res_code and res_ops parameters.
16423         Update call accordingly.
16424         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
16425         and ops with a single res_op parameter.  Update calls to
16426         maybe_build_generic_op and maybe_push_res_to_seq.
16427         (fold_stmt_1): Update calls to gimple_simplify and
16428         replace_stmt_with_simplification.
16429         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
16430         and gimple_simplified_result_is_gimple_val.
16431         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
16432         gimple_simplify.
16433         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
16434         with a gimple_match_op *.
16435         (vn_nary_build_or_lookup): Likewise.  Update call to
16436         vn_nary_build_or_lookup_1.
16437         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
16438         gimple_match_op *.  Update calls to the gimple_resimplify routines
16439         and to gimple_simplified_result_is_gimple_val.
16440         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
16441         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
16442         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
16443         (visit_nary_op): Likewise.
16444         (visit_reference_op_load): Likewise.
16446 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16448         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
16449         modifier for printing the step amount.
16451 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
16453         PR target/78849
16454         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
16455         types.
16457 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
16459         * doc/sourcebuild.texi (Endianness): New subsubsection.
16461 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16463         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16464         <prefetch_dynamic_strides>: New const bool field.
16465         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16466         prefetch_dynamic_strides.
16467         (exynosm1_prefetch_tune): Likewise.
16468         (thunderxt88_prefetch_tune): Likewise.
16469         (thunderx_prefetch_tune): Likewise.
16470         (thunderx2t99_prefetch_tune): Likewise.
16471         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
16472         false.
16473         (aarch64_override_options_internal): Update to set
16474         PARAM_PREFETCH_DYNAMIC_STRIDES.
16475         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16476         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16477         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16478         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16479         prefetch-dynamic-strides setting.
16481 2018-05-23  Luis Machado  <luis.machado@linaro.org>
16483         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16484         <minimum_stride>: New const int field.
16485         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16486         minimum_stride field defaulting to -1.
16487         (exynosm1_prefetch_tune): Likewise.
16488         (thunderxt88_prefetch_tune): Likewise.
16489         (thunderx_prefetch_tune): Likewise.
16490         (thunderx2t99_prefetch_tune): Likewise.
16491         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
16492         <default_opt_level>: Set to 3.
16493         (aarch64_override_options_internal): Update to set
16494         PARAM_PREFETCH_MINIMUM_STRIDE.
16495         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16496         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16497         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16498         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16499         stride is constant and is below the minimum stride threshold.
16501 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16503         * config/arm/arm-cpus.in (mode26): Delete.
16504         (armv4): Delete mode26 reference.
16505         * config/arm/arm.c (arm_configure_build_target): Delete use of
16506         isa_bit_mode26.
16508 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16510         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
16511         New insn pattern.
16512         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
16513         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
16514         for non-SSE modes.
16515         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
16516         (floatunsdidf2): Ditto.
16518 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
16520         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
16521         (fixuns_trunc<mode>si2_avx512f): Ditto.
16522         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
16523         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
16524         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
16526 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16528         PR rtl-optimization/79985
16529         * df-scan.c (df_insn_refs_collect): Remove special case for
16530         global registers and asm statements.
16532 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
16534         * extend.texi (Global Register Variables): Rewrite the bullet list.
16535         Note that the register is available for allocation. Note that access
16536         via inline asm must use constraints. Add note about async-signal
16537         handlers. Remove paragraph about automagic register selection.
16539 2018-05-23  Richard Biener  <rguenther@suse.de>
16541         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
16542         of fixed offset from memset VN.
16544 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
16546         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
16547         first_interp field.
16548         (alloc_cand_and_find_basis): Initialize first_interp field.
16549         (slsr_process_mul): Modify first_interp field.
16550         (slsr_process_add): Likewise.
16551         (slsr_process_cast): Modify first_interp field for each new
16552         interpretation.
16553         (slsr_process_copy): Likewise.
16554         (dump_candidate): Dump first_interp field.
16555         (replace_mult_candidate): Process all interpretations, not just
16556         subsequent ones.
16557         (replace_rhs_if_not_dup): Likewise.
16558         (replace_one_candidate): Likewise.
16560 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
16562         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
16563         Add new boolean.
16564         (aarch64_needs_frame_chain): New function.
16565         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
16567 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
16569         PR target/84882
16570         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
16571         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
16572         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
16573         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
16574         as true for strict-align.
16575         (aarch64_can_inline_p): Perform checks even when callee has no
16576         attributes to check for strict alignment.
16577         * doc/extend.texi (AArch64 Function Attributes): Document
16578         no-strict-align.
16579         * doc/invoke.texi: (AArch64 Options): Likewise.
16581 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
16583         PR tree-optimization/85853
16584         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
16585         the handling of the root of the node to...
16586         (vect_slp_analyze_node_operations_1): ...this new function,
16587         and run the whole thing with the child nodes' def types
16588         set according to their SLP node's def type.
16590 2018-05-23  Richard Biener  <rguenther@suse.de>
16592         PR middle-end/85874
16593         * tree-data-ref.c (create_runtime_alias_checks): Defer
16594         and ignore overflow warnings.
16596 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
16598         PR tree-optimization/85822
16599         * tree-vrp.c (is_masked_range_test): Fix handling of negative
16600         constants.
16602 2018-05-23  Richard Biener  <rguenther@suse.de>
16604         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16605         memset constants via native_interpret_expr.
16607 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
16609         PR target/85345
16610         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
16611         attribute.
16612         (cgraph_node::create_alias): Likewise.
16613         (cgraph_node::get_availability): Check ifunc_resolver instead
16614         of looking up ifunc attribute.
16615         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
16616         * varasm.c (do_assemble_alias): Likewise.
16617         (assemble_alias): Likewise.
16618         (default_binds_local_p_3): Likewise.
16619         * cgraph.h (cgraph_node): Add ifunc_resolver.
16620         (cgraph_node::only_called_directly_or_aliased_p): Return false
16621         for IFUNC resolver.
16622         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16623         attribute.
16624         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16625         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16626         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16627         instead of looking up ifunc attribute.
16629 2018-05-22  Luis Machado  <luis.machado@linaro.org>
16631         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16633 2018-05-22  Martin Sebor  <msebor@redhat.com>
16635         PR middle-end/85359
16636         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16637         only when expasion succeeds.
16638         (expand_builtin_strcmp): Same.
16639         (expand_builtin_strncmp): Same.
16641 2018-05-22  Martin Sebor  <msebor@redhat.com>
16643         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16645 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16646             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16648         * config/aarch64/aarch64-ldpstp.md: Replace uses of
16649         aarch64_mem_pair_operand with memory_operand and delete operand swapping
16650         code.
16651         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16652         Add check for legitimate_address.
16653         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16654         (aarch64_swap_ldrstr_operands): New.
16655         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16656         Define prototype.
16658 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16659             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16661         * config/aarch64/aarch64.md: New patterns to generate stp
16662         and ldp.
16663         (store_pair_sw, store_pair_dw): New patterns to generate stp for
16664         single words and double words.
16665         (load_pair_sw, load_pair_dw): Likewise.
16666         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16667         Delete.
16668         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16669         Delete.
16670         * config/aarch64/aarch64-ldpstp.md: Modify peephole
16671         for different mode ldpstp and add peephole for merged zero stores.
16672         Likewise for loads.
16673         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16674         Add size check.
16675         (aarch64_gen_store_pair): Rename calls to match new patterns.
16676         (aarch64_gen_load_pair): Rename calls to match new patterns.
16677         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16678         (load_pair<DREG:mode><DREG2:mode>): ... This.
16679         (store_pair<mode>): Rename to...
16680         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16681         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16682         New mode iterators.
16683         (V_INT_EQUIV): Handle SImode.
16684         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16685         New predicate.
16687 2018-05-22  Martin Sebor  <msebor@redhat.com>
16689         PR c/85623
16690         * calls.c (maybe_warn_nonstring_arg): Use string length to set
16691         or ajust the presumed bound on an operation to avoid unnecessary
16692         warnings.
16694 2018-05-22  Martin Sebor  <msebor@redhat.com>
16696         PR tree-optimization/85826
16697         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16698         assuming that a DECL necesarily has a constant size.
16700 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
16702         PR middle-end/85862
16703         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16705 2018-05-22  Richard Biener  <rguenther@suse.de>
16707         PR tree-optimization/85834
16708         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16709         non-constant and non-zero memset arguments.
16711 2018-05-22  Martin Liska  <mliska@suse.cz>
16713         PR ipa/85607
16714         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16716 2018-05-22  Richard Biener  <rguenther@suse.de>
16718         PR tree-optimization/85863
16719         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16720         comparisons when vectype is specified.
16721         (vectorizable_condition): Do not specify vectype for
16722         vect_is_simple_cond when SLP vectorizing.
16724 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
16726         PR target/85657
16727         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16728         define __ibm128 as long double.
16729         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16730         as a distinct type when IEEE 128-bit support is enabled.
16731         (init_float128_ieee): Fix up conversions between IFmode and IEEE
16732         128-bit types to use the correct functions.
16733         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16734         convert between 128-bit floating point types that have different
16735         modes but the same representation, instead of using gen_lowpart to
16736         makean alias.
16737         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16738         KFmode.
16739         (IFKF_reg): New attributes to give the register constraints for
16740         IFmode and KFmode.
16741         (extend<mode>tf2_internal): New insns to mark an explicit
16742         conversion between 128-bit floating point types that have a
16743         different mode but share the same representation.
16745 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
16747         PR tree-optimization/85814
16748         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16749         a null return from get_strinfo when unsharing the next
16750         strinfo in the chain.
16752 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
16754         PR gcc/84923
16755         * varasm.c (weak_finish): Clean up weak_decls.
16757 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16759         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16760         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16761         UNSPEC_UADALP values.
16762         * config/aarch64/iterators.md (ABAL): New int iterator.
16763         (ABDL2): Likewise.
16764         (ADALP): Likewise.
16765         (sur): Add mappings for the above.
16766         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16767         New define_insn.
16768         (aarch64_<sur>abal<mode>_4): Likewise.
16769         (aarch64_<sur>adalp<mode>_3): Likewise.
16770         (<sur>sadv16qi): New define_expand.
16772 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
16774         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16775         (*movdf_internal): Ditto.
16776         (*rcpsf2_sse): Ditto.
16777         (*rsqrtsf2_sse): Ditto.
16778         (*sqrt<mode>2_sse): Ditto.
16780 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
16782         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16783         eor3q<mode>4.
16784         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16785         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16786         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16787         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16788         vbcaxq_s64): New.
16789         * config/aarch64/arm_neon.h: Likewise.
16790         * config/aarch64/iterators.md (VQ_I): New.
16792 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
16794         * config.gcc: Add arc/t-multilib-linux to tmake_file for
16795         arc*-*-linux*.
16796         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16797         MULTILIB_DIRNAMES
16799 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16801         * config/nds32/constraints.md (S): New constraint.
16802         * config/nds32/nds32.md (call_internal): Use constraint S.
16803         (call_value_internal): Likewise.
16804         (sibcall_internal): Likewise.
16805         (sibcall_value_internal): Likewise.
16807 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16808             Chung-Ju Wu  <jasonwucj@gmail.com>
16810         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16811         into consideration.
16813 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16814             Chung-Ju Wu  <jasonwucj@gmail.com>
16816         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16817         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16818         (nds32_rtx_costs_impl): Simplify.
16819         (nds32_address_cost_impl): Simplify.
16820         (nds32_init_rtx_costs): New function.
16821         (nds32_rtx_costs_speed_prefer): Likewise.
16822         (nds32_rtx_costs_size_prefer): Likewise.
16823         (nds32_address_cost_speed_prefer): Likewise.
16824         (nds32_address_cost_speed_fwprop): Likewise.
16825         (nds32_address_cost_size_prefer): Likewise.
16826         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16827         * config/nds32/nds32.c (nds32_option_override): Use
16828         nds32_init_rtx_costs function.
16830 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16832         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16833         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16834         (TARGET_PIPELINE_N8): Likewise.
16835         (TARGET_PIPELINE_N10): Likewise.
16836         (TARGET_PIPELINE_N13): Likewise.
16837         (TARGET_PIPELINE_GRAYWOLF): Likewise.
16839 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
16841         * config/nds32/nds32-fpu.md: Update copyright year.
16843 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16845         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16847 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16849         * config/nds32/nds32.c
16850         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16851         * config/nds32/nds32.opt (minline-asm-r15): New option.
16853 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16855         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16856         MASK_HW_ABS.
16857         * config/nds32/nds32.md (abssi2): New pattern.
16859 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
16861         * config/i386/i386.md (rex64namesuffix): New mode attribute.
16862         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16863         Merge insn pattern from sse_cvtsi2ss<round_name> and
16864         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16865         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16866         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16867         using SWI48 mode iterator.
16868         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16869         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16870         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16871         pattern from sse_cvttss2si<round_saeonly_name>
16872         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16873         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16874         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16875         using SWI48 mode iterator.
16876         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16877         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16878         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16879         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16880         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16881         using SWI48 mode iterator.
16882         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16883         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16884         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16885         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16886         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16887         SWI48 mode iterator.
16888         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16889         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16890         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16891         pattern from sse_cvttsd2si<round_saeonly_name>
16892         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16894 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16896         * config/nds32/nds32-md-auxiliary.c
16897         (nds32_valid_smw_lwm_base_p): Refine.
16898         (nds32_output_smw_single_word): Refine.
16899         (nds32_output_smw_double_word): New.
16900         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16902 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16904         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16905         (nds32_output_stack_pop): Refine.
16906         (nds32_expand_unaligned_load): Refine.
16907         (nds32_expand_unaligned_store): Refine.
16909 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16910             Chung-Ju Wu  <jasonwucj@gmail.com>
16912         * config/nds32/constants.md: Add TP_REGNUM constant.
16913         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16914         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16915         UNSPEC_ADD32.
16916         * config/nds32/nds32-doubleword.md: Consider flag_pic.
16917         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16918         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16919         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16920         and PIC code generation.
16921         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16922         code generation.
16923         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16924         optimization.
16925         * config/nds32/nds32.md: Support TLS and PIC.
16926         * config/nds32/nds32.c: Support TLS and PIC.
16927         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16928         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16929         predicate.
16931 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16933         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16934         mode with E_ prefix.
16936 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16937             Chung-Ju Wu  <jasonwucj@gmail.com>
16939         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16940         * config/nds32/nds32-md-auxiliary.c
16941         (symbolic_reference_mentioned_p): New.
16942         (nds32_legitimize_ict_address): New.
16943         (nds32_expand_ict_move): New.
16944         (nds32_indirect_call_referenced_p): New.
16945         (nds32_symbol_binds_local_p): Delete.
16946         (nds32_long_call_p): Modify.
16947         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16948         * config/nds32/nds32-protos.h
16949         (symbolic_reference_mentioned_p): Declare.
16950         (nds32_legitimize_ict_address): Declare.
16951         (nds32_expand_ict_move): Declare.
16952         (nds32_indirect_call_referenced_p): Declare.
16953         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16954         (nds32_relax_group): Use nds32_ict_const_p as condition.
16955         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16956         (nds32_asm_file_start): Output ict_model directive in asm code.
16957         (nds32_legitimate_address_p): Consider indirect call.
16958         (nds32_print_operand): Consider indirect call.
16959         (nds32_print_operand_address): Consider indirect call.
16960         (nds32_insert_attributes): Handle "indirect_call" attribute.
16961         (TARGET_LEGITIMATE_ADDRESS_P): Define.
16962         (TARGET_LEGITIMATE_CONSTANT_P): Define.
16963         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16964         (TARGET_DELEGITIMIZE_ADDRESS): Define.
16965         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16966         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16967         (TARGET_ICT_MODEL_SMALL): Define.
16968         (TARGET_ICT_MODEL_LARGE): Define.
16969         * config/nds32/nds32.md (movsi): Consider ict model.
16970         (call, call_value): Consider ict model.
16971         (sibcall, sibcall_value): Consider ict model.
16972         * config/nds32/nds32.opt (mict-model): New option.
16973         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16974         model.
16976 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
16977             Monk Chiang  <sh.chiang04@gmail.com>
16978             Jim Wilson <jimw@sifive.com>
16980         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16981         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
16982         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16983         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16984         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
16985         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16986         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16987         compute save_libcall_adjustment properly.
16988         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16989         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16990         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16991         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16992         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16993         (ABI_SPEC): Handle mabi=ilp32e.
16994         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16995         (RVE): Add RVE mask.
16996         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16997         <-march>: Add rv32e as an example.
16999 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
17001         PR c++/82899
17002         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
17003         (intra_create_variable_infos): Handle C++ constructors.
17005 2018-05-18  Martin Liska  <mliska@suse.cz>
17007         * passes.def: Remove a redundant pass.
17009 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
17011         PR bootstrap/85838
17012         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
17014 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17016         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
17017         (ARMv4): Update.
17018         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
17019         (ARMv6m): Update.
17020         (armv2, armv2a, armv3, armv3m): Delete architectures.
17021         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
17022         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
17023         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
17024         Delete cpus.
17025         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
17026         (*mulsidi3adddi): Likewise.
17027         (mulsidi3): Likewise.
17028         (*mulsidi3_nov6): Likewise.
17029         (umulsidi3): Likewise.
17030         (umulsidi3_nov6): Likewise.
17031         (umaddsidi4): Likewise.
17032         (*umulsidi3adddi): Likewise.
17033         (smulsi3_highpart): Likewise.
17034         (*smulsi3_highpart_nov6): Likewise.
17035         (umulsi3_highpart): Likewise.
17036         (*umulsi3_highpart_nov6): Likewise.
17037         * config/arm/arm.h (arm_arch3m): Delete.
17038         * config/arm/arm.c (arm_arch3m): Delete.
17039         (arm_option_override_internal): Update armv3-related comment.
17040         (arm_configure_build_target): Delete use of isa_bit_mode32.
17041         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
17042         (arm_rtx_costs_internal): Delete check of arm_arch3m.
17043         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
17044         (mulsa3): Likewise.
17045         (mulusa3): Likewise.
17046         * config/arm/arm-protos.h (arm_arch3m): Delete.
17047         * config/arm/arm-tables.opt: Regenerate.
17048         * config/arm/arm-tune.md: Likewise.
17049         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
17050         deleted architectures.
17052 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17054         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
17055         (armv5t, armv5te): New features.
17056         (ARMv5, ARMv5e): Delete fgroups.
17057         (ARMv5t, ARMv5te): Adjust for above changes.
17058         (ARMv6m): Likewise.
17059         (armv5, armv5e): Delete arches.
17060         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
17061         arm_arch5.
17062         (*call_reg_arm): Likewise.
17063         (*call_value_reg_armv5): Likewise.
17064         (*call_value_reg_arm): Likewise.
17065         (*call_symbol): Likewise.
17066         (*call_value_symbol): Likewise.
17067         (*sibcall_insn): Likewise.
17068         (*sibcall_value_insn): Likewise.
17069         (clzsi2): Likewise.
17070         (prefetch): Likewise.
17071         (define_split and define_peephole2 dependent on arm_arch5):
17072         Likewise.
17073         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
17074         arm_arch5e.
17075         (TARGET_ARM_QBIT): Likewise.
17076         (TARGET_DSP_MULTIPLY): Likewise.
17077         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
17078         (arm_arch5, arm_arch5e): Delete.
17079         (arm_arch5t, arm_arch5te): Declare.
17080         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
17081         (arm_arch5t): Declare.
17082         (arm_option_reconfigure_globals): Update for the above.
17083         (arm_options_perform_arch_sanity_checks): Update comment, replace
17084         use of arm_arch5 with arm_arch5t.
17085         (use_return_insn): Likewise.
17086         (arm_emit_call_insn): Likewise.
17087         (output_return_instruction): Likewise.
17088         (arm_final_prescan_insn): Likewise.
17089         (arm_coproc_builtin_available): Likewise.
17090         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
17091         arm_arch5e with arm_arch5t and arm_arch5te.
17092         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
17093         (arm_arch5t, arm_arch5te): Declare.
17094         * config/arm/arm-tables.opt: Regenerate.
17095         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
17096         * config/arm/t-multilib: Likewise.
17097         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
17098         instead of arm_arch5.
17099         (*call_reg_thumb1): Likewise.
17100         (*call_value_reg_thumb1_v5): Likewise.
17101         (*call_value_reg_thumb1): Likewise.
17102         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
17103         unreachable path.
17104         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
17106 2018-05-18  Martin Liska  <mliska@suse.cz>
17108         PR gcov-profile/84846
17109         * doc/gcov.texi: Document -t option of gcov tool.
17111 2018-05-18  Martin Liska  <mliska@suse.cz>
17113         PR gcov-profile/84846
17114         * gcov.c (print_usage): Add new -t option.
17115         (process_args): Handle the option.
17116         (generate_results): Use stdout as output when requested by
17117         the option.
17119 2018-05-18  Martin Liska  <mliska@suse.cz>
17121         PR gcov-profile/84846
17122         * coverage.c (coverage_init): Write PWD to .gcno file.
17123         * doc/gcov.texi: Document how working directory is printed.
17124         * gcov-dump.c (dump_gcov_file): Print PWD.
17125         * gcov.c (output_intermediate_file): Likewise.
17126         (read_graph_file): Read PWD string.
17127         (output_lines): Print PWD.
17129 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17131         PR middle-end/85817
17132         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
17133         for retval and return false if all args to phi are zero.
17135 2018-05-18  Richard Biener  <rguenther@suse.de>
17137         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
17138         method.
17139         (evrp_dom_walker::before_dom_children): Call it.
17141 2018-05-18  Richard Biener  <rguenther@suse.de>
17143         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
17144         results when processing array refs with variable index.
17146 2018-05-18  Toon Moene  <toon@moene.org>
17148         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
17149         directly after that of -floop-interchange. Indicate that both
17150         options are enabled by default when specifying -O3.
17152 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17154         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
17155         iterator.  Delete separate integer-mode vec_set<mode> expander.
17156         (aarch64_simd_vec_setv2di): Delete.
17157         (vec_setv2di): Delete.
17158         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
17159         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
17160         the "w, r" alternative.
17162 2018-05-18  Martin Liska  <mliska@suse.cz>
17164         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
17165         * tree-pass.h (make_pass_lower_switch_O0): New function.
17166         * tree-switch-conversion.c (node_has_low_bound): Remove.
17167         (node_has_high_bound): Likewise.
17168         (node_is_bounded): Likewise.
17169         (class pass_lower_switch): Make it a template type and create
17170         two instances.
17171         (pass_lower_switch::execute): Add template argument.
17172         (make_pass_lower_switch): New function.
17173         (make_pass_lower_switch_O0): New function.
17174         (do_jump_if_equal): Remove.
17175         (emit_case_nodes): Simplify to just handle all 3 cases and leave
17176         all the hard work to tree optimization passes.
17178 2018-05-18  Martin Liska  <mliska@suse.cz>
17180         * dbgcnt.c (limit_low): Renamed from limit.
17181         (limit_high): New variable.
17182         (dbg_cnt_is_enabled): Check for upper limit.
17183         (dbg_cnt): Adjust dumping.
17184         (dbg_cnt_set_limit_by_index): Add new argument for high
17185         value.
17186         (dbg_cnt_set_limit_by_name): Likewise.
17187         (dbg_cnt_process_single_pair): Parse new format.
17188         (dbg_cnt_process_opt): Use strtok.
17189         (dbg_cnt_list_all_counters): Remove 'value' and add
17190         'limit_high'.
17191         * doc/invoke.texi: Document changes.
17193 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
17195         * doc/sourcebuild.texi (scalar_all_fma): Document.
17196         * tree.def (FMA_EXPR): Delete.
17197         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
17198         * internal-fn.c (ternary_direct): New macro.
17199         (expand_ternary_optab_fn): Likewise.
17200         (direct_ternary_optab_supported_p): Likewise.
17201         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
17202         * builtins.c (fold_builtin_fma): Delete.
17203         (fold_builtin_3): Don't call it.
17204         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
17205         * expr.c (expand_expr_real_2): Likewise.
17206         * fold-const.c (operand_equal_p): Likewise.
17207         (fold_ternary_loc): Likewise.
17208         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
17209         * gimple.c (DEFTREECODE): Likewise.
17210         * gimplify.c (gimplify_expr): Likewise.
17211         * optabs-tree.c (optab_for_tree_code): Likewise.
17212         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
17213         * tree-eh.c (operation_could_trap_p): Likewise.
17214         (stmt_could_throw_1_p): Likewise.
17215         * tree-inline.c (estimate_operator_cost): Likewise.
17216         * tree-pretty-print.c (dump_generic_node): Likewise.
17217         (op_code_prio): Likewise.
17218         * tree-ssa-loop-im.c (stmt_cost): Likewise.
17219         * tree-ssa-operands.c (get_expr_operands): Likewise.
17220         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
17221         * fold-const-call.h (fold_fma): Delete.
17222         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
17223         CFN_FNMA and CFN_FNMS.
17224         (fold_fma): Delete.
17225         * genmatch.c (combined_fn): New enum.
17226         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
17227         (commutative_op): New function.
17228         (commutate): Use it.  Handle more than 2 operands.
17229         (dt_operand::gen_gimple_expr): Use commutative_op.
17230         (parser::parse_expr): Allow :c to be used with non-binary
17231         operators if the commutative operand is known.
17232         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
17233         CFN_FMS, CFN_FNMA and CFN_FNMS.
17234         (backprop::process_assign_use): Remove FMA_EXPR handling.
17235         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
17236         (gen_hsa_fma): New function.
17237         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
17238         IFN_FNMA and IFN_FNMS.
17239         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
17240         * gimple-fold.h (follow_all_ssa_edges): Declare.
17241         * gimple-fold.c (follow_all_ssa_edges): New function.
17242         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
17243         gimple_build interface and use follow_all_ssa_edges to fold the result.
17244         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
17245         instead of checking for optabs directly.
17246         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
17247         rather than FMA_EXPRs.
17248         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
17249         call to IFN_FMA instead of an FMA_EXPR.
17251 2018-05-17  Jim Wilson  <jimw@sifive.com>
17253         * expr.c (do_tablejump): When converting index to Pmode, if we have a
17254         sign extended promoted subreg, and the range does not have the sign bit
17255         set, then do a sign extend.
17257         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
17258         test, check for sign extended subreg and/or constant operands, and
17259         do a sign extend in that case.
17261 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
17263         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
17264         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
17265         Add untyped.
17266         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
17267         Change logics_shift_reg to logics_shift_imm.
17268         (thunderx2t99_fp_loadpair_basic): Delete.
17269         (thunderx2t99_fp_storepair_basic): Delete.
17270         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
17271         (thunderx2t99_asimd_polynomial): Delete.
17272         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
17273         and neon_fp_mul_d_scalar_q.
17274         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
17275         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
17276         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
17277         (thunderx2t99_asimd_lut): Add missing tbl types.
17278         (thunderx2t99_asimd_ext): Delete.
17279         (thunderx2t99_asimd_load1_1_mult): Delete.
17280         (thunderx2t99_asimd_load1_2_mult): Delete.
17281         (thunderx2t99_asimd_load1_ldp): New.
17282         (thunderx2t99_asimd_load1): New.
17283         (thunderx2t99_asimd_load2): Add missing *load2* types.
17284         (thunderx2t99_asimd_load3): New.
17285         (thunderx2t99_asimd_load4): New.
17286         (thunderx2t99_asimd_store1_1_mult): Delete.
17287         (thunderx2t99_asimd_store1_2_mult): Delete.
17288         (thunderx2t99_asimd_store2_mult): Delete.
17289         (thunderx2t99_asimd_store2_onelane): Delete.
17290         (thunderx2t99_asimd_store_stp): New.
17291         (thunderx2t99_asimd_store1): New.
17292         (thunderx2t99_asimd_store2): New.
17293         (thunderx2t99_asimd_store3): New.
17294         (thunderx2t99_asimd_store4): New.
17296 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
17298         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
17299         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
17301 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
17302             Segher Boessenkool  <segher@kernel.crashing.org>
17304         PR target/85698
17305         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
17306         operand.
17308 2018-05-17  Richard Biener  <rguenther@suse.de>
17310         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
17311         for pruning loop and prune defs feeding only already visited PHIs.
17313 2018-05-17  Richard Biener  <rguenther@suse.de>
17315         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
17317 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
17318             Richard Biener  <rguenther@suse.de>
17320         PR tree-optimization/85793
17321         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
17322         for VMAT_ELEMENTWISE.
17324 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17326         * internal-fn.h (lookup_internal_fn): Declare
17327         * internal-fn.c (lookup_internal_fn): New function.
17328         * gimple.c (gimple_build_call_from_tree): Handle calls to
17329         internal functions.
17330         * gimple-pretty-print.c (dump_gimple_call): Print "." before
17331         internal function names.
17332         * tree-pretty-print.c (dump_generic_node): Likewise.
17333         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
17335 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
17337         * gimple-fold.h (gimple_build): Make the function forms take
17338         combined_fn rather than built_in_function.
17339         (gimple_simplify): Likewise.
17340         * gimple-match-head.c (gimple_simplify): Likewise.
17341         * gimple-fold.c (gimple_build): Likewise.
17342         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
17343         rather than gimple_build_call_internal.
17344         (get_initial_defs_for_reduction): Likewise.
17345         (vect_create_epilog_for_reduction): Likewise.
17346         (vectorizable_live_operation): Likewise.
17348 2018-05-17  Martin Liska  <mliska@suse.cz>
17350         * gimple-ssa-sprintf.c (format_directive): Do not use
17351         space in between 'G_' and '('.
17353 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
17355         PR target/85323
17356         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
17357         even if the mask is not all ones.
17359         PR target/85323
17360         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
17361         vector.
17362         (ix86_gimple_fold_builtin): Likewise.
17364         PR target/85323
17365         * config/i386/i386.c: Include tree-vector-builder.h.
17366         (ix86_vector_shift_count): New function.
17367         (ix86_fold_builtin): Fold shift builtins by scalar count.
17368         (ix86_gimple_fold_builtin): Likewise.
17370         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
17371         _mm512_setzero): New intrinsics.
17373 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
17374             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17376         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
17377         code generation for cases where splatting a value is not useful.
17378         * simplify-rtx.c (simplify_ternary_operation): Simplify
17379         vec_merge across a vec_duplicate and a paradoxical subreg forming
17380         a vector mode to a vec_concat.
17382 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
17384         * config.gcc: Support "goldmont-plus".
17385         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17386         "goldmont-plus".
17387         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17388         PROCESSOR_GOLDMONT_PLUS.
17389         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
17390         (processor_target_table): Add "goldmont-plus".
17391         (PTA_GOLDMONT_PLUS): Define.
17392         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
17393         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
17394         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
17395         (fold_builtin_cpu): Add "goldmont-plus".
17396         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
17397         (ix86_option_override_internal): Add "goldmont-plus".
17398         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
17399         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
17400         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
17401         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
17403 2018-05-17  Richard Biener  <rguenther@suse.de>
17405         PR tree-optimization/85757
17406         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
17407         remove defs that only feed that PHI from further processing.
17409 2018-05-16  Jim Wilson  <jimw@sifive.com>
17411         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
17412         asterisk to name.
17413         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
17415 2018-05-16  Mark Wielaard  <mark@klomp.org>
17417         * dwarf2out.c (count_index_strings): New function.
17418         (output_indirect_strings): Call count_index_strings and generate
17419         header for dwarf_version >= 5.
17421 2018-05-16  Mark Wielaard  <mark@klomp.org>
17423         * dwarf2out.c (dwarf_FORM): New function.
17424         (set_indirect_string): Use dwarf_FORM.
17425         (reset_indirect_string): Likewise.
17426         (size_of_die): Likewise.
17427         (value_format): Likewise.
17428         (output_die): Likewise.
17429         (add_skeleton_AT_string): Likewise.
17430         (output_macinfo_op): Likewise.
17431         (index_string): Likewise.
17432         (output_index_string_offset): Likewise.
17433         (output_index_string): Likewise.
17434         (count_index_strings): Likewise.
17436 2018-05-16  Carl Love  <cel@us.ibm.com>
17438         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17439         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
17441 2018-05-16  Martin Jambor  <mjambor@suse.cz>
17443         * ipa-prop.c (ipa_free_all_edge_args): Remove.
17444         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
17446 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
17448         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
17449         (fnma<mode>4): Likewise.
17450         (fms<mode>4): Likewise.
17451         (fnms<mode>4): Likewise.
17452         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
17453         (aarch64_fnma<mode>4): Likewise.
17454         (aarch64_fms<mode>4): Likewise.
17455         (aarch64_fnms<mode>4): Likewise.
17456         (aarch64_fnmadd<mode>4): Likewise.
17458 2018-05-16  Jason Merrill  <jason@redhat.com>
17460         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
17462 2018-05-16  Richard Biener  <rguenther@suse.de>
17464         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
17465         (dump_stmt_cost): Declare.
17466         (add_stmt_cost): Dump cost we add.
17467         (add_stmt_costs): New function.
17468         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
17469         No longer exported.
17470         (vect_analyze_stmt): Adjust prototype.
17471         (vectorizable_condition): Likewise.
17472         (vectorizable_live_operation): Likewise.
17473         (vectorizable_reduction): Likewise.
17474         (vectorizable_induction): Likewise.
17475         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
17476         cost vector to pass to vectorizable_ and record afterwards.
17477         (vect_model_reduction_cost): Take cost vector argument and adjust.
17478         (vect_model_induction_cost): Likewise.
17479         (vectorizable_reduction): Likewise.
17480         (vectorizable_induction): Likewise.
17481         (vectorizable_live_operation): Likewise.
17482         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
17483         SLP_TREE_NUMBER_OF_VEC_STMTS.
17484         (vect_analyze_slp_cost_1): Remove.
17485         (vect_analyze_slp_cost): Likewise.
17486         (vect_slp_analyze_node_operations): Take visited args and
17487         a target cost vector.  Avoid processing already visited stmt sets.
17488         (vect_slp_analyze_operations): Use a local cost vector to gather
17489         costs and register those of non-discarded instances.
17490         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
17491         (vect_schedule_slp_instance): Remove copying of
17492         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
17493         zero.
17494         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
17495         adding cost.  Record cost entry location.
17496         (vect_prologue_cost_for_slp_op): Function to compute cost of
17497         a constant or invariant generated for SLP vect in the prologue,
17498         split out from vect_analyze_slp_cost_1.
17499         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
17500         (vect_model_promotion_demotion_cost): Likewise.
17501         (vect_model_store_cost): Likewise, make static.
17502         (vect_model_load_cost): Likewise.
17503         (vectorizable_bswap): Add cost vector arg and adjust.
17504         (vectorizable_call): Likewise.
17505         (vectorizable_simd_clone_call): Likewise.
17506         (vectorizable_conversion): Likewise.
17507         (vectorizable_assignment): Likewise.
17508         (vectorizable_shift): Likewise.
17509         (vectorizable_operation): Likewise.
17510         (vectorizable_store): Likewise.
17511         (vectorizable_load): Likewise.
17512         (vectorizable_condition): Likewise.
17513         (vectorizable_comparison): Likewise.
17514         (can_vectorize_live_stmts): Likewise.
17515         (vect_analyze_stmt): Likewise.
17516         (vect_transform_stmt): Adjust calls to vectorizable_*.
17517         * tree-vectorizer.c: Include gimple-pretty-print.h.
17518         (dump_stmt_cost): New function.
17520 2018-05-16  Richard Biener  <rguenther@suse.de>
17522         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
17523         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
17524         * tree-ssa-dse.c: Include tree-ssa-loop.h.
17525         (check_name): New callback.
17526         (dse_classify_store): Track cycles via a visited bitmap of PHI
17527         defs and simplify handling of in-loop and across loop dead stores
17528         and properly fail for loop-variant refs.  Handle byte-tracking with
17529         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
17530         limiting the walk.
17532 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
17534         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
17535         (vect_get_mask_type_for_stmt): Likewise.
17536         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
17537         split out from...
17538         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
17539         to determine the statement's vector type and the vector type that
17540         should be used for calculating nunits.  Deal with cases in which
17541         the type has to be deferred.
17542         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
17543         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
17544         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
17545         (vect_determine_vf_for_stmt): New functions, split out from...
17546         (vect_determine_vectorization_factor): ...here.
17547         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
17548         (vect_get_mask_type_for_stmt): New functions, split out from
17549         vect_determine_vectorization_factor.
17551 2018-05-16  Richard Biener  <rguenther@suse.de>
17553         * tree-cfg.c (verify_gimple_assign_ternary): Properly
17554         verify the [VEC_]COND_EXPR embedded comparison.
17556 2018-05-15  Martin Sebor  <msebor@redhat.com>
17558         PR tree-optimization/85753
17559         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
17560         RECORD_TYPE in addition to ARRAY_TYPE.
17562 2018-05-15  Martin Sebor  <msebor@redhat.com>
17564         PR middle-end/85643
17565         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17567 2018-05-15  Richard Biener  <rguenther@suse.de>
17569         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
17570         add by_clobber_p one.  Change algorithm to collect all defs
17571         representing uses we need to walk and try reducing them to
17572         a single one before failing.
17573         (dse_dom_walker::dse_optimize_stmt): Adjust.
17575 2018-05-13  Mark Wielaard  <mark@klomp.org>
17577         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
17578         (size_of_loc_descr): Likewise.
17579         (output_loc_operands): Likewise.
17580         (output_loc_operands_raw): Likewise.
17581         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
17582         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17583         (hash_loc_operands): Likewise.
17584         (compare_loc_operands): Likewise.
17586 2018-05-14  Mark Wielaard  <mark@klomp.org>
17588         * dwarf2out.c (count_index_addrs): New function.
17589         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
17591 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17593         PR tree-optimization/83648
17594         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
17595         return value as malloc candidate.
17597 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17599         PR ipa/85734
17600         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
17601         param as true in call to suggest_attribute.
17603 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
17605         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
17606         -mreadonly-in-sdata.
17608 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17610         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
17611         New pattern.
17612         (aarch64_crypto_aesd_fused): Likewise.
17614 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
17616         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17617         (movsi_aarch64): Likewise.
17618         (load_pairsi): Likewise.
17619         (load_pairdi): Likewise.
17620         (store_pairsi): Likewise.
17621         (store_pairdi): Likewise.
17622         (load_pairsf): Likewise.
17623         (load_pairdf): Likewise.
17624         (store_pairsf): Likewise.
17625         (store_pairdf): Likewise.
17626         (zero_extend): Likewise.
17627         (trunc): Swap alternatives.
17628         (fcvt_target): Add '?' to prefer w over r.
17630 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
17632         PR target/85756
17633         * config/i386/i386.md: Disallow non-commutative arithmetics in
17634         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17635         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
17636         in the peephole2 before it.
17638 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
17640         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17641         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17642         (ix86_handle_option): Handle -mcldemote.
17643         * config.gcc: New header.
17644         * config/i386/cldemoteintrin.h: New file.
17645         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17646         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17647         -mcldemote.
17648         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17649         OPTION_MASK_ISA_CLDEMOTE.
17650         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17651         (ix86_valid_target_attribute_inner_p): Ditto.
17652         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17653         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17654         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17655         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17656         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17657         (cldemote): New.
17658         * config/i386/i386.opt: Add -mcldemote.
17659         * config/i386/x86intrin.h: New header.
17660         * doc/invoke.texi: Add -mcldemote.
17662 2018-05-14  Richard Biener  <rguenther@suse.de>
17664         * doc/match-and-simplify.texi: Adjust :s documentation.
17666 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
17668         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17669         intended memcpy size.
17670         (REORDER_45): Likewise.
17672 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
17674         * sort.cc: New file.
17675         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17676         * vec.c (qsort_chk): Use gcc_qsort.
17677         * Makefile.in (OBJS-libcommon): Add sort.o.
17678         (build/sort.o): New target.  Use it...
17679         (BUILD_RTL): ... here, and...
17680         (build/gencfn-macros): ... here, and...
17681         (build/genmatch): ... here.
17683 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17684             Chung-Ju Wu  <jasonwucj@gmail.com>
17686         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17687         * config/nds32/nds32-graywolf.md: New file.
17688         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17689         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17690         pipeline.
17691         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17692         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17693         * config/nds32/nds32.md (pipeline_model): Add graywolf.
17694         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17695         * config/nds32/pipelines.md: Include n15 settings.
17697 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17698             Chung-Ju Wu  <jasonwucj@gmail.com>
17700         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17701         * config/nds32/nds32-n13.md: New file.
17702         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17703         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17704         pipeline.
17705         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17706         * config/nds32/nds32.md (pipeline_model): Add n13.
17707         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17708         * config/nds32/pipelines.md: Include n13 settings.
17710 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17711             Chung-Ju Wu  <jasonwucj@gmail.com>
17713         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17714         * config/nds32/nds32-n10.md: New file.
17715         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17716         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17717         pipeline.
17718         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17719         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17720         * config/nds32/nds32.md (pipeline_model): Add n10.
17721         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17722         * config/nds32/pipelines.md: Include n10 settings.
17724 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
17725             Kito Cheng  <kito.cheng@gmail.com>
17726             Chung-Ju Wu  <jasonwucj@gmail.com>
17728         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17729         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17730         Add enum values for DSP extension instructions.
17731         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17732         New constraints.
17733         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17734         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17735         New code iterators.
17736         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17737         * config/nds32/nds32-dspext.md: New file for DSP implementation.
17738         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17739         * config/nds32/nds32-intrinsic.md: Likewise.
17740         * config/nds32/nds32_intrinsic.h: Likewise.
17741         * config/nds32/nds32-md-auxiliary.c: Likewise.
17742         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17743         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17744         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17745         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17746         * config/nds32/nds32-protos.h: New declarations for DSP extension.
17747         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17748         TYPE_DMAC in switch statement.
17749         * config/nds32/nds32.c: New checking and implementation for DSP
17750         extension instructions.
17751         * config/nds32/nds32.h: Likewise.
17752         * config/nds32/nds32.md: Likewise.
17753         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17754         * config/nds32/predicates.md: Implement new predicates for DSP
17755         extension.
17757 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
17759         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17760         Reformat alternatives and attributes so it is easier to identify
17761         which constraints/attributes go with which instruction.
17762         (mov<mode>_hardfloat32, FMOVE64): Likewise.
17763         (mov<mode>_softfloat32, FMOVE64): Likewise.
17764         (mov<mode>_hardfloat64, FMOVE64): Likewise.
17765         (mov<mode>_softfloat64, FMOVE64): Likewise.
17767 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17769         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17770         subsection.
17771         (Basic PowerPC Built-in Functions): The new name of the
17772         subsection previously known as "PowerPC Built-in Functions".
17773         (Basic PowerPC Built-in Functions Available on all Configurations):
17774         New subsubsection.
17775         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17776         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17777         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17778         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17780 2018-05-11  Martin Jambor  <mjambor@suse.cz>
17782         PR ipa/85655
17783         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17784         single const.
17786 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17788         PR target/85733
17789         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17791 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
17793         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17794         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17795         (ix86_handle_option): Handle -mwaitpkg.
17796         * config.gcc: New header.
17797         * config/i386/cpuid.h (bit_WAITPKG): New bit.
17798         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17799         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17800         function type.
17801         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17802         OPTION_MASK_ISA_WAITPKG.
17803         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17804         (ix86_option_override_internal): Add PTA_WAITPKG.
17805         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17806         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17807         IX86_BUILTIN_TPAUSE.
17808         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17809         __builtin_ia32_umwait and __builtin_ia32_tpause.
17810         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17811         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17812         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17813         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17814         UNSPECV_TPAUSE): New.
17815         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17816         * config/i386/i386.opt: Add -mwaitpkg.
17817         * config/i386/waitpkgintrin.h: New file.
17818         * config/i386/x86intrin.h: New header.
17819         * doc/invoke.texi: Add -mwaitpkg.
17821 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17823         PR target/85606
17824         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17825         equivalent.
17826         (cortex-m0): Use armv6s-m isa.
17827         (cortex-m0plus): Likewise.
17828         (cortex-m1): Likewise.
17829         (cortex-m0.small-multiply): Likewise.
17830         (cortex-m0plus.small-multiply): Likewise.
17831         (cortex-m1.small-multiply): Likewise.
17833 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
17834             Jakub Jelinek  <jakub@redhat.com>
17836         PR tree-optimization/85692
17837         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17838         source permute as well.
17840 2018-05-11  Martin Liska  <mliska@suse.cz>
17842         PR sanitizer/85556
17843         * doc/extend.texi: Document LLVM style format for no_sanitize
17844         attribute.
17846 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17848         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17849         mode_supports_vsx_dform_quad to mode_supports_dq_form.
17850         (mode_supports_vsx_dform_quad): Likewise.
17851         (mode_supports_vmx_dform): Move these functions to be next to the
17852         other mode_supports functions.
17853         (mode_supports_dq_form): Likewise.
17854         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17855         mode_supports_dq_form.
17856         (reg_offset_addressing_ok_p): Likewise.
17857         (offsettable_ok_by_alignment): Likewise.
17858         (rs6000_legitimate_offset_address_p): Likewise.
17859         (legitimate_lo_sum_address_p): Likewise.
17860         (rs6000_legitimize_address): Likewise.
17861         (rs6000_legitimize_reload_address): Likewise.
17862         (rs6000_secondary_reload_inner): Likewise.
17863         (rs6000_preferred_reload_class): Likewise.
17864         (rs6000_output_move_128bit): Likewise.
17866 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17868         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17869         Generate SImode target register for null target.
17870         <case IX86_BUILTIN_XGETBV>: Ditto.
17871         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17872         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17874 2018-05-10  Carl Love  <cel@us.ibm.com>
17876         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17877         dcbtt and dcbtstt if operands[2] is 0.
17879 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17881         PR target/85693
17882         * config/i386/sse.md (usadv64qi): New expander.
17884 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
17886         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17887         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17888         -maltivec=be support.
17889         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17890         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17891         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17892         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17893         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17894         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17895         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17896         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17897         altivec_vsumsws): Adjust.
17898         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17899         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17900         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17901         support.
17902         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17903         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17904         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17905         (altivec_lve<VI_char>x): Delete expand.
17906         (*altivec_lve<VI_char>x_internal): Rename to...
17907         (altivec_lve<VI_char>x): ... this.
17908         (altivec_lvxl_<mode>): Delete expand.
17909         (*altivec_lvxl_<mode>_internal): Rename to ...
17910         (altivec_lvxl_<mode>): ... this.
17911         (altivec_stvxl_<mode>): Delete expand.
17912         (*altivec_stvxl_<mode>_internal): Rename to ...
17913         (altivec_stvxl_<mode>): ... this.
17914         (altivec_stve<VI_char>x): Delete expand.
17915         (*altivec_stve<VI_char>x_internal): Rename to ...
17916         (altivec_stve<VI_char>x): ... this.
17917         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17918         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17919         reduc_plus_scal_<mode>): Adjust.
17920         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17921         comment.
17922         (rs6000_cpu_cpp_builtins): Adjust.
17923         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17924         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17925         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17926         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17927         -maltivec=be support.
17928         (rs6000_split_vec_extract_var): Adjust.
17929         (rs6000_split_v4si_init): Adjust.
17930         (swap_selector_for_mode): Delete.
17931         (altivec_expand_lvx_be, altivec_expand_stvx_be,
17932         altivec_expand_stvex_be): Delete.
17933         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17934         -maltivec=be support.
17935         (rs6000_gimple_fold_builtin): Ditto.
17936         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17937         Adjust.
17938         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17939         (TARGET_DIRECT_MOVE_64BIT): Adjust.
17940         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17941         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17942         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17943         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17944         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17945         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17946         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17947         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17948         anonymous split): Adjust.
17949         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17950         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17952 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
17954         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17955         when --with-gxx-include-dir is also specified.
17956         * configure: Regenerate.
17958 2018-05-09  Jim Wilson  <jimw@sifive.com>
17960         PR target/84797
17961         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17962         * config/riscv/t-withmultilib: New.
17963         * config/riscv/withmultilib.h: New.
17964         * doc/install.texi: Document RISC-V --with-multilib-list support.
17966 2018-05-09  Richard Biener  <rguenther@suse.de>
17968         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17969         vector.
17970         (vect_bb_vectorization_profitable_p): Adjust.  Compute
17971         actual scalar cost using the cost vector and the add_stmt_cost
17972         machinery.
17974 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17976         PR rtl-optimization/85645
17977         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17978         in the REG_CFA_REGISTER note for LR, don't leave it empty.
17980 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17982         PR rtl-optimization/85645
17983         * shrink-wrap.c (spread_components): Return a boolean saying if
17984         anything was changed.
17985         (try_shrink_wrapping_separate): Iterate spread_components until
17986         nothing changes anymore.
17988 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17990         PR rtl-optimization/85645
17991         * regrename.c (build_def_use): Also kill the chains that include the
17992         destination of a REG_CFA_REGISTER note.
17994 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17996         PR rtl-optimization/85645
17997         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17998         insn that has a REG_CFA_REGISTER note.
18000 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
18002         * cfgexpand.c (expand_clobber): New function.
18003         (expand_gimple_stmt_1): Use it.
18004         * tree-vect-stmts.c (vect_clobber_variable): New function,
18005         split out from...
18006         (vectorizable_simd_clone_call): ...here.
18007         (vectorizable_store): Emit a clobber either side of an
18008         IFN_STORE_LANES sequence.
18009         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
18011 2018-05-09  Tom de Vries  <tom@codesourcery.com>
18013         PR target/85626
18014         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
18015         (define_insn "trap_if_false"): Add exit after trap.
18017 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
18019         PR rtl-optimization/85638
18020         * bb-reorder.c: Include common/common-target.h.
18021         (create_forwarder_block): New function extracted from...
18022         (fix_up_crossing_landing_pad): ...here.  Rename into...
18023         (dw2_fix_up_crossing_landing_pad): ...this.
18024         (sjlj_fix_up_crossing_landing_pad): New function.
18025         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
18026         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
18027         from both partitions and exit the loop after one iteration.
18029 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18031         Revert:
18032         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18033         subsection.
18034         (Basic PowerPC Built-in Functions): The new name of the
18035         subsection previously known as "PowerPC Built-in Functions".
18036         (Basic PowerPC Built-in Functions Available on all Configurations):
18037         New subsubsection.
18038         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18039         subsubsection.
18040         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18041         subsubsection.
18042         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18043         subsubsection.
18044         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18045         subsubsection.
18047 2018-05-08  Jim Wilson  <jimw@sifive.com>
18049         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
18050         (LD_EMUL_SUFFIX): New.
18051         (LINK_SPEC): Use it.
18053 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18055         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18056         subsection.
18057         (Basic PowerPC Built-in Functions): The new name of the
18058         subsection previously known as "PowerPC Built-in Functions".
18059         (Basic PowerPC Built-in Functions Available on all Configurations):
18060         New subsubsection.
18061         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18062         subsubsection.
18063         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18064         subsubsection.
18065         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18066         subsubsection.
18067         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18068         subsubsection.
18070 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
18072         PR target/85683
18073         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
18074         after cmpelim optimization.
18076 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
18078         * config.gcc: Support "goldmont".
18079         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
18080         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18081         PROCESSOR_GOLDMONT.
18082         * config/i386/i386.c (m_GOLDMONT): Define.
18083         (processor_target_table): Add "goldmont".
18084         (PTA_GOLDMONT): Define.
18085         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
18086         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
18087         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
18088         (fold_builtin_cpu): Add "goldmont".
18089         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
18090         (ix86_option_override_internal): Add "goldmont".
18091         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
18092         (processor_type): Add PROCESSOR_GOLDMONT.
18093         * config/i386/i386.md: Add CPU "glm".
18094         * config/i386/glm.md: New file.
18095         * config/i386/x86-tune.def: Add m_GOLDMONT.
18096         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
18098 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
18100         PR target/85572
18101         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
18102         E_V4DImode.
18103         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
18104         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
18105         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
18107         PR target/85317
18108         * config/i386/i386.c (ix86_fold_builtin): Handle
18109         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
18111         PR target/85480
18112         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
18113         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
18115 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
18117         PR target/85658
18118         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
18119         (check_arch): Likewise.
18120         (check_fpu): Return the result rather than printing it.
18121         (end arch): Fix operator precedence.
18122         (end cpu): Likewise.
18123         (END): Print the result from check_fpu.
18125 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
18126             Alan Hayward  <alan.hayward@arm.com>
18127             David Sherwood  <david.sherwood@arm.com>
18129         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
18130         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
18131         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
18132         (*fcmuo<mode>_and): New patterns.
18134 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
18136         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
18137         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
18138         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
18139         (cmp_op, sve_imm_con): New code attributes.
18140         (SVE_COND_INT_CMP, imm_con): Delete.
18141         (cmp_op): Remove above unspecs from int attribute.
18142         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
18143         to...
18144         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
18145         comparison-specific unspecs.
18146         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
18147         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
18148         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
18149         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
18150         (*vec_fcm<cmp_op><mode>): Rename to...
18151         (*fcm<cmp_op><mode>): ...this and adjust likewise.
18152         (*vec_fcmuo<mode>): Rename to...
18153         (*fcmuo<mode>): ...this and adjust likewise.
18154         (*pred_fcm<cmp_op><mode>): New pattern.
18155         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
18156         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
18157         functions.
18158         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
18159         and UNORDERED.
18160         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
18161         (aarch64_emit_sve_predicated_cond): New function.
18162         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
18163         (aarch64_emit_unspec_cond_or): Replace with...
18164         (aarch64_emit_sve_or_conds): ...this new function.  Use
18165         aarch64_emit_sve_ptrue_op for the individual comparisons and
18166         aarch64_emit_binop to OR them together.
18167         (aarch64_emit_inverted_unspec_cond): Replace with...
18168         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
18169         aarch64_emit_sve_ptrue_op for the comparison and
18170         aarch64_emit_unop to invert the result.
18171         (aarch64_expand_sve_vec_cmp_float): Update after the above
18172         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
18174 2018-05-07  Nathan Sidwell  <nathan@acm.org>
18176         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
18177         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
18178         (Backwards Compatibility): Likewise.
18180 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18182         PR bootstrap/85681
18183         Revert:
18184         2018-05-07  Luis Machado  <luis.machado@linaro.org>
18186         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18187         <prefetch_dynamic_strides>: New const bool field.
18188         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18189         prefetch_dynamic_strides.
18190         (exynosm1_prefetch_tune): Likewise.
18191         (thunderxt88_prefetch_tune): Likewise.
18192         (thunderx_prefetch_tune): Likewise.
18193         (thunderx2t99_prefetch_tune): Likewise.
18194         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18195         to false.
18196         (aarch64_override_options_internal): Update to set
18197         PARAM_PREFETCH_DYNAMIC_STRIDES.
18198         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18199         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18200         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18201         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18202         prefetch-dynamic-strides setting.
18204         2018-05-07  Luis Machado  <luis.machado@linaro.org>
18206         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18207         <minimum_stride>: New const int field.
18208         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18209         minimum_stride field.
18210         (exynosm1_prefetch_tune): Likewise.
18211         (thunderxt88_prefetch_tune): Likewise.
18212         (thunderx_prefetch_tune): Likewise.
18213         (thunderx2t99_prefetch_tune): Likewise.
18214         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18215         (aarch64_override_options_internal): Update to set
18216         PARAM_PREFETCH_MINIMUM_STRIDE.
18217         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18218         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18219         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18220         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18221         stride is constant and is below the minimum stride threshold.
18223 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18225         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
18226         to 512.
18228 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18230         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18231         <prefetch_dynamic_strides>: New const bool field.
18232         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18233         prefetch_dynamic_strides.
18234         (exynosm1_prefetch_tune): Likewise.
18235         (thunderxt88_prefetch_tune): Likewise.
18236         (thunderx_prefetch_tune): Likewise.
18237         (thunderx2t99_prefetch_tune): Likewise.
18238         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
18239         to false.
18240         (aarch64_override_options_internal): Update to set
18241         PARAM_PREFETCH_DYNAMIC_STRIDES.
18242         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18243         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18244         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18245         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18246         prefetch-dynamic-strides setting.
18248 2018-05-07  Luis Machado  <luis.machado@linaro.org>
18250         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18251         <minimum_stride>: New const int field.
18252         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18253         minimum_stride field.
18254         (exynosm1_prefetch_tune): Likewise.
18255         (thunderxt88_prefetch_tune): Likewise.
18256         (thunderx_prefetch_tune): Likewise.
18257         (thunderx2t99_prefetch_tune): Likewise.
18258         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
18259         (aarch64_override_options_internal): Update to set
18260         PARAM_PREFETCH_MINIMUM_STRIDE.
18261         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18262         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18263         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18264         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18265         stride is constant and is below the minimum stride threshold.
18267 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
18269         PR c++/85659
18270         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
18271         the type is addressable.  Don't force op into register if it has
18272         BLKmode.
18274 2018-05-05  Roland McGrath  <mcgrathr@google.com>
18276         PR other/77609
18277         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
18278         any section for which we don't know a specific type it should have,
18279         regardless of name.  Previously this was done only for the exact
18280         names ".init_array", ".fini_array", and ".preinit_array".
18281         (default_elf_asm_named_section): Add comment about
18282         relationship with default_section_type_flags and SECTION_NOTYPE.
18283         (get_section): Don't consider it a type conflict if one side has
18284         SECTION_NOTYPE and the other doesn't, as long as neither has the
18285         SECTION_BSS et al used in the default_section_type_flags logic.
18287 2018-05-05  Tom de Vries  <tom@codesourcery.com>
18289         PR target/85653
18290         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
18291         (workaround_barsyncs): New function.
18292         (nvptx_reorg): Use workaround_barsyncs.
18293         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
18294         (define_expand "nvptx_membar_cta"): New define_expand.
18295         (define_insn "*nvptx_membar_cta"): New insn.
18297 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
18299         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
18300         To improve optimization opportunities.
18301         * builtin-types.def: The new needed builtin types for the above.
18303 2018-05-04  Richard Biener  <rguenther@suse.de>
18305         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
18306         * gimple-ssa-store-merging.c
18307         (imm_store_chain_info::output_merged_store): Remove redundant create,
18308         release split_store vector contents on failure.
18309         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
18310         scalar stmt vector on cache hit.
18312 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
18314         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
18315         Xilinx FP support.
18316         * config.gcc (powerpc-xilinx-eabi*): Remove.
18317         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
18318         support.
18319         (fusion_addis_mem_combo_load): Ditto.
18320         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
18321         FP support.
18322         (rs6000_cpu_cpp_builtins): Ditto.
18323         * config/rs6000/rs6000-linux.c
18324         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
18325         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
18326         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
18327         support.
18328         (rs6000_setup_reg_addr_masks): Ditto.
18329         (rs6000_init_hard_regno_mode_ok): Ditto.
18330         (rs6000_option_override_internal): Ditto.
18331         (legitimate_lo_sum_address_p): Ditto.
18332         (rs6000_legitimize_address): Ditto.
18333         (rs6000_legitimize_reload_address): Ditto.
18334         (rs6000_legitimate_address_p): Ditto.
18335         (abi_v4_pass_in_fpr): Ditto.
18336         (setup_incoming_varargs): Ditto.
18337         (rs6000_gimplify_va_arg): Ditto.
18338         (rs6000_split_multireg_move): Ditto.
18339         (rs6000_savres_strategy): Ditto.
18340         (rs6000_emit_prologue_components): Ditto.
18341         (rs6000_emit_epilogue_components): Ditto.
18342         (rs6000_emit_prologue): Ditto.
18343         (rs6000_emit_epilogue): Ditto.
18344         (rs6000_elf_file_end): Ditto.
18345         (rs6000_function_value): Ditto.
18346         (rs6000_libcall_value): Ditto.
18347         * config/rs6000/rs6000.h: Ditto.
18348         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
18349         (TARGET_MINMAX): ... this.  New.
18350         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
18351         * config/rs6000/rs6000.md: Remove Xilinx FP support.
18352         (*movsi_internal1_single): Delete.
18353         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
18354         mfpu=, mxilinx-fpu): Delete.
18355         * config/rs6000/singlefp.h: Delete.
18356         * config/rs6000/sysv4.h: Remove Xilinx FP support.
18357         * config/rs6000/t-rs6000: Ditto.
18358         * config/rs6000/t-xilinx: Delete.
18359         * config/rs6000/titan.md: Adjust for fp_type removal.
18360         * config/rs6000/vsx.md: Remove Xilinx FP support.
18361         (VStype_simple): Delete.
18362         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
18363         * config/rs6000/xfpu.h: Delete.
18364         * config/rs6000/xfpu.md: Delete.
18365         * config/rs6000/xilinx.h: Delete.
18366         * config/rs6000/xilinx.opt: Delete.
18367         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
18368         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
18370 2018-05-04  Tom de Vries  <tom@codesourcery.com>
18372         PR libgomp/85639
18373         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
18374         if ignore == 0.
18376 2018-05-04  Richard Biener  <rguenther@suse.de>
18378         PR middle-end/85627
18379         * tree-complex.c (update_complex_assignment): We are always in SSA form.
18380         (expand_complex_div_wide): Likewise.
18381         (expand_complex_operations_1): Likewise.
18382         (expand_complex_libcall): Preserve EH info of the original stmt.
18383         (tree_lower_complex): Handle removed blocks.
18384         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
18385         on complex multiplication and division libcall builtins.
18387 2018-05-04  Richard Biener  <rguenther@suse.de>
18389         PR middle-end/85574
18390         * fold-const.c (negate_expr_p): Restrict negation of operand
18391         zero of a division to when we know that can happen without
18392         overflow.
18393         (fold_negate_expr_1): Likewise.
18395 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
18397         PR libstdc++/85466
18398         * real.h (real_nextafter): Declare.
18399         * real.c (real_nextafter): New function.
18400         * fold-const-call.c (fold_const_nextafter): New function.
18401         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
18402         CASE_CFN_NEXTTOWARD.
18403         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
18404         even when arg1_mode is different from arg0_mode.
18406 2018-05-03  Nathan Sidwell  <nathan@acm.org>
18408         * doc/extend.texi (Deprecated Features): Remove
18409         -ffriend-injection.
18410         (Backwards Compatibility): Likewise.
18411         * doc/invoke.texi (C++ Language Options): Likewise.
18412         (C++ Dialect Options): Likewise.
18414 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
18416         PR target/85530
18417         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
18418         _mm512_mask_mullox_epi64): New intrinsics.
18420 2018-05-03  Tom de Vries  <tom@codesourcery.com>
18422         PR testsuite/85106
18423         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18424         dump files): Add offload-tree.
18426 2018-05-03  Richard Biener  <rguenther@suse.de>
18428         PR tree-optimization/85615
18429         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
18430         to loops not nested in BBs loop father to avoid creating multi-entry
18431         loops.
18433 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18435         PR tree-optimization/70291
18436         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
18437         arguments.  Change return type to tree.  Emit libcall as a new
18438         statement rather than replacing existing one when inplace_p is true.
18439         (expand_complex_multiplication_components): New function.
18440         (expand_complex_multiplication): Expand floating-point complex
18441         multiplication using the above.
18442         (expand_complex_division): Rename inner_type parameter to type.
18443         Update expand_complex_libcall call-site.
18444         (expand_complex_operations_1): Update expand_complex_multiplication
18445         and expand_complex_division call-sites.
18447 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18449         PR target/85582
18450         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18451         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18452         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
18453         the highest significant bit of the shift count mask is clear.  In
18454         check whether and[sq]i3 is needed verify that all significant bits
18455         of the shift count other than the highest are set.
18457 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18459         PR libgomp/82428
18460         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
18461         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
18462         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
18463         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
18464         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
18465         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
18466         __builtin_goacc_parlevel_size.
18468 2018-05-02  Richard Biener  <rguenther@suse.de>
18470         PR tree-optimization/85597
18471         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
18472         do not use split vect_get_vec_defs call but call vect_get_slp_defs
18473         directly.
18475 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18477         PR testsuite/85106
18478         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
18479         dump files): Add ltrans-tree.
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 wpa-ipa.
18487 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
18489         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
18490         powerpc*-*-linux*paired* target.
18491         * config/rs6000/750cl.h: Delete.
18492         * config/rs6000/paired.h: Delete.
18493         * config/rs6000/paired.md: Delete.
18494         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
18495         float support.
18496         * config/rs6000/rs6000-builtin.def: Remove paired float support.
18497         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
18498         comment.  Remove paired float support.
18499         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
18500         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
18501         VECTOR_PAIRED.
18502         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
18503         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
18504         declarations.
18505         * config/rs6000/rs6000.c: Remove paired float support.
18506         (paired_expand_vector_init, paired_expand_vector_move,
18507         paired_emit_vector_compare, paired_emit_vector_cond_expr,
18508         (paired_expand_lv_builtin, paired_expand_stv_builtin,
18509         paired_expand_builtin, paired_expand_predicate_builtin,
18510         paired_init_builtins): Delete.
18511         * config/rs6000/rs6000.h: Remove paired float support.
18512         * config/rs6000/rs6000.md: Remove paired float support.
18513         (move_from_CR_ov_bit): Delete.
18514         * config/rs6000/rs6000.opt (mpaired): Delete.
18515         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
18516         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
18518 2018-05-02  Richard Biener  <rguenther@suse.de>
18520         PR middle-end/85567
18521         * gimplify.c (gimplify_save_expr): When in SSA form allow
18522         SAVE_EXPRs to compute to SSA vars.
18524 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
18526         PR target/85582
18527         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
18528         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
18529         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
18530         clobber operands[2], instead use a new pseudo.  Formatting fixes.
18532 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
18534         PR tree-optimization/85586
18535         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
18536         exit early for statements in the same group if the accesses are
18537         not strided.
18539 2018-05-02  Tom de Vries  <tom@codesourcery.com>
18541         PR lto/85451
18542         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
18543         error message.
18545 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
18547         PR tree-optimization/85143
18548         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
18550 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18552         PR lto/85451
18553         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
18554         not found" error message.
18556 2018-05-01  Tom de Vries  <tom@codesourcery.com>
18558         PR other/83786
18559         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
18560         * vec.c (test_ordered_remove_if): New function.
18561         (vec_c_tests): Call test_ordered_remove_if.
18562         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
18563         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
18564         * tree-vect-patterns.c (vect_pattern_recog_1): Use
18565         VEC_ORDERED_REMOVE_IF.
18567 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18569         PR tree-optimization/82665
18570         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
18571         pointer subtraction where arguments come from a memchr call.
18573 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
18575         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
18576         --push-state --as-needed and --pop-state instead of --as-needed and
18577         --no-as-needed if ld supports it.
18578         * configure: Regenerated.
18580         PR web/85578
18581         * doc/install.texi2html: Replace _002d with - and _002a with * in
18582         generated html files using sed.
18584 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18586         PR c++/85523
18587         * gcc-rich-location.c (blank_line_before_p): New function.
18588         (use_new_line): New function.
18589         (gcc_rich_location::add_fixit_insert_formatted): New function.
18590         * gcc-rich-location.h
18591         (gcc_rich_location::add_fixit_insert_formatted): New function.
18593 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18595         * selftest.c (assert_streq): Rename "expected" and "actual" to
18596         "val1" and "val2".  Extend NULL-handling to cover both inputs
18597         symmetrically, while still requiring both to be non-NULL for a pass.
18598         * selftest.h (assert_streq): Rename "expected" and "actual" to
18599         "val1" and "val2".
18600         (ASSERT_EQ): Likewise.
18601         (ASSERT_EQ_AT): Likewise.
18602         (ASSERT_KNOWN_EQ): Likewise.
18603         (ASSERT_KNOWN_EQ_AT): Likewise.
18604         (ASSERT_NE): Likewise.
18605         (ASSERT_MAYBE_NE): Likewise.
18606         (ASSERT_MAYBE_NE_AT): Likewise.
18607         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
18608         the assertion to pass.
18609         (ASSERT_STREQ_AT): Likewise.
18611 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18613         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
18614         interaction with -pie.
18616 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18618         * selftest.h: Fix alphabetization of per-source-file selftest
18619         declarations.
18621 2018-04-30  Jason Merrill  <jason@redhat.com>
18623         PR c++/61982 - dead stores to destroyed objects.
18624         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18625         of clobber.
18627 2018-04-30  Jason Merrill  <jason@redhat.com>
18629         * tree.c (build_clobber): New.
18630         * tree.h: Declare it.
18631         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18633 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18635         * diagnostic-show-locus.c (layout::layout): Update for
18636         location_get_source_line returning a char_span.
18637         (struct char_span): Move to input.h.
18638         (struct correction): Update for fields in char_span becoming
18639         private.
18640         (struct source_line): Update for location_get_source_line
18641         returning a char_span.
18642         (layout::print_line): Likewise.
18643         * edit-context.c (edited_file::print_content): Likewise.
18644         (edited_file::print_diff_hunk): Likewise.
18645         (edited_file::print_run_of_changed_lines): Likewise.
18646         (edited_file::get_num_lines): Likewise.
18647         (edited_line::edited_line): Likewise.
18648         * final.c (asm_show_source): Likewise.
18649         * input.c (location_get_source_line): Convert return type
18650         from const char * to char_span, losing the final "line_len"
18651         param.
18652         (dump_location_info): Update for the above.
18653         (get_substring_ranges_for_loc): Likewise.  Use a char_span
18654         when handling the literal within the line.
18655         (test_reading_source_line): Update for location_get_source_line
18656         returning a char_span.
18657         * input.h (class char_span): Move here from
18658         diagnostic-show-locus.c, converting from a struct to a class.
18659         Make data members private.
18660         (char_span::operator bool): New.
18661         (char_span::length): New.
18662         (char_span::get_buffer): New.
18663         (char_span::operator[]): New.
18664         (char_span::subspan): Make const.
18665         (char_span::xstrdup): New.
18666         (location_get_source_line): Convert return type from const char *
18667         to char_span, losing the final "line_size" param.
18669 2018-04-30  Jan Hubicka  <jh@suse.cz>
18671         * lto-wrapper.c (ltrans_priorities): New static var.
18672         (cmp_priority): New.
18673         (run_gcc): Read priorities and if doing parallel build order
18674         the Makefile by them.
18676 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18678         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18680 2018-04-30  Richard Biener  <rguenther@suse.de>
18682         * tree-cfg.c (verify_address): Remove base argument, add
18683         flag whether to check TREE_ADDRESSABLE and do that.
18684         (verify_expr): Remove.
18685         (verify_types_in_gimple_reference): Add pieces from verify_expr.
18686         (verify_gimple_assign_single): Likewise.
18687         (verify_gimple_switch): Likewise.
18688         (verify_expr_location_1): Dereference tp once.  Add (disabled)
18689         piece from verify_expr.
18690         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18692 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18694         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18696 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18698         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18699         (small_data_pattern): Likewise.
18700         (arc_rewrite_small_data): Likewise.
18701         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18702         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18703         (get_symbol_alignment): New function.
18704         (legitimate_small_data_address_p): Likewise.
18705         (legitimate_scaled_address): Update, call
18706         legitimate_small_data_address_p.
18707         (output_sdata): New static variable.
18708         (arc_print_operand): Update how we handle small data operands.
18709         (arc_print_operand_address): Likewise.
18710         (arc_legitimate_address_p): Update, use
18711         legitimate_small_data_address_p.
18712         (arc_rewrite_small_data_p): Remove.
18713         (arc_rewrite_small_data_1): Likewise.
18714         (arc_rewrite_small_data): Likewise.
18715         (small_data_pattern): Likewise.
18716         (compact_sda_memory_operand): Update to use
18717         legitimate_small_data_address_p and get_symbol_alignment.
18718         (prepare_move_operands): Don't rewite sdata pattern.
18719         (prepare_extend_operands): Remove.
18720         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18721         pattern.
18722         (zero_extendqisi2): Likewise.
18723         (zero_extendhisi2): Likewise.
18724         (extendqihi2): Likewise.
18725         (extendqisi2): Likewise.
18726         (extendhisi2): Likewise.
18727         (addsi3): Likewise.
18728         (subsi3): Likewise.
18729         (andsi3): Likewise.
18730         * config/arc/constraints.md (Usd): Change it to memory constraint.
18732 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18734         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18735         as source of std instructions.
18736         * config/arc/arc.md (movsi_insn): Update pattern predicate to
18737         allow 6-bit constants as source for store instructions.
18738         (movdi_insn): Update instruction pattern to allow 6-bit constants
18739         as source for store instructions.
18741 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18743         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18745 2018-04-30  Nathan Sidwell  <nathan@acm.org>
18746             Sandra Loosemore <sandra@codesourcery.com>
18748         * dumpfile.c (dump_open): Allow '-' for stdout.
18749         * doc/invoke.texi (Developer Options): Document dump filename
18750         determination early.  Document stdin/stdout selection.
18752 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
18754         Microblaze Target: PIC data text relative
18756         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18757         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18758         Add declaration.
18759         * config/microblaze/microblaze.h (microblaze_constant_address_p):
18760         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18761         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18762         New addressing mode for data-text relative position indepenedent code.
18763         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18764         'ADDRESS_SYMBOLIC_TXT_REL'.
18765         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18766         (microblaze_legitimate_pic_operand): Exclude function calls from
18767         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18768         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18769         addresses cases.
18770         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18771         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18772         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18773         for 'address + offset'.
18774         (microblaze_expand_prologue): Add new function prologue call for
18775         'r20' assignation.
18776         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18777         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18778         table in case of TARGET_PIC_DATA_TEXT_REL.
18779         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18780         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18781         Add new macros 'UNSPEC_TEXT',
18782         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18783         + exclude function calls from 'UNSPEC_PLT' in case of data text
18784         relative mode.
18785         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18786         new target hook for generating address diff vector tables in case of
18787         flag_pic.
18788         * doc/tm.texi : Regenerate.
18789         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18790         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18791         of addr diff vector generation.
18792         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18793         target hook definition.
18794         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18795         Add default function for generate_pic_addr_diff_vec -> flag_pic.
18796         * doc/invoke.texi (Add new pic option): Add new microblaze pic
18797         option for data text relative.
18799 2018-04-30  Richard Biener  <rguenther@suse.de>
18801         * tree-chrec.h (evolution_function_is_constant_p): Remove
18802         redundant check.
18803         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18805 2018-04-30  Richard Biener  <rguenther@suse.de>
18807         PR bootstrap/85571
18808         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18810 2018-04-30  Richard Biener  <rguenther@suse.de>
18812         PR tree-optimization/28364
18813         PR tree-optimization/85275
18814         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18815         copying first exit test.
18817 2018-04-28  Mark Wielaard  <mark@klomp.org>
18819         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18820         dwarf_version >= 5.
18821         (dwarf_AT): Handle DW_AT_addr_base.
18822         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18824 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
18826         PR target/84431
18827         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18828         (*ashl<dwi>3_doubleword_mask_1): Ditto.
18829         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18830         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18832 2018-04-28  Richard Biener  <rguenther@suse.de>
18834         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18835         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18836         to reflect use.  Only add interesting stmts.
18838 2018-04-27  Martin Jambor  <mjambor@suse.cz>
18840         PR ipa/85549
18841         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18842         the jump function allows for passing through aggregate values.
18844 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
18846         * input.h (in_system_header_at): Convert from macro to inline
18847         function.
18848         (from_macro_expansion_at): Likewise.
18849         (from_macro_definition_at): Likewise.
18851 2018-04-27  Jeff Law  <law@redhat.com>
18853         * config.gcc: Mark tile* targets as deprecated/obsolete.
18855 2018-04-27  Richard Biener  <rguenther@suse.de>
18857         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18858         fix for ILP32.
18860 2018-04-27  Richard Biener  <rguenther@suse.de>
18862         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18864 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
18866         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18867         with Yd constraint. Set "preferred_for_speed" attribute from
18868         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18869         with Yd constraint.
18870         (*movdi_internal): Ditto.
18871         (movti_interunit splitters): Remove
18872         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18873         (movdi_interunit splitters): Ditto.
18874         * config/i386/constraints.md (Ye): Remove.
18875         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18877 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18879         PR target/85512
18880         * config/aarch64/constraints.md (Usg): Limit to 31.
18881         (Usj): Limit to 63.
18883 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
18885         PR tree-optimization/85529
18886         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18887         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18888         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18889         zero extension or masking of the MSB bit.
18890         (optimize_range_tests): Add FIRST_BB argument, pass it through
18891         to optimize_range_tests_var_bound.
18892         (maybe_optimize_range_tests, reassociate_bb): Adjust
18893         optimize_range_tests callers.
18895 2018-04-26  Richard Biener  <rguenther@suse.de>
18896             Jakub Jelinek  <jakub@redhat.com>
18898         * cgraph.h (symbol_table): Just declare debug method here.
18899         * symtab.c (symbol_table::debug): Define.
18901 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
18903         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18905 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
18907         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18908         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18909         (*movdi_internal): Substitute Yi and Yj constraint with x
18910         and Ym and Yn constraint with y constraint.  Update "isa"
18911         attribute and set "preferred_for_speed" attribute from
18912         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18913         (*movsi_internal): Ditto.
18914         (*movdf_internal): Ditto.
18915         (*movsf_internal): Ditto.
18916         (*zero_extendsidi2): Ditto.
18917         * config/i386/sse.md (vec_set<mode>_0): Ditto.
18918         (sse2_loadld): Ditto.
18919         (*vec_extract<ssevecmodelower>_0): Ditto.
18920         (*vec_extractv4si_0_zext_sse4): Ditto.
18921         (vec_concatv2di): Ditto.
18922         (*vec_dup<mode>): Ditto.
18923         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18924         * config/i386/constraints.md (Yi): Remove.
18925         (Yj): Remove.
18926         (Ym): Remove.
18927         (Yn): Remove.
18929 2018-04-26  Nathan Sidwell  <nathan@acm.org>
18931         * dumpfile.c (dump_open): New.
18932         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18933         (dump_finish): Detect stdio/stderr by value not name.
18935 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
18937         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18939 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18941         PR target/84952
18942         * config/nvptx/nvptx.c (verify_neutering_jumps)
18943         (verify_neutering_labels): New function
18944         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18946 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18948         PR target/84025
18949         * config/nvptx/nvptx.c (needs_neutering_p): New function.
18950         (nvptx_single): Use needs_neutering_p to skip over insns that do not
18951         need neutering.
18953 2018-04-26  Richard Biener <rguenther@suse.de>
18954             Tom de Vries  <tom@codesourcery.com>
18956         PR lto/85422
18957         * lto-streamer-out.c (output_function): Fixup loops if required to match
18958         discovery done in the reader.
18960 2018-04-26  Richard Biener  <rguenther@suse.de>
18962         PR tree-optimization/85116
18963         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18964         have a loop exit from the single latch predecessor.  Remove
18965         case of header with just condition.
18966         (ch_base::copy_headers): Exclude infinite loops from any
18967         processing.
18968         (pass_ch::execute): Record exits.
18970 2018-04-26  Richard Biener  <rguenther@suse.de>
18972         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18973         prologue cost vector and pass it to vect_get_load_cost.
18974         (vect_get_peeling_costs_all_drs): Likewise.
18975         (vect_peeling_hash_get_lowest_cost): Likewise.
18976         (vect_enhance_data_refs_alignment): Likewise.
18978 2018-04-26  Richard Biener  <rguenther@suse.de>
18980         PR middle-end/85450
18981         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18982         checking of integer<->pointer conversions.
18983         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18984         sign-/zero-extending pointer types.
18985         (expand_omp_for_static_chunk): Likewise.
18987 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
18988             Jean Lee  <xiaoyur347@gmail.com>
18990         * config/mips/mips.c (mips_asan_shadow_offset): New function.
18991         (TARGET_ASAN_SHADOW_OFFSET): Define.
18992         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18993         true for -fsanitize=address.
18995 2018-04-25  Mark Wielaard  <mark@klomp.org>
18997         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18998         shorter ones.
19000 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
19002         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
19003         than "alu", remove explicit "memory" and "imm_disp" attributes.
19004         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
19006         PR middle-end/85414
19007         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
19008         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
19009         gen_lowpart_no_emit.
19011 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
19013         PR target/85473
19014         * config/i386/i386.c (ix86_expand_builtin): Change memory
19015         operand to XI, extend p0 to Pmode.
19016         * config/i386/i386.md: Change unspec volatile and operand
19017         1 mode to XI, change operand 0 mode to P.
19019 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19021         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
19022         GET_MODE_MASK before any checking.
19023         (nds32_can_use_bset_p): Likewise.
19024         (nds32_can_use_btgl_p): Likewise.
19026 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19028         * config/nds32/nds32-doubleword.md: New define_split pattern for
19029         illegal register number.
19031 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19033         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
19035 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19037         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
19039 2018-04-25  Richard Biener  <rguenther@suse.de>
19041         * lto-streamer.h (LTO_major_version): Bump to 8.
19043 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
19045         * BASE-VER: Set to 9.0.0.
19047 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
19049         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
19050         in __abskf2 and __powikf2.
19052 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19054         PR target/85512
19055         * config/aarch64/constraints.md (Usg, Usj): New constraints.
19056         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
19057         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
19058         Use the above on operand 2.  Reindent.
19059         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
19061 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
19063         PR target/85485
19064         * common/config/i386/i386-common.c (ix86_handle_option): Don't
19065         handle OPT_mcet.
19066         * config/i386/i386.opt (mcet): Removed.
19067         * doc/install.texi: Remove -mcet documentation.
19068         * doc/invoke.texi: Likewise.
19070 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
19072         PR target/85485
19073         * doc/install.texi: Remove -mcet from bootstrap-cet.
19075 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
19077         PR target/85511
19078         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
19079         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
19080         if TARGET_64BIT.
19082         PR target/85503
19083         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
19084         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
19085         containing a CONST_VECTOR.
19087 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
19089         * doc/install.texi: Update newlib dependency for nvptx.
19091 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
19093         PR target/85508
19094         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
19095         instead of INTVAL when shifting x left.
19097 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
19099         PR tree-optimization/85478
19100         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
19101         vect_grouped_store_supported for single element vectors.
19103 2018-04-24  Richard Biener  <rguenther@suse.de>
19105         PR target/85491
19106         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
19107         load cost increase to the case of non-constant step.
19109 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
19111         PR target/84828
19112         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
19113         destination if any_malformed_asm.
19115 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
19117         PR middle-end/85496
19118         * expr.c (store_field): In the bitfield case, if the value comes from
19119         a function call and is returned in registers by means of a PARALLEL,
19120         do not change the mode of the temporary unless BLKmode and VOIDmode.
19122 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
19124         PR rtl-optimization/85423
19125         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
19126         dependencies to debug insns when the previous insn is non-debug.
19128 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
19130         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
19131         enums into a single definition.
19132         (fls): Fix predicates and printing.
19133         (seti): Likewise.
19135 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
19137         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
19138         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
19139         and short u6 immediate.
19140         (check_if_valid_sleep_operand): Remove.
19141         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
19143 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19145         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
19146         flag_always_save_lp condition.
19147         * config/nds32/nds32.opt (malways-save-lp): New option.
19149 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19151         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
19152         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
19153         * config/nds32/nds32.h
19154         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
19155         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
19157 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19159         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
19160         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
19162 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
19163             Chung-Ju Wu  <jasonwucj@gmail.com>
19165         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
19166         Declare.
19167         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
19168         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
19170 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19172         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
19174 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19176         * config/nds32/nds32-protos.h (nds32_data_alignment,
19177         nds32_local_alignment): Declare.
19178         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
19179         nds32_local_alignment): New functions.
19180         (TARGET_CONSTANT_ALIGNMENT): Define.
19181         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
19183 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19185         * config/nds32/nds32.c
19186         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
19187         (TARGET_MODES_TIEABLE_P): Likewise.
19189 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
19191         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
19192         level Ofast and Og.
19194 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
19195             Chung-Ju Wu  <jasonwucj@gmail.com>
19197         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19198         for unaligned access.
19199         * config/nds32/nds32-intrinsic.c: Implementation of expanding
19200         unaligned access.
19201         * config/nds32/nds32-intrinsic.md: Likewise.
19202         * config/nds32/nds32_intrinsic.h: Likewise.
19203         * config/nds32/nds32.h (nds32_builtins): Likewise.
19204         * config/nds32/nds32.opt (munaligned-access): New option.
19205         * config/nds32/nds32.c (nds32_asm_file_start): Display
19206         flag_unaligned_access status.
19208 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
19210         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
19211         -mno-relax is present.
19212         * config/riscv/linux.h (LINK_SPEC): Ditto.
19214 2018-04-20  Martin Sebor  <msebor@redhat.com>
19216         PR c/85365
19217         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
19218         for null pointers.
19219         (gimple_fold_builtin_stxcpy_chk): Same.
19220         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
19222 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
19224         PR target/85456
19225         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
19226         __powikf2 when long double is IEEE 128-bit.
19228 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
19230         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
19231         step to make sure stack always aligned.
19233 2018-04-20  Carl Love  <cel@us.ibm.com>
19235         PR target/83402
19236         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
19237         size check for arg0.
19239 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
19240             Tom de Vries  <tom@codesourcery.com>
19242         PR target/85445
19243         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
19244         Emit insns for calls too.
19245         (nvptx_find_par): Always look for worker-level predecessor insn.
19246         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
19247         calls.
19248         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
19249         (nvptx_process_pars): Propagate frames for calls.
19251 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
19253         PR target/85469
19254         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
19255         Removed.
19256         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
19257         (ix86_handle_option): Don't handle OPT_mibt.
19258         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
19259         __SHSTK__.
19260         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
19261         has_ibt and ibt.
19262         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
19263         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
19264         (ix86_target_macros): Define __CET__ with flag_cf_protection
19265         for -fcf-protection.
19266         * config/i386/i386.c (isa2_opts): Remove -mibt.
19267         * config/i386/i386.h (TARGET_IBT): Removed.
19268         (TARGET_IBT_P): Likewise.
19269         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
19270         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
19271         * config/i386/i386.opt (mcet): Update help message.
19272         (mshstk): Likewise.
19273         (mibt): Removed.
19274         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
19275         -mcet as an alias for -mshstk.
19277 2018-04-20  Richard Biener <rguenther@suse.de>
19279         PR middle-end/85475
19280         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
19281         complexity by forcing a single use of the multiply operand.
19283 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19285         ipa/85449
19286         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
19287         recursion dependency to only apply to non-clones.
19289 2018-04-20  Martin Jambor  <mjambor@suse.cz>
19291         ipa/85447
19292         * ipa-cp.c (create_specialized_node): Check that clones of
19293         self-recursive edges exist during IPA-CP.
19295 2018-04-19  Toon Moene  <toon@moene.org>
19297         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
19298         by -O3.
19300 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19302         PR tree-optimization/85467
19303         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
19304         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
19305         VECTOR_CST element to type.
19307 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19309         PR target/85397
19310         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
19311         * config/i386/i386.md (builtin_setjmp_setup): Removed.
19312         (builtin_longjmp): Likewise.
19313         (save_stack_nonlocal): New pattern.
19314         (restore_stack_nonlocal): Likewise.
19316 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19318         PR target/85404
19319         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19320         Replace ASM_OUTPUT_LABEL with fprintf.
19322 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
19324         PR target/85417
19325         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
19326         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
19327         * config/i386/i386-c.c (ix86_target_macros_internal): Also
19328         define __IBT__ and __SHSTK__ for -fcf-protection.
19329         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
19330         TARGET_IBT.
19331         (ix86_trampoline_init): Likewise.
19332         (x86_output_mi_thunk): Likewise.
19333         (ix86_notrack_prefixed_insn_p): Likewise.
19334         (ix86_option_override_internal): Don't disallow -fcf-protection.
19335         * config/i386/i386.md (rdssp<mode>): Also enable for
19336         -fcf-protection.
19337         (incssp<mode>): Likewise.
19338         (nop_endbr): Likewise.
19339         * config/i386/i386.opt (mcet): Change help message to built-in
19340         functions only.
19341         (mibt): Likewise.
19342         (mshstk): Likewise.
19343         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
19344         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
19345         enable CET built-in functions.
19347 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
19349         * common/config/i386/i386-common.c
19350         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
19351         OPTION_MASK_ISA_MOVDIRI_UNSET,
19352         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
19353         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
19354         * config.gcc (movdirintrin.h): New header.
19355         * config/i386/cpuid.h (bit_MOVDIRI,
19356         bit_MOVDIR64B): New bits.
19357         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
19358         and -mmvodir64b.
19359         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
19360         (VOID, PVOID, PCVOID)): New function types.
19361         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
19362         __builtin_ia32_directstoreu_u64,
19363         __builtin_ia32_movdir64b): New builtins.
19364         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
19365         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
19366         and -mmovdiri.
19367         (ix86_valid_target_attribute_inner_p): Ditto.
19368         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
19369         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
19370         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
19371         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
19372         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
19373         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
19374         (movdiri<mode>, movdir64b_<mode>): New.
19375         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
19376         * config/i386/immintrin.h: Include movdirintrin.h.
19377         * config/i386/movdirintrin.h: New file.
19378         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
19380 2018-04-19  Richard Biener  <rguenther@suse.de>
19382         PR middle-end/85455
19383         * cfg.c (clear_bb_flags): When loop state says we have
19384         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
19386 2018-04-19  Richard Biener  <rguenther@suse.de>
19388         PR tree-optimization/84737
19389         * tree-vect-data-refs.c (vect_copy_ref_info): New function
19390         copying restrict info.
19391         (vect_setup_realignment): Use it.
19392         * tree-vectorizer.h (vect_copy_ref_info): Declare.
19393         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
19394         the first DR to all generated stores.
19395         (vectorizable_load): Likewise for loads.
19397 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
19399         PR tree-optimization/85446
19400         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
19401         the integral and pointer types to have the same precision.
19403         * doc/install.texi: Document --disable-cet being the default and
19404         --enable-cet=auto.
19406 2018-04-18  Martin Liska  <mliska@suse.cz>
19408         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
19409         style.
19411 2018-04-18  Martin Liska  <mliska@suse.cz>
19413         Revert
19414         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
19416         PR ipa/83983
19417         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19418         arguments if they are comparable.
19420 2018-04-18  Martin Liska  <mliska@suse.cz>
19422         Revert
19423         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
19425         PR lto/84805
19426         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19427         incomplete types.
19429 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
19431         PR target/85388
19432         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
19433         ENDBR after calling __morestack.
19435 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
19437         PR jit/85384
19438         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
19439         by using gcc_base_ver to generate a gcc_driver_version, and use
19440         it when generating GCC_DRIVER_NAME.
19441         * configure: Regenerate.
19443 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
19445         PR target/81084
19446         * config.gcc: Obsolete powerpc*-*-*spe*.
19448 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19450         PR debug/84637
19451         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
19452         (stabstr_D): Change type of unum from unsigned int to
19453         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
19454         type.
19456 2018-04-17  Jim Wilson  <jimw@sifive.com>
19458         PR 84856
19459         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
19460         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
19461         Set arg_pointer_offset after using pretend_args_size.
19463 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19465         PR rtl-optimization/85431
19466         * dse.c (record_store): Ignore zero width stores.
19468         PR sanitizer/85230
19469         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
19470         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
19471         __builtin_stack_restore rather than after it.
19472         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
19473         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
19474         argument instead of virtual_dynamic_stack_rtx.
19476 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19478         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
19479         New prototype.
19480         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19481         Add note to error message to explain internal mapping of overloaded
19482         built-in function name to non-overloaded built-in function name.
19483         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
19484         function.
19486 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
19488         PR target/85424
19489         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
19490         where the inputs overlap with the output.
19492 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19494         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
19495         (=v, v) alternative and explicit "memory" attribute.
19496         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
19497         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19498         attributes.
19499         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
19500         "sselog1" type instead of "sselog".
19501         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
19502         "sselog".  Remove explicit "memory" attribute.
19503         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
19504         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19505         attributes.
19506         (vec_extract_hi_v32hi): Merge all alternatives into one, use
19507         "sselog1" type instead of "sselog".  Remove explicit "memory"
19508         attribute.
19509         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
19510         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19511         attribute.
19512         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
19513         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
19514         attributes.
19515         (vec_extract_hi_v64qi): Merge all alternatives into one, use
19516         "sselog1" type instead of "sselog".  Remove explicit "memory"
19517         attribute.
19518         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
19519         use "sselog1" type instead of "sselog".  Remove explicit "memory"
19520         attribute.
19522         PR target/85430
19523         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
19525         PR middle-end/85414
19526         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
19527         on a SUBREG.
19529 2018-04-17  Martin Jambor  <mjambor@suse.cz>
19531         PR ipa/85421
19532         * ipa-cp.c (create_specialized_node): Call
19533         expand_all_artificial_thunks if necessary.
19535 2018-04-17  Martin Liska  <mliska@suse.cz>
19537         PR lto/85405
19538         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
19539         in message, remote space in between '_G' and '('.
19541 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
19543         PR target/85281
19544         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
19545         avx512f_vmcmp<mode>3<round_saeonly_name>,
19546         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
19547         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
19548         avx512f_rndscale<mode><round_saeonly_name>,
19549         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
19550         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
19551         Use %<iptr>2 instead of %2 for -masm=intel.
19552         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
19553         avx512f_vcvttss2usi<round_saeonly_name>,
19554         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
19555         -masm=intel.
19556         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
19557         avx512f_vcvttsd2usi<round_saeonly_name>,
19558         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
19559         Use %q1 instead of %1 for -masm=intel.
19560         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
19561         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
19562         of %3 for -masm=intel.
19563         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
19564         -masm=intel.
19565         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
19566         -masm=intel.
19567         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
19568         -masm=intel.
19569         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
19570         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
19571         %g1.
19572         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
19573         -masm=intel.
19574         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
19575         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
19576         %g1 and one with %0 and %1.
19577         (avx512er_vmrcp28<mode><round_saeonly_name>,
19578         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
19579         %1 for -masm=intel.
19580         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
19581         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
19582         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
19583         of %0 and %{%4%} for -masm=intel.
19584         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
19585         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
19586         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
19587         order of %0 and %{%5%}%{z%} for -masm=intel.
19589 2018-04-17  Jan Hubicka  <jh@suse.cz>
19591         PR lto/85405
19592         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
19594 2018-04-17  Martin Liska  <mliska@suse.cz>
19596         PR ipa/85329
19597         * multiple_target.c (create_dispatcher_calls): Set apostrophes
19598         for target_clone error message.  Make default implementation
19599         clone to be a local declaration.
19600         (separate_attrs): Add new argument and check for an empty
19601         string.
19602         (expand_target_clones): Handle it.
19603         (ipa_target_clone): Make redirection just for target_clones
19604         functions.
19606 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
19607             Tom de Vries  <tom@codesourcery.com>
19609         PR middle-end/84955
19610         * omp-expand.c (expand_oacc_for): Add dummy false branch for
19611         tiled basic blocks without omp continue statements.
19613 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
19615         PR target/83660
19616         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19617         vec_extract expression as having side effects to make sure it gets
19618         a cleanup point.
19620 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
19622         PR target/85403
19623         * config/i386/i386.c (get_builtin_code_for_version): Check
19624         error_mark_node.
19626 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
19628         PR target/84331
19629         * config.gcc: Support "skylake".
19630         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19631         PROCESSOR_SKYLAKE.
19632         * config/i386/i386.c (m_SKYLAKE): Define.
19633         (processor_target_table): Add "skylake".
19634         (ix86_option_override_internal): Add "skylake".
19635         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19636         PROCESSOR_CANNONLAKE.
19637         (get_builtin_code_for_version): Fix priority for
19638         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19639         PROCESSOR_SKYLAKE-AVX512.
19640         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19641         (processor_type): Add PROCESSOR_SKYLAKE.
19643 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
19644             Jason Merrill  <jason@redhat.com>
19646         PR c++/85112
19647         * convert.c (convert_to_integer_1): Use direct recursion for
19648         enumeral types and types with a precision less than the number
19649         of bits in their mode.
19651 2018-04-16  Julia Koval  <julia.koval@intel.com>
19653         PR target/84413
19654         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19655         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19657 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
19659         PR target/85293
19660         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19661         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19662         and -mno-direct-move.
19664 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
19666         PR target/83402
19667         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19668         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19669         Ensure negative shifts result in {0}.
19671 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
19673         PR rtl-optimization/79916
19674         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19675         regs (if any) to define how to gnerate SD moves when LRA is in
19676         progress.
19678 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
19680         PR rtl-optimization/85393
19681         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19682         * except.c (expand_dw2_landing_pad_for_region): Make static.
19683         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19684         a label and unconditional jump to old_bb, rather than
19685         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19686         basic block.
19688         PR rtl-optimization/85376
19689         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19690         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19691         instead of a specific value.
19693 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19694             Bin Cheng  <bin.cheng@arm.com>
19696         PR tree-optimization/82965
19697         PR tree-optimization/83991
19698         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19699         by_profile_only parameter.
19700         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19701         information if the loop was predicted to iterate too many times.
19702         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19704 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19706         PR lto/71991
19707         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19708         always inline.
19710 2018-04-13  Martin Liska  <mliska@suse.cz>
19711             Jakub Jelinek  <jakub@redhat.com>
19713         PR middle-end/81657
19714         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19715         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19716         * builtins.c (expand_builtin_memory_copy_args): Use
19717         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19718         handle dest_addr == pc_rtx.
19720 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
19722         PR target/85291
19723         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19724         asked to not generate direct moves.
19725         (fix_trunc<mode>si2_stfiwx): Similar.
19726         (fix_trunc<mode>si2_internal): Similar.
19728 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19730         PR debug/83157
19731         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19732         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19733         lookup if dest in some wider mode is known to be const0_rtx and
19734         if so, record permanent equivalence for it to be ZERO_EXTEND of
19735         the narrower mode destination.
19737 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19739         * lto-streamer-out.c (output_function): Revert 259346.
19740         * omp-expand.c (expand_oacc_for): Likewise.
19742 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
19744         PR rtl-optimization/85354
19745         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19746         * sel-sched.c (sel_global_init): ... here.
19748 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
19750         PR target/85238
19751         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19752         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19753         mode for PE-COFF targets.
19754         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19755         (i386_pe_asm_lto_end): Likewise.
19756         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19757         (TARGET_ASM_LTO_END): Likewise.
19758         * config/i386/winnt.c (saved_debug_info_level): New static variable.
19759         (i386_pe_asm_lto_start): New function.
19760         (i386_pe_asm_lto_end): Likewise.
19762 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19763             Richard Biener  <rguenther@suse.de>
19765         PR middle-end/84955
19766         * lto-streamer-out.c (output_function): Fix CFG loop state before
19767         streaming out.
19768         * omp-expand.c (expand_oacc_for): Handle calls to internal
19769         functions like regular functions.
19771 2018-04-12  Richard Biener  <rguenther@suse.de>
19773         PR lto/85371
19774         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19775         for the early LTO debug to properly generate references to it
19776         during DIE emission.  Do not re-use that for the skeleton for
19777         split-dwarf.
19778         (dwarf2out_early_finish): Likewise.
19780 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19782         PR target/85328
19783         * config/i386/sse.md
19784         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19785         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19786         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19787         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19788         and output is a reg, avoid creating invalid lowpart subreg, but
19789         instead split into a 512-bit move.  Don't split if not AVX512VL,
19790         input is xmm16+ reg and output is a mem.
19791         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19792         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19793         xmm16+ reg and output is a mem.
19795 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19797         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19798         also for flag_dwarf2_cfi_asm.
19800 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19802         PR rtl-optimization/85342
19803         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19804         a bool scalar var inside of the loop instead.  Don't try to update
19805         recog_data.operand after failed apply_change_group.
19807 2018-04-12  Tom de Vries  <tom@codesourcery.com>
19809         PR target/85296
19810         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19811         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
19812         array with flexible array member as array without given dimension.
19813         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19814         argument for undefined param to true.
19816 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
19818         PR target/85321
19819         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19820         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19821         from PowerPC section.
19822         * config/rs6000/sysv4.opt (mcall-): Improve help text.
19823         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19824         help text that is too long.
19825         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19826         help text that is too long.
19827         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19828         help text that is too long.
19830 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
19832         * config/alpha/alpha.md (stack_probe_internal): Rename
19833         from "probe_stack".  Update all callers.
19835 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19837         PR rtl-optimization/84566
19838         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19839         sched_macro_fuse_insns.
19841 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19843         PR target/84301
19844         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19845         (compute_block_dependences): ... from here.
19847 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19849         PR tree-optimization/85331
19850         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19851         from int to HOST_WIDE_INT.
19853 2018-04-11  Martin Jambor  <mjambor@suse.cz>
19855         PR ipa/84149
19856         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19857         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19858         not the same as the source val.
19859         (cgraph_edge_brings_value_p): New parameter.
19860         (gather_edges_for_value): Pass destination value to
19861         cgraph_edge_brings_value_p.
19862         (perhaps_add_new_callers): Likewise.
19863         (get_info_about_necessary_edges): Likewise and exclude values brought
19864         only by self-recursive edges.
19865         (create_specialized_node): Redirect only clones of self-calling edges.
19866         (+self_recursive_pass_through_p): New function.
19867         (find_more_scalar_values_for_callers_subset): Use it.
19868         (find_aggregate_values_for_callers_subset): Likewise.
19869         (known_aggs_to_agg_replacement_list): Removed.
19870         (decide_whether_version_node): Re-calculate known constants for all
19871         remaining context clones.
19873 2018-04-11  Richard Biener  <rguenther@suse.de>
19875         PR lto/85339
19876         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19877         from early DWARF output.
19878         (dwarf2out_early_finish): Output line info unconditionally into
19879         early DWARF and add reference to it.
19881 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19883         PR target/85281
19884         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19885         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19886         other than V2DFmode using iptr mode attribute.
19887         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19889 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19891         PR rtl-optimization/84659
19892         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19894 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19896         PR debug/85302
19897         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19898         SIZEP is NULL.
19899         (output_loc_list): Pass address of a dummy size variable even in the
19900         locview handling loop.
19901         (index_location_lists): Add comment on why skip_loc_list_entry can't
19902         call size_of_locs.
19904 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19906         PR target/85261
19907         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19908         into register.
19910 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
19912         PR target/85321
19913         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19914         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19915         and -mstring-compare-inline-limit.
19917 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19919         PR target/85287
19920         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19921         for stack clash protection in a register whenever we need it to be in
19922         a register.
19924 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19926         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19927         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19929 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19931         PR target/85321
19932         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19933         the help text.
19934         (mlong-double-): Ditto.
19935         * config/rs6000/sysv4.opt (msdata=): Ditto.
19936         (mtls-size=): Ditto.
19938 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19940         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19941         erroneous entries for
19942         "vector int vec_ldl (int, long int *)", and
19943         "vector unsigned int vec_ldl (int, unsigned long int *)".
19944         Add comments and entries for
19945         "vector bool char vec_ldl (int, bool char *)",
19946         "vector bool short vec_ldl (int, bool short *)",
19947         "vector bool int vec_ldl (int, bool int *)",
19948         "vector bool long long vec_ldl (int, bool long long *)",
19949         "vector pixel vec_ldl (int, pixel *)",
19950         "vector long long vec_ldl (int, long long *)",
19951         "vector unsigned long long vec_ldl (int, unsigned long long *)".
19952         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19953         type tree bool_long_long_type_node and correct definition of
19954         bool_V2DI_type_node to make reference to this new type tree.
19955         (rs6000_mangle_type): Replace erroneous reference to
19956         bool_long_type_node with bool_long_long_type_node.
19957         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19958         comments to emphasize sign distinctions for char and int types and
19959         replace RS6000_BTI_bool_long constant with
19960         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
19961         use of RS6000_BTI_pixel.
19962         (bool_long_type_node): Remove this macro definition.
19963         (bool_long_long_type_node): New macro definition
19965 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19967         PR rtl-optimization/85300
19968         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19969         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19970         simplify_unary_operation fails.
19972 2018-04-10  Martin Liska  <mliska@suse.cz>
19974         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19975         cgraph_edge and ipa_ref.
19977 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19979         PR target/85177
19980         PR target/85255
19981         * config/i386/sse.md
19982         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19983         computation of the VEC_MERGE selector from mask.
19984         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19985         Fix decoding of the VEC_MERGE selector into mask.
19987 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19989         PR tree-optimization/85286
19990         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19992 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19994         * final.c (final_1): Set insn_last_address as well as
19995         insn_current_address.
19997 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19999         PR target/85173
20000         * explow.c (emit_stack_probe): Call validize_mem on memory location
20001         before passing it to gen_probe_stack.  Create address operand and
20002         legitimize it for the probe_stack_address case.
20004 2018-04-09  Jan Hubicka  <jh@suse.cz>
20006         PR lto/85078
20007         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
20008         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
20009         * tree.c (free_lang_data_in_type): Fix handling of binfos;
20010         walk basetypes.
20011         (free_lang_data): Rebuild type inheritance graph.
20013 2018-04-09  Martin Sebor  <msebor@redhat.com>
20015         * invoke.texi (-finline-small-functions): Mention other optimization
20016         options.
20017         (-findirect-inlining, -fpartial-inlining): Same.
20018         (-finline-functions-called-once): Same.
20019         (-freorder-blocks-and-partition): Same.
20021 2018-04-09  Jan Hubicka  <jh@suse.cz>
20023         PR rtl/84058
20024         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
20025         jumps; choose last target that matches the criteria (i.e.
20026         no partition changes for non-crossing jumps).
20027         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
20028         support for redirecting crossing jumps to non-crossing.
20030 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
20032         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
20033         also for naked functions.
20035 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
20037         * config/arc/arc.md (add_shift): New pattern.
20038         (add_shift2): Likewise.
20039         (sub_shift): Likewise.
20040         (sub_shift_cmp0_noout): Likewise.
20041         (compare_si_ashiftsi): Likewise.
20042         (xbfu_cmp0_noout): New combine pattern.
20043         (xbfu_cmp0"): Likewise.
20044         (movsi_set_cc_insn): Place the predicable variant first.
20045         (commutative_binary_cmp0_noout): Remove clobber.
20046         (commutative_binary_cmp0): New pattern.
20047         (noncommutative_binary_cmp0): Likewise.
20048         (noncommutative_binary_cmp0_noout): Likewise.
20049         (noncommutative_binary_comparison_result_used): Removed.
20050         (rsub_cmp0): New pattern.
20051         (rsub_cmp0_noout): Likewise.
20052         (extzvsi): Changed, keep only meaningful variants.
20053         (SQH, SEZ): New iterators.
20054         (SQH_postfix): New mode attribute.
20055         (SEZ_prefix): New code attribute.
20056         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
20057         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
20058         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
20059         of numerical value.
20060         (noncommutative_operator): Check the availability of barrel
20061         shifter option.
20063 2018-04-09  Richard Biener  <rguenther@suse.de>
20065         PR tree-optimization/85284
20066         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
20067         Only use the niter constraining form of simple_iv when the exit
20068         is always executed.
20070 2018-04-09  Tom de Vries  <tom@codesourcery.com>
20072         PR target/84041
20073         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20074         (define_expand "*memory_barrier"): New define_expand.
20075         (define_insn "memory_barrier"): New insn.
20077 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20079         PR rtl-optimization/80463
20080         PR rtl-optimization/83972
20081         PR rtl-optimization/83480
20083         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
20084         correct producer for the insn.
20085         (tidy_control_flow): Fixup seqnos in case of debug insns.
20087 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20089         PR rtl-optimization/83913
20091         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
20092         different sched-times when merging exprs.
20094 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20096         PR rtl-optimization/83962
20098         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
20099         tidy_fallthru_edge and tidy_control_flow.
20101 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20103         PR rtl-optimization/83530
20105         * sel-sched.c (force_next_insn): New global variable.
20106         (remove_insn_for_debug): When force_next_insn is true, also leave only
20107         next insn in the ready list.
20108         (sel_sched_region): When the region wasn't scheduled, make another pass
20109         over it with force_next_insn set to 1.
20111 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
20113         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
20114         into tm_file.
20115         * config/nds32/constants.md (unspec_volatile_element): Add enum values
20116         for interrupt control.
20117         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
20118         functions for interrupt control.
20119         * config/nds32/nds32-intrinsic.md: Likewise.
20120         * config/nds32/nds32_intrinsic.h: Likewise.
20121         * config/nds32/nds32.h (nds32_builtins): Likewise.
20123 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
20125         * config/nds32/nds32.c (nds32_init_machine_status,
20126         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
20127         strict_aligned_p field.
20128         (nds32_expand_to_rtl_hook): New function.
20129         (TARGET_EXPAND_TO_RTL_HOOK): Define.
20130         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
20132 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20133             Chung-Ju Wu  <jasonwucj@gmail.com>
20135         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
20136         * config/nds32/nds32-n7.md: New file.
20137         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
20138         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
20139         pipeline.
20140         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
20141         * config/nds32/nds32.md (pipeline_model): Add n7.
20142         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
20143         * config/nds32/pipelines.md: Include n7 settings.
20145 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20146             Chung-Ju Wu  <jasonwucj@gmail.com>
20148         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
20149         * config/nds32/nds32-e8.md: New file.
20150         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
20151         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
20152         pipeline.
20153         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
20154         * config/nds32/nds32.md (pipeline_model): Add e8.
20155         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
20156         * config/nds32/pipelines.md: Include e8 settings.
20158 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20159             Chung-Ju Wu  <jasonwucj@gmail.com>
20161         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
20162         * config/nds32/nds32-n8.md: New file.
20163         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
20164         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
20165         pipeline.
20166         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
20167         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
20168         * config/nds32/nds32.md (pipeline_model): Add n8.
20169         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
20170         * config/nds32/pipelines.md: Include n8 settings.
20172 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
20173             Chung-Ju Wu  <jasonwucj@gmail.com>
20175         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
20176         * config/nds32/nds32-n9-2r1w.md: New file.
20177         * config/nds32/nds32-n9-3r2w.md: New file.
20178         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
20179         nds32_register_ports): New or modify for cpu n9.
20180         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
20181         pipeline.
20182         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
20183         * config/nds32/nds32-utils.c: New file.
20184         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
20185         TARGET_MUL_SLOW): Define.
20186         * config/nds32/nds32.md (pipeline_model): New attribute.
20187         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
20188         New options that support cpu n9.
20189         * config/nds32/pipelines.md: Include n9 settings.
20190         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
20192 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
20194         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
20195         information if necessary.
20196         (output_cond_branch_compare_zero): Likewise.
20197         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
20198         (nds32_target_alignment): Refine for alignment.
20199         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
20200         (FUNCTION_BOUNDARY): Modify.
20201         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
20202         align case.
20203         * config/nds32/nds32.opt (malways-align, malign-functions): New.
20205 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
20207         * config/nds32/constants.md (unspec_volatile_element): Add values for
20208         TLB operation and data prefetch.
20209         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
20210         functions for TLB operation and data prefetch.
20211         * config/nds32/nds32-intrinsic.md: Likewise.
20212         * config/nds32/nds32_intrinsic.h: Likewise.
20213         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
20214         (nds32_print_operand): Likewise.
20215         * config/nds32/nds32.h (nds32_builtins): Likewise.
20217 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
20218         Andrew Pinski <pinsika@gcc.gnu.org>
20220         PR middle-end/82976
20221         * match.pd: Use constant_boolean_node of correct type instead of
20222         boolean_true_node or boolean_false_node for simplifying
20223         pointer comparisons to zero.
20225 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20227         PR tree-optimization/80021
20228         * tree.c (verify_type_variant): Make error call in verify_variant_match
20229         translatable and remove final full stop.
20231 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20233         * config/nds32/constants.md (unspec_volatile_element): Add
20234         UNSPEC_VOLATILE_EH_RETURN.
20235         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
20236         nds32_output_stack_pop): Support dwarf exception handling process.
20237         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
20238         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
20239         exception handling process.
20240         (nds32_compute_stack_frame): Likewise.
20241         (nds32_return_addr_rtx): Likewise.
20242         (nds32_initial_elimination_offset): Likewise.
20243         (nds32_expand_prologue): Likewise.
20244         (nds32_expand_epilogue): Likewise.
20245         (nds32_dynamic_chain_address): New function.
20246         * config/nds32/nds32.h (machine_function): Add fields for dwarf
20247         exception handling.
20248         (DYNAMIC_CHAIN_ADDRESS): Define.
20249         (EH_RETURN_DATA_REGNO): Define.
20250         (EH_RETURN_STACKADJ_RTX): Define.
20251         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
20252         patterns for dwarf exception handling.
20254 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20256         * config/nds32/nds32.h: Clean up obsolete macros.
20258 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20260         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20261         Add enum values for particular instructions.
20262         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20263         particular intrinsic functions.
20264         * config/nds32/nds32-intrinsic.md: Likewise.
20265         * config/nds32/nds32_intrinsic.h: Likewise.
20266         * config/nds32/nds32.h (nds32_builtins): Likewise.
20267         * config/nds32/nds32.md (type): Add pbsad and pbsada.
20268         (btst, ave): New patterns for particular instructions.
20270 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20272         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
20273         Add enum values for atomic load/store and memory sync.
20274         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
20275         and memory sync.
20276         * config/nds32/nds32-intrinsic.md: Likewise.
20277         * config/nds32/nds32_intrinsic.h: Likewise.
20278         * config/nds32/nds32.h (nds32_builtins): Likewise.
20280 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
20282         PR tree-optimization/85257
20283         * fold-const.c (native_encode_vector): If not all elts could fit
20284         and off is -1, return 0 rather than offset.
20285         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
20286         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
20287         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
20288         adjust buffer in native_interpret_expr call.
20290 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20292         * config/nds32/constants.md (unspec_volatile_element): Add cache
20293         control enum values.
20294         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
20295         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
20296         * config/nds32/nds32.c (nds32_cctl_names): New.
20297         (nds32_print_operand): Handle cache control register names.
20298         * config/nds32/nds32.h (nds32_builtins): New enum values.
20299         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
20300         macros.
20301         * config/nds32/nds32.md (type): Add mmu.
20302         * config/nds32/pipelines.md (simple_insn): Add mmu.
20304 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20306         * config/nds32/nds32.md (type): Remove call.
20307         * config/nds32/pipelines.md (simple_insn): Likewise.
20309 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20311         * config/nds32/constants.md (unspec_volatile_element): Add
20312         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
20313         UNSPEC_VOLATILE_FMFCFG.
20314         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
20315         description for fmfcfg and fmfcsr.
20316         (bdesc_1arg): Add fmtcsr.
20317         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
20318         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
20319         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
20320         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
20321         unspec_fmfcfg): New patterns.
20322         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
20323         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
20324         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
20325         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
20326         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
20327         __nds32__fmfcfg): Define.
20329 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
20331         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
20332         intrinsic register names.
20333         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
20334         intrinsic register enum values and macros.
20336 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
20338         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
20339         for load/store addressing form.
20340         (nds32_print_operand_address): Likewise.
20342 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
20344         PR target/85196
20345         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
20346         based on LABEL_REF.  Remove useless assertion.
20347         (pic_address_needs_scratch): Fix formatting.
20348         (sparc_legitimize_pic_address): Minor tweaks.
20349         (sparc_delegitimize_address): Adjust assertion accordingly.
20350         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
20351         into symbolic_operand.
20352         (movsi_high_pic_label_ref): Likewise.
20353         (movsi_lo_sum_pic_label_ref): Likewise.
20354         (movdi_pic_label_ref): Likewise.
20355         (movdi_high_pic_label_ref): Likewise.
20356         (movdi_lo_sum_pic_label_ref): Likewise.
20358 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
20360         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
20361         custom LIB_SPEC setup.
20363 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
20364             Kito Cheng  <kito.cheng@gmail.com>
20366         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
20367         * config/riscv/freebsd.h: New.
20369 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20371         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
20372         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
20373         file.
20375 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20376             Kito Cheng  <kito.cheng@gmail.com>
20378         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
20379         nds32_output_call, nds32_symbol_binds_local_p): New functions.
20380         * config/nds32/nds32-protos.h (nds32_output_call,
20381         nds32_output_return): Declare.
20382         * config/nds32/nds32.md: Refine all the call and return patterns.
20384 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20386         PR debug/85252
20387         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
20388         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
20390         PR rtl-optimization/84872
20391         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
20392         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
20393         EDGE_CROSSING edge.
20395 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
20397         * expr.c (copy_blkmode_to_reg): Revert 254862.
20398         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
20400 2018-04-06  Richard Biener  <rguenther@suse.de>
20402         PR middle-end/85244
20403         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
20404         after seeing a component reference with an adjacent field.  Treat
20405         refs to arrays at struct end of external decls similar to
20406         refs to unconstrained commons.
20408 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
20410         PR sanitizer/85213
20411         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
20412         look through SAVE_EXPRs with non-side-effects argument.  Adjust
20413         recursive calls.
20414         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
20415         save_p here.
20417 2018-04-06  Richard Biener  <rguenther@suse.de>
20419         PR middle-end/85180
20420         * alias.c (find_base_term): New wrapper around find_base_term
20421         unwinding CSELIB_VAL_PTR changes.
20422         (find_base_term): Do not restore CSELIB_VAL_PTR during the
20423         recursion.
20425 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20427         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
20428         instructions.
20429         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
20430         constant definitions.
20431         ("nop"): lr 0,0 -> nopr r0
20432         ("nop_lr0", "nop_lr1"): New insn definitions.
20434 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
20436         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
20437         NDS32_V3PUSH_AVAILABLE_P macro.
20439 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
20440             Chung-Ju Wu  <jasonwucj@gmail.com>
20442         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
20443         (nds32*-*-*): Add float and fpu_config into supported_defaults.
20444         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
20445         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
20446         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
20447         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
20448         * config/nds32/constraints.md: New constraints and checking for hard
20449         float configuration.
20450         * config/nds32/iterators.md: New mode iterator and attribute for hard
20451         float configuration.
20452         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
20453         patterns.
20454         * config/nds32/nds32-fpu.md: New file.
20455         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
20456         deal with hard float code generation.
20457         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
20458         ARCH_V3S.
20459         (abi_type, float_reg_number): New enum type.
20460         * config/nds32/nds32-predicates.c: New predicates for hard float.
20461         * config/nds32/nds32-protos.h: Declare functions for hard float.
20462         * config/nds32/nds32.c: Implementation for hard float configuration.
20463         * config/nds32/nds32.h: Definitions for hard float configuration.
20464         * config/nds32/nds32.md: Include hard float machine description and
20465         modify patterns for hard float configuration.
20466         * config/nds32/nds32.opt: New options for hard float configuration.
20467         * config/nds32/predicates.md: New predicates for hard float
20468         configuration.
20470 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20472         * common/config/nds32/nds32-common.c
20473         (nds32_option_optimization_table): Enable -mreleax-hint by default.
20475 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
20477         PR middle-end/85195
20478         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
20479         CONSTRUCTOR_ELT (ctor, ...)->value.
20481 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
20483         PR target/85193
20484         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
20486 2018-04-05  Tom de Vries  <tom@codesourcery.com>
20488         PR target/85204
20489         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
20490         cond jump.
20492 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20493             Kito Cheng  <kito.cheng@gmail.com>
20495         * config/nds32/constraints.md (U33): Fine-tune checking condition.
20496         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
20497         * config/nds32/nds32.h (nds32_16bit_address_type): Add
20498         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
20500 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
20501             Kito Cheng  <kito.cheng@gmail.com>
20503         * config/nds32/constraints.md (Ufe): New memory constraint.
20504         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
20505         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
20506         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
20507         operands.
20508         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
20509         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
20511 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20513         * config/nds32/nds32.md: Use optimize_size in the condition for
20514         alu-shift instructions.
20516 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20518         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
20520 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20522         * config/nds32/nds32.md (negsi2): Refine pattern.
20524 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20525             Chung-Ju Wu  <jasonwucj@gmail.com>
20527         * config/nds32/iterators.md (shift_rotate): New code iterator.
20528         (shift): New code attribute.
20529         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
20530         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
20531         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
20532         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
20533         bit-wise operations.
20534         (andsi3, *andsi3): Ditto.
20535         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
20536         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
20537         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
20538         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
20539         nds32_ior_operand, nds32_xor_operand): New predicates.
20541 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20543         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
20544         (addsi3, subsi3): ... this.
20546 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20548         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
20550 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20552         * config/nds32/nds32.md: Adjust indention.
20554 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
20556         * config/nds32/nds32.md (feature): New attribute.
20558 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20560         * config/nds32/nds32.md (subtype): New attribute.
20562 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20564         PR target/85203
20565         * config/arm/arm-builtins.c (arm_expand_builtin): Change
20566         expansion to perform a bitwise AND of the argument followed by a
20567         boolean negation of the result.
20569 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
20571         PR rtl-optimization/84878
20572         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
20573         the basic block.  Assert the use reference is not artificial and that
20574         it has an associated insn.
20576 2018-04-04  Michael Matz  <matz@suse.de>
20578         * builtins.c (compute_objsize): Pass correct operand
20579         to array_at_struct_end_p.
20581 2018-04-04  Richard Biener  <rguenther@suse.de>
20583         PR lto/85176
20584         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
20585         from contexts for DINFO_LEVEL_TERSE and below.
20587 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20589         * config/nds32/nds32-doubleword.md (move_<mode>): Require
20590         resiter_operand condition.
20591         * config/nds32/nds32.md (*move<mode>): Ditto.
20593 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20594             Monk Chiang  <sh.chiang04@gmail.com>
20596         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
20598 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20600         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
20602 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20603             Kito Cheng  <kito.cheng@gmail.com>
20605         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
20606         nds32_cond_code_str, output_cond_branch,
20607         output_cond_branch_compare_zero, nds32_expand_cbranch,
20608         nds32_expand_cstore, nds32_expand_movcc,
20609         nds32_output_cbranchsi4_equality_zero,
20610         nds32_output_cbranchsi4_equality_reg,
20611         nds32_output_cbranchsi4_equality_reg_or_const_int,
20612         nds32_output_cbranchsi4_greater_less_zero: New functions.
20613         * config/nds32/nds32-protos.h (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): Declare.
20619         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20620         nds32_rimm11s_operand): New predicates.
20621         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20622         * config/nds32/nds32.md: Rewrite all the branch and conditional move
20623         patterns.
20625 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20627         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20628         * config/nds32/nds32.md: Ditto.
20629         * config/nds32/pipelines.md: Ditto.
20631 2018-04-04  Richard Biener  <rguenther@suse.de>
20633         PR tree-optimization/85168
20634         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20635         propagating abnormals.
20637 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20639         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20641 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20642             Kito Cheng  <kito.cheng@gmail.com>
20644         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20645         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20646         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20647         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20648         * config/nds32/nds32.md (sibcall_internal): New.
20649         (sibcall_register): Remove.
20650         (sibcall_immediate): Remove.
20651         (sibcall_value_internal): New.
20652         (sibcall_value_register): Remove.
20653         (sibcall_value_immediate): Remove.
20654         * config/nds32/predicates.md (nds32_general_register_operand): New.
20655         (nds32_call_address_operand): New.
20657 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20659         PR rtl-optimization/85167
20660         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20661         bb_defs if *split_p, instead preinitialize it to NULL.
20663         PR tree-optimization/85156
20664         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20665         evaluating the argument multiple times.
20667 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
20669         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20670         than vector.
20671         (_mm_cvtpd_ps): Likewise.
20672         (_mm_cvttpd_epi32): Likewise.
20673         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20674         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20675         vector, pixel, and bool following altivec.h include.
20677 2018-04-03  Martin Sebor  <msebor@redhat.com>
20679         * doc/extend.texi (Common Function Attributes): Clarify.
20680         (const attribute): Likewise.
20681         (pure attribute): Likewise.
20683 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20685         PR target/85169
20686         * config/i386/i386.c (ix86_expand_vector_set): Use
20687         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
20689 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
20691         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20692         instructions when changing rounding bits to preserve precision bits
20693         in the x87 control word.
20695 2018-04-03  Martin Liska  <mliska@suse.cz>
20697         PR tree-optimization/82491
20698         * rtl.h (strip_offset_and_add): Replace += suboffset with
20699         poly_uint64 () + suboffset.
20701 2018-03-29  Martin Liska  <mliska@suse.cz>
20702             Martin Jambor  <mjambor@suse.cz>
20704         PR ipa/84947
20705         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20706         param_type is not an integral or pointer type.
20708 2018-04-03  Richard Biener  <rguenther@suse.de>
20710         * sese.h (recompute_all_dominators): Remove.
20712 2018-04-02  Martin Sebor  <msebor@redhat.com>
20714         * doc/invoke.texi (-Wrestrict): Fix typos.
20716 2018-04-02  Jim Wilson  <jimw@sifive.com>
20718         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20719         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20720         (<optab>di3, <optab>si3_extend): Likewise.
20721         (<optab>si3_mask, <optab>si3_mask_1): New.
20722         (<optab>di3_mask, <optab>di3_mask_1): New.
20723         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20724         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20725         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20727 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
20729         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20730         example.
20732 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20734         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20735         (nds32_canonicalize_comparison): New function.
20737 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20738             Kito Cheng  <kito.cheng@gmail.com>
20739             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20741         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20742         * config/nds32/constants.md (unspec_volatile_element): Add
20743         UNSPEC_VOLATILE_RELAX_GROUP.
20744         * config/nds32/nds32-relax-opt.c: New file.
20745         * config/nds32/nds32-predicates.c
20746         (nds32_symbol_load_store_p): New function.
20747         * config/nds32/nds32-protos.h
20748         (nds32_symbol_load_store_p): Declare function.
20749         (make_pass_nds32_relax_opt): Declare new rtl pass function.
20750         * config/nds32/nds32.c
20751         (nds32_register_pass): New function to register pass.
20752         (nds32_register_passes): New function to register passes.
20753         * config/nds32/nds32.md (relax_group): New pattern.
20754         * config/nds32/nds32.opt (mrelax-hint): New option.
20755         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20757 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
20759         * config/nds32/t-nds32: Modify files dependency.
20761 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20763         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20764         (PROFILE_HOOK): Define its implementation.
20766 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20768         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20769         type and 32-bit size.
20771 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
20773         PR middle-end/85090
20774         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20775         (V_128_256): New mode iterator.
20776         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20777         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20778         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20779         of V.
20780         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20781         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20783 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
20785         PR target/83315
20786         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20787         NaN inputs correctly.
20789 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
20791         PR target/80546
20792         * config/rs6000/vsx.md (??r): New mode attribute.
20793         (*vsx_mov<mode>_64bit): Use it.
20794         (*vsx_mov<mode>_32bit): Likewise.
20796 2018-03-30  Martin Sebor  <msebor@redhat.com>
20798         PR tree-optimization/84818
20799         * builtins.c (check_access): Use warning_n.
20801 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20803         PR target/83822
20804         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20805         condition.
20806         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20807         condition.
20809 2018-03-30  Julia Koval  <julia.koval@intel.com>
20811         PR target/84413
20812         * x86-tune.def (movx, partial_reg_dependency): Enable for
20813         m_SKYLAKE_AVX512.
20815 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
20817         PR inline-asm/84985
20818         * lra-constraints.c (process_alt_operands): Move setting
20819         this_alternative_matches below.
20821 2018-03-29  Martin Liska  <mliska@suse.cz>
20823         PR lto/84995.
20824         * doc/invoke.texi: Document how LTO works with debug info.
20825         Describe auto-load support of binutils.  Mention 'x86-64'
20826         as valid option value of -march option.
20828 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
20830         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20832         PR c/85094
20833         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20834         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20835         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20836         checking.
20838 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20840         PR target/84912
20841         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20842         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20843         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20844         for RS6000_BTM_POWERPC64.
20845         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20846         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20847         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20848         definition.
20849         (DIVDE): Use it.
20850         (DIVDEU): Likewise.
20852 2018-03-28  Carl Love  <cel@us.ibm.com>
20854         Revert
20855         2017-09-27  Carl Love  <cel@us.ibm.com>
20857         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20858         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20859         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20860         fctiw instruction.
20862 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20864         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20865         instead of __vector bool.
20866         (_mm_max_pu8): Likewise.
20867         (_mm_min_pi16): Likewise.
20869 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20871         PR target/84912
20872         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20873         (DIVWEUO): Likewise.
20874         (DIVDEO): Likewise.
20875         (DIVDEUO): Likewise.
20876         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20877         DIVWEUO and DIVDEUO.
20878         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20879         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20880         (div_extend): Likewise.
20881         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20882         builtin function.
20883         (__builtin_divweuo): Likewise.
20884         (__builtin_divdeo): Likewise.
20885         (__builtin_divdeuo): Likewise.
20887 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20889         PR target/85095
20890         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20891         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20893         PR tree-optimization/82004
20894         * gimple-match-head.c (optimize_pow_to_exp): New function.
20895         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20896         Don't fold to exp if optimize_pow_to_exp is false.
20898 2018-03-28  Martin Liska  <mliska@suse.cz>
20900         PR other/84819
20901         * calls.c (initialize_argument_information): Fix trailing space.
20902         * common.opt: Fix typo and provide better explanation for
20903         -fsanitize-coverage option.
20904         * config/i386/i386.opt: Fix typo.
20906 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20907             Martin Liska  <mliska@suse.cz>
20909         PR sanitizer/85081
20910         * gimplify.c (asan_poison_variable): Don't do the check for
20911         gimplify_omp_ctxp here.
20912         (gimplify_decl_expr): Do it here.
20913         (gimplify_target_expr): Likewise.
20915 2018-03-28  Martin Liska  <mliska@suse.cz>
20917         PR target/84988
20918         * config/i386/i386.c (ix86_function_arg_advance): Do not call
20919         chkp_type_bounds_count if MPX is not enabled.
20921 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20923         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20925 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
20927         PR target/84914
20928         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20929         function to create the function decl for complex long double
20930         multiply and divide for -mabi=ieeelongdouble.
20931         (init_float128_ieee): Call it.
20933 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
20935         PR target/85044
20936         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20937         -fcf-protection=branch -mibt.
20938         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20940 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20942         PR target/81863
20943         * config/arm/arm.c (arm_valid_symbolic_address): Handle
20944         arm_word_relocations.
20946 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
20948         PR target/85056
20949         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20950         extern array declarations.
20952 2018-03-27  Richard Biener  <rguenther@suse.de>
20954         PR middle-end/84067
20955         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20956         explicit single_use checks.
20958 2018-03-27  Richard Biener  <rguenther@suse.de>
20960         PR tree-optimization/85082
20961         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20962         Valueize the VUSE.
20964 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20966         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20967         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20968         Turn on fasynchronous-unwind-tables and funwind-tables.
20970 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
20972         PR target/85073
20973         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20974         (*bmi_blsr_<mode>_ccz): Ditto.
20976 2018-03-26  Tom de Vries  <tom@codesourcery.com>
20978         PR tree-optimization/85063
20979         * omp-general.c (offloading_function_p): New function.  Factor out
20980         of ...
20981         * omp-offload.c (pass_omp_target_link::gate): ... here.
20982         * omp-general.h (offloading_function_p): Declare.
20983         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20984         with attribute omp declare target for offloading functions.
20986 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
20988         PR tree-optimization/84005
20989         * tree-data-ref.h (get_base_for_alignment): Declare.
20990         * tree-data-ref.c (get_base_for_alignment_1): New function.
20991         (get_base_for_alignment): Likewise.
20992         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20993         get_base_for_alignment to find a suitable base object, instead
20994         of always using drb->base_address.
20996 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20998         PR inline-asm/85022
20999         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
21000         known size by default.
21002 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
21004         PR inline-asm/85030
21005         * lra-constraints.c (process_alt_operands): Don't match BLKmode
21006         and non BLKmode operands.
21008 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21010         PR target/85026
21011         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
21012         Clean up attributes.
21014 2018-03-23  Richard Biener  <rguenther@suse.de>
21016         PR debug/85020
21017         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
21018         we are going to emit early debug for LTO.
21020 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
21022         PR inline-asm/85034
21023         * function.c (match_asm_constraints_1): Don't optimize if input
21024         doesn't satisfy general_operand predicate for output's mode.
21026         PR inline-asm/85022
21027         * alias.c (write_dependence_p): Don't require for x_canonicalized
21028         non-VOIDmode if x has VOIDmode.
21030         PR sanitizer/85029
21031         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
21032         just don't try to optimize it rather than assert it never happens.
21034 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21036         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
21037         macro expansions for definition of ST_INTERNAL_<mode> and
21038         LD_INTERNAL_<mode> builtins.
21039         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
21040         Remove prototype.
21041         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
21042         function.
21043         (altivec_expand_st_builtin): Likewise.
21044         (altivec_expand_builtin): Remove calls to deleted functions.
21045         (rs6000_address_for_altivec): Delete this function.
21046         * config/rs6000/vector.md: Remove expands for
21047         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
21049 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
21051         PR target/84826
21052         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
21053         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
21054         re-computing once computed.
21055         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
21056         (arm_init_machine_status): Initialize
21057         machine->static_chain_stack_bytes.
21059 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21061         PR target/84760
21062         * doc/extend.texi: Add four new prototypes for vec_ld.
21063         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
21064         definitions for more logical presentation.
21065         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
21066         entries for V1TI variants of __builtin_altivec_ld builtin.
21067         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
21068         handling of V1TI variant of LVX icode pattern.
21069         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
21070         (rs6000_gimple_fold_builtin): Likewise.
21071         (altivec_init_builtins): Add code to define
21072         __builtin_altivec_lvx_v1ti function.
21074 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
21076         PR inline-asm/84941
21077         * function.c (match_asm_constraints_1): Don't do the optimization
21078         if input isn't a REG, SUBREG, MEM or constant.
21080 2018-03-22  Tom de Vries  <tom@codesourcery.com>
21082         PR tree-optimization/84956
21083         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
21084         bb_has_abnormal_pred.
21086 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
21088         PR sanitizer/85018
21089         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
21090         DECL_INITIAL (decl) to decl at the end.
21091         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
21092         adjust the comment.
21094 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
21096         * doc/extend.texi (__builtin_tgmath): Document when complex
21097         integer types are treated as _Complex _Float64.
21099 2018-03-21  Tom de Vries  <tom@codesourcery.com>
21101         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
21103 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
21105         PR tree-optimization/84960
21106         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
21107         if it is ENTRY block, move them into single succ of ENTRY in that case.
21109 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
21111         PR tree-optimization/84811
21112         * poly-int.h (poly_span_traits): Remove the T3 parameter and
21113         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
21114         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
21115         (known_subrange_p): Update accordingly.  Cast each value involved
21116         in the size comparison, rather than casting the result of the
21117         subtraction.
21119 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
21121         PR tree-optimization/84982
21122         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
21123         by flipping the least significant bit rather than all bits from
21124         bitpos to bitpos + bitsize - 1.
21126 2018-03-21  Nathan Sidwell  <nathan@acm.org>
21128         * doc/extend.texi (Deprecated Features): Remove mention of
21129         long-deleted deprecations.
21131 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21133         PR jit/84288
21134         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
21135         * configure: Regenerate.
21137 2018-03-21  Tom de Vries  <tom@codesourcery.com>
21139         PR tree-optimization/83126
21140         * tree-parloops.c (num_phis): New function.
21141         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
21143 2018-03-21  Nathan Sidwell  <nathan@acm.org>
21145         * doc/extend.texi (Deprecated Features): Update deprecated flags,
21146         mention anon-struct/union members and trailing attributes.
21148 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
21150         PR tree-optimization/84969
21151         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
21152         builtin memset partitions if they set different rhs values.
21154 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
21156         PR rtl-optimization/84989
21157         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
21158         VEC_DUPLICATE with scalar result mode.
21160 2018-03-21  Martin Liska  <mliska@suse.cz>
21162         PR ipa/84963
21163         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
21164         not intended return statement.
21166 2018-03-21  Martin Liska  <mliska@suse.cz>
21168         PR target/84988
21169         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
21170         (chkp_find_bound_slots_1): Limit number of iterations.
21172 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
21174         PR target/84838
21175         * Minor grammar fixes for x86 options.
21177 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21179         PR debug/84875
21180         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
21181         holding REG_CFA_RESTORE notes, instead turn them into a USE.
21183 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
21185         PR target/83789
21186         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
21187         (altivec_lvx_<mode>_1op): Likewise.
21188         (altivec_stvx_<mode>_2op): Likewise.
21189         (altivec_stvx_<mode>_1op): Likewise.
21190         (altivec_lvx_<VM2:mode>): New define_expand.
21191         (altivec_stvx_<VM2:mode>): Likewise.
21192         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
21193         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21194         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
21195         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
21196         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
21197         (rs6000_gen_lvx): Likewise.
21198         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
21199         (altivec_expand_stv_builtin): Likewise.
21200         (altivec_expand_builtin): Likewise.
21201         * config/rs6000/vector.md: Likewise.
21203 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21205         PR target/82518
21206         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
21207         BYTES_BIG_ENDIAN.
21209 2018-03-20  Richard Biener  <rguenther@suse.de>
21211         PR target/84986
21212         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
21213         sign-conversions as zero, fall back to standard scalar_stmt
21214         cost for the rest.
21216 2018-03-20  Martin Liska  <mliska@suse.cz>
21218         PR ipa/84825
21219         * predict.c (rebuild_frequencies): Handle case when we have
21220         PROFILE_ABSENT, but flag_guess_branch_prob is false.
21222 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21224         PR target/84990
21225         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
21226         flag_section_anchors.
21227         * varasm.c (use_blocks_for_decl_p): Remove hack for
21228         dw2_force_const_mem.
21230         PR target/84845
21231         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
21232         to ...
21233         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
21234         be created, use lowpart_subreg of operands[0] rather than operands[0]
21235         itself.
21236         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
21237         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
21238         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
21239         and n constraint instead of aarch64_shift_imm_di and Usd.
21240         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
21241         (*aarch64_<optab>_reg_minus<mode>3): ... this.
21243 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
21245         PR target/82989
21246         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
21247         to favor GPR over NEON registers.
21248         (<shift>di3_neon): Likewise.
21250 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21252         PR target/84952
21253         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
21254         (nvptx_process_pars): Emit bar.sync asap and alap.
21256 2018-03-20  Tom de Vries  <tom@codesourcery.com>
21258         PR target/84954
21259         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
21260         seen_label if seen_label is already set.
21262 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
21264         PR target/84945
21265         * config/i386/i386.c (fold_builtin_cpu): For features above 31
21266         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
21267         Use 1U instead of 1.  Formatting fixes.
21269         PR c/84953
21270         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
21271         instead of TREE_TYPE (s1) for the return value.
21273 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
21275         PR tree-optimization/84946
21276         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
21277         bitsize + bitsize in poly_uint64 rather than poly_int64.
21279         PR sanitizer/78651
21280         * dwarf2asm.c: Include fold-const.c.
21281         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
21282         of decl rather than decl itself.
21284         PR rtl-optimization/84643
21285         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
21287 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
21289         PR sanitizer/78651
21290         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
21291         calling assemble_variable.
21293 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
21295         PR target/81647
21296         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
21297         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
21299 2018-03-19  Jim Wilson  <jimw@sifive.com>
21301         PR bootstrap/84856
21302         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
21303         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
21304         (riscv_first_stack_step): Likewise.
21305         (riscv_option_override): Use STACK_BOUNDARY instead of
21306         MIN_STACK_BOUNDARY.
21307         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
21308         MIN_STACK_BOUNDARY.
21309         (BIGGEST_ALIGNMENT): Set to 128.
21310         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
21311         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
21312         STACK_BOUNDARY.
21314 2018-03-19  Richard Biener  <rguenther@suse.de>
21316         PR tree-optimization/84933
21317         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
21318         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
21320 2018-03-19  Richard Biener  <rguenther@suse.de>
21322         PR tree-optimization/84859
21323         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
21324         (cond_if_else_store_replacement): Perform sinking operation on
21325         single-store BBs regardless of MAX_STORES_TO_SINK setting.
21326         Generalize what a BB with a single eligible store is.
21328 2018-03-19  Richard Biener  <rguenther@suse.de>
21330         PR tree-optimization/84929
21331         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
21332         chrec_is_positive against non-chrec arg.
21334 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
21336         PR target/84711
21337         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
21339 2018-03-18  Martin Liska  <mliska@suse.cz>
21341         PR rtl-optimization/84635
21342         * regrename.c (build_def_use): Use matches_mode only when
21343         matches >= 0.
21345 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
21347         PR tree-optimization/84913
21348         * tree-vect-loop.c (vectorizable_reduction): Don't try to
21349         vectorize chains of COND_EXPRs.
21351 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21353         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
21355 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21357         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
21359 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
21361         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21363 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21364             Kito Cheng  <kito.cheng@gmail.com>
21366         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
21367         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
21368         (nds32_adjust_reg_alloc_order): New function.
21369         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
21371 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
21373         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
21374         nds32_print_operand, nds32_print_operand_address): Use
21375         HOST_WIDE_INT_PRINT_DEC instead.
21377 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
21379         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
21381 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
21383         PR target/84902
21384         * config/i386/i386.c (initial_ix86_tune_features,
21385         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
21386         unsigned long long.
21387         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
21388         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
21389         rather than 1u << ix86_tune.  Formatting fix.
21390         (ix86_option_override_internal): Change ix86_arch_mask from
21391         unsigned int to unsigned HOST_WIDE_INT, initialize to
21392         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
21393         (ix86_function_specific_restore): Likewise.
21395 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21397         PR target/84899
21398         * postreload.c (reload_combine_recognize_pattern): Perform
21399         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
21400         truncate_int_for_mode the result for the destination's mode.
21402         PR c/84909
21403         * hsa-gen.c (mem_type_for_type): Fix comment typo.
21404         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
21405         Likewise.
21406         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21407         Likewise.
21409 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
21411         PR target/84876
21412         * lra-assigns.c (lra_split_hard_reg_for): Don't use
21413         regno_allocno_class_array and sorted_pseudos.
21414         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
21415         insns where regno is used.
21417 2018-03-16  Martin Liska  <mliska@suse.cz>
21419         PR ipa/84833
21420         * multiple_target.c (create_dispatcher_calls): Redirect
21421         reference in the symbol table.
21423 2018-03-16  Martin Liska  <mliska@suse.cz>
21425         PR ipa/84722
21426         * multiple_target.c (create_dispatcher_calls): Redirect also
21427         an alias.
21429 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21431         PR c++/79937
21432         PR c++/82410
21433         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
21434         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
21435         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
21437 2018-03-16  Julia Koval  <julia.koval@intel.com>
21439         * doc/invoke.texi (Skylake Server): Add CLWB.
21440         Cannonlake): Remove CLWB.
21442 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
21444         PR tree-optimization/84841
21445         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
21446         1 << 3.
21447         (FLOAT_ONE_CONST_TYPE): Define.
21448         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
21449         (sort_by_operand_rank): Put entries with higher constant_type last
21450         rather than first to match comments.
21452 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
21454         * config/nios2/nios2.md (movsi_internal): Fix thinko in
21455         split predicate.
21457 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21459         PR c++/79085
21460         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
21461         check and use address of target always.
21463 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
21465         PR target/84574
21466         * config/i386/i386.c (indirect_thunk_needed): Update comments.
21467         (indirect_thunk_bnd_needed): Likewise.
21468         (indirect_thunks_used): Likewise.
21469         (indirect_thunks_bnd_used): Likewise.
21470         (indirect_return_needed): New.
21471         (indirect_return_bnd_needed): Likewise.
21472         (output_indirect_thunk_function): Add a bool argument for
21473         function return.
21474         (output_indirect_thunk_function): Don't generate alias for
21475         function return thunk.
21476         (ix86_code_end): Call output_indirect_thunk_function to generate
21477         function return thunks.
21478         (ix86_output_function_return): Set indirect_return_bnd_needed
21479         and indirect_return_needed instead of indirect_thunk_bnd_needed
21480         and indirect_thunk_needed.
21482 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
21484         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
21485         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
21486         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
21488 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
21489             Paul Hua <paul.hua.gm@gmail.com>
21491         PR c/84852
21492         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
21494 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
21496         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
21497         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
21498         resp. SFmode cases.
21500 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
21502         PR target/84711
21503         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
21504         instead of GET_MODE_SIZE when comparing Units.
21506 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
21508         PR target/68256
21509         * varasm.c (hash_section): Return an unchangeble hash value
21510         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
21511         Return !aarch64_can_use_per_function_literal_pools_p ().
21513 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
21515         PR target/84860
21516         * optabs.c (emit_conditional_move): Pass address of cmode's copy
21517         rather than address of cmode as last argument to prepare_cmp_insn.
21519 2018-03-15  Julia Koval  <julia.koval@intel.com>
21521         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
21522         F_AVX512VNNI, F_AVX512BITALG): New.
21524 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
21526         PR target/83451
21527         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
21528         insn for floating-point loads and stores.
21530 2018-03-14  Carl Love  <cel@us.ibm.com>
21532         * config/rs6000/rs6000-c.c: Add macro definitions for
21533         ALTIVEC_BUILTIN_VEC_PERMXOR.
21534         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
21535         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
21536         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
21537         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
21538         UNSPEC_VPERMXOR.
21539         * config/doc/extend.texi: Add prototypes for vec_permxor.
21541 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
21543         PR c/84852
21544         * diagnostic-show-locus.c (class layout_point): Convert m_line
21545         from int to linenum_type.
21546         (line_span::comparator): Use linenum "compare" function when
21547         comparing line numbers.
21548         (test_line_span): New function.
21549         (layout_range::contains_point): Convert param "row" from int to
21550         linenum_type.
21551         (layout_range::intersects_line_p): Likewise.
21552         (layout::will_show_line_p): Likewise.
21553         (layout::print_source_line): Likewise.
21554         (layout::should_print_annotation_line_p): Likewise.
21555         (layout::print_annotation_line): Likewise.
21556         (layout::print_leading_fixits): Likewise.
21557         (layout::annotation_line_showed_range_p): Likewise.
21558         (struct line_corrections): Likewise for field m_row.
21559         (line_corrections::line_corrections): Likewise for param "row".
21560         (layout::print_trailing_fixits): Likewise.
21561         (layout::get_state_at_point): Likewise.
21562         (layout::get_x_bound_for_row): Likewise.
21563         (layout::print_line): Likewise.
21564         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
21565         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
21566         * input.c (selftest::test_linenum_comparisons): New function.
21567         (selftest::input_c_tests): Call it.
21568         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
21569         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
21570         * selftest.h (ASSERT_GT): New macro.
21571         (ASSERT_GT_AT): New macro.
21572         (ASSERT_LT): New macro.
21573         (ASSERT_LT_AT): New macro.
21575 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
21577         PR rtl-optimization/84780
21578         * combine.c (distribute_links): Don't make a link based on pc_rtx.
21580 2018-03-14  Martin Liska  <mliska@suse.cz>
21582         * tree.c (record_node_allocation_statistics): Use
21583         get_stats_node_kind.
21584         (get_stats_node_kind): New function extracted from
21585         record_node_allocation_statistics.
21586         (free_node): Use get_stats_node_kind.
21588 2018-03-14  Richard Biener  <rguenther@suse.de>
21590         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
21591         that the value-set of ANTIC_IN doesn't grow.
21593         Revert
21594         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21595         member.
21596         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21597         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21599 2018-03-14  Julia Koval  <julia.koval@intel.com>
21601         * config.gcc (icelake-client, icelake-server): New.
21602         (icelake): Remove.
21603         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
21604         (initial_ix86_arch_features): Ditto.
21605         (PTA_SKYLAKE): Add SGX.
21606         (PTA_ICELAKE): Remove.
21607         (PTA_ICELAKE_CLIENT): New.
21608         (PTA_ICELAKE_SERVER): New.
21609         (ix86_option_override_internal): Split up icelake on icelake client and
21610         icelake server.
21611         (get_builtin_code_for_version): Ditto.
21612         (fold_builtin_cpu): Ditto.
21613         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
21614         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
21615         * config/i386/i386.h (processor_type): Ditto.
21616         * doc/invoke.texi: Ditto.
21618 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
21620         PR sanitizer/83392
21621         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21622         INTEGER_CST offset, add it together with bitpos / 8 and
21623         sign extend based on POINTER_SIZE.
21625         PR target/84844
21626         Revert
21627         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
21629         PR target/78090
21630         * config/i386/constraints.md (Yc): New register constraint.
21631         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21632         Use Yc constraint for alternative 2 of operand 0.  Remove
21633         preferred_for_speed attribute.
21635 2018-03-14  Richard Biener  <rguenther@suse.de>
21637         PR tree-optimization/84830
21638         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21639         with the old one to avoid oscillations.
21641 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
21643         PR target/83712
21644         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21645         pseudos.
21646         (assign_by_spills): Return a flag of reload assignment failure.
21647         Do not process the reload assignment failures.  Do not spill other
21648         reload pseudos if they has the same reg class.  Update n if
21649         necessary.
21650         (lra_assign): Add a return arg.  Set up from the result of
21651         assign_by_spills call.
21652         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21653         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21654         usage_insns if it is not NULL.
21655         (spill_hard_reg_in_range): New function.
21656         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21657         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21658         function prototypes.
21659         (lra_assign): Change prototype.
21660         * lra.c (lra): Add code to deal with fails by splitting hard reg
21661         live ranges.
21663 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
21665         * config/riscv/riscv.opt (mrelax): New option.
21666         * config/riscv/riscv.c (riscv_file_start): Emit ".option
21667         "norelax" when riscv_mrelax is disabled.
21668         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21670 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21672         PR target/84743
21673         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21674         reassociation for int modes.
21676 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21678         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21679         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21680         for big-endian.
21681         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21682         * config/aarch64/aarch64-sve.md
21683         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21684         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21685         (*extend<mode><Vwide>2): Rename to...
21686         (aarch64_sve_extend<mode><Vwide>2): ...this.
21687         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21688         renaming the old pattern to...
21689         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
21690         unsigned packs.
21691         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21692         define_expand, renaming the old pattern to...
21693         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21694         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21695         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21696         account when deciding which SVE instruction the optab should use.
21697         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21699 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21701         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21702         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21703         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21704         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21705         (tlsdesc_small_<mode>): Turn a define_expand and use
21706         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
21707         (tlsdesc_small_advsimd_<mode>): ...this.
21708         (tlsdesc_small_sve_<mode>): New pattern.
21710 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21712         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21713         (UNSPEC_UMUL_HIGHPART): New constants.
21714         (MUL_HIGHPART): New int iteraor.
21715         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21716         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21717         define_expand.
21718         (*<su>mul<mode>3_highpart): New define_insn.
21720 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21722         PR lto/84805
21723         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21724         incomplete types.
21726 2018-03-13  Martin Liska  <mliska@suse.cz>
21728         PR ipa/84658.
21729         * (sem_item_optimizer::sem_item_optimizer): Initialize new
21730         vector.
21731         (sem_item_optimizer::~sem_item_optimizer): Release it.
21732         (sem_item_optimizer::merge_classes): Register variable aliases.
21733         (sem_item_optimizer::fixup_pt_set): New function.
21734         (sem_item_optimizer::fixup_points_to_sets): Likewise.
21735         * ipa-icf.h: Declare new variables and functions.
21737 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
21739         PR middle-end/84834
21740         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21741         integer_pow2p@2 and test integer_pow2p in condition.
21742         (A < 0 ? C : 0): Similarly for @1.
21744         PR middle-end/84831
21745         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21746         characters starting at p contain '\0' character, don't look beyond
21747         that.
21749         PR target/84827
21750         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21751         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21753         PR target/84828
21754         * reg-stack.c (change_stack): Change update_end var from int to
21755         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21756         also call set_block_for_insn on the newly added insns and rescan.
21758         PR target/84786
21759         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21760         on the last operand.
21762         PR c++/84704
21763         * tree.c (stabilize_reference_1): Return save_expr (e) for
21764         STATEMENT_LIST even if it doesn't have side-effects.
21766 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
21768         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21770 2018-03-12  Renlin Li  <renlin.li@arm.com>
21772         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21773         aarch64_output_scalar_simd_mov_immediate.
21775 2018-03-12  Martin Sebor  <msebor@redhat.com>
21777         PR tree-optimization/83456
21778         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21779         for perfectly overlapping calls to memcpy.
21780         (gimple_fold_builtin_memory_chk): Same.
21781         (gimple_fold_builtin_strcpy): Handle no-warning.
21782         (gimple_fold_builtin_stxcpy_chk): Same.
21783         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21785 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21787         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21788         parameter.  Use it for SFmode.
21789         (rs6000_function_arg_advance_1): Adjust.
21790         (rs6000_function_arg): Adjust.
21791         (rs6000_gimplify_va_arg): Pass false for that new parameter.
21793 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21795         PR rtl-optimization/84169
21796         PR rtl-optimization/84780
21797         * combine.c (can_combine_p): Check for a 2-insn combination whether
21798         the destination register is used between the two insns, too.
21800 2018-03-12  Richard Biener  <rguenther@suse.de>
21802         PR tree-optimization/84803
21803         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21804         for refs DR analysis didn't process.
21806 2018-03-12  Richard Biener  <rguenther@suse.de>
21808         PR tree-optimization/84777
21809         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21810         force-vectorize loops ignore whether we are optimizing for size.
21812 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
21814         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21815         (TARGET_MD_ASM_ADJUST): Define.
21817 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
21818             Kito Cheng  <kito.cheng@gmail.com>
21819             Chung-Ju Wu  <jasonwucj@gmail.com>
21821         * config/nds32/nds32.c (nds32_compute_stack_frame,
21822         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21823         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21824         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21825         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21826         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21827         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21828         * config/nds32/nds32.md (prologue, epilogue): Use macro
21829         NDS32_V3PUSH_AVAILABLE_P to do checking.
21831 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
21833         PR debug/58150
21834         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21835         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21836         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
21837         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
21838         addition of most attributes on !orig_type_die or the attribute not
21839         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21841 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21842             Chung-Ju Wu  <jasonwucj@gmail.com>
21844         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21845         __NDS32_VH__ macro.
21846         * config/nds32/nds32.opt (mvh): New option.
21848 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21849             Chung-Ju Wu  <jasonwucj@gmail.com>
21851         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21852         function.
21853         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21854         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21855         definition.
21857 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21858             Chung-Ju Wu  <jasonwucj@gmail.com>
21860         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21861         function.
21862         * config/nds32/nds32-multiple.md (strlensi): New pattern.
21863         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21865 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
21866             Kito Cheng  <kito.cheng@gmail.com>
21867             Chung-Ju Wu  <jasonwucj@gmail.com>
21869         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21870         UNSPEC_FFMISM and UNSPEC_FLMISM.
21871         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21872         for ffb, ffmism and flmism.
21873         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21874         (unspec_ffmism): Ditto.
21875         (unspec_flmism): Ditto.
21876         (nds32_expand_builtin_impl): Check if string extension is available.
21877         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21878         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21880 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
21882         Reverting patch:
21883         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21885         PR target/83712
21886         * lra-assigns.c (assign_by_spills): Return a flag of reload
21887         assignment failure.  Do not process the reload assignment
21888         failures.  Do not spill other reload pseudos if they has the same
21889         reg class.
21890         (lra_assign): Add a return arg.  Set up from the result of
21891         assign_by_spills call.
21892         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21893         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21894         usage_insns if it is not NULL.
21895         (spill_hard_reg_in_range): New function.
21896         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21897         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21898         function prototypes.
21899         (lra_assign): Change prototype.
21900         * lra.c (lra): Add code to deal with fails by splitting hard reg
21901         live ranges.
21903 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
21905         PR target/84807
21906         * config/i386/i386.opt: Replace Enforcment with Enforcement.
21908 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
21910         PR debug/84620
21911         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21912         (dw_val_node): Add val_symbolic_view.
21913         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21914         (symview_upper_bound): New.
21915         (new_line_info_table): Initialize symviews_since_reset.
21916         (dwarf2out_source_line): Count symviews_since_reset and set
21917         symview_upper_bound.
21918         (dw_val_equal_p): Handle symview.
21919         (add_AT_symview): New.
21920         (print_dw_val): Handle symview.
21921         (attr_checksum, attr_checksum_ordered): Likewise.
21922         (same_dw_val_p, size_of_die): Likewise.
21923         (value_format, output_die): Likewise.
21924         (add_high_low_attributes): Use add_AT_symview for entry_view.
21925         (dwarf2out_finish): Reset symview_upper_bound, clear
21926         zero_view_p.
21928 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
21930         PR target/83969
21931         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21932         Add strict argument and use it.
21933         (rs6000_split_multireg_move): Update for new strict argument.
21934         (mem_operand_gpr): Disallow all non-offsettable addresses.
21935         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21937 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
21939         PR target/84772
21940         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21941         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21942         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21944         PR c++/84767
21945         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21946         decl, use remap_type if we want to use the type.
21948 2018-03-09  Martin Sebor  <msebor@redhat.com>
21950         PR tree-optimization/84526
21951         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21952         Remove dead code.
21953         (builtin_access::generic_overlap): Be prepared to handle non-array
21954         base objects.
21956 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
21958         PR rtl-optimization/84682
21959         * lra-constraints.c (process_address_1): Check is_address flag
21960         for address constraints.
21961         (process_alt_operands): Likewise.
21962         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21963         preprocess_constraints.
21964         * recog.h (preprocess_constraints): Add oploc parameter.
21965         Adjust callers.
21966         * recog.c (preprocess_constraints): Test address_operand for
21967         CT_ADDRESS constraints.
21969 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21971         PR target/83712
21972         * lra-assigns.c (assign_by_spills): Return a flag of reload
21973         assignment failure.  Do not process the reload assignment
21974         failures.  Do not spill other reload pseudos if they has the same
21975         reg class.
21976         (lra_assign): Add a return arg.  Set up from the result of
21977         assign_by_spills call.
21978         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21979         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21980         usage_insns if it is not NULL.
21981         (spill_hard_reg_in_range): New function.
21982         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21983         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21984         function prototypes.
21985         (lra_assign): Change prototype.
21986         * lra.c (lra): Add code to deal with fails by splitting hard reg
21987         live ranges.
21989 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21991         PR target/83193
21992         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21993         Accept complain bool parameter.  Only emit errors if it is true.
21994         (arm_parse_cpu_option_name): Likewise.
21995         (arm_target_thumb_only): Adjust callers of the above.
21996         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21997         prototype to take a default true bool parameter.
21998         (arm_parse_arch_option_name): Likewise.
22000 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
22001             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
22003         PR jit/64089
22004         PR jit/84288
22005         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
22006         * configure: Regenerate.
22007         * configure.ac ("linker --version-script option"): New.
22008         ("linker soname option"): New.
22010 2018-03-09  Richard Biener  <rguenther@suse.de>
22012         PR tree-optimization/84775
22013         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
22014         immediate uses of predicate stmts and mark them modified.
22016         Revert
22017         PR tree-optimization/84178
22018         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22019         to caller.
22020         (version_loop_for_if_conversion): Delay update_ssa call.
22021         (tree_if_conversion): Delay update_ssa until after predicate
22022         insertion.
22024 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
22026         PR target/84763
22027         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
22028         when the function accesses prior frames.
22030 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
22032         PR debug/84456
22033         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
22034         gen_llsym, otherwise call maybe_gen_llsym.
22036         PR inline-asm/84742
22037         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
22038         has ',' character inside of it.
22040 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22042         PR target/84748
22043         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
22044         as clobbering CC_REGNUM.
22046 2018-03-08  Richard Biener  <rguenther@suse.de>
22048         PR middle-end/84552
22049         * tree-scalar-evolution.c: Include tree-into-ssa.h.
22050         (follow_copies_to_constant): Do not follow SSA names registered
22051         for update.
22053 2018-03-08  Richard Biener  <rguenther@suse.de>
22055         PR tree-optimization/84178
22056         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22057         to caller.
22058         (version_loop_for_if_conversion): Delay update_ssa call.
22059         (tree_if_conversion): Delay update_ssa until after predicate
22060         insertion.
22062 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
22064         PR tree-optimization/84178
22065         * tree-if-conv.c (release_bb_predicate): Remove the
22066         the assertion that the stmts have NULL use_ops.
22067         Discard the statements, asserting that they haven't
22068         yet been added to a BB.
22070 2018-03-08  Richard Biener  <rguenther@suse.de>
22072         PR tree-optimization/84746
22073         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
22074         (phi_translate): Pass in destination ANTIC_OUT set.
22075         (phi_translate_1): Likewise.  For a simplified result lookup
22076         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
22077         (phi_translate_set): Adjust.
22078         (do_pre_regular_insertion): Likewise.
22079         (do_pre_partial_partial_insertion): Likewise.
22081 2018-03-08  Martin Liska  <mliska@suse.cz>
22083         PR gcov-profile/84735
22084         * doc/gcov.texi: Document usage of profile files.
22085         * gcov-io.h: Document changes in the format.
22087 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
22089         PR debug/84404
22090         PR debug/84408
22091         * dwarf2out.c (struct dw_line_info_table): Update comments for
22092         view == -1.
22093         (FORCE_RESET_NEXT_VIEW): New.
22094         (FORCE_RESETTING_VIEW_P): New.
22095         (RESETTING_VIEW_P): Check for -1 too.
22096         (ZERO_VIEW_P): Likewise.
22097         (new_line_info_table): Force-reset next view.
22098         (dwarf2out_begin_function): Likewise.
22099         (dwarf2out_source_line): Simplify zero_view_p initialization.
22100         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
22101         view directly.  Omit view when omitting .loc at line 0.
22103 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
22105         PR tree-optimization/84740
22106         * tree-switch-conversion.c (process_switch): Call build_constructors
22107         only if info.phi_count is non-zero.
22109         PR tree-optimization/84739
22110         * tree-tailcall.c (find_tail_calls): Check call arguments against
22111         DECL_ARGUMENTS (current_function_decl) rather than
22112         DECL_ARGUMENTS (func) when checking for tail recursion.
22114 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
22116         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
22117         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
22118         Volker Reichelt's entry and add entries for people that perform
22119         GCC fuzzy testing and report numerous bugs.
22121 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
22123         PR target/82411
22124         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
22125         readonly data in sdata, if that is disabled.
22126         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
22127         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
22128         -mreadonly-in-sdata option.
22130 2018-03-07  Martin Sebor  <msebor@redhat.com>
22132         PR tree-optimization/84468
22133         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
22134         basic block when looking for nul assignment.
22136 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
22138         PR target/84277
22139         * except.h (output_function_exception_table): Adjust prototype.
22140         * except.c (output_function_exception_table): Remove FNNAME parameter
22141         and add SECTION parameter.  Ouput one part of the table at a time.
22142         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
22143         the first part of the exception table and emit unwind directives.
22144         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
22145         (i386_pe_seh_cold_init): Likewise.
22146         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
22147         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
22148         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
22149         (ix86_output_call_insn): Emit a nop in one more case for SEH.
22150         * config/i386/winnt.c: Include except.h.
22151         (struct seh_frame_state): Add reg_offset, after_prologue and
22152         in_cold_section fields.
22153         (i386_pe_seh_end_prologue): Set seh->after_prologue.
22154         (i386_pe_seh_cold_init): New function.
22155         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
22156         to seh->in_cold_section.
22157         (seh_emit_push): Record the offset of the push.
22158         (seh_emit_save): Record the offet of the save.
22159         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
22160         Test seh->after_prologue to disregard the epilogue.
22161         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
22162         (i386_pe_end_cold_function): New function.
22164 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
22166         PR fortran/84565
22167         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
22168         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
22170         PR c++/84704
22171         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
22172         on tmp_var.
22173         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
22174         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
22176         PR middle-end/84723
22177         * multiple_target.c: Include tree-inline.h and intl.h.
22178         (expand_target_clones): Diagnose and fail if node->definition and
22179         !tree_versionable_function_p (node->decl).
22181 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
22183         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
22184         sprint_ul.
22185         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
22186         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
22187         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
22189 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
22191         PR target/84710
22192         * combine.c (try_combine): Use reg_or_subregno instead of handling
22193         just paradoxical SUBREGs and REGs.
22195 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
22197         * config/arc/arc.c (arc_finalize_pic): Remove function.
22198         (arc_must_save_register): We use single base PIC register, remove
22199         checks to save/restore the PIC register.
22200         (arc_expand_prologue): Likewise.
22201         * config/arc/arc-protos.h (arc_set_default_type_attributes):
22202         Remove.
22203         (arc_verify_short): Likewise.
22204         (arc_attr_type): Likewise.
22205         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
22206         (walk_stores): Likewise.
22207         (arc_address_cost): Make it static.
22208         (arc_verify_short): Likewise.
22209         (branch_dest): Likewise.
22210         (arc_attr_type): Likewise.
22211         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
22212         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
22213         (arc_final_prescan_insn): Remove inserting the nops due to
22214         hardware hazards.  It is done in reorg step.
22215         (insn_length_variant_t): Remove.
22216         (insn_length_parameters_t): Likewise.
22217         (arc_insn_length_parameters): Likewise.
22218         (arc_get_insn_variants): Likewise.
22219         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
22221 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
22223         PR inline-asm/84683
22224         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
22225         assertion failure.
22227         PR tree-optimization/84687
22228         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
22229         on new_node->decl.
22230         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
22232 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22234         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
22235         Rename to ppc_speculation_barrier.
22236         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
22237         __builtin_ppc_speculation_barrier.
22239 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22241         PR target/84700
22242         * combine.c (combine_simplify_rtx): Don't try to simplify if
22243         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
22244         are equal to x.
22246 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
22248         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
22249         to 32 bytes when compiling for POWER9.
22251 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
22253         PR target/84564
22254         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
22255         regparm >= 3 with no arg reg available also for calls with
22256         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
22258         PR target/84524
22259         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
22260         orig,vex.
22261         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
22263 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
22265         PR target/84264
22266         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
22268 2018-03-05  Richard Biener  <rguenther@suse.de>
22270         PR tree-optimization/84486
22271         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
22272         When inserting a __builtin_assume_aligned call set the LHS
22273         SSA name alignment info accordingly.
22275 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
22277         PR tree-optimization/84114
22278         * config/aarch64/aarch64.c (aarch64_reassociation_width)
22279         Avoid reassociation of FLOAT_MODE addition.
22281 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
22283         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
22284         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
22285         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
22286         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
22287         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
22288         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
22289         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
22290         and -mwbnoinvd.
22291         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
22292         __builtin_ia32_wbinvd): New builtins.
22293         (SPECIAL_ARGS2): New.
22294         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
22295         (SPECIAL_ARGS2): New.
22296         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
22297         (ix86_valid_target_attribute_inner_p): Ditto.
22298         (ix86_init_mmx_sse_builtins): Add special_args2.
22299         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
22300         TARGET_WBNOINVD_P): New.
22301         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
22302         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
22303         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
22304         * config/i386/immintrin.h (_wbinvd): New intrinsic.
22305         * config/i386/pconfigintrin.h: New file.
22306         * config/i386/wbnoinvdintrin.h: Ditto.
22307         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
22308         wbnoinvdintrin.h.
22309         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
22311 2018-03-05  Richard Biener  <rguenther@suse.de>
22313         PR tree-optimization/84670
22314         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22315         member.
22316         (BB_VISITED_WITH_VISITED_SUCCS): New define.
22317         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22318         (compute_antic_aux): Only assert the number of values in ANTIC_IN
22319         doesn't grow if all successors (recursively) were visited at least
22320         once.
22322 2018-03-05  Richard Biener  <rguenther@suse.de>
22324         PR tree-optimization/84650
22325         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
22326         if executed in the loop pipeline.
22328 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
22330         * doc/configfiles.texi (Configuration Files): Move info about
22331         conditionalizing $target-protos.h to...
22332         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
22333         differs from $target-protos.h.
22335 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
22336             Chung-Ju Wu  <jasonwucj@gmail.com>
22338         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
22339         * config/nds32/nds32-multiple.md (setmemsi): Define.
22340         * config/nds32/nds32-memory-manipulation.c
22341         (nds32_gen_dup_4_byte_to_word_value): New.
22342         (emit_setmem_word_loop): New.
22343         (emit_setmem_byte_loop): New.
22344         (nds32_expand_setmem_loop): New.
22345         (nds32_expand_setmem_loop_v3m): New.
22346         (nds32_expand_setmem_unroll): New.
22347         (nds32_expand_setmem): New.
22349 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22350             Chung-Ju Wu  <jasonwucj@gmail.com>
22352         * config/nds32/nds32-memory-manipulation.c
22353         (nds32_emit_load_store): New.
22354         (nds32_emit_post_inc_load_store): New.
22355         (nds32_emit_mem_move): New.
22356         (nds32_emit_mem_move_block): New.
22357         (nds32_expand_movmemsi_loop_unknown_size): New.
22358         (nds32_expand_movmemsi_loop_known_size): New.
22359         (nds32_expand_movmemsi_loop): New.
22360         (nds32_expand_movmemsi_unroll): New.
22361         (nds32_expand_movmemqi): Rename ...
22362         (nds32_expand_movmemsi): ... to this.
22363         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
22364         (movmemsi): ... to this.
22365         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
22366         (nds32_expand_movmemsi): ... to this.
22368 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22369             Monk Chiang  <sh.chiang04@gmail.com>
22370             Chung-Ju Wu  <jasonwucj@gmail.com>
22372         * config/nds32/nds32-protos.h
22373         (nds32_expand_load_multiple): New arguments.
22374         (nds32_expand_store_multiple): Ditto.
22375         (nds32_valid_multiple_load_store): Rename ...
22376         (nds32_valid_multiple_load_store_p): ... to this.
22377         * config/nds32/nds32-memory-manipulation.c
22378         (nds32_expand_load_multiple): Refine implementation.
22379         (nds32_expand_store_multiple): Ditto.
22380         * config/nds32/nds32-multiple.md
22381         (load_multiple): Update nds32_expand_load_multiple interface.
22382         (store_multiple): Update nds32_expand_store_multiple interface.
22383         * config/nds32/nds32-predicates.c
22384         (nds32_valid_multiple_load_store): Rename ...
22385         (nds32_valid_multiple_load_store_p): ... to this and refine
22386         implementation.
22387         * config/nds32/predicates.md
22388         (nds32_load_multiple_and_update_address_operation): New predicate.
22389         (nds32_store_multiple_and_update_address_operation): New predicate.
22391 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
22392             Chung-Ju Wu  <jasonwucj@gmail.com>
22394         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
22395         (combo): New attribute.
22396         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
22398 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
22400         * config/nds32/nds32.opt: Change -mcmodel= default value.
22402 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
22403             Monk Chiang  <sh.chiang04@gmail.com>
22404             Chung-Ju Wu  <jasonwucj@gmail.com>
22406         * config/nds32/constants.md (unspec_element): New enum.
22407         * config/nds32/constraints.md (Umw): New constraint.
22408         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
22409         * config/nds32/nds32-intrinsic.md: Likewise.
22410         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
22411         (nds32_valid_smw_lwm_base_p): New.
22412         (nds32_output_smw_single_word): New.
22413         (nds32_output_lmw_single_word): New.
22414         (nds32_expand_unaligned_load): New.
22415         (nds32_expand_unaligned_store): New.
22416         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
22417         (nds32_output_smw_single_word): Declare.
22418         (nds32_output_lmw_single_word): Declare.
22419         (nds32_expand_unaligned_load): Declare.
22420         (nds32_expand_unaligned_store): Declare.
22421         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
22422         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
22423         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
22424         NDS32_BUILTIN_UASTORE_DW.
22425         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
22426         predicate.
22428 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
22429             Kito Cheng  <kito.cheng@gmail.com>
22430             Chung-Ju Wu  <jasonwucj@gmail.com>
22432         * config/nds32/nds32-intrinsic.c
22433         (nds32_expand_builtin_null_ftype_reg): Delete.
22434         (nds32_expand_builtin_reg_ftype_imm): Ditto.
22435         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
22436         (nds32_read_argument): New.
22437         (nds32_legitimize_target): Ditto.
22438         (nds32_legitimize_argument): Ditto.
22439         (nds32_check_constant_argument): Ditto.
22440         (nds32_expand_unop_builtin): Ditto.
22441         (nds32_expand_unopimm_builtin): Ditto.
22442         (nds32_expand_binop_builtin): Ditto.
22443         (nds32_builtin_decl_impl): Ditto.
22444         (builtin_description): Ditto.
22445         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
22446         (nds32_init_builtins_impl): Ditto.
22447         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
22448         (nds32_builtin_decl): New.
22449         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
22450         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
22452 2018-03-02  Jeff Law  <law@redhat.com>
22454         * reorg.c (stop_search_p): Handle DEBUG_INSN.
22455         (redundant_insn, fill_simple_delay_slots): Likewise.
22456         (fill_slots_from_thread): Likewise.
22457         * resource.c (mark_referenced_resources): Likewise.
22458         (mark_set_resources, find_dead_or_set_registers): Likewise.
22460 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22462         * substring-locations.h (format_warning_va): Formatting fix for
22463         ATTRIBUTE_GCC_DIAG.
22464         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
22465         argument.
22466         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
22467         * substring-locations.c: Include intl.h.
22468         (format_warning_va): Turned into small wrapper around
22469         format_warning_n_va, renamed to ...
22470         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
22471         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
22472         use ngettext.
22473         (format_warning_at_substring_n): New function.
22474         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
22475         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
22476         format_warning_at_substring with just a shorter name instead of
22477         const function pointer.
22478         (fmtwarn_n): New function.
22479         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
22480         appropriate, get rid of all the fmtstr temporaries, move conditionals
22481         with G_() wrapped string literals directly into fmtwarn arguments,
22482         cast dir.len to (int), formatting fixes.
22484 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
22486         * doc/invoke.texi: Remove "Cilk Plus" references.
22488 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22489             Richard Biener  <rguenther@suse.de>
22491         PR ipa/84628
22492         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
22493         for error or warning attributes if CALL_FROM_THUNK_P is set.
22494         Formatting fixes.
22496 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22498         PR target/56540
22499         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
22500         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
22502         PR target/56540
22503         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
22504         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
22506         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
22507         instead of -1U in last predictors element's probability member.
22509 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
22511         PR ipa/83983
22512         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22513         arguments if they are comparable.
22515 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22517         PR tree-optimization/84634
22518         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
22519         masks and masked_loop_p with a single loop_masks, making sure it's
22520         null for bb vectorization.
22522 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
22524         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
22525         (vect_analyze_data_ref_access): Use loop->safe_len rather than
22526         loop->force_vectorize to check whether there is no alias.
22528 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
22530         PR target/84614
22531         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
22532         prototypes.
22533         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
22534         comments.
22535         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
22536         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
22537         instead of a loop around prev_real_insn.
22538         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
22539         prev_real_insn.
22541         PR inline-asm/84625
22542         * config/i386/i386.c (ix86_print_operand): Use conditional
22543         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
22544         zero vector.
22546 2018-03-02  Richard Biener  <rguenther@suse.de>
22548         PR tree-optimization/84427
22549         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
22550         (bitmap_set_subtract_values): Rewrite to handle multiple
22551         exprs per value.
22552         (clean): Likewise.
22553         (prune_clobbered_mems): Likewise.
22554         (phi_translate): Take edge instead of pred/phiblock.
22555         (phi_translate_1): Likewise.
22556         (phi_translate_set): Likewise.  Insert all translated
22557         exprs for a value into the set, keeping possibly multiple
22558         expressions per value.
22559         (compute_antic_aux): Adjust for phi_translate changes.
22560         When intersecting union the expressions and prune those
22561         not in the final value set, keeping possibly multiple
22562         expressions per value.  Do not use value-insertion
22563         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
22564         all expressions.  Add verification that the value-sets
22565         only shrink during iteration.
22566         (compute_partial_antic_aux): Adjust for the phi_translate changes.
22567         (do_pre_regular_insertion): Likewise.
22568         (do_pre_partial_partial_insertion): Likewise.
22570 2018-03-02  Richard Biener  <rguenther@suse.de>
22572         PR target/82005
22573         * config/darwin.c (saved_debug_info_level): New static global.
22574         (darwin_asm_lto_start): Disable debug info generation for LTO out.
22575         (darwin_asm_lto_end): Restore debug info generation settings.
22577 2018-03-01  Martin Liska  <mliska@suse.cz>
22579         PR sanitizer/82484
22580         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
22581         volatile arguments.
22583 2018-03-01  Richard Biener  <rguenther@suse.de>
22585         PR debug/84645
22586         * dwarf2out.c (gen_variable_die): Properly handle late VLA
22587         type annotation with LTO when debug was disabled at compile-time.
22589 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
22591         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
22592         XINT with INTVAL.
22593         (mips_final_postscan_insn): Likewise.
22595 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
22597         PR rtl-optimization/84528
22598         * alias.c (init_alias_target): Add commentary.
22599         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
22600         a unique base value if the frame pointer is not eliminated
22601         to the stack pointer.
22603 2018-03-01  Tom de Vries  <tom@codesourcery.com>
22605         PR rtl-optimization/83327
22606         * lra-int.h (hard_regs_spilled_into): Declare.
22607         * lra.c (hard_regs_spilled_into): Define.
22608         (init_reg_info): Init hard_regs_spilled_into.
22609         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22610         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
22611         (process_bb_lives): Handle hard_regs_spilled_into.
22612         (lra_create_live_ranges_1): Before doing liveness propagation, clear
22613         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
22615 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
22617         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22618         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22619         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22620         * config/rs6000/aix72.h: New file.
22622 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
22624         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22625         instead of warning_at with conditional singular and plural messages
22626         where possible.
22628         PR target/52991
22629         * stor-layout.c (update_alignment_for_field): For
22630         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22631         && !DECL_PACKED (field), do the alignment update, just use
22632         only desired_align instead of MAX (type_align, desired_align)
22633         as the alignment.
22634         (place_field): Don't do known_align < desired_align handling
22635         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22636         is non-NULL, instead do it after rli->prev_field handling and
22637         only if not within a bitfield word.  For DECL_PACKED (field)
22638         use type_align of BITS_PER_UNIT.
22640 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
22642         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22643         superfluous parentheses and trailing spaces.
22645 2018-02-28  Richard Biener  <rguenther@suse.de>
22647         PR tree-optimization/84584
22648         * graphite-scop-detection.c (scop_detection::add_scop): Discard
22649         SCoPs with fake exit edge.
22651 2018-02-28  Martin Liska  <mliska@suse.cz>
22653         PR testsuite/84597
22654         * timevar.c (timer::print): Fix format to properly print 100%
22655         values.
22657 2018-02-28  Richard Biener  <rguenther@suse.de>
22659         PR middle-end/84607
22660         * genmatch.c (capture_info::walk_match): Do not mark
22661         captured expressions without operands as expr_p given
22662         they act more like predicates and should be subject to
22663         "lost tail" side-effect preserving.
22665 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
22667         PR rtl-optimization/81611
22668         * auto-inc-dec.c (attempt_change): Move dead note from
22669         mem_insn if it's the next use of regno
22670         (find_address): Take address use of reg holding
22671         non-incremented value.  Add parm to limit search to the named
22672         reg only.
22673         (merge_in_block): Attempt to use a mem insn that is the next
22674         use of the original regno.
22676 2018-02-27  Martin Sebor  <msebor@redhat.com>
22678         PR c++/83871
22679         * doc/invoke.texi (-Wmissing-attributes): New option.
22680         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22682 2018-02-27  Martin Sebor  <msebor@redhat.com>
22684         PR translation/84207
22685         * diagnostic-core.h (warning_n, error_n, inform_n): Change
22686         n argument to unsigned HOST_WIDE_INT.
22687         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22688         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
22689         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22690         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22692 2018-02-27  Richard Biener  <rguenther@suse.de>
22694         PR tree-optimization/84512
22695         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22696         Do not use the estimate returned from record_stmt_cost for
22697         the scalar iteration cost but sum properly using add_stmt_cost.
22699 2018-02-27  Richard Biener  <rguenther@suse.de>
22701         PR tree-optimization/84466
22702         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22703         Adjust last change to less strictly validate use operands.
22705 2018-02-27  Martin Liska  <mliska@suse.cz>
22707         PR gcov-profile/84548
22708         * gcov.c (process_file): Allow partial overlap and consider it
22709         also as group functions.
22710         (output_lines): Properly calculate range of lines for a group.
22712 2018-02-27  Martin Liska  <mliska@suse.cz>
22714         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22715         'ggc' suffixes.  Change first column width.
22716         (timer::print): Fix formatting of the column.
22718 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
22720         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22721         preserve inline entry blocks for the sake of debug inline
22722         entry point markers alone.
22723         (remove_unused_locals): Suggest in comments a better place to
22724         force the preservation of inline entry blocks that are
22725         otherwise unused, but do not preserve them.
22727 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22729         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22731 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22733         PR target/84039
22734         * config/i386/constraints.md (Bs): Replace
22735         ix86_indirect_branch_register with
22736         TARGET_INDIRECT_BRANCH_REGISTER.
22737         (Bw): Likewise.
22738         * config/i386/i386.md (indirect_jump): Likewise.
22739         (tablejump): Likewise.
22740         (*sibcall_memory): Likewise.
22741         (*sibcall_value_memory): Likewise.
22742         Peepholes of indirect call and jump via memory: Likewise.
22743         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22744         (*sibcall_value_GOT_32): Likewise.
22745         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22746         (GOT_memory_operand): Likewise.
22747         (call_insn_operand): Likewise.
22748         (sibcall_insn_operand): Likewise.
22749         (GOT32_symbol_operand): Likewise.
22750         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22752 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22754         PR rtl-optimization/83496
22755         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22756         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
22757         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22758         redundant insn, if any.
22759         (relax_delay_slots): Likewise.
22760         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22762 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
22764         PR tree-optimization/83965
22765         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22766         that grouped statements are part of a reduction chain.  Return
22767         true if the statement is not marked as a reduction itself but
22768         is part of a group.
22769         (vect_recog_dot_prod_pattern): Don't check whether the statement
22770         is part of a group here.
22771         (vect_recog_sad_pattern): Likewise.
22772         (vect_recog_widen_sum_pattern): Likewise.
22774 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22776         PR debug/84545
22777         * final.c (rest_of_clean_state): Also look for calls inside sequences.
22779 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22781         PR target/84530
22782         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22783         the bool argument.
22784         (ix86_output_indirect_function_return): New prototype.
22785         (ix86_split_simple_return_pop_internal): Likewise.
22786         * config/i386/i386.c (indirect_return_via_cx): New.
22787         (indirect_return_via_cx_bnd): Likewise.
22788         (indirect_thunk_name): Handle return va CX_REG.
22789         (output_indirect_thunk_function): Create alias for
22790         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22791         (ix86_output_indirect_jmp): Remove the bool argument.
22792         (ix86_output_indirect_function_return): New function.
22793         (ix86_split_simple_return_pop_internal): Likewise.
22794         * config/i386/i386.md (*indirect_jump): Don't pass false
22795         to ix86_output_indirect_jmp.
22796         (*tablejump_1): Likewise.
22797         (simple_return_pop_internal): Change it to define_insn_and_split.
22798         Call ix86_split_simple_return_pop_internal to split it for
22799         -mfunction-return=.
22800         (simple_return_indirect_internal): Call
22801         ix86_output_indirect_function_return instead of
22802         ix86_output_indirect_jmp.
22804 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
22806         PR bootstrap/84405
22807         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22808         memset and value initialization afterwards.
22810 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
22812         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22814 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22816         PR target/84521
22817         * common/config/aarch64/aarch64-common.c
22818         (aarch_option_optimization_table[]): Switch
22819         off fomit-frame-pointer
22821 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22822             Chung-Ju Wu  <jasonwucj@gmail.com>
22824         * config/nds32/nds32-multiple.md (load_multiple): Disallow
22825         volatile memory.
22826         (store_multiple): Ditto.
22828 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22830         * config.gcc: Add --with-cpu support for nds32 target.
22831         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22832         * config/nds32/nds32.opt: Add -mcpu= option.
22834 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
22836         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22837         isel=yes): Warn for these deprecated options.
22839 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
22841         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22842         ISA_2_5_MASKS_EMBEDDED.
22844 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
22846         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22847         p->max as pointers rather than using iterative_hash_expr.
22849 2018-02-23  Carl Love  <cel@us.ibm.com>
22851         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22852         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22853         BU_P8V_OVERLOAD_2.
22854         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22855         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22856         P8V_BUILTIN_VEC_VUNSIGNED2.
22858 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
22860         PR target/81572
22861         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22862         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22863         LRA_UNKNOWN_ALT.
22864         * lra-constraints.c (curr_insn_transform): Set up
22865         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
22866         LRA_UNKNOWN_ALT.
22867         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22868         * lra-eliminations.c (spill_pseudos): Ditto.
22869         (process_insn_for_elimination): Ditto.
22870         * lra-lives.c (reg_early_clobber_p): Use the new macros.
22871         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22872         LRA_NON_CLOBBERED_ALT.
22874 2018-02-22  Martin Sebor  <msebor@redhat.com>
22876         PR tree-optimization/84480
22877         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22878         to maybe_diag_stxncpy_trunc.  Call it.
22879         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22880         from gimple_fold_builtin_strcpy.  Print inlining stack.
22881         (handle_builtin_stxncpy): Print inlining stack.
22882         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22884 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
22886         PR target/84176
22887         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22888         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22889         and -fcheck-pointer-bounds are used together.
22890         (indirect_thunk_prefix): New enum.
22891         (indirect_thunk_need_prefix): New function.
22892         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
22893         "_nt" instead of "_bnd" for NOTRACK prefix.
22894         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22895         (output_indirect_thunk_function): Likewise.
22896         (): Likewise.
22897         (ix86_code_end): Update output_indirect_thunk_function calls.
22898         (ix86_output_indirect_branch_via_reg): Replace
22899         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22900         (ix86_output_indirect_branch_via_push): Likewise.
22901         (ix86_output_function_return): Likewise.
22902         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22903         incompatible with -fcf-protection=branch and
22904         -fcheck-pointer-bounds.
22906 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22908         PR target/83335
22909         * config/aarch64/aarch64.c (aarch64_print_address_internal):
22910         Change gcc_assert call to output_operand_lossage.
22912 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22914         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22916 2018-02-22  DJ Delorie  <dj@redhat.com>
22917             Sebastian Perta  <sebastian.perta@renesas.com>
22918             Oleg Endo  <olegendo@gcc.gnu.org>
22920         * config/rx/rx.c (rx_rtx_costs): New function.
22921         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22923 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22925         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22927 2018-02-22  Martin Liska  <mliska@suse.cz>
22929         PR driver/83193
22930         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22931         Add "native" as a possible value.
22933 2018-02-22  Martin Liska  <mliska@suse.cz>
22935         PR driver/83193
22936         * config/i386/i386.c (ix86_option_override_internal):
22937         Add "native" as a possible value for -march and -mtune.
22939 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
22941         PR target/84502
22942         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22943         to all type variants.
22945         PR tree-optimization/84503
22946         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22947         width as info->bitpos + info->bitsize - start.
22948         (merged_store_group::merge_overlapping): Simplify width computation.
22949         (check_no_overlap): New function.
22950         (imm_store_chain_info::try_coalesce_bswap): Compute expected
22951         start + width and last_order of the group, fail if check_no_overlap
22952         fails.
22953         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22954         to group if check_no_overlap fails.
22956 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22958         * config/rs6000/altivec.md: Delete contraint arguments to
22959         define_expand, define_split, and define_peephole2, and in
22960         define_insn_and_split if always unused.
22961         * config/rs6000/darwin.md: Ditto.
22962         * config/rs6000/dfp.md: Ditto.
22963         * config/rs6000/rs6000.md: Ditto.
22964         * config/rs6000/sync.md: Ditto.
22965         * config/rs6000/vector.md: Ditto.
22966         * config/rs6000/vsx.md: Ditto.
22968 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22970         * config/rs6000/altivec.md: Write output control strings as braced
22971         blocks instead of double-quoted strings.
22972         * config/rs6000/darwin.md: Ditto.
22973         * config/rs6000/rs6000.md: Ditto.
22974         * config/rs6000/vector.md: Ditto.
22975         * config/rs6000/vsx.md: Ditto.
22977 2018-02-21  Jason Merrill  <jason@redhat.com>
22979         PR c++/84314 - ICE with templates and fastcall attribute.
22980         * attribs.c (build_type_attribute_qual_variant): Remove assert.
22982 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22984         * ipa-cp.c (determine_versionability): Fix comment typos.
22986 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22988         PR c/84229
22989         * ipa-cp.c (determine_versionability): Do not version functions caling
22990         va_arg_pack.
22992 2018-02-21  Martin Liska  <mliska@suse.cz>
22994         PR driver/83193
22995         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22996         Add "native" as a possible value.
22997         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
22998         the macro when native cpu detection is available.
23000 2018-02-21  Martin Liska  <mliska@suse.cz>
23002         PR driver/83193
23003         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
23004         Add "native" as a possible value.
23005         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
23006         when native cpu detection is available.
23008 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
23009             Martin Sebor  <msebor@redhat.com>
23011         PR tree-optimization/84478
23012         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
23013         false.
23014         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
23015         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
23016         support which is conservatively correct, for 2 only stay conservative
23017         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
23018         argument to the 2 argument get_range_strlen, adjust 6 arg
23019         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
23020         false.
23021         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
23022         (gimple_fold_builtin_strlen): Pass true as last argument to
23023         get_range_strlen.
23025 2018-02-20  Martin Sebor  <msebor@redhat.com>
23027         PR middle-end/84095
23028         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
23029         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
23030         (builtin_memref::builtin_memref): Factor out parts into
23031         set_base_and_offset and call it.
23033 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
23035         PR middle-end/84406
23036         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
23037         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
23038         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
23039         search at the associated MODE_INT.
23041 2018-02-20  Jeff Law  <law@redhat.com>
23043         PR middle-end/82123
23044         PR tree-optimization/81592
23045         PR middle-end/79257
23046         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
23047         for range data rather than using global data.
23048         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
23049         range data rather than using global data.
23050         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
23051         pass it to children as needed.
23052         (struct directive::fmtresult): Similarly.
23053         (struct directive::set_width): Similarly.
23054         (struct directive::set_precision): Similarly.
23055         (format_integer, format_directive, parse_directive): Similarly.
23056         (format_none): Accept unnamed vr_values parameter.
23057         (format_percent, format_floating, format_character): Similarly.
23058         (format_string, format_plain): Similarly.
23059         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
23060         the EVRP range analyzer for range data rather than using global data.
23061         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
23062         gimple-ssa-evrp-analyze.h
23063         (class sprintf_dom_walker): Add after_dom_children member function.
23064         Add evrp_range_analyzer member.
23065         (sprintf_dom_walker::before_dom_children): Call into the EVRP
23066         range analyzer as needed.
23067         (sprintf_dom_walker::after_dom_children): New member function.
23068         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
23069         if not optimizing.
23070         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
23071         (evrp_range_analyzer::pop_to_marker): Likewise.
23073 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
23075         PR tree-optimization/84419
23076         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
23077         with the required type if its current type is compatible but
23078         different.
23080 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
23082         PR middle-end/82004
23083         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
23084         after vectorization.
23086 2018-02-20  Martin Liska  <mliska@suse.cz>
23088         PR driver/83193
23089         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
23090         possible values if we don't have a hint.
23092 2018-02-20  Martin Liska  <mliska@suse.cz>
23094         PR c/84310
23095         PR target/79747
23096         * final.c (shorten_branches): Build align_tab array with one
23097         more element.
23098         * opts.c (finish_options): Add alignment option limit check.
23099         (MAX_CODE_ALIGN): Likewise.
23100         (MAX_CODE_ALIGN_VALUE): Likewise.
23101         * doc/invoke.texi: Document maximum allowed option value for
23102         all -falign-* options.
23104 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
23106         PR target/84146
23107         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
23108         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
23109         * var-tracking.c (emit_note_insn_var_location): Remove all references
23110         to NOTE_INSN_CALL_ARG_LOCATION.
23111         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
23112         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
23113         Use copy_rtx_if_shared.
23114         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
23115         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
23116         (dwarf2out_var_location): Remove handling of
23117         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
23118         on call_insn.
23119         * final.c (final_scan_insn): Remove all references to
23120         NOTE_INSN_CALL_ARG_LOCATION.
23121         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
23122         before dumping final insns.
23123         * except.c (emit_note_eh_region_end): Remove all references to
23124         NOTE_INSN_CALL_ARG_LOCATION.
23125         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
23126         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
23127         * config/arc/arc.c (hwloop_optimize): Likewise.
23128         * config/arm/arm.c (create_fix_barrier): Likewise.
23129         * config/s390/s390.c (s390_chunkify_start): Likewise.
23130         * config/sh/sh.c (find_barrier): Likewise.
23131         * config/i386/i386.c (rest_of_insert_endbranch,
23132         ix86_seh_fixup_eh_fallthru): Likewise.
23133         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
23134         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
23135         * config/frv/frv.c (frv_function_prologue): Likewise.
23136         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
23137         reg note.
23138         (note_outside_basic_block_p): Remove all references to
23139         NOTE_INSN_CALL_ARG_LOCATION.
23140         * gengtype.c (adjust_field_rtx_def): Likewise.
23141         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
23142         Likewise.
23143         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
23144         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
23146         PR c++/84444
23147         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
23148         is ADDR_EXPR.
23150         PR tree-optimization/84452
23151         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
23152         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
23153         is NULL.
23155 2018-02-19  Martin Liska  <mliska@suse.cz>
23157         PR sanitizer/82183
23158         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
23160 2018-02-19  Martin Liska  <mliska@suse.cz>
23161             Richard Sandiford  <richard.sandiford@linaro.org>
23163         PR tree-optimization/82491
23164         * gimple-fold.c (get_base_constructor): Make earlier bail out
23165         to prevent ubsan.
23167 2018-02-19  Carl Love  <cel@us.ibm.com>
23169         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
23170         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
23171         BU_P8V_OVERLOAD_1.
23172         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
23173         P8V_BUILTIN_VEC_NEG.
23175 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
23177         * config/rl78/rl78.md (movdf): New define expand.
23179 2018-02-19  Martin Liska  <mliska@suse.cz>
23181         PR other/80589
23182         * doc/invoke.texi: Fix typo.
23183         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
23185 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
23187         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
23188         handle rs6000_single_float and rs6000_double_float specially for
23189         e500 family CPUs.
23191 2018-02-16  Jeff Law  <law@redhat.com>
23193         * config/rx/rx.c (add_pop_cfi_notes): New function.;
23194         (pop_regs): Use it.
23196 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
23198         PR ipa/84425
23199         * ipa-inline.c (inline_small_functions): Fix a typo.
23201 2018-02-16  Nathan Sidwell  <nathan@acm.org>
23203         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
23205 2018-02-16  Carl Love  <cel@us.ibm.com>
23207         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
23208         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
23209         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
23210         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
23211         expansion to P8V_BUILTIN_VEC_FLOAT2.
23213 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
23215         PR rtl-optimization/70023
23216         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
23217         src_regno into account.
23219 2018-02-16  Carl Love  <cel@us.ibm.com>
23221         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
23222         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
23223         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
23224         * config/rs6000/rs6000.c: Remove case statements for
23225         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
23226         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
23227         and P9V_BUILTIN_VEC_VINSERT4B.
23228         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
23229         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
23230         * config/rs6000/vsx.md:
23231         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
23232         vec_insert4b.
23234 2018-02-16  Carl Love  <cel@us.ibm.com>
23236         * config/rs6000/altivec.h: Add builtin names vec_extract4b
23237         vec_insert4b.
23238         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
23239         definitions.
23240         * config/rs6000/rs6000-c.c: Add the definitions for
23241         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
23242         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
23243         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
23244         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
23245         definition for insert4b and define insn *insert3b_internal.
23246         * doc/extend.texi: Add documentation for vec_extract4b.
23248 2018-02-16  Nathan Sidwell  <nathan@acm.org>
23250         * doc/extend.texi (Backwards Compatibility): Mention friend
23251         injection.  Note for-scope is deprecated.
23252         * doc/invoke.texi (-ffriend-injection): Deprecate.
23254 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
23256         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
23257         that moved to I2, also allow destinations that are a paradoxical
23258         subreg (instead of a normal reg).
23260 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
23262         PR target/83831
23263         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
23264         to QImode.
23266 2018-02-16  Richard Biener  <rguenther@suse.de>
23268         PR tree-optimization/84037
23269         PR tree-optimization/84016
23270         PR target/82862
23271         * config/i386/i386.c (ix86_builtin_vectorization_cost):
23272         Adjust vec_construct for the fact we need additional higher latency
23273         128bit inserts for AVX256 and AVX512 vector builds.
23274         (ix86_add_stmt_cost): Scale vector construction cost for
23275         elementwise loads.
23277 2018-02-16  Richard Biener  <rguenther@suse.de>
23279         PR tree-optimization/84417
23280         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
23281         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
23282         (non_rewritable_lvalue_p): Likewise, use poly-ints.
23284 2018-02-16  Martin Liska  <mliska@suse.cz>
23286         PR sanitizer/84307
23287         * internal-fn.def (ASAN_CHECK): Set proper flags.
23288         (ASAN_MARK): Likewise.
23290 2018-02-16  Julia Koval  <julia.koval@intel.com>
23292         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
23293         from PTA_CANNONLAKE.
23295 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
23297         PR target/84272
23298         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
23299         Use ++iter rather than iter++ for std::list iterators.
23300         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
23301         defer deleting them until all nodes in the forest are processed.  Do
23302         free even leaf nodes.  Change to_process into auto_vec.
23304         PR bootstrap/84405
23305         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
23306         * vec.h (vec_default_construct): Use memset instead of placement new
23307         if BROKEN_VALUE_INITIALIZATION is defined.
23308         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
23309         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
23310         is defined.
23312         PR rtl-optimization/83723
23313         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
23314         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
23315         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
23316         recursive calls.
23317         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
23318         callers.
23319         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
23321 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
23323         PR rtl-optimization/81443
23324         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
23325         from inner REGs to paradoxical SUBREGs.
23327 2018-02-16  Richard Biener  <rguenther@suse.de>
23329         PR tree-optimization/84399
23330         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23331         For operands we can analyze at their definition make sure we can
23332         analyze them at each use as well.
23334 2018-02-16  Richard Biener  <rguenther@suse.de>
23336         PR tree-optimization/84190
23337         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
23338         volatile accesses if the decl isn't volatile.
23340 2018-02-15  Jason Merrill  <jason@redhat.com>
23342         PR c++/84314 - ICE with templates and fastcall attribute.
23343         * attribs.c (build_type_attribute_qual_variant): Don't clobber
23344         TYPE_CANONICAL on an existing type.
23346 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
23348         PR tree-optimization/84383
23349         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
23350         dstoff nor call operand_equal_p if dstbase is NULL.
23352         PR tree-optimization/84334
23353         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
23354         also a CONSTANT_CLASS_P, punt.
23356 2018-02-14  Jim Wilson  <jimw@sifive.com>
23358         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
23359         first SMALL_OPERAND check.  New local min_second_step.  Move assert
23360         to where locals are set.  Add TARGET_RVC support.
23361         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
23363 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
23365         * doc/invoke.texi: Correct -Wformat-overflow code sample.
23367 2018-02-14  Martin Sebor  <msebor@redhat.com>
23369         PR tree-optimization/83698
23370         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
23371         arrays constrain the offset range to their bounds.
23372         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
23373         (builtin_access::overlap): Avoid setting the size of overlap if it's
23374         already been set.
23375         (maybe_diag_overlap): Also consider arrays when deciding what values
23376         of offsets to include in diagnostics.
23378 2018-02-14  Martin Sebor  <msebor@redhat.com>
23380         PR c/84108
23381         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
23382         that correspond to the kind of a declaration.
23384 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
23386         PR target/83984
23387         * config/pa/pa.md: Load address of PIC label using the linkage table
23388         if the label is nonlocal.
23390 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23392         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
23393         warning message if user requests -maltivec=be.
23394         * doc/invoke.texi: Document deprecation of -maltivec=be.
23396 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
23398         PR target/84220
23399         * config/rs6000/rs6000-c.c: Update definitions for
23400         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
23401         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
23403 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23405         PR target/84239
23406         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
23407         add _get_ssp intrinsics. Remove argument from
23408         __builtin_ia32_rdssp[d|q].
23409         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
23410         * config/i386/i386-builtin.def: Remove argument from
23411         __builtin_ia32_rdssp[d|q].
23412         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
23413         ix86_expand_special_args_builtin for _rdssp[d|q].
23414         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
23415         Clear register before usage.
23416         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
23417         Add documentation for new _get_ssp and _inc_ssp intrinsics.
23419 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
23421         PR tree-optimization/84357
23422         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
23423         operand 1 of an ARRAY_REF too.
23425 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
23427         PR target/83831
23428         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
23429         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23430         declarations.
23431         (set_of_reg): New struct.
23432         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
23433         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
23434         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
23435         functions.
23436         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
23437         Split into bitclr, bitset, bitinvert patterns if appropriate.
23438         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
23439         use rx_fuse_in_memory_bitop.
23440         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
23441         to named insn, correct maximum insn length.
23443 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
23445         PR target/79242
23446         * machmode.def: Define a complex mode for PARTIAL_INT.
23447         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
23448         MODE_PARTIAL_INT.
23449         * doc/rtl.texi: Document CSPImode.
23450         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
23451         handling.
23452         (msp430_hard_regno_nregs_with_padding): Likewise.
23454 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
23456         PR target/84279
23457         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
23459 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
23461         PR rtl-optimization/84169
23462         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
23463         we generated a parallel as new i3 and we split that to new i2 and i3
23464         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
23465         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
23466         those to i2, not i1.  Partially rewrite this scan code.
23468 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23470         PR c/82210
23471         * stor-layout.c (place_field): For variable length fields, adjust
23472         offset_align afterwards not just based on the field's alignment,
23473         but also on the size.
23475         PR middle-end/84309
23476         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
23477         of exps and logs in the use_exp2 case.
23479 2018-02-13  Jeff Law  <law@redhat.com>
23481         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
23482         entry for "vector".
23484         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
23485         ARGS as unused.
23487 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
23489         PR debug/84342
23490         PR debug/84319
23491         * common.opt (gas-loc-support, gas-locview-support): New.
23492         (ginline-points, ginternal-reset-location-views): New.
23493         * doc/invoke.texi: Document them.  Use @itemx where intended.
23494         (gvariable-location-views): Adjust.
23495         * target.def (reset_location_view): New.
23496         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
23497         (TARGET_RESET_LOCATION_VIEW): New.
23498         * doc/tm.texi: Rebuilt.
23499         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
23500         (dwarf2out_default_as_locview_support): New.
23501         (output_asm_line_debug_info): Use option variables.
23502         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
23503         (output_loc_list): Likewise.
23504         (add_high_low_attributes): Check option variables.
23505         Don't output entry view attribute in strict mode.
23506         (gen_inlined_subroutine_die): Check option variables.
23507         (dwarf2out_inline_entry): Likewise.
23508         (init_sections_and_labels): Likewise.
23509         (dwarf2out_early_finish): Likewise.
23510         (maybe_reset_location_view): New, from...
23511         (dwarf2out_var_location): ... here.  Call it.
23512         * debug.h (dwarf2out_default_as_loc_support): Declare.
23513         (dwarf2out_default_as_locview_support): Declare.
23514         * hooks.c (hook_int_rtx_insn_0): New.
23515         * hooks.h (hook_int_rtx_insn_0): Declare.
23516         * toplev.c (process_options): Take -gas-loc-support and
23517         -gas-locview-support from dwarf2out.  Enable
23518         -gvariable-location-views by default only with locview
23519         assembler support.  Enable -ginternal-reset-location-views by
23520         default only if the target defines the corresponding hook.
23521         Enable -ginline-points by default if location views are
23522         enabled; force it disabled if statement frontiers are
23523         disabled.
23524         * tree-inline.c (expand_call_inline): Check option variables.
23525         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
23527 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
23529         PR tree-optimization/84321
23530         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
23531         handling.  Also check whether the anti-range contains any values
23532         that satisfy the mask; switch to a VR_RANGE if not.
23534 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
23536         PR sanitizer/84340
23537         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
23539 2018-02-13  Martin Jambor  <mjambor@suse.cz>
23541         PR c++/83990
23542         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
23543         of call statements, also set location of a load to a temporary.
23545 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
23547         * config/rl78/rl78.c (add_vector_labels): New function.
23548         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
23549         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
23550         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
23551         which checks that no arguments are passed.
23552         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
23553         * doc/extend.texi: Documentation for the new attribute.
23555 2018-02-13  Andreas Schwab  <schwab@suse.de>
23557         * config/riscv/linux.h (CPP_SPEC): Define.
23559 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23561         PR target/84335
23562         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
23563         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
23564         OPTION_MASK_ISA_AES as first argument to def_builtin_const
23565         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
23566         instead of OPTION_MASK_ISA_PCLMUL as first argument to
23567         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
23568         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
23569         temporarily for AES and PCLMUL builtins.
23571         PR tree-optimization/84339
23572         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
23573         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
23574         Formatting fixes.
23576         PR middle-end/84309
23577         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
23578         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
23579         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
23580         inline function.
23581         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
23582         inline function.
23583         * omp-simd-clone.h: New file.
23584         * omp-simd-clone.c: Include omp-simd-clone.h.
23585         (expand_simd_clones): No longer static.
23586         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
23587         cgraph.h and omp-simd-clone.h.
23588         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
23589         (vect_recog_widen_shift_pattern): Formatting fix.
23590         (vect_pattern_recog_1): Don't check optab for calls.
23592         PR target/84336
23593         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
23594         operands[2] into a REG before using gen_lowpart on it.
23596 2018-02-12  Jeff Law  <law@redhat.com>
23598         PR target/83760
23599         * config/sh/sh.c (find_barrier): Consider a sibling call
23600         a barrier as well.
23602         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
23603         successfully back substituting a reg.
23605 2018-02-12  Richard Biener  <rguenther@suse.de>
23607         PR tree-optimization/84037
23608         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
23609         parameter, move visited init to caller.
23610         (vect_slp_analyze_operations): Separate cost from validity
23611         check, initialize visited once for all instances.
23612         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
23613         for all instances.
23614         * tree-vect-stmts.c (vect_model_simple_cost): Make early
23615         out an assert.
23616         (vect_model_promotion_demotion_cost): Likewise.
23617         (vectorizable_bswap): Guard cost modeling with !slp_node
23618         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23619         SLP stmts.
23620         (vectorizable_call): Likewise.
23621         (vectorizable_conversion): Likewise.
23622         (vectorizable_assignment): Likewise.
23623         (vectorizable_shift): Likewise.
23624         (vectorizable_operation): Likewise.
23625         (vectorizable_store): Likewise.
23626         (vectorizable_load): Likewise.
23627         (vectorizable_condition): Likewise.
23628         (vectorizable_comparison): Likewise.
23630 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
23632         PR sanitizer/84307
23633         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23634         (ASAN_MARK): Fix fnspec to account for return value, change pointer
23635         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23637 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23639         PR middle-end/83665
23640         * params.def (inline-min-speedup): Increase from 8 to 15.
23641         (max-inline-insns-auto): Decrease from 40 to 30.
23642         * ipa-split.c (consider_split): Add some buffer for function to
23643         be considered inlining candidate.
23644         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23645         default values.
23647 2018-02-12  Richard Biener  <rguenther@suse.de>
23649         PR tree-optimization/84037
23650         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23651         matched stmts if we cannot swap the non-matched ones.
23653 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23655         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23656         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23657         _mm_maskz_scalef_round_ss): New intrinsics.
23658         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23659         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23660         __builtin_ia32_scalefss_round): Remove.
23661         (__builtin_ia32_scalefsd_mask_round,
23662         __builtin_ia32_scalefss_mask_round): New intrinsics.
23663         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23664         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23665         ((match_operand:VF_128 2 "<round_nimm_predicate>"
23666         "<round_constraint>")): Changed to ...
23667         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23668         "<round_scalar_constraint>")): ... this.
23669         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23670         %0, %1, %2<round_op3>}"): Changed to ...
23671         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23672         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23673         %2<round_scalar_mask_op3>}"): ... this.
23674         * config/i386/subst.md (round_scalar_nimm_predicate): New.
23676 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23678         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23679         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23680         (_mm_maskz_sqrt_round_ss): New intrinsics.
23681         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23682         (__builtin_ia32_sqrtsd_mask_round)
23683         (__builtin_ia32_sqrtss_mask_round): New builtins.
23684         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23685         (__builtin_ia32_sqrtss_round): Remove.
23686         (__builtin_ia32_sqrtsd_mask_round)
23687         (__builtin_ia32_sqrtss_mask_round): New builtins.
23688         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23689         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23690         ((match_operand:VF_128 1 "vector_operand"
23691         "xBm,<round_constraint>")): Changed to ...
23692         ((match_operand:VF_128 1 "vector_operand"
23693         "xBm,<round_scalar_constraint>")): ... this.
23694         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23695         %0, %2, %<iptr>1<round_op3>}): Changed to ...
23696         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23697         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23698         %<iptr>1<round_scalar_mask_op3>}): ... this.
23699         ((set_attr "prefix" "<round_prefix>")): Changed to ...
23700         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23702 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
23704         PR target/84266
23705         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23706         Cast vec_cmpeq result to correct type.
23707         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23708         Cast vec_cmpgt result to correct type.
23710 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
23712         * final.c (final_scan_insn_1): Renamed from...
23713         (final_scan_insn): ... this.  New wrapper, to recover
23714         seen from the outermost call in recursive ones.
23715         * config/sparc/sparc.c (output_return): Drop seen from call.
23716         (output_sibcall): Likewise.
23717         * config/visium/visium.c (output_branch): Likewise.
23719 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
23721         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23722         function label.
23724 2018-02-10  Alan Modra  <amodra@gmail.com>
23726         PR target/84300
23727         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23728         Specify LR as an input.
23730 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
23732         PR sanitizer/83987
23733         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23734         remove_member_access_dummy_vars): New functions.
23735         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23736         lower_omp_1, execute_lower_omp): Use them.
23738         PR rtl-optimization/84308
23739         * shrink-wrap.c (spread_components): Release todo vector.
23741 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
23743         PR rtl-optimization/57193
23744         * ira-color.c (struct allocno_color_data): Add member
23745         conflict_allocno_hard_prefs.
23746         (update_conflict_allocno_hard_prefs): New.
23747         (bucket_allocno_compare_func): Add a preference based on
23748         conflict_allocno_hard_prefs.
23749         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23750         (color_allocnos): Remove a dead code.  Initiate
23751         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
23753 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23755         PR target/84226
23756         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23757         constraint from =wa to wa.  Avoid a subreg on the output operand,
23758         instead use a pseudo and subreg it in a move.
23759         (p9_xxbrd_<mode>): Changed to ...
23760         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23761         (p9_xxbrd_v2df): New expander.
23762         (p9_xxbrw_<mode>): Changed to ...
23763         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23764         (p9_xxbrw_v4sf): New expander.
23766 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23768         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23770 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
23772         PR target/83926
23773         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23774         multiply in 32-bit mode.
23775         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23776         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23777         mode.
23779 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23781         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23782         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23783         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23784         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23786 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
23788         PR lto/84213
23789         * dwarf2out.c (is_trivial_indirect_ref): New function.
23790         (dwarf2out_late_global_decl): Do not generate a location
23791         attribute for variables that have a non-trivial DECL_VALUE_EXPR
23792         and that are not defined in the current unit.
23794 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23796         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23797         instead of a libcall for UNORDERED.
23799 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
23801         PR target/82641
23802         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23803         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23805 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23807         PR target/PR84295
23808         * config/s390/s390.c (s390_set_current_function): Invoke
23809         s390_indirect_branch_settings also if fndecl didn't change.
23811 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23813         * config/rs6000/rs6000.md (blockage): Set length to zero.
23815 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23817         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23819 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23821         PR sanitizer/84285
23822         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23823         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23824         -static-lib*san.
23826         PR debug/84252
23827         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23828         PARALLEL incoming that failed vt_get_decl_and_offset check.
23830         PR middle-end/84237
23831         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23832         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23833         TREE_READONLY bit.
23834         (get_variable_section): For decls in named .bss* sections pass true as
23835         second argument to bss_initializer_p.
23837 2018-02-09  Marek Polacek  <polacek@redhat.com>
23838             Jakub Jelinek  <jakub@redhat.com>
23840         PR c++/83659
23841         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23842         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
23843         Sync some changes from cxx_fold_indirect_ref.
23845 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23847         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23848         markers.
23849         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23850         (BLOCK_INLINE_ENTRY_LABEL): New.
23851         (dwarf2out_var_location): Disregard inline entry markers.
23852         (inline_entry_data): New struct.
23853         (inline_entry_data_hasher): New hashtable type.
23854         (inline_entry_data_hasher::hash): New.
23855         (inline_entry_data_hasher::equal): New.
23856         (inline_entry_data_table): New variable.
23857         (add_high_low_attributes): Add DW_AT_entry_pc and
23858         DW_AT_GNU_entry_view attributes if a pending entry is found
23859         in inline_entry_data_table.  Add old entry_pc attribute only
23860         if debug nonbinding markers are disabled.
23861         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23862         markers are enabled.
23863         (block_within_block_p, dwarf2out_inline_entry): New.
23864         (dwarf2out_finish): Check that no entries remained in
23865         inline_entry_data_table.
23866         * final.c (reemit_insn_block_notes): Handle inline entry notes.
23867         (final_scan_insn, notice_source_line): Likewise.
23868         (rest_of_clean_state): Skip inline entry markers.
23869         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23870         markers.
23871         * gimple.c (gimple_build_debug_inline_entry): New.
23872         * gimple.h (enum gimple_debug_subcode): Add
23873         GIMPLE_DEBUG_INLINE_ENTRY.
23874         (gimple_build_debug_inline_entry): Declare.
23875         (gimple_debug_inline_entry_p): New.
23876         (gimple_debug_nonbind_marker_p): Adjust.
23877         * insn-notes.def (INLINE_ENTRY): New.
23878         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23879         inline entry marker notes.
23880         (print_insn): Likewise.
23881         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23882         (INSN_DEBUG_MARKER_KIND): Likewise.
23883         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23884         * tree-inline.c (expand_call_inline): Build and insert
23885         debug_inline_entry stmt.
23886         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23887         inline entry blocks early, if nonbind markers are enabled.
23888         (dump_scope_block): Dump fragment info.
23889         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23890         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23891         (gimple_build_debug_inline_entry): New.
23892         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23893         Enable/disable inline entry points too.
23894         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23895         (DEBUG_INSN): Describe inline entry markers.
23897         * common.opt (gvariable-location-views): New.
23898         (gvariable-location-views=incompat5): New.
23899         * config.in: Rebuilt.
23900         * configure: Rebuilt.
23901         * configure.ac: Test assembler for view support.
23902         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23903         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23904         * dwarf2out.c (var_loc_view): New typedef.
23905         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23906         (dwarf2out_locviews_in_attribute): New.
23907         (dwarf2out_locviews_in_loclist): New.
23908         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23909         (enum dw_line_info_opcode): Add LI_adv_address.
23910         (struct dw_line_info_table): Add view.
23911         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23912         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23913         (zero_view_p): New variable.
23914         (ZERO_VIEW_P): New macro.
23915         (output_asm_line_debug_info): New.
23916         (struct var_loc_node): Add view.
23917         (add_AT_view_list, AT_loc_list): New.
23918         (add_var_loc_to_decl): Add view param.  Test it against last.
23919         (new_loc_list): Add view params.  Record them.
23920         (AT_loc_list_ptr): Handle loc and view lists.
23921         (view_list_to_loc_list_val_node): New.
23922         (print_dw_val): Handle dw_val_class_view_list.
23923         (size_of_die): Likewise.
23924         (value_format): Likewise.
23925         (loc_list_has_views): New.
23926         (gen_llsym): Set vl_symbol too.
23927         (maybe_gen_llsym, skip_loc_list_entry): New.
23928         (dwarf2out_maybe_output_loclist_view_pair): New.
23929         (output_loc_list): Output view list or entries too.
23930         (output_view_list_offset): New.
23931         (output_die): Handle dw_val_class_view_list.
23932         (output_dwarf_version): New.
23933         (output_compilation_unit_header): Use it.
23934         (output_skeleton_debug_sections): Likewise.
23935         (output_rnglists, output_line_info): Likewise.
23936         (output_pubnames, output_aranges): Update version comments.
23937         (output_one_line_info_table): Output view numbers in asm comments.
23938         (dw_loc_list): Determine current endview, pass it to new_loc_list.
23939         Call maybe_gen_llsym.
23940         (loc_list_from_tree_1): Adjust.
23941         (add_AT_location_description): Create view list attribute if
23942         needed, check it's absent otherwise.
23943         (convert_cfa_to_fb_loc_list): Adjust.
23944         (maybe_emit_file): Call output_asm_line_debug_info for test.
23945         (dwarf2out_var_location): Reset views as needed.  Precompute
23946         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
23947         attribute.  Set view.
23948         (new_line_info_table): Reset next view.
23949         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23950         (dwarf2out_source_line): Likewise.  Output view resets and labels to
23951         the assembler, or select appropriate line info opcodes.
23952         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23953         (optimize_string_length): Catch it.  Adjust.
23954         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
23955         dw_val_class_view_list, and remove it if no longer needed.
23956         (hash_loc_list): Hash view numbers.
23957         (loc_list_hasher::equal): Compare them.
23958         (optimize_location_lists): Check whether a view list symbol is
23959         needed, and whether the locview attribute is present, and
23960         whether they match.  Remove the locview attribute if no longer
23961         needed.
23962         (index_location_lists): Call skip_loc_list_entry for test.
23963         (dwarf2out_finish): Call output_asm_line_debug_info for test.
23964         Use output_dwarf_version.
23965         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23966         (struct dw_val_node): Add val_view_list.
23967         * final.c (SEEN_NEXT_VIEW): New.
23968         (set_next_view_needed): New.
23969         (clear_next_view_needed): New.
23970         (maybe_output_next_view): New.
23971         (final_start_function): Rename to...
23972         (final_start_function_1): ... this.  Take pointer to FIRST,
23973         add SEEN parameter.  Emit param bindings in the initial view.
23974         (final_start_function): Reintroduce SEEN-less interface.
23975         (final): Rename to...
23976         (final_1): ... this.  Take SEEN parameter.  Output final pending
23977         next view at the end.
23978         (final): Reintroduce seen-less interface.
23979         (final_scan_insn): Output pending next view before switching
23980         sections or ending a block.  Mark the next view as needed when
23981         outputting variable locations.  Notify debug backend of section
23982         changes, and of location view changes.
23983         (rest_of_handle_final): Adjust.
23984         * toplev.c (process_options): Autodetect value for debug variable
23985         location views option.  Warn on incompat5 without -gdwarf-5.
23986         * doc/invoke.texi (gvariable-location-views): New.
23987         (gvariable-location-views=incompat5): New.
23988         (gno-variable-location-views): New.
23990 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
23992         PR tree-optimization/84136
23993         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23994         that the result of find_edge is non-NULL.
23996 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23998         PR target/83008
23999         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
24000         storing integer register in SImode.  Fix cost of 256 and 512
24001         byte aligned SSE register store.
24003 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
24005         * config/i386/i386.c (ix86_multiplication_cost): Fix
24006         multiplication cost for TARGET_AVX512DQ.
24008 2018-02-08  Marek Polacek  <polacek@redhat.com>
24010         PR tree-optimization/84238
24011         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
24012         get_range_strlen.
24014 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24016         PR tree-optimization/84265
24017         * tree-vect-stmts.c (vectorizable_store): Don't treat
24018         VMAT_CONTIGUOUS accesses as grouped.
24019         (vectorizable_load): Likewise.
24021 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24023         PR tree-optimization/81635
24024         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
24025         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
24026         (test_round_for_mask): New functions.
24027         (wide_int_cc_tests): Call test_round_for_mask.
24028         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
24029         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
24030         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
24031         range returned by get_range_info.
24033 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
24035         PR ipa/81360
24036         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
24037         * symtab.c: Include builtins.h
24038         (symtab_node::output_to_lto_symbol_table_p): Move here
24039         from lto-streamer-out.c:output_symbol_p.
24040         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
24041         (output_symbol_p): Move all logic to symtab.c
24042         (produce_symtab): Update.
24044 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24046         * config/s390/s390-opts.h (enum indirect_branch): Define.
24047         * config/s390/s390-protos.h (s390_return_addr_from_memory)
24048         (s390_indirect_branch_via_thunk)
24049         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
24050         (enum s390_indirect_branch_type): Define.
24051         * config/s390/s390.c (struct s390_frame_layout, struct
24052         machine_function): Remove.
24053         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
24054         (indirect_branch_table_label_no, indirect_branch_table_name):
24055         Define variables.
24056         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
24057         (enum s390_indirect_branch_option): Define.
24058         (s390_return_addr_from_memory): New function.
24059         (s390_handle_string_attribute): New function.
24060         (s390_attribute_table): Add new attribute handler.
24061         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
24062         (s390_indirect_branch_via_thunk): New function.
24063         (s390_indirect_branch_via_inline_thunk): New function.
24064         (s390_function_ok_for_sibcall): When jumping via thunk disallow
24065         sibling call optimization for non z10 compiles.
24066         (s390_emit_call): Force indirect branch target to be a single
24067         register.  Add r1 clobber for non-z10 compiles.
24068         (s390_emit_epilogue): Emit return jump via return_use expander.
24069         (s390_reorg): Handle JUMP_INSNs as execute targets.
24070         (s390_option_override_internal): Perform validity checks for the
24071         new command line options.
24072         (s390_indirect_branch_attrvalue): New function.
24073         (s390_indirect_branch_settings): New function.
24074         (s390_set_current_function): Invoke s390_indirect_branch_settings.
24075         (s390_output_indirect_thunk_function):  New function.
24076         (s390_code_end): Implement target hook.
24077         (s390_case_values_threshold): Implement target hook.
24078         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
24079         macros.
24080         * config/s390/s390.h (struct s390_frame_layout)
24081         (struct machine_function): Move here from s390.c.
24082         (TARGET_INDIRECT_BRANCH_NOBP_RET)
24083         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
24084         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
24085         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
24086         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
24087         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
24088         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
24089         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
24090         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
24091         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
24092         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
24093         (mnemonic attribute): Add values which aren't recognized
24094         automatically.
24095         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
24096         pattern for branch conversion.  Fix mnemonic attribute.
24097         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
24098         indirect branch via thunk if requested.
24099         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
24100         ("*indirect_jump"): Disable for branch conversion using out of
24101         line thunks.
24102         ("indirect_jump_via_thunk<mode>_z10")
24103         ("indirect_jump_via_thunk<mode>")
24104         ("indirect_jump_via_inlinethunk<mode>_z10")
24105         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
24106         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
24107         ("casesi_jump_via_inlinethunk<mode>_z10")
24108         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
24109         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
24110         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
24111         ("*indirect2_jump"): Disable for branch conversion.
24112         ("casesi_jump"): Turn into expander and expand patterns for branch
24113         conversion.
24114         ("return_use"): New expander.
24115         ("*return"): Emit return via thunk and rename it to ...
24116         ("*return<mode>"): ... this one.
24117         * config/s390/s390.opt: Add new options and and enum for the
24118         option values.
24120 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24122         * lra-constraints.c (match_reload): Unconditionally use
24123         gen_lowpart_SUBREG, rather than selecting between that
24124         and equivalent gen_rtx_SUBREG code.
24126 2018-02-08  Richard Biener  <rguenther@suse.de>
24128         PR tree-optimization/84233
24129         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
24130         changed flag instead of boguously re-using phi_inserted.
24132 2018-02-08  Martin Jambor  <mjambor@suse.cz>
24134         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
24135         static local variables.
24137 2018-02-08  Richard Biener  <rguenther@suse.de>
24139         PR tree-optimization/84278
24140         * tree-vect-stmts.c (vectorizable_store): When looking for
24141         smaller vector types to perform grouped strided loads/stores
24142         make sure the mode is supported by the target.
24143         (vectorizable_load): Likewise.
24145 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
24147         * config/aarch64/aarch64.c (aarch64_components_for_bb):
24148         Increase LDP/STP opportunities by adding adjacent callee-saves.
24150 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
24152         PR rtl-optimization/84068
24153         PR rtl-optimization/83459
24154         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
24156 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
24158         PR tree-optimization/84224
24159         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
24160         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
24161         non-zero arguments.
24163 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
24165         PR target/84113
24166         * config/rs6000/altivec.md (*restore_world): Remove LR use.
24167         * config/rs6000/predicates.md (restore_world_operation): Adjust op
24168         count, remove one USE.
24170 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24172         * doc/install.texi (Configuration): Document the
24173         --with-long-double-format={ibm,ieee} PowerPC configuration
24174         options.
24176         PR target/84154
24177         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
24178         Convert from define_expand to be define_insn_and_split.  Rework
24179         float/double/_Float128 conversions to QI/HI/SImode to work with
24180         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
24181         conversions to QI/HImode types did a store and then a load to
24182         truncate the value.  For conversions to VSX registers, don't split
24183         the insn, instead emit the code directly.  Use the code iterator
24184         any_fix to combine signed and unsigned conversions.
24185         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
24186         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
24187         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24188         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
24189         (fix_<mode>di2_hw): Likewise.
24190         (fixuns_<mode>di2_hw): Likewise.
24191         (fix_<mode>si2_hw): Likewise.
24192         (fixuns_<mode>si2_hw): Likewise.
24193         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
24194         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
24195         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
24196         fix<uns>_trunc<SFDF:mode>si2_p8.
24197         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
24198         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
24199         (fix<uns>_<mode>_mem): Likewise.
24200         (fctiw<u>z_<mode>_mem): Likewise.
24201         (fix<uns>_<mode>_mem): Likewise.
24202         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
24203         the register allocator from doing a direct move to the GPRs to do
24204         a store, and instead use the ISA 3.0 store byte/half-word from
24205         vector register instruction.  For IEEE 128-bit floating point,
24206         also optimize stores of 32-bit ints.
24207         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
24209 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
24211         * genextract.c (push_pathstr_operand): New function to support
24212         [a-zA-Z].
24213         (walk_rtx): Call push_pathstr_operand.
24214         (print_path): Support [a-zA-Z].
24216 2018-02-07  Richard Biener  <rguenther@suse.de>
24218         PR tree-optimization/84037
24219         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
24220         (cse_and_gimplify_to_preheader): Declare.
24221         (vect_get_place_in_interleaving_chain): Likewise.
24222         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24223         ivexpr_map.
24224         (_loop_vec_info::~_loop_vec_info): Delete it.
24225         (cse_and_gimplify_to_preheader): New function.
24226         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
24227         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
24228         (vectorizable_load): Likewise.  For grouped stores always base
24229         the IV on the first element.
24230         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
24231         condition before gimplifying.
24233 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24235         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
24236         *DIV_EXPR and *MOD_EXPR.
24238 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
24240         PR target/84248
24241         * config/i386/i386.c (ix86_option_override_internal): Mask out
24242         the CF_SET bit when checking -fcf-protection.
24244 2018-02-07  Tom de Vries  <tom@codesourcery.com>
24246         PR libgomp/84217
24247         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
24248         enough.
24250 2018-02-07  Richard Biener  <rguenther@suse.de>
24252         PR tree-optimization/84204
24253         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
24254         this place.
24256         PR tree-optimization/84205
24257         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
24258         special-case isl_ast_op_zdiv_r.
24260         PR tree-optimization/84223
24261         * graphite-scop-detection.c (gather_bbs::before_dom_children):
24262         Only add conditions from within the region.
24263         (gather_bbs::after_dom_children): Adjust.
24265 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
24267         PR target/84209
24268         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
24269         * config/avr/avr.md: Only post-reload split REG-REG moves if
24270         either register is GENERAL_REG_P.
24272 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
24274         PR tree-optimization/84235
24275         * tree-ssa-scopedtables.c
24276         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
24277         if the subtraction is performed in floating point type where NaNs are
24278         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
24279         build 1.  Formatting fix.
24281 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
24283         PR target/84146
24284         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
24285         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
24286         and skip it regardless of bb boundaries.  Use CALL_P macro,
24287         don't test INSN_P (insn) together with CALL_P or JUMP_P check
24288         unnecessarily, formatting fix.
24290 2018-02-06  Michael Collison  <michael.collison@arm.com>
24292         * config/arm/thumb2.md:
24293         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
24294         (*thumb_mov_notscc): Ditto.
24296 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
24298         PR target/84154
24299         * config/rs6000/rs6000.md (su code attribute): Use "u" for
24300         unsigned_fix, not "s".
24302 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24304         * configure.ac (gcc_fn_eh_frame_ro): New function.
24305         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
24306         correct .eh_frame permissions.
24307         * configure: Regenerate.
24309 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
24311         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
24312         irrelevant options.
24314 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24316         * config/rs6000/rs6000.c (rs6000_option_override_internal):
24317         Display warning message for -mno-speculate-indirect-jumps.
24319 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
24321         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
24322         Undocumented.
24323         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
24325 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
24327         PR tree-optimization/84225
24328         * tree-eh.c (find_trapping_overflow): Only call
24329         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
24331 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24333         PR target/84145
24334         * config/i386/i386.c: Reimplement the check of possible options
24335         -mibt/-mshstk conbination. Change error messages.
24336         * doc/invoke.texi: Fix a typo: remove extra '='.
24338 2018-02-06  Marek Polacek  <polacek@redhat.com>
24340         PR tree-optimization/84228
24341         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
24343 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
24345         PR target/82641
24346         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
24347         emitted arch directives.
24348         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
24349         __ARM_FEATURE_COPROC before changing architectures.
24351 2018-02-06  Richard Biener  <rguenther@suse.de>
24353         * config/i386/i386.c (print_reg): Fix typo.
24354         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
24356 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
24358         * configure: Regenerate.
24360 2018-02-05  Martin Sebor  <msebor@redhat.com>
24362         PR tree-optimization/83369
24363         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
24364         inlining context.
24366 2018-02-05  Martin Liska  <mliska@suse.cz>
24368         * doc/invoke.texi: Cherry-pick upstream r323995.
24370 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
24372         * ira.c (ira_init_register_move_cost): Adjust comment.
24374 2018-02-05  Martin Liska  <mliska@suse.cz>
24376         PR gcov-profile/84137
24377         * doc/gcov.texi: Fix typo in documentation.
24379 2018-02-05  Martin Liska  <mliska@suse.cz>
24381         PR gcov-profile/83879
24382         * doc/gcov.texi: Document necessity of --dynamic-list-data when
24383         using dlopen functionality.
24385 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
24387         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
24388         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
24389         _mm_maskz_range_ss, _mm_mask_range_round_ss,
24390         _mm_maskz_range_round_ss): New intrinsics.
24391         (__builtin_ia32_rangesd128_round)
24392         (__builtin_ia32_rangess128_round): Remove.
24393         (__builtin_ia32_rangesd128_mask_round,
24394         __builtin_ia32_rangess128_mask_round): New builtins.
24395         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
24396         __builtin_ia32_rangess128_round): Remove.
24397         (__builtin_ia32_rangesd128_mask_round,
24398         __builtin_ia32_rangess128_mask_round): New builtins.
24399         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
24400         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
24401         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
24402         "<round_saeonly_constraint>")): Changed to ...
24403         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
24404         "<round_saeonly_scalar_constraint>")): ... this.
24405         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
24406         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
24407         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
24408         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
24409         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
24411 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
24413         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
24414         options.
24415         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
24416         Remove all values except native, 8540 and 8548.
24418 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
24420         * config/i386/i386.c (ix86_output_function_return): Pass
24421         INVALID_REGNUM, instead of -1, as invalid register number to
24422         indirect_thunk_name and output_indirect_thunk.
24424 2018-02-02  Julia Koval  <julia.koval@intel.com>
24426         * config.gcc: Add -march=icelake.
24427         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
24428         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
24429         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
24430         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
24431         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
24432         (processor_target_table): Add icelake.
24433         (ix86_option_override_internal): Handle new PTAs.
24434         (get_builtin_code_for_version): Handle icelake.
24435         (M_INTEL_COREI7_ICELAKE): New.
24436         (fold_builtin_cpu): Handle icelake.
24437         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
24438         * doc/invoke.texi: Add -march=icelake.
24440 2018-02-02  Julia Koval  <julia.koval@intel.com>
24442         * config/i386/i386.c (ix86_option_override_internal): Change flags type
24443         to wide_int_bitmask.
24444         * wide-int-bitmask.h: New.
24446 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24448         PR target/84066
24449         * config/i386/i386.md: Replace Pmode with word_mode in
24450         builtin_setjmp_setup and builtin_longjmp to support x32.
24452 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
24454         PR target/56010
24455         PR target/83743
24456         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
24457         #include "opts.h".
24458         (rs6000_supported_cpu_names): New static variable.
24459         (linux_cpu_translation_table): Likewise.
24460         (elf_platform) <cpu>: Define new static variable and use it.
24461         Translate kernel AT_PLATFORM name to canonical name if needed.
24462         Error if platform name is unknown.
24464 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
24466         PR target/84089
24467         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
24469 2018-02-01  Jeff Law  <law@redhat.com>
24471         PR target/84128
24472         * config/i386/i386.c (release_scratch_register_on_entry): Add new
24473         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
24474         the scratch if RELEASE_VIA_POP is false.
24475         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
24476         If we have to save a temporary register, decrement SIZE appropriately.
24477         Pass new arguments to release_scratch_register_on_entry.
24478         (ix86_adjust_stack_and_probe): Likewise.
24479         (ix86_emit_probe_stack_range): Pass new arguments to
24480         release_scratch_register_on_entry.
24482 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
24484         PR rtl-optimization/84157
24485         * combine.c (change_zero_ext): Use REG_P predicate in
24486         front of HARD_REGISTER_P predicate.
24488 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
24490         * config/avr/avr.c (avr_option_override): Move disabling of
24491         -fdelete-null-pointer-checks to...
24492         * common/config/avr/avr-common.c (avr_option_optimization_table):
24493         ...here.
24495 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24497         PR tree-optimization/81635
24498         * tree-data-ref.c (split_constant_offset_1): For types that
24499         wrap on overflow, try to use range info to prove that wrapping
24500         cannot occur.
24502 2018-02-01  Renlin Li  <renlin.li@arm.com>
24504         PR target/83370
24505         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
24506         TAILCALL_ADDR_REGS.
24507         (aarch64_register_move_cost): Likewise.
24508         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
24509         TAILCALL_ADDR_REGS.
24510         (REG_CLASS_NAMES): Likewise.
24511         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
24512         TAILCALL_ADDR_REGS. Remove IP registers.
24513         * config/aarch64/aarch64.md (Ucs): Update register constraint.
24515 2018-02-01  Richard Biener  <rguenther@suse.de>
24517         * domwalk.h (dom_walker::dom_walker): Add additional constructor
24518         for specifying RPO order and allow NULL for that.
24519         * domwalk.c (dom_walker::dom_walker): Likewise.
24520         (dom_walker::walk): Handle NULL RPO order.
24521         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
24522         in RPO order.
24523         (rewrite_update_dom_walker): Likewise.
24524         (mark_def_dom_walker): Likewise.
24526 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24528         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
24529         (aarch64_maybe_expand_sve_subreg_move): Declare.
24530         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
24531         * config/aarch64/predicates.md (aarch64_any_register_operand): New
24532         predicate.
24533         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
24534         that are semantically a reverse operation.
24535         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
24536         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
24537         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
24538         functions.
24539         (aarch64_can_change_mode_class): For big-endian, forbid changes
24540         between two SVE modes if they have different element sizes.
24542 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24544         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
24545         the TImode handling for big-endian targets.
24547 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24549         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
24550         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
24551         not just bytes.
24552         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
24553         Remove BSWAP handing for big-endian targets and use the form of
24554         LD1RQ appropariate for the mode.
24556 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24558         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
24559         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
24560         duplicated element.
24562 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24564         PR tearget/83845
24565         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
24566         check for operands that need to go through aarch64_sve_reload_be.
24568 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
24570         PR tree-optimization/81661
24571         PR tree-optimization/84117
24572         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
24573         * tree-eh.c: Include gimplify.h.
24574         (find_trapping_overflow, replace_trapping_overflow,
24575         rewrite_to_non_trapping_overflow): New functions.
24576         * tree-vect-loop.c: Include tree-eh.h.
24577         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
24578         * tree-data-ref.c: Include tree-eh.h.
24579         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
24581 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
24583         PR rtl-optimization/84123
24584         * combine.c (change_zero_ext): Check if hard register satisfies
24585         can_change_dest_mode before calling gen_lowpart_SUBREG.
24587 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
24589         PR target/82444
24590         * ira.c (ira_init_register_move_cost): Remove assert.
24592 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24594         PR rtl-optimization/84071
24595         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
24596         * doc/tm.texi: Regenerate.
24598 2018-01-31  Richard Biener  <rguenther@suse.de>
24600         PR tree-optimization/84132
24601         * tree-data-ref.c (analyze_miv_subscript): Properly
24602         check whether evolution_function_is_affine_multivariate_p
24603         before calling gcd_of_steps_may_divide_p.
24605 2018-01-31  Julia Koval  <julia.koval@intel.com>
24607         PR target/83618
24608         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
24609         * config/i386/i386.md (rdpid_rex64) New.
24610         (rdpid): Make 32bit only.
24612 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
24614         PR lto/84105
24615         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
24616         an IDENTIFIER_NODE for FUNCTION_TYPE's.
24618 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24620         Revert
24621         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24623         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24625 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24627         PR rtl-optimization/84071
24628         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24629         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24631 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24633         * config/arc/arc.c (arc_handle_aux_attribute): New function.
24634         (arc_attribute_table): Add 'aux' attribute.
24635         (arc_in_small_data_p): Consider aux like variables.
24636         (arc_is_aux_reg_p): New function.
24637         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24638         (arc_get_aux_arg): New function.
24639         (prepare_move_operands): Handle aux-register access.
24640         (arc_handle_aux_attribute): New function.
24641         * doc/extend.texi (ARC Variable attributes): Add subsection.
24643 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24645         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24646         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24647         (arc_attribute_table): Add 'uncached' attribute.
24648         (arc_print_operand): Print '.di' flag for uncached memory
24649         accesses.
24650         (arc_in_small_data_p): Do not consider for small data the uncached
24651         types.
24652         (arc_is_uncached_mem_p): New function.
24653         * config/arc/predicates.md (compact_store_memory_operand): Check
24654         for uncached memory accesses.
24655         (nonvol_nonimm_operand): Likewise.
24656         * doc/extend.texi (ARC Type Attribute): New subsection.
24658 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
24660         PR c/84100
24661         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24662         falign-loops=): Add Optimization flag.
24664 2018-01-30  Jeff Law  <law@redhat.com>
24666         PR target/84064
24667         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24668         INT_REGISTERS_SAVED.  Check it prior to calling
24669         get_scratch_register_on_entry.
24670         (ix86_adjust_stack_and_probe): Similarly.
24671         (ix86_emit_probe_stack_range): Similarly.
24672         (ix86_expand_prologue): Corresponding changes.
24674 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24676         PR target/40411
24677         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24678         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24680 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
24682         PR target/84112
24683         * lra-constraints.c (curr_insn_transform): Process AND in the
24684         address.
24686 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
24688         PR rtl-optimization/83986
24689         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24690         dependence against last_pending_memory_flush in addition to
24691         pending_jump_insns.
24693 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
24695         PR tree-optimization/81611
24696         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24697         copies.
24699 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24701         PR target/83758
24702         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24703         a reg rtx.
24705 2018-01-30  Richard Biener  <rguenther@suse.de>
24706             Jakub Jelinek  <jakub@redhat.com>
24708         PR tree-optimization/84111
24709         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24710         inner loops added during recursion, as they don't have up-to-date
24711         SSA form.
24713 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24715         PR ipa/81360
24716         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24717         (can_inline_edge_by_limits_p): ... here.
24718         (can_early_inline_edge_p, check_callers,
24719         update_caller_keys, update_callee_keys, recursive_inlining,
24720         add_new_edges_to_heap, speculation_useful_p,
24721         inline_small_functions,
24722         inline_small_functions, flatten_function,
24723         inline_to_all_callers_1): Update.
24725 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24727         * profile-count.c (profile_count::combine_with_ipa_count): Handle
24728         zeros correctly.
24730 2018-01-30  Richard Biener  <rguenther@suse.de>
24732         PR tree-optimization/83008
24733         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24734         invariant and constant vector uses in stmts when they need
24735         more than one stmt.
24737 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24739         PR bootstrap/84017
24740         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24741         * configure: Regenerate.
24743 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24745         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24746         pattern.
24747         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24748         Use gen_rtx_REG rather than gen_lowpart.
24750 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24752         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24753         rather than 0 when creating partial subregs.
24755 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24757         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24758         of usage.
24760 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
24762         PR target/81550
24763         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24764         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24765         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24766         flags.  This restores the settings used before the 2017-07-24.
24767         Turning off pre increment/decrement/modify allows IVOPTS to
24768         optimize DF/SF loops where the index is an int.
24770 2018-01-29  Richard Biener  <rguenther@suse.de>
24771             Kelvin Nilsen  <kelvin@gcc.gnu.org>
24773         PR bootstrap/80867
24774         * tree-vect-stmts.c (vectorizable_call): Don't call
24775         targetm.vectorize_builtin_md_vectorized_function if callee is
24776         NULL.
24778 2018-01-22  Carl Love  <cel@us.ibm.com>
24780         * doc/extend.tex: Fix typo in second arg in
24781         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24783 2018-01-29  Richard Biener  <rguenther@suse.de>
24785         PR tree-optimization/84086
24786         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24787         (flush_ssaname_freelist): When SSA names were released reset
24788         the SCEV hash table.
24790 2018-01-29  Richard Biener  <rguenther@suse.de>
24792         PR tree-optimization/84057
24793         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24794         removed paths when removing edges.
24796 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
24798         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24799         -mfunction-return=@var{choice}.
24801 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
24803         PR diagnostic/84034
24804         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24805         Handle CR like TAB.
24806         (layout::print_source_line): Likewise.
24807         (test_get_line_width_without_trailing_whitespace): Add test cases.
24809 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
24811         PR middle-end/84040
24812         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24813         debug insns.
24815 2018-01-26  Jim Wilson  <jimw@sifive.com>
24817         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24819         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24820         specified.
24822 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24824         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24825         and CMP + SUB-immediate -> SUBS.
24827 2018-01-26  Martin Sebor  <msebor@redhat.com>
24829         PR tree-optimization/83896
24830         * tree-ssa-strlen.c (get_string_len): Rename...
24831         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
24832         Avoid assuming length is constant.
24833         (handle_char_store): Use HOST_WIDE_INT for string length.
24835 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
24837         PR target/81763
24838         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24839         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24841 2018-01-26  Richard Biener  <rguenther@suse.de>
24843         PR rtl-optimization/84003
24844         * dse.c (record_store): Only record redundant stores when
24845         the earlier store aliases at least all accesses the later one does.
24847 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
24849         PR rtl-optimization/83985
24850         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24851         REG_CFA_RESTORE insns.
24852         (delete_unmarked_insns): Don't ignore separate shrink wrapping
24853         REG_CFA_RESTORE insns here.
24855         PR c/83989
24856         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24857         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24859 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24861         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24862         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24863         (arc_init): Likewise.
24864         (arc_override_options): Likewise.
24865         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24866         value.
24867         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24868         support.
24869         * config/arc/arc.h (TARGET_DBNZ): Define.
24870         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24871         properly set the tune attribute.
24872         (dbnz): Use TARGET_DBNZ guard.
24873         * config/arc/arc.opt (mtune): Add core3 option.
24875 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24877         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24878         recognize new pic like addresses.
24879         (arc_delegitimize_address): Clean up.
24881 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24883         * config/arc/arc-arches.def: Option mrf16 valid for all
24884         architectures.
24885         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24886         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24887         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24888         * config/arc/arc-tables.opt: Regenerate.
24889         * config/arc/arc.c (arc_conditional_register_usage): Handle
24890         reduced register file case.
24891         (arc_file_start): Set must have build attributes.
24892         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24893         mrf16 option value.
24894         * config/arc/arc.opt (mrf16): Add new option.
24895         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24896         * config/arc/genmultilib.awk: Handle new mrf16 option.
24897         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24898         * config/arc/t-multilib: Regenerate.
24899         * doc/invoke.texi (ARC Options): Document mrf16 option.
24901 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24903         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24904         * config/arc/arc.c (arc_handle_secure_attribute): New function.
24905         (arc_attribute_table): Add 'secure_call' attribute.
24906         (arc_print_operand): Print secure call operand.
24907         (arc_function_ok_for_sibcall): Don't optimize tail calls when
24908         secure.
24909         (arc_is_secure_call_p): New function.  * config/arc/arc.md
24910         (call_i): Add support for sjli instruction.
24911         (call_value_i): Likewise.
24912         * config/arc/constraints.md (Csc): New constraint.
24914 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24915             John Eric Martin  <John.Martin@emmicro-us.com>
24917         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24918         * config/arc/arc.c (_arc_jli_section): New struct.
24919         (arc_jli_section): New type.
24920         (rc_jli_sections): New static variable.
24921         (arc_handle_jli_attribute): New function.
24922         (arc_attribute_table): Add jli_always and jli_fixed attribute.
24923         (arc_file_end): New function.
24924         (TARGET_ASM_FILE_END): Define.
24925         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24926         (arc_add_jli_section): New function.
24927         (jli_call_scan): Likewise.
24928         (arc_reorg): Call jli_call_scan.
24929         (arc_output_addsi): Remove 'S' from printing asm operand.
24930         (arc_is_jli_call_p): New function.
24931         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24932         operand.
24933         (movhi_insn): Likewise.
24934         (movsi_insn): Likewise.
24935         (movsi_set_cc_insn): Likewise.
24936         (loadqi_update): Likewise.
24937         (load_zeroextendqisi_update): Likewise.
24938         (load_signextendqisi_update): Likewise.
24939         (loadhi_update): Likewise.
24940         (load_zeroextendhisi_update): Likewise.
24941         (load_signextendhisi_update): Likewise.
24942         (loadsi_update): Likewise.
24943         (loadsf_update): Likewise.
24944         (movsicc_insn): Likewise.
24945         (bset_insn): Likewise.
24946         (bxor_insn): Likewise.
24947         (bclr_insn): Likewise.
24948         (bmsk_insn): Likewise.
24949         (bicsi3_insn): Likewise.
24950         (cmpsi_cc_c_insn): Likewise.
24951         (movsi_ne): Likewise.
24952         (movsi_cond_exec): Likewise.
24953         (clrsbsi2): Likewise.
24954         (norm_f): Likewise.
24955         (normw): Likewise.
24956         (swap): Likewise.
24957         (divaw): Likewise.
24958         (flag): Likewise.
24959         (sr): Likewise.
24960         (kflag): Likewise.
24961         (ffs): Likewise.
24962         (ffs_f): Likewise.
24963         (fls): Likewise.
24964         (call_i): Remove 'S' asm letter, add jli instruction.
24965         (call_value_i): Likewise.
24966         * config/arc/arc.op (mjli-always): New option.
24967         * config/arc/constraints.md (Cji): New constraint.
24968         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24969         operand.
24970         (subsf3_fpx): Likewise.
24971         (mulsf3_fpx): Likewise.
24972         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24973         asm operand.
24974         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24975         function attrbutes.
24976         * doc/invoke.texi (ARC): Document mjli-always option.
24978 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
24980         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24981         avoid addition with 0 and use incw and decw where possible.
24983 2018-01-26  Richard Biener  <rguenther@suse.de>
24985         PR tree-optimization/81082
24986         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24987         association if it requires casting to unsigned.
24988         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24989         from fold_plusminus_mult_expr to catch important cases late when
24990         range info is available.
24992 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24994         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24995         * configure.ac (hidden_linkonce): New test.
24996         * configure: Regenerate.
24997         * config.in: Regenerate.
24999 2018-01-26  Julia Koval  <julia.koval@intel.com>
25001         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
25002         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
25003         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
25004         _mm_mask_bitshuffle_epi64_mask): Fix type.
25005         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
25006         USI_FTYPE_V4DI_V4DI_USI): Remove.
25007         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
25008         __builtin_ia32_vpshufbitqmb256_mask,
25009         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
25010         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
25011         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
25013 2018-01-26  Alan Modra  <amodra@gmail.com>
25015         PR target/84033
25016         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
25017         UNSPEC_VBPERMQ.  Sort other unspecs.
25019 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
25021         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
25023 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
25025         PR middle-end/83055
25026         * predict.c (drop_profile): Do not push/pop cfun; update also
25027         node->count.
25028         (handle_missing_profiles): Fix logic looking for zero profiles.
25030 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
25032         PR middle-end/83977
25033         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
25034         on functions with #pragma omp declare simd or functions with simd
25035         attribute.
25036         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
25037         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
25038         Remove trailing \n from warning_at calls.
25040 2018-01-25  Tom de Vries  <tom@codesourcery.com>
25042         PR target/84028
25043         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
25044         for neutered workers.
25046 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
25048         PR target/68467
25049         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
25050         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
25052 2018-01-24  Jeff Law  <law@redhat.com>
25054         PR target/83994
25055         * i386.c (get_probe_interval): Move to earlier point.
25056         (ix86_compute_frame_layout): If -fstack-clash-protection and
25057         the frame is larger than the probe interval, then use pushes
25058         to save registers rather than reg->mem moves.
25059         (ix86_expand_prologue): Remove conditional for int_registers_saved
25060         assertion.
25062 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
25064         PR target/84014
25065         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
25066         min/max for never referenced object.
25068 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
25070         PR middle-end/83977
25071         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
25072         here.
25073         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
25074         attributes from DECL_ATTRIBUTES (decl) without affecting
25075         DECL_ATTRIBUTES (current_function_decl).
25076         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
25077         functions with non-NULL DECL_ABSTRACT_ORIGIN.
25079 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
25081         PR tree-optimization/83979
25082         * fold-const.c (fold_comparison): Use constant_boolean_node
25083         instead of boolean_{true,false}_node.
25085 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
25087         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
25088         with zero counts.
25090 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25092         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25093         Simplify the clause that sets the length attribute.
25094         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25095         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
25096         clause that sets the length attribute.
25097         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25099 2018-01-24  Tom de Vries  <tom@codesourcery.com>
25101         PR target/83589
25102         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
25103         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
25104         Add strict parameter.
25105         (prevent_branch_around_nothing): Insert dummy insn between branch to
25106         label and label with no ptx insn inbetween.
25107         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
25109 2018-01-24  Tom de Vries  <tom@codesourcery.com>
25111         PR target/81352
25112         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
25113         for neutered threads in warp.
25114         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
25116 2018-01-24  Richard Biener  <rguenther@suse.de>
25118         PR tree-optimization/83176
25119         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
25120         operands.
25122 2018-01-24  Richard Biener  <rguenther@suse.de>
25124         PR tree-optimization/82819
25125         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
25126         code generating pluses that are no-ops in the target precision.
25128 2018-01-24  Richard Biener  <rguenther@suse.de>
25130         PR middle-end/84000
25131         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
25133 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25135         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
25136         to merge probabilities.
25137         * predict.c (probably_never_executed): Also mark as cold functions
25138         with global 0 profile and guessed local profile.
25139         * profile-count.c (profile_probability::combine_with_count): New
25140         member function.
25141         * profile-count.h (profile_probability::operator*,
25142         profile_probability::operator*=, profile_probability::operator/,
25143         profile_probability::operator/=): Reduce precision to adjusted
25144         and set value to guessed on contradictory divisions.
25145         (profile_probability::combine_with_freq): Remove.
25146         (profile_probability::combine_wiht_count): Declare.
25147         (profile_count::force_nonzero):: Set to adjusted.
25148         (profile_count::probability_in):: Set quality to adjusted.
25149         * tree-ssa-tail-merge.c (replace_block_by): Use
25150         combine_with_count.
25152 2018-01-23  Andrew Waterman  <andrew@sifive.com>
25153             Jim Wilson  <jimw@sifive.com>
25155         * config/riscv/riscv.c (riscv_stack_boundary): New.
25156         (riscv_option_override): Set riscv_stack_boundary.  Handle
25157         riscv_preferred_stack_boundary_arg.
25158         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
25159         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
25160         (STACK_BOUNDARY): Set to riscv_stack_boundary.
25161         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
25162         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
25163         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
25165 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
25167         PR target/83905
25168         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
25169         of struct ix86_frame.
25170         (ix86_expand_epilogue): Likewise.  Add a local variable for
25171         the reg_save_offset field in struct ix86_frame.
25173 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
25175         PR tree-optimization/82604
25176         * tree-loop-distribution.c (enum partition_kind): New enum item
25177         PKIND_PARTIAL_MEMSET.
25178         (partition_builtin_p): Support above new enum item.
25179         (generate_code_for_partition): Ditto.
25180         (compute_access_range): Differentiate cases that equality can be
25181         proven at all loops, the innermost loops or no loops.
25182         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
25183         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
25184         (finalize_partitions, distribute_loop): Don't fuse partition of
25185         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
25186         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
25187         parloop is enabled.
25189 2018-01-23  Martin Liska  <mliska@suse.cz>
25191         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
25192         order to ignore the predictor.
25193         (PRED_POLYMORPHIC_CALL): Likewise.
25194         (PRED_RECURSIVE_CALL): Likewise.
25196 2018-01-23  Martin Liska  <mliska@suse.cz>
25198         * tree-profile.c (tree_profiling): Print function header to
25199         aware reader which function we are working on.
25200         * value-prof.c (gimple_find_values_to_profile): Do not print
25201         not interesting value histograms.
25203 2018-01-23  Martin Liska  <mliska@suse.cz>
25205         * profile-count.h (enum profile_quality): Add
25206         profile_uninitialized as the first value. Do not number values
25207         as they are zero based.
25208         (profile_count::verify): Update sanity check.
25209         (profile_probability::verify): Likewise.
25211 2018-01-23  Nathan Sidwell  <nathan@acm.org>
25213         * doc/invoke.texi (ffor-scope): Deprecate.
25215 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25217         PR tree-optimization/83510
25218         * domwalk.c (set_all_edges_as_executable): New function.
25219         (dom_walker::dom_walker): Convert bool param
25220         "skip_unreachable_blocks" to enum reachability.  Move setup of
25221         edge flags to set_all_edges_as_executable and only do it when
25222         reachability is REACHABLE_BLOCKS.
25223         * domwalk.h (enum dom_walker::reachability): New enum.
25224         (dom_walker::dom_walker): Convert bool param
25225         "skip_unreachable_blocks" to enum reachability.
25226         (set_all_edges_as_executable): New decl.
25227         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
25228         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
25229         "reachability".
25230         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
25231         but converting true to REACHABLE_BLOCKS.
25232         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
25233         * tree-vrp.c
25234         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
25235         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
25236         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
25237         REACHABLE_BLOCKS.
25238         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
25239         if check_all_array_refs will be called.
25241 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25243         * tree.c (selftest::test_location_wrappers): Add more test
25244         coverage.
25246 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
25248         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
25249         (selftest::test_bit_in_range): Likewise.
25251 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25253         PR testsuite/83888
25254         * doc/sourcebuild.texi (vect_float): Say that the selector
25255         only describes the situation when -funsafe-math-optimizations is on.
25256         (vect_float_strict): Document.
25258 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
25260         PR tree-optimization/83965
25261         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
25262         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
25263         instead of checking only for a reduction.
25264         (vect_recog_widen_sum_pattern): Likewise.
25266 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25268         * predict.c (probably_never_executed): Only use precise profile info.
25269         (compute_function_frequency): Skip after inlining hack since we now
25270         have quality checking.
25272 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
25274         * profile-count.h (profile_probability::very_unlikely,
25275         profile_probability::unlikely, profile_probability::even): Set
25276         precision to guessed.
25278 2018-01-23  Richard Biener  <rguenther@suse.de>
25280         PR tree-optimization/83963
25281         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
25282         Properly terminate dominator walk when crossing the exit edge not
25283         when visiting its source block.
25285 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
25287         PR c++/83918
25288         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
25289         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
25291 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
25293         PR tree-optimization/83957
25294         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
25295         semicolon after for body surrounded by braces.
25297         PR tree-optimization/83081
25298         * profile-count.h (profile_probability::split): New method.
25299         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
25300         Use profile_probability::split.
25301         (do_compare_rtx_and_jump): Fix adjustment of probabilities
25302         when splitting a single conditional jump into 2.
25304 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
25306         PR tree-optimization/69452
25307         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
25308         decl.
25310 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25312         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
25313         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
25314         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
25316 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25318         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
25319         declaration.
25320         * config/rl78/rl78.md (movdi): New define_expand.
25321         * config/rl78/rl78.c (rl78_split_movdi): New function.
25323 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
25325         PR target/83862
25326         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
25327         no longer used.
25328         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
25329         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
25330         128-bit to produce an UNSPEC move to get the double word with the
25331         signbit and then a shift directly to do signbit.
25332         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
25333         implementation with a new version that just does either a direct
25334         move or a regular move.  Move memory interface to separate insns.
25335         Move insns so they are next to the expander.
25336         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
25337         with signbit move.  Split big and little endian case.
25338         (signbit<mode>2_dm_mem_le): Likewise.
25339         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
25340         (signbit<mode>2_dm2): Likewise.
25342 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25344         * config/rl78/rl78.md (anddi3): New define_expand.
25346 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25348         * config/rl78/rl78.md (umindi3): New define_expand.
25350 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25352         * config/rl78/rl78.md (smindi3): New define_expand.
25354 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25356         * config/rl78/rl78.md (smaxdi3): New define_expand.
25358 2018-01-22  Carl Love  <cel@us.ibm.com>
25360         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
25361         LVX_V1TI): Add macro expansion.
25362         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
25363         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
25364         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
25365         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
25366         Change check to determine if the instruction is a byte reversing
25367         entry.  Fix typo in comment.
25368         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
25369         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
25370         Add def_builtin calls for new builtins.
25371         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
25372         Add define_insn expansion.
25374 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25376         * config/rl78/rl78.md (umaxdi3): New define_expand.
25378 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
25380         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
25381         for non-QImode registers.
25383 2018-01-22  Richard Biener  <rguenther@suse.de>
25385         PR tree-optimization/83963
25386         * graphite-scop-detection.c (scop_detection::get_sese): Delay
25387         including the loop exit block.
25388         (scop_detection::merge_sese): Likewise.
25389         (scop_detection::add_scop): Do it here instead.
25391 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25393         * doc/sourcebuild.texi (arm_softfloat): Document.
25395 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
25397         PR gcc/77734
25398         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
25399         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
25400         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
25402 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25403             David Edelsohn  <dje.gcc@gmail.com>
25405         PR target/83946
25406         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25407         Change "crset eq" to "crset 2".
25408         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25409         (*call_indirect_aix<mode>_nospec): Likewise.
25410         (*call_value_indirect_aix<mode>_nospec): Likewise.
25411         (*call_indirect_elfv2<mode>_nospec): Likewise.
25412         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
25413         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
25414         change assembly output from . to $.
25415         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25416         (indirect_jump<mode>_nospec): Change assembly output from . to $.
25417         (*tablejump<mode>_internal1_nospec): Likewise.
25419 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
25421         PR target/80870
25422         * config/sh/sh_optimize_sett_clrt.cc:
25423         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
25425 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
25427         PR tree-optimization/83940
25428         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
25429         offset_dt to vect_constant_def rather than vect_unknown_def_type.
25430         (vect_check_load_store_mask): Add a mask_dt_out parameter and
25431         use it to pass back the definition type.
25432         (vect_check_store_rhs): Likewise rhs_dt_out.
25433         (vect_build_gather_load_calls): Add a mask_dt argument and use
25434         it instead of a call to vect_is_simple_use.
25435         (vectorizable_store): Update calls to vect_check_load_store_mask
25436         and vect_check_store_rhs.  Use the dt returned by the latter instead
25437         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
25438         instead of calls to vect_is_simple_use.  Pass the scalar rather
25439         than the vector operand to vect_is_simple_use when handling
25440         second and subsequent copies of an rhs value.
25441         (vectorizable_load): Update calls to vect_check_load_store_mask
25442         and vect_build_gather_load_calls.  Use the cached mask_dt and
25443         gs_info.offset_dt instead of calls to vect_is_simple_use.
25445 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
25447         PR middle-end/83945
25448         * tree-emutls.c: Include gimplify.h.
25449         (lower_emutls_2): New function.
25450         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
25451         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
25452         it before further processing.
25454         PR target/83930
25455         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
25456         UINTVAL (trueop1) instead of INTVAL (op1).
25458 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
25460         PR debug/81570
25461         PR debug/83728
25462         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
25463         INCOMING_FRAME_SP_OFFSET if not defined.
25464         (scan_trace): Add ENTRY argument.  If true and
25465         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
25466         emit a note to adjust the CFA offset.
25467         (create_cfi_notes): Adjust scan_trace callers.
25468         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
25469         INCOMING_FRAME_SP_OFFSET in the CIE.
25470         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
25471         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
25472         Likewise.
25473         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
25474         * doc/tm.texi: Regenerated.
25476 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25478         PR rtl-optimization/83147
25479         * lra-constraints.c (remove_inheritance_pseudos): Use
25480         lra_substitute_pseudo_within_insn.
25482 2018-01-19  Tom de Vries  <tom@codesourcery.com>
25483             Cesar Philippidis  <cesar@codesourcery.com>
25485         PR target/83920
25486         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
25488 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
25490         PR target/83790
25491         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
25492         spaces for function labels.
25494 2018-01-19  Martin Liska  <mliska@suse.cz>
25496         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
25497         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
25498         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
25499         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
25500         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
25501         (PRED_CONST_RETURN): Change from 69 to 65.
25502         (PRED_NULL_RETURN): Change from 91 to 71.
25503         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
25504         (PRED_LOOP_GUARD): Change from 66 to 73.
25506 2018-01-19  Martin Liska  <mliska@suse.cz>
25508         * predict.c (predict_insn_def): Add new assert.
25509         (struct branch_predictor): Change type to signed integer.
25510         (test_prediction_value_range): Amend test to cover
25511         PROB_UNINITIALIZED.
25512         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
25513         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
25514         (PRED_LOOP_ITERATIONS_MAX): Likewise.
25515         (PRED_LOOP_IV_COMPARE): Likewise.
25516         * predict.h (PROB_UNINITIALIZED): Define new constant.
25518 2018-01-19  Martin Liska  <mliska@suse.cz>
25520         * predict.c (dump_prediction): Add new format for
25521         analyze_brprob.py script which is enabled with -details
25522         suboption.
25523         * profile-count.h (precise_p): New function.
25525 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25527         PR tree-optimization/83922
25528         * tree-vect-loop.c (vect_verify_full_masking): Return false if
25529         there are no statements that need masking.
25530         (vect_active_double_reduction_p): New function.
25531         (vect_analyze_loop_operations): Use it when handling phis that
25532         are not in the loop header.
25534 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
25536         PR tree-optimization/83914
25537         * tree-vect-loop.c (vectorizable_induction): Don't convert
25538         init_expr or apply the peeling adjustment for inductions
25539         that are nested within the vectorized loop.
25541 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25543         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
25544         instead of NEG.
25546 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25548         PR sanitizer/81715
25549         PR testsuite/83882
25550         * function.h (gimplify_parameters): Add gimple_seq * argument.
25551         * function.c: Include gimple.h and options.h.
25552         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
25553         for the added local temporaries if needed.
25554         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
25555         if there are any parameter cleanups, wrap whole body into a
25556         try/finally with the cleanups.
25558 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
25560         PR target/82964
25561         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
25562         Use GET_MODE_CLASS for scalar floating point.
25564 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25566         PR ipa/82256
25567         patch by PaX Team
25568         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
25569         Fix call of call_cgraph_insertion_hooks.
25571 2018-01-18  Martin Sebor  <msebor@redhat.com>
25573         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
25575 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25577         PR ipa/83619
25578         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
25579         frequencies.
25581 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
25583         PR other/70268
25584         * common.opt: (-ffile-prefix-map): New option.
25585         * opts.c (common_handle_option): Defer it.
25586         * opts-global.c (handle_common_deferred_options): Handle it.
25587         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
25588         * file-prefix-map.h: New file.
25589         (remap_debug_filename, add_debug_prefix_map): ...here.
25590         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
25591         * final.c (debug_prefix_map, add_debug_prefix_map
25592         remap_debug_filename): Move to...
25593         * file-prefix-map.c: New file.
25594         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
25595         generalize, get rid of alloca(), use strrchr() instead of strchr().
25596         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
25597         Implement in terms of add_prefix_map().
25598         (remap_macro_filename, remap_debug_filename): Implement in term of
25599         remap_filename().
25600         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
25601         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
25602         * dbxout.c: Include file-prefix-map.h.
25603         * varasm.c: Likewise.
25604         * vmsdbgout.c: Likewise.
25605         * xcoffout.c: Likewise.
25606         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
25607         * doc/cppopts.texi (-fmacro-prefix-map): Document.
25608         * doc/invoke.texi (-ffile-prefix-map): Document.
25609         (-fdebug-prefix-map): Update description.
25611 2018-01-18  Martin Liska  <mliska@suse.cz>
25613         * config/i386/i386.c (indirect_thunk_name): Document that also
25614         lfence is emitted.
25615         (output_indirect_thunk): Document why both instructions
25616         (pause and lfence) are generated.
25618 2018-01-18  Richard Biener  <rguenther@suse.de>
25620         PR tree-optimization/83887
25621         * graphite-scop-detection.c
25622         (scop_detection::get_nearest_dom_with_single_entry): Remove.
25623         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25624         (scop_detection::merge_sese): Re-implement with a flood-fill
25625         algorithm that properly finds a SESE region if it exists.
25627 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25629         PR c/61240
25630         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25631         pointer_diff optimizations use view_convert instead of convert.
25633 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25635         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25636         Generate different code for -mno-speculate-indirect-jumps.
25637         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25638         (*call_indirect_aix<mode>): Disable for
25639         -mno-speculate-indirect-jumps.
25640         (*call_indirect_aix<mode>_nospec): New define_insn.
25641         (*call_value_indirect_aix<mode>): Disable for
25642         -mno-speculate-indirect-jumps.
25643         (*call_value_indirect_aix<mode>_nospec): New define_insn.
25644         (*sibcall_nonlocal_sysv<mode>): Generate different code for
25645         -mno-speculate-indirect-jumps.
25646         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25648 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
25650         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25651         long double type, set the flags for noting the default long double
25652         type, even if we don't pass or return a long double type.
25654 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
25656         PR ipa/83051
25657         * ipa-inline.c (flatten_function): Do not overwrite final inlining
25658         failure.
25660 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
25662         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25663         support for merge[hl].
25664         (fold_mergehl_helper): New helper function.
25665         (tree-vector-builder.h): New #include for tree_vector_builder usage.
25666         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25667         (altivec_vmrglw_direct): Add xxmrglw insn.
25669 2018-01-17  Andrew Waterman  <andrew@sifive.com>
25671         * config/riscv/riscv.c (riscv_conditional_register_usage): If
25672         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25674 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
25676         PR lto/83121
25677         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25678         call the lto_location_cache before reading the
25679         DECL_SOURCE_LOCATION of the types.
25681 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
25682             Richard Sandiford  <richard.sandiford@linaro.org>
25684         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25685         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25686         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25687         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25688         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25689         Add declaration.
25690         * config/aarch64/constraints.md (aarch64_movti_operand):
25691         Limit immediates.
25692         * config/aarch64/predicates.md (Uti): Add new constraint.
25694 2018-01-17  Carl Love  <cel@us.ibm.com>
25696         * config/rs6000/vsx.md (define_expand xl_len_r,
25697         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25698         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25699         lxvll.
25700         (define_expand, define_insn): Move the shift left from  the
25701         define_insn to the define_expand for lxvl and stxvl instructions.
25702         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25703         and XL_LEN_R definitions to PURE.
25705 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
25707         * config/i386/i386.c (indirect_thunk_name): Declare regno
25708         as unsigned int.  Compare regno with INVALID_REGNUM.
25709         (output_indirect_thunk): Ditto.
25710         (output_indirect_thunk_function): Ditto.
25711         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
25712         in the call to output_indirect_thunk_function.
25714 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
25716         PR middle-end/83884
25717         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25718         rather than the size of inner_type to determine the stack slot size
25719         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25721 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
25723         PR target/83546
25724         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25725         to PTA_SILVERMONT.
25727 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
25729         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25730         endian Linux systems to optionally enable multilibs for selecting
25731         the long double type if the user configured an explicit type.
25732         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25733         have no long double multilibs if not defined.
25734         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25735         warn if the user used -mabi={ieee,ibm}longdouble and we built
25736         multilibs for long double.
25737         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25738         appropriate multilib option.
25739         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25740         multilib options.
25741         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25742         for building long double multilibs.
25743         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25745 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
25747         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25748         copies.
25750         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25751         64 bits.
25752         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25753         128 bits.
25755         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25756         variables.
25758         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25759         return value.
25761 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25763         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25764         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25766 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25768         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25769         different rtl trees depending on TARGET_64BIT.
25770         (rs6000_gen_lvx): Likewise.
25772 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25774         * config/visium/visium.md (nop): Tweak comment.
25775         (hazard_nop): Likewise.
25777 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25779         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25780         -mspeculate-indirect-jumps.
25781         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25782         for -mno-speculate-indirect-jumps.
25783         (*call_indirect_elfv2<mode>_nospec): New define_insn.
25784         (*call_value_indirect_elfv2<mode>): Disable for
25785         -mno-speculate-indirect-jumps.
25786         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25787         (indirect_jump): Emit different RTL for
25788         -mno-speculate-indirect-jumps.
25789         (*indirect_jump<mode>): Disable for
25790         -mno-speculate-indirect-jumps.
25791         (*indirect_jump<mode>_nospec): New define_insn.
25792         (tablejump): Emit different RTL for
25793         -mno-speculate-indirect-jumps.
25794         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25795         (tablejumpsi_nospec): New define_expand.
25796         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25797         (tablejumpdi_nospec): New define_expand.
25798         (*tablejump<mode>_internal1): Disable for
25799         -mno-speculate-indirect-jumps.
25800         (*tablejump<mode>_internal1_nospec): New define_insn.
25801         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25802         option.
25804 2018-01-16  Artyom Skrobov tyomitch@gmail.com
25806         * caller-save.c (insert_save): Drop unnecessary parameter.  All
25807         callers updated.
25809 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25810             Richard Biener  <rguenth@suse.de>
25812         PR libgomp/83590
25813         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25814         return early, inline manually is_gimple_sizepos.  Make sure if we
25815         call gimplify_expr we don't end up with a gimple constant.
25816         * tree.c (variably_modified_type_p): Don't return true for
25817         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
25818         * gimplify.h (is_gimple_sizepos): Remove.
25820 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25822         PR tree-optimization/83857
25823         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25824         vectorizable_live_operation for pure SLP statements.
25825         (vectorizable_live_operation): Handle PHIs.
25827 2018-01-16  Richard Biener  <rguenther@suse.de>
25829         PR tree-optimization/83867
25830         * tree-vect-stmts.c (vect_transform_stmt): Precompute
25831         nested_in_vect_loop_p since the scalar stmt may get invalidated.
25833 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25835         PR c/83844
25836         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25837         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25838         If off is not INTEGER_CST, issue a may not be aligned warning
25839         rather than isn't aligned.  Use isn%'t rather than isn't.
25840         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25841         into MULT_EXPR.
25842         <case MULT_EXPR>: Improve the case when bottom and one of the
25843         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25844         operand, in that case check if the other operand is multiple of
25845         bottom divided by the INTEGER_CST operand.
25847 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25849         PR target/83858
25850         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25851         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25852         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25853         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25854         * config/pa/pa.c (pa_function_arg_advance): Likewise.
25855         (pa_function_arg, pa_arg_partial_bytes): Likewise.
25856         (pa_function_arg_size): New function.
25858 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25860         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25861         in a separate statement.
25863 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25865         PR tree-optimization/83847
25866         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25867         group gathers and scatters.
25869 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25871         PR rtl-optimization/86620
25872         * params.def (max-sched-ready-insns): Bump minimum value to 1.
25874         PR rtl-optimization/83213
25875         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25876         to last if both are JUMP_INSNs.
25878         PR tree-optimization/83843
25879         * gimple-ssa-store-merging.c
25880         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25881         store_immediate_info for bswap/nop orig_stores.
25883 2018-01-15  Andrew Waterman  <andrew@sifive.com>
25885         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25886         !TARGET_MUL.
25887         <UDIV>: Increase cost if !TARGET_DIV.
25889 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
25891         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25892         (define_attr "cr_logical_3op"): New.
25893         (cceq_ior_compare): Adjust.
25894         (cceq_ior_compare_complement): Adjust.
25895         (*cceq_rev_compare): Adjust.
25896         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25897         (is_cracked_insn): Adjust.
25898         (insn_must_be_first_in_group): Adjust.
25899         * config/rs6000/40x.md: Adjust.
25900         * config/rs6000/440.md: Adjust.
25901         * config/rs6000/476.md: Adjust.
25902         * config/rs6000/601.md: Adjust.
25903         * config/rs6000/603.md: Adjust.
25904         * config/rs6000/6xx.md: Adjust.
25905         * config/rs6000/7450.md: Adjust.
25906         * config/rs6000/7xx.md: Adjust.
25907         * config/rs6000/8540.md: Adjust.
25908         * config/rs6000/cell.md: Adjust.
25909         * config/rs6000/e300c2c3.md: Adjust.
25910         * config/rs6000/e500mc.md: Adjust.
25911         * config/rs6000/e500mc64.md: Adjust.
25912         * config/rs6000/e5500.md: Adjust.
25913         * config/rs6000/e6500.md: Adjust.
25914         * config/rs6000/mpc.md: Adjust.
25915         * config/rs6000/power4.md: Adjust.
25916         * config/rs6000/power5.md: Adjust.
25917         * config/rs6000/power6.md: Adjust.
25918         * config/rs6000/power7.md: Adjust.
25919         * config/rs6000/power8.md: Adjust.
25920         * config/rs6000/power9.md: Adjust.
25921         * config/rs6000/rs64.md: Adjust.
25922         * config/rs6000/titan.md: Adjust.
25924 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25926         * config/i386/predicates.md (indirect_branch_operand): Rewrite
25927         ix86_indirect_branch_register logic.
25929 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25931         * config/i386/constraints.md (Bs): Update
25932         ix86_indirect_branch_register check.  Don't check
25933         ix86_indirect_branch_register with GOT_memory_operand.
25934         (Bw): Likewise.
25935         * config/i386/predicates.md (GOT_memory_operand): Don't check
25936         ix86_indirect_branch_register here.
25937         (GOT32_symbol_operand): Likewise.
25939 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25941         * config/i386/predicates.md (constant_call_address_operand):
25942         Rewrite ix86_indirect_branch_register logic.
25943         (sibcall_insn_operand): Likewise.
25945 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25947         * config/i386/constraints.md (Bs): Replace
25948         ix86_indirect_branch_thunk_register with
25949         ix86_indirect_branch_register.
25950         (Bw): Likewise.
25951         * config/i386/i386.md (indirect_jump): Likewise.
25952         (tablejump): Likewise.
25953         (*sibcall_memory): Likewise.
25954         (*sibcall_value_memory): Likewise.
25955         Peepholes of indirect call and jump via memory: Likewise.
25956         * config/i386/i386.opt: Likewise.
25957         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25958         (GOT_memory_operand): Likewise.
25959         (call_insn_operand): Likewise.
25960         (sibcall_insn_operand): Likewise.
25961         (GOT32_symbol_operand): Likewise.
25963 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25965         PR middle-end/83837
25966         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25967         type rather than type addr's type points to.
25968         (expand_omp_atomic_mutex): Likewise.
25969         (expand_omp_atomic): Likewise.
25971 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25973         PR target/83839
25974         * config/i386/i386.c (output_indirect_thunk_function): Use
25975         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25976         for  __x86_return_thunk.
25978 2018-01-15  Richard Biener  <rguenther@suse.de>
25980         PR middle-end/83850
25981         * expmed.c (extract_bit_field_1): Fix typo.
25983 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25985         PR target/83687
25986         * config/arm/iterators.md (VF): New mode iterator.
25987         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25988         Remove integer-related logic from pattern.
25989         (neon_vabd<mode>_3): Likewise.
25991 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25993         PR middle-end/82694
25994         * common.opt (fstrict-overflow): No longer an alias.
25995         (fwrapv-pointer): New option.
25996         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25997         also for pointer types based on flag_wrapv_pointer.
25998         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25999         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
26000         opts->x_flag_wrapv got set.
26001         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
26002         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
26003         POINTER_TYPE_OVERFLOW_UNDEFINED.
26004         * match.pd: Likewise in address comparison pattern.
26005         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
26007 2018-01-15  Richard Biener  <rguenther@suse.de>
26009         PR lto/83804
26010         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
26011         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
26012         Reset type names to their identifier if their TYPE_DECL doesn't
26013         have linkage (and thus is used for ODR and devirt).
26014         (save_debug_info_for_decl): Remove.
26015         (save_debug_info_for_type): Likewise.
26016         (add_tree_to_fld_list): Adjust.
26017         * tree-pretty-print.c (dump_generic_node): Make dumping of
26018         type names more robust.
26020 2018-01-15  Richard Biener  <rguenther@suse.de>
26022         * BASE-VER: Bump to 8.0.1.
26024 2018-01-14  Martin Sebor  <msebor@redhat.com>
26026         PR other/83508
26027         * builtins.c (check_access): Avoid warning when the no-warning bit
26028         is set.
26030 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
26032         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
26033         * ira-color (allocno_hard_regs_compare): Likewise.
26035 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
26037         PR target/83013
26038         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
26039         Use .pushsection/.popsection.
26041 2018-01-14  Martin Sebor  <msebor@redhat.com>
26043         PR c++/81327
26044         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
26046 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
26048         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
26049         entry from extra_headers.
26050         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
26051         extra_headers, make the list bitwise identical to the i?86-*-* one.
26053 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26055         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
26056         -mcmodel=large with -mindirect-branch=thunk,
26057         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
26058         -mfunction-return=thunk-extern.
26059         * doc/invoke.texi: Document -mcmodel=large is incompatible with
26060         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
26061         -mfunction-return=thunk and -mfunction-return=thunk-extern.
26063 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26065         * config/i386/i386.c (print_reg): Print the name of the full
26066         integer register without '%'.
26067         (ix86_print_operand): Handle 'V'.
26068         * doc/extend.texi: Document 'V' modifier.
26070 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26072         * config/i386/constraints.md (Bs): Disallow memory operand for
26073         -mindirect-branch-register.
26074         (Bw): Likewise.
26075         * config/i386/predicates.md (indirect_branch_operand): Likewise.
26076         (GOT_memory_operand): Likewise.
26077         (call_insn_operand): Likewise.
26078         (sibcall_insn_operand): Likewise.
26079         (GOT32_symbol_operand): Likewise.
26080         * config/i386/i386.md (indirect_jump): Call convert_memory_address
26081         for -mindirect-branch-register.
26082         (tablejump): Likewise.
26083         (*sibcall_memory): Likewise.
26084         (*sibcall_value_memory): Likewise.
26085         Disallow peepholes of indirect call and jump via memory for
26086         -mindirect-branch-register.
26087         (*call_pop): Replace m with Bw.
26088         (*call_value_pop): Likewise.
26089         (*sibcall_pop_memory): Replace m with Bs.
26090         * config/i386/i386.opt (mindirect-branch-register): New option.
26091         * doc/invoke.texi: Document -mindirect-branch-register option.
26093 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26095         * config/i386/i386-protos.h (ix86_output_function_return): New.
26096         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
26097         set function_return_type.
26098         (indirect_thunk_name): Add ret_p to indicate thunk for function
26099         return.
26100         (output_indirect_thunk_function): Pass false to
26101         indirect_thunk_name.
26102         (ix86_output_indirect_branch_via_reg): Likewise.
26103         (ix86_output_indirect_branch_via_push): Likewise.
26104         (output_indirect_thunk_function): Create alias for function
26105         return thunk if regno < 0.
26106         (ix86_output_function_return): New function.
26107         (ix86_handle_fndecl_attribute): Handle function_return.
26108         (ix86_attribute_table): Add function_return.
26109         * config/i386/i386.h (machine_function): Add
26110         function_return_type.
26111         * config/i386/i386.md (simple_return_internal): Use
26112         ix86_output_function_return.
26113         (simple_return_internal_long): Likewise.
26114         * config/i386/i386.opt (mfunction-return=): New option.
26115         (indirect_branch): Mention -mfunction-return=.
26116         * doc/extend.texi: Document function_return function attribute.
26117         * doc/invoke.texi: Document -mfunction-return= option.
26119 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26121         * config/i386/i386-opts.h (indirect_branch): New.
26122         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
26123         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
26124         with local indirect jump when converting indirect call and jump.
26125         (ix86_set_indirect_branch_type): New.
26126         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
26127         (indirectlabelno): New.
26128         (indirect_thunk_needed): Likewise.
26129         (indirect_thunk_bnd_needed): Likewise.
26130         (indirect_thunks_used): Likewise.
26131         (indirect_thunks_bnd_used): Likewise.
26132         (INDIRECT_LABEL): Likewise.
26133         (indirect_thunk_name): Likewise.
26134         (output_indirect_thunk): Likewise.
26135         (output_indirect_thunk_function): Likewise.
26136         (ix86_output_indirect_branch_via_reg): Likewise.
26137         (ix86_output_indirect_branch_via_push): Likewise.
26138         (ix86_output_indirect_branch): Likewise.
26139         (ix86_output_indirect_jmp): Likewise.
26140         (ix86_code_end): Call output_indirect_thunk_function if needed.
26141         (ix86_output_call_insn): Call ix86_output_indirect_branch if
26142         needed.
26143         (ix86_handle_fndecl_attribute): Handle indirect_branch.
26144         (ix86_attribute_table): Add indirect_branch.
26145         * config/i386/i386.h (machine_function): Add indirect_branch_type
26146         and has_local_indirect_jump.
26147         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
26148         to true.
26149         (tablejump): Likewise.
26150         (*indirect_jump): Use ix86_output_indirect_jmp.
26151         (*tablejump_1): Likewise.
26152         (simple_return_indirect_internal): Likewise.
26153         * config/i386/i386.opt (mindirect-branch=): New option.
26154         (indirect_branch): New.
26155         (keep): Likewise.
26156         (thunk): Likewise.
26157         (thunk-inline): Likewise.
26158         (thunk-extern): Likewise.
26159         * doc/extend.texi: Document indirect_branch function attribute.
26160         * doc/invoke.texi: Document -mindirect-branch= option.
26162 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
26164         PR ipa/83051
26165         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
26167 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
26169         * ipa-inline.c (want_inline_small_function_p): Return false if
26170         inlining has already failed with CIF_FINAL_ERROR.
26171         (update_caller_keys): Call want_inline_small_function_p before
26172         can_inline_edge_p.
26173         (update_callee_keys): Likewise.
26175 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26177         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
26178         New function.
26179         (rs6000_quadword_masked_address_p): Likewise.
26180         (quad_aligned_load_p): Likewise.
26181         (quad_aligned_store_p): Likewise.
26182         (const_load_sequence_p): Add comment to describe the outer-most loop.
26183         (mimic_memory_attributes_and_flags): New function.
26184         (rs6000_gen_stvx): Likewise.
26185         (replace_swapped_aligned_store): Likewise.
26186         (rs6000_gen_lvx): Likewise.
26187         (replace_swapped_aligned_load): Likewise.
26188         (replace_swapped_load_constant): Capitalize argument name in
26189         comment describing this function.
26190         (rs6000_analyze_swaps): Add a third pass to search for vector loads
26191         and stores that access quad-word aligned addresses and replace
26192         with stvx or lvx instructions when appropriate.
26193         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
26194         New function prototype.
26195         (rs6000_quadword_masked_address_p): Likewise.
26196         (rs6000_gen_lvx): Likewise.
26197         (rs6000_gen_stvx): Likewise.
26198         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
26199         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
26200         when memory address is aligned.
26201         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
26202         this split to select lvx instruction when memory address is aligned.
26203         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
26204         instruction when memory address is aligned.
26205         (*vsx_le_perm_load_v16qi): Likewise.
26206         (four unnamed splitters): Modify to select the stvx instruction
26207         when memory is aligned.
26209 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
26211         * predict.c (determine_unlikely_bbs): Handle correctly BBs
26212         which appears in the queue multiple times.
26214 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26215             Alan Hayward  <alan.hayward@arm.com>
26216             David Sherwood  <david.sherwood@arm.com>
26218         * tree-vectorizer.h (vec_lower_bound): New structure.
26219         (_loop_vec_info): Add check_nonzero and lower_bounds.
26220         (LOOP_VINFO_CHECK_NONZERO): New macro.
26221         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
26222         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
26223         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
26224         fields.  Make seg_len the distance travelled, not including the
26225         access size.
26226         (dr_direction_indicator): Declare.
26227         (dr_zero_step_indicator): Likewise.
26228         (dr_known_forward_stride_p): Likewise.
26229         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
26230         tree-ssanames.h.
26231         (runtime_alias_check_p): Allow runtime alias checks with
26232         variable strides.
26233         (operator ==): Compare access_size and align.
26234         (prune_runtime_alias_test_list): Rework for new distinction between
26235         the access_size and seg_len.
26236         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
26237         segment lengths.
26238         (get_segment_min_max): New function.
26239         (create_intersect_range_checks): Use it.
26240         (dr_step_indicator): New function.
26241         (dr_direction_indicator): Likewise.
26242         (dr_zero_step_indicator): Likewise.
26243         (dr_known_forward_stride_p): Likewise.
26244         * tree-loop-distribution.c (data_ref_segment_size): Return
26245         DR_STEP * (niters - 1).
26246         (compute_alias_check_pairs): Update call to the dr_with_seg_len
26247         constructor.
26248         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
26249         (vect_preserves_scalar_order_p): New function, split out from...
26250         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
26251         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
26252         (vect_vfa_access_size): New function.
26253         (vect_vfa_align): Likewise.
26254         (vect_compile_time_alias): Take access_size_a and access_b arguments.
26255         (dump_lower_bound): New function.
26256         (vect_check_lower_bound): Likewise.
26257         (vect_small_gap_p): Likewise.
26258         (vectorizable_with_step_bound_p): Likewise.
26259         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
26260         depencies if the vectorization factor is 1.  Convert the checks
26261         for nonzero steps into checks on the bounds of DR_STEP.  Try using
26262         a bunds check for variable steps if the minimum required step is
26263         relatively small. Update calls to the dr_with_seg_len
26264         constructor and to vect_compile_time_alias.
26265         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
26266         function.
26267         (vect_loop_versioning): Call it.
26268         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
26269         when retrying.
26270         (vect_estimate_min_profitable_iters): Account for any bounds checks.
26272 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26273             Alan Hayward  <alan.hayward@arm.com>
26274             David Sherwood  <david.sherwood@arm.com>
26276         * doc/sourcebuild.texi (vect_scatter_store): Document.
26277         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
26278         optabs.
26279         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
26280         Document.
26281         * genopinit.c (main): Add supports_vec_scatter_store and
26282         supports_vec_scatter_store_cached to target_optabs.
26283         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
26284         IFN_MASK_SCATTER_STORE.
26285         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
26286         functions.
26287         * internal-fn.h (internal_store_fn_p): Declare.
26288         (internal_fn_stored_value_index): Likewise.
26289         * internal-fn.c (scatter_store_direct): New macro.
26290         (expand_scatter_store_optab_fn): New function.
26291         (direct_scatter_store_optab_supported_p): New macro.
26292         (internal_store_fn_p): New function.
26293         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
26294         IFN_MASK_SCATTER_STORE.
26295         (internal_fn_mask_index): Likewise.
26296         (internal_fn_stored_value_index): New function.
26297         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
26298         for scatter stores.
26299         * optabs-query.h (supports_vec_scatter_store_p): Declare.
26300         * optabs-query.c (supports_vec_scatter_store_p): New function.
26301         * tree-vectorizer.h (vect_get_store_rhs): Declare.
26302         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
26303         true for scatter stores.
26304         (vect_gather_scatter_fn_p): Handle scatter stores too.
26305         (vect_check_gather_scatter): Consider using scatter stores if
26306         supports_vec_scatter_store_p.
26307         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
26308         scatter stores too.
26309         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26310         internal_fn_stored_value_index.
26311         (check_load_store_masking): Handle scatter stores too.
26312         (vect_get_store_rhs): Make public.
26313         (vectorizable_call): Use internal_store_fn_p.
26314         (vectorizable_store): Handle scatter store internal functions.
26315         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
26316         when deciding whether the end of the group has been reached.
26317         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
26318         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
26319         (mask_scatter_store<mode>): New insns.
26321 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26322             Alan Hayward  <alan.hayward@arm.com>
26323             David Sherwood  <david.sherwood@arm.com>
26325         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
26326         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
26327         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
26328         function.
26329         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
26330         Use vect_truncate_gather_scatter_offset if we can't treat the
26331         operation as a normal gather load or scatter store.
26332         (get_group_load_store_type): Take the gather_scatter_info
26333         as argument.  Try using a gather load or scatter store for
26334         single-element groups.
26335         (get_load_store_type): Update calls to get_group_load_store_type
26336         and vect_use_strided_gather_scatters_p.
26338 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26339             Alan Hayward  <alan.hayward@arm.com>
26340             David Sherwood  <david.sherwood@arm.com>
26342         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
26343         optional tree argument.
26344         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
26345         null target hooks.
26346         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
26347         but continue to use the current value as a fallback.
26348         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
26349         to compare the updates.
26350         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
26351         (get_load_store_type): Use it when handling a strided access.
26352         (vect_get_strided_load_store_ops): New function.
26353         (vect_get_data_ptr_increment): Likewise.
26354         (vectorizable_load): Handle strided gather loads.  Always pass
26355         a step to vect_create_data_ref_ptr and bump_vector_ptr.
26357 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26358             Alan Hayward  <alan.hayward@arm.com>
26359             David Sherwood  <david.sherwood@arm.com>
26361         * doc/md.texi (gather_load@var{m}): Document.
26362         (mask_gather_load@var{m}): Likewise.
26363         * genopinit.c (main): Add supports_vec_gather_load and
26364         supports_vec_gather_load_cached to target_optabs.
26365         * optabs-tree.c (init_tree_optimization_optabs): Use
26366         ggc_cleared_alloc to allocate target_optabs.
26367         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
26368         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
26369         functions.
26370         * internal-fn.h (internal_load_fn_p): Declare.
26371         (internal_gather_scatter_fn_p): Likewise.
26372         (internal_fn_mask_index): Likewise.
26373         (internal_gather_scatter_fn_supported_p): Likewise.
26374         * internal-fn.c (gather_load_direct): New macro.
26375         (expand_gather_load_optab_fn): New function.
26376         (direct_gather_load_optab_supported_p): New macro.
26377         (direct_internal_fn_optab): New function.
26378         (internal_load_fn_p): Likewise.
26379         (internal_gather_scatter_fn_p): Likewise.
26380         (internal_fn_mask_index): Likewise.
26381         (internal_gather_scatter_fn_supported_p): Likewise.
26382         * optabs-query.c (supports_at_least_one_mode_p): New function.
26383         (supports_vec_gather_load_p): Likewise.
26384         * optabs-query.h (supports_vec_gather_load_p): Declare.
26385         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
26386         and memory_type field.
26387         (NUM_PATTERNS): Bump to 15.
26388         * tree-vect-data-refs.c: Include internal-fn.h.
26389         (vect_gather_scatter_fn_p): New function.
26390         (vect_describe_gather_scatter_call): Likewise.
26391         (vect_check_gather_scatter): Try using internal functions for
26392         gather loads.  Recognize existing calls to a gather load function.
26393         (vect_analyze_data_refs): Consider using gather loads if
26394         supports_vec_gather_load_p.
26395         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
26396         (vect_get_gather_scatter_offset_type): Likewise.
26397         (vect_convert_mask_for_vectype): Likewise.
26398         (vect_add_conversion_to_patterm): Likewise.
26399         (vect_try_gather_scatter_pattern): Likewise.
26400         (vect_recog_gather_scatter_pattern): New pattern recognizer.
26401         (vect_vect_recog_func_ptrs): Add it.
26402         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
26403         internal_fn_mask_index and internal_gather_scatter_fn_p.
26404         (check_load_store_masking): Take the gather_scatter_info as an
26405         argument and handle gather loads.
26406         (vect_get_gather_scatter_ops): New function.
26407         (vectorizable_call): Check internal_load_fn_p.
26408         (vectorizable_load): Likewise.  Handle gather load internal
26409         functions.
26410         (vectorizable_store): Update call to check_load_store_masking.
26411         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
26412         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
26413         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
26414         (aarch64_gather_scale_operand_d): New predicates.
26415         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
26416         (mask_gather_load<mode>): New insns.
26418 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26419             Alan Hayward  <alan.hayward@arm.com>
26420             David Sherwood  <david.sherwood@arm.com>
26422         * optabs.def (fold_left_plus_optab): New optab.
26423         * doc/md.texi (fold_left_plus_@var{m}): Document.
26424         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
26425         * internal-fn.c (fold_left_direct): Define.
26426         (expand_fold_left_optab_fn): Likewise.
26427         (direct_fold_left_optab_supported_p): Likewise.
26428         * fold-const-call.c (fold_const_fold_left): New function.
26429         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
26430         * tree-parloops.c (valid_reduction_p): New function.
26431         (gather_scalar_reductions): Use it.
26432         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
26433         (vect_finish_replace_stmt): Declare.
26434         * tree-vect-loop.c (fold_left_reduction_fn): New function.
26435         (needs_fold_left_reduction_p): New function, split out from...
26436         (vect_is_simple_reduction): ...here.  Accept reductions that
26437         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
26438         (vect_force_simple_reduction): Also store the reduction type in
26439         the assignment's STMT_VINFO_REDUC_TYPE.
26440         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
26441         (merge_with_identity): New function.
26442         (vect_expand_fold_left): Likewise.
26443         (vectorize_fold_left_reduction): Likewise.
26444         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
26445         scalar phi in place for it.  Check for target support and reject
26446         cases that would reassociate the operation.  Defer the transform
26447         phase to vectorize_fold_left_reduction.
26448         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
26449         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
26450         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
26452 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26454         * tree-if-conv.c (predicate_mem_writes): Remove redundant
26455         call to ifc_temp_var.
26457 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26458             Alan Hayward  <alan.hayward@arm.com>
26459             David Sherwood  <david.sherwood@arm.com>
26461         * target.def (legitimize_address_displacement): Take the original
26462         offset as a poly_int.
26463         * targhooks.h (default_legitimize_address_displacement): Update
26464         accordingly.
26465         * targhooks.c (default_legitimize_address_displacement): Likewise.
26466         * doc/tm.texi: Regenerate.
26467         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
26468         as an argument, moving assert of ad->disp == ad->disp_term to...
26469         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
26470         Try calling targetm.legitimize_address_displacement before expanding
26471         the address rather than afterwards, and adjust for the new interface.
26472         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
26473         Match the new hook interface.  Handle SVE addresses.
26474         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
26475         new hook interface.
26477 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26479         * Makefile.in (OBJS): Add early-remat.o.
26480         * target.def (select_early_remat_modes): New hook.
26481         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
26482         * doc/tm.texi: Regenerate.
26483         * targhooks.h (default_select_early_remat_modes): Declare.
26484         * targhooks.c (default_select_early_remat_modes): New function.
26485         * timevar.def (TV_EARLY_REMAT): New timevar.
26486         * passes.def (pass_early_remat): New pass.
26487         * tree-pass.h (make_pass_early_remat): Declare.
26488         * early-remat.c: New file.
26489         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
26490         function.
26491         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
26493 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26494             Alan Hayward  <alan.hayward@arm.com>
26495             David Sherwood  <david.sherwood@arm.com>
26497         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
26498         vfm1 with a bound_epilog parameter.
26499         (vect_do_peeling): Update calls accordingly, and move the prologue
26500         call earlier in the function.  Treat the base bound_epilog as 0 for
26501         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
26502         this base when peeling for gaps.
26503         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
26504         with fully-masked loops.
26505         (vect_estimate_min_profitable_iters): Handle the single peeled
26506         iteration in that case.
26508 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26509             Alan Hayward  <alan.hayward@arm.com>
26510             David Sherwood  <david.sherwood@arm.com>
26512         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
26513         single-element interleaving even if the size is not a power of 2.
26514         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
26515         accesses for single-element interleaving if the group size is
26516         not a power of 2.
26518 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26519             Alan Hayward  <alan.hayward@arm.com>
26520             David Sherwood  <david.sherwood@arm.com>
26522         * doc/md.texi (fold_extract_last_@var{m}): Document.
26523         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
26524         * optabs.def (fold_extract_last_optab): New optab.
26525         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
26526         * internal-fn.c (fold_extract_direct): New macro.
26527         (expand_fold_extract_optab_fn): Likewise.
26528         (direct_fold_extract_optab_supported_p): Likewise.
26529         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
26530         * tree-vect-loop.c (vect_model_reduction_cost): Handle
26531         EXTRACT_LAST_REDUCTION.
26532         (get_initial_def_for_reduction): Do not create an initial vector
26533         for EXTRACT_LAST_REDUCTION reductions.
26534         (vectorizable_reduction): Leave the scalar phi in place for
26535         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
26536         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
26537         epilogue code for EXTRACT_LAST_REDUCTION and defer the
26538         transform phase to vectorizable_condition.
26539         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
26540         split out from...
26541         (vect_finish_stmt_generation): ...here.
26542         (vect_finish_replace_stmt): New function.
26543         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
26544         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
26545         pattern.
26546         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
26548 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26549             Alan Hayward  <alan.hayward@arm.com>
26550             David Sherwood  <david.sherwood@arm.com>
26552         * doc/md.texi (extract_last_@var{m}): Document.
26553         * optabs.def (extract_last_optab): New optab.
26554         * internal-fn.def (EXTRACT_LAST): New internal function.
26555         * internal-fn.c (cond_unary_direct): New macro.
26556         (expand_cond_unary_optab_fn): Likewise.
26557         (direct_cond_unary_optab_supported_p): Likewise.
26558         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
26559         loops using EXTRACT_LAST.
26560         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
26561         (extract_last_<mode>): ...this optab.
26562         (vec_extract<mode><Vel>): Update accordingly.
26564 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26565             Alan Hayward  <alan.hayward@arm.com>
26566             David Sherwood  <david.sherwood@arm.com>
26568         * target.def (empty_mask_is_expensive): New hook.
26569         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
26570         * doc/tm.texi: Regenerate.
26571         * targhooks.h (default_empty_mask_is_expensive): Declare.
26572         * targhooks.c (default_empty_mask_is_expensive): New function.
26573         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
26574         if the target says that empty masks are expensive.
26575         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
26576         New function.
26577         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
26579 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26580             Alan Hayward  <alan.hayward@arm.com>
26581             David Sherwood  <david.sherwood@arm.com>
26583         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
26584         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
26585         (vect_use_loop_mask_for_alignment_p): New function.
26586         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
26587         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
26588         niters_skip argument.  Make sure that the first niters_skip elements
26589         of the first iteration are inactive.
26590         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
26591         Update call to vect_set_loop_masks_directly.
26592         (get_misalign_in_elems): New function, split out from...
26593         (vect_gen_prolog_loop_niters): ...here.
26594         (vect_update_init_of_dr): Take a code argument that specifies whether
26595         the adjustment should be added or subtracted.
26596         (vect_update_init_of_drs): Likewise.
26597         (vect_prepare_for_masked_peels): New function.
26598         (vect_do_peeling): Skip prologue peeling if we're using a mask
26599         instead.  Update call to vect_update_inits_of_drs.
26600         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26601         mask_skip_niters.
26602         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
26603         alignment.  Do not include the number of peeled iterations in
26604         the minimum threshold in that case.
26605         (vectorizable_induction): Adjust the start value down by
26606         LOOP_VINFO_MASK_SKIP_NITERS iterations.
26607         (vect_transform_loop): Call vect_prepare_for_masked_peels.
26608         Take the number of skipped iterations into account when calculating
26609         the loop bounds.
26610         * tree-vect-stmts.c (vect_gen_while_not): New function.
26612 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26613             Alan Hayward  <alan.hayward@arm.com>
26614             David Sherwood  <david.sherwood@arm.com>
26616         * doc/sourcebuild.texi (vect_fully_masked): Document.
26617         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26618         default value to 0.
26619         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26620         split out from...
26621         (vect_analyze_loop_2): ...here. Don't check the vectorization
26622         factor against the number of loop iterations if the loop is
26623         fully-masked.
26625 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26626             Alan Hayward  <alan.hayward@arm.com>
26627             David Sherwood  <david.sherwood@arm.com>
26629         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26630         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26631         (dump_groups): Update accordingly.
26632         (iv_use::mem_type): New member variable.
26633         (address_p): New function.
26634         (record_use): Add a mem_type argument and initialize the new
26635         mem_type field.
26636         (record_group_use): Add a mem_type argument.  Use address_p.
26637         Remove obsolete null checks of base_object.  Update call to record_use.
26638         (find_interesting_uses_op): Update call to record_group_use.
26639         (find_interesting_uses_cond): Likewise.
26640         (find_interesting_uses_address): Likewise.
26641         (get_mem_type_for_internal_fn): New function.
26642         (find_address_like_use): Likewise.
26643         (find_interesting_uses_stmt): Try find_address_like_use before
26644         calling find_interesting_uses_op.
26645         (addr_offset_valid_p): Use the iv mem_type field as the type
26646         of the addressed memory.
26647         (add_autoinc_candidates): Likewise.
26648         (get_address_cost): Likewise.
26649         (split_small_address_groups_p): Use address_p.
26650         (split_address_groups): Likewise.
26651         (add_iv_candidate_for_use): Likewise.
26652         (autoinc_possible_for_pair): Likewise.
26653         (rewrite_groups): Likewise.
26654         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26655         (determine_group_iv_cost): Update after split of USE_ADDRESS.
26656         (get_alias_ptr_type_for_ptr_address): New function.
26657         (rewrite_use_address): Rewrite address uses in calls that were
26658         identified by find_address_like_use.
26660 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26661             Alan Hayward  <alan.hayward@arm.com>
26662             David Sherwood  <david.sherwood@arm.com>
26664         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26665         TARGET_MEM_REFs.
26666         * gimple-expr.h (is_gimple_addressable: Likewise.
26667         * gimple-expr.c (is_gimple_address): Likewise.
26668         * internal-fn.c (expand_call_mem_ref): New function.
26669         (expand_mask_load_optab_fn): Use it.
26670         (expand_mask_store_optab_fn): Likewise.
26672 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26673             Alan Hayward  <alan.hayward@arm.com>
26674             David Sherwood  <david.sherwood@arm.com>
26676         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26677         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26678         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26679         (cond_umax@var{mode}): Document.
26680         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26681         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26682         (cond_umin_optab, cond_umax_optab): New optabs.
26683         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26684         (COND_IOR, COND_XOR): New internal functions.
26685         * internal-fn.h (get_conditional_internal_fn): Declare.
26686         * internal-fn.c (cond_binary_direct): New macro.
26687         (expand_cond_binary_optab_fn): Likewise.
26688         (direct_cond_binary_optab_supported_p): Likewise.
26689         (get_conditional_internal_fn): New function.
26690         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26691         Cope with reduction statements that are vectorized as calls rather
26692         than assignments.
26693         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26694         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26695         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26696         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26697         (UNSPEC_COND_EOR): New unspecs.
26698         (optab): Add mappings for them.
26699         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26700         (sve_int_op, sve_fp_op): New int attributes.
26702 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26703             Alan Hayward  <alan.hayward@arm.com>
26704             David Sherwood  <david.sherwood@arm.com>
26706         * optabs.def (while_ult_optab): New optab.
26707         * doc/md.texi (while_ult@var{m}@var{n}): Document.
26708         * internal-fn.def (WHILE_ULT): New internal function.
26709         * internal-fn.h (direct_internal_fn_supported_p): New override
26710         that takes two types as argument.
26711         * internal-fn.c (while_direct): New macro.
26712         (expand_while_optab_fn): New function.
26713         (convert_optab_supported_p): Likewise.
26714         (direct_while_optab_supported_p): New macro.
26715         * wide-int.h (wi::udiv_ceil): New function.
26716         * tree-vectorizer.h (rgroup_masks): New structure.
26717         (vec_loop_masks): New typedef.
26718         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26719         and fully_masked_p.
26720         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26721         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26722         (vect_max_vf): New function.
26723         (slpeel_make_loop_iterate_ntimes): Delete.
26724         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26725         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26726         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26727         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26728         internal-fn.h, stor-layout.h and optabs-query.h.
26729         (vect_set_loop_mask): New function.
26730         (add_preheader_seq): Likewise.
26731         (add_header_seq): Likewise.
26732         (interleave_supported_p): Likewise.
26733         (vect_maybe_permute_loop_masks): Likewise.
26734         (vect_set_loop_masks_directly): Likewise.
26735         (vect_set_loop_condition_masked): Likewise.
26736         (vect_set_loop_condition_unmasked): New function, split out from
26737         slpeel_make_loop_iterate_ntimes.
26738         (slpeel_make_loop_iterate_ntimes): Rename to..
26739         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
26740         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26741         (vect_do_peeling): Update call accordingly.
26742         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26743         loops.
26744         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26745         mask_compare_type, can_fully_mask_p and fully_masked_p.
26746         (release_vec_loop_masks): New function.
26747         (_loop_vec_info): Use it to free the loop masks.
26748         (can_produce_all_loop_masks_p): New function.
26749         (vect_get_max_nscalars_per_iter): Likewise.
26750         (vect_verify_full_masking): Likewise.
26751         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26752         retries, and free the mask rgroups before retrying.  Check loop-wide
26753         reasons for disallowing fully-masked loops.  Make the final decision
26754         about whether use a fully-masked loop or not.
26755         (vect_estimate_min_profitable_iters): Do not assume that peeling
26756         for the number of iterations will be needed for fully-masked loops.
26757         (vectorizable_reduction): Disable fully-masked loops.
26758         (vectorizable_live_operation): Likewise.
26759         (vect_halve_mask_nunits): New function.
26760         (vect_double_mask_nunits): Likewise.
26761         (vect_record_loop_mask): Likewise.
26762         (vect_get_loop_mask): Likewise.
26763         (vect_transform_loop): Handle the case in which the final loop
26764         iteration might handle a partial vector.  Call vect_set_loop_condition
26765         instead of slpeel_make_loop_iterate_ntimes.
26766         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26767         (check_load_store_masking): New function.
26768         (prepare_load_store_mask): Likewise.
26769         (vectorizable_store): Handle fully-masked loops.
26770         (vectorizable_load): Likewise.
26771         (supportable_widening_operation): Use vect_halve_mask_nunits for
26772         booleans.
26773         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26774         (vect_gen_while): New function.
26775         * config/aarch64/aarch64.md (umax<mode>3): New expander.
26776         (aarch64_uqdec<mode>): New insn.
26778 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26779             Alan Hayward  <alan.hayward@arm.com>
26780             David Sherwood  <david.sherwood@arm.com>
26782         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26783         (reduc_xor_scal_optab): New optabs.
26784         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26785         (reduc_xor_scal_@var{m}): Document.
26786         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26787         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26788         internal functions.
26789         * fold-const-call.c (fold_const_call): Handle them.
26790         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26791         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26792         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26793         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26794         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26795         (UNSPEC_XORV): New unspecs.
26796         (optab): Add entries for them.
26797         (BITWISEV): New int iterator.
26798         (bit_reduc_op): New int attributes.
26800 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26801             Alan Hayward  <alan.hayward@arm.com>
26802             David Sherwood  <david.sherwood@arm.com>
26804         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26805         * internal-fn.def (VEC_SHL_INSERT): New internal function.
26806         * optabs.def (vec_shl_insert_optab): New optab.
26807         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26808         (duplicate_and_interleave): Likewise.
26809         * tree-vect-loop.c: Include internal-fn.h.
26810         (neutral_op_for_slp_reduction): New function, split out from
26811         get_initial_defs_for_reduction.
26812         (get_initial_def_for_reduction): Handle option 2 for variable-length
26813         vectors by loading the neutral value into a vector and then shifting
26814         the initial value into element 0.
26815         (get_initial_defs_for_reduction): Replace the code argument with
26816         the neutral value calculated by neutral_op_for_slp_reduction.
26817         Use gimple_build_vector for constant-length vectors.
26818         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26819         but the first group_size elements have a neutral value.
26820         Use duplicate_and_interleave otherwise.
26821         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26822         Update call to get_initial_defs_for_reduction.  Handle SLP
26823         reductions for variable-length vectors by creating one vector
26824         result for each scalar result, with the elements associated
26825         with other scalar results stubbed out with the neutral value.
26826         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26827         Require IFN_VEC_SHL_INSERT for double reductions on
26828         variable-length vectors, or SLP reductions that have
26829         a neutral value.  Require can_duplicate_and_interleave_p
26830         support for variable-length unchained SLP reductions if there
26831         is no neutral value, such as for MIN/MAX reductions.  Also require
26832         the number of vector elements to be a multiple of the number of
26833         SLP statements when doing variable-length unchained SLP reductions.
26834         Update call to vect_create_epilog_for_reduction.
26835         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26836         and remove initial values.
26837         (duplicate_and_interleave): Make public.
26838         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26839         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26841 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26842             Alan Hayward  <alan.hayward@arm.com>
26843             David Sherwood  <david.sherwood@arm.com>
26845         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26846         (can_duplicate_and_interleave_p): New function.
26847         (vect_get_and_check_slp_defs): Take the vector of statements
26848         rather than just the current one.  Remove excess parentheses.
26849         Restriction rejectinon of vect_constant_def and vect_external_def
26850         for variable-length vectors to boolean types, or types for which
26851         can_duplicate_and_interleave_p is false.
26852         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26853         (duplicate_and_interleave): New function.
26854         (vect_get_constant_vectors): Use gimple_build_vector for
26855         constant-length vectors and suitable variable-length constant
26856         vectors.  Use duplicate_and_interleave for other variable-length
26857         vectors.  Don't defer the update when inserting new statements.
26859 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26860             Alan Hayward  <alan.hayward@arm.com>
26861             David Sherwood  <david.sherwood@arm.com>
26863         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26864         min_profitable_iters doesn't go negative.
26866 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26867             Alan Hayward  <alan.hayward@arm.com>
26868             David Sherwood  <david.sherwood@arm.com>
26870         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26871         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26872         * optabs.def (vec_mask_load_lanes_optab): New optab.
26873         (vec_mask_store_lanes_optab): Likewise.
26874         * internal-fn.def (MASK_LOAD_LANES): New internal function.
26875         (MASK_STORE_LANES): Likewise.
26876         * internal-fn.c (mask_load_lanes_direct): New macro.
26877         (mask_store_lanes_direct): Likewise.
26878         (expand_mask_load_optab_fn): Handle masked operations.
26879         (expand_mask_load_lanes_optab_fn): New macro.
26880         (expand_mask_store_optab_fn): Handle masked operations.
26881         (expand_mask_store_lanes_optab_fn): New macro.
26882         (direct_mask_load_lanes_optab_supported_p): Likewise.
26883         (direct_mask_store_lanes_optab_supported_p): Likewise.
26884         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26885         parameter.
26886         (vect_load_lanes_supported): Likewise.
26887         * tree-vect-data-refs.c (strip_conversion): New function.
26888         (can_group_stmts_p): Likewise.
26889         (vect_analyze_data_ref_accesses): Use it instead of checking
26890         for a pair of assignments.
26891         (vect_store_lanes_supported): Take a masked_p parameter.
26892         (vect_load_lanes_supported): Likewise.
26893         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26894         vect_store_lanes_supported and vect_load_lanes_supported.
26895         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26896         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26897         parameter.  Don't allow gaps for masked accesses.
26898         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
26899         and vect_load_lanes_supported.
26900         (get_load_store_type): Take a masked_p parameter and update
26901         call to get_group_load_store_type.
26902         (vectorizable_store): Update call to get_load_store_type.
26903         Handle IFN_MASK_STORE_LANES.
26904         (vectorizable_load): Update call to get_load_store_type.
26905         Handle IFN_MASK_LOAD_LANES.
26907 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26908             Alan Hayward  <alan.hayward@arm.com>
26909             David Sherwood  <david.sherwood@arm.com>
26911         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26912         modes for SVE.
26913         * config/aarch64/aarch64-protos.h
26914         (aarch64_sve_struct_memory_operand_p): Declare.
26915         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26916         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26917         (VPRED, vpred): Handle SVE structure modes.
26918         * config/aarch64/constraints.md (Utx): New constraint.
26919         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26920         (aarch64_sve_struct_nonimmediate_operand): New predicates.
26921         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26922         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26923         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26924         structure modes.  Split into pieces after RA.
26925         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26926         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26927         New patterns.
26928         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26929         SVE structure modes.
26930         (aarch64_classify_address): Likewise.
26931         (sizetochar): Move earlier in file.
26932         (aarch64_print_operand): Handle SVE register lists.
26933         (aarch64_array_mode): New function.
26934         (aarch64_sve_struct_memory_operand_p): Likewise.
26935         (TARGET_ARRAY_MODE): Redefine.
26937 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26938             Alan Hayward  <alan.hayward@arm.com>
26939             David Sherwood  <david.sherwood@arm.com>
26941         * target.def (array_mode): New target hook.
26942         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26943         * doc/tm.texi: Regenerate.
26944         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26945         * hooks.c (hook_optmode_mode_uhwi_none): New function.
26946         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26947         targetm.array_mode.
26948         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
26949         type sizes.
26951 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26952             Alan Hayward  <alan.hayward@arm.com>
26953             David Sherwood  <david.sherwood@arm.com>
26955         * fold-const.c (fold_binary_loc): Check the argument types
26956         rather than the result type when testing for a vector operation.
26958 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26960         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26961         * doc/tm.texi: Regenerate.
26963 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26964             Alan Hayward  <alan.hayward@arm.com>
26965             David Sherwood  <david.sherwood@arm.com>
26967         * doc/invoke.texi (-msve-vector-bits=): Document new option.
26968         (sve): Document new AArch64 extension.
26969         * doc/md.texi (w): Extend the description of the AArch64
26970         constraint to include SVE vectors.
26971         (Upl, Upa): Document new AArch64 predicate constraints.
26972         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26973         enum.
26974         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26975         (msve-vector-bits=): New option.
26976         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26977         SVE when these are disabled.
26978         (sve): New extension.
26979         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26980         modes.  Adjust their number of units based on aarch64_sve_vg.
26981         (MAX_BITSIZE_MODE_ANY_MODE): Define.
26982         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26983         aarch64_addr_query_type.
26984         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26985         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26986         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26987         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26988         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26989         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26990         (aarch64_simd_imm_zero_p): Delete.
26991         (aarch64_check_zero_based_sve_index_immediate): Declare.
26992         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26993         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26994         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26995         (aarch64_sve_float_mul_immediate_p): Likewise.
26996         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26997         rather than an rtx.
26998         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26999         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
27000         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
27001         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
27002         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
27003         (aarch64_regmode_natural_size): Likewise.
27004         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
27005         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
27006         left one place.
27007         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
27008         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
27009         for VG and the SVE predicate registers.
27010         (V_ALIASES): Add a "z"-prefixed alias.
27011         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
27012         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
27013         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
27014         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
27015         (REG_CLASS_NAMES): Add entries for them.
27016         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
27017         and the predicate registers.
27018         (aarch64_sve_vg): Declare.
27019         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
27020         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
27021         (REGMODE_NATURAL_SIZE): Define.
27022         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
27023         SVE macros.
27024         * config/aarch64/aarch64.c: Include cfgrtl.h.
27025         (simd_immediate_info): Add a constructor for series vectors,
27026         and an associated step field.
27027         (aarch64_sve_vg): New variable.
27028         (aarch64_dbx_register_number): Handle VG and the predicate registers.
27029         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
27030         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
27031         (VEC_ANY_DATA, VEC_STRUCT): New constants.
27032         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
27033         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
27034         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
27035         (aarch64_get_mask_mode): New functions.
27036         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
27037         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
27038         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
27039         predicate modes and predicate registers.  Explicitly restrict
27040         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
27041         to store a vector mode if it is recognized by
27042         aarch64_classify_vector_mode.
27043         (aarch64_regmode_natural_size): New function.
27044         (aarch64_hard_regno_caller_save_mode): Return the original mode
27045         for predicates.
27046         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
27047         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
27048         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
27049         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
27050         functions.
27051         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
27052         does not overlap dest if the function is frame-related.  Handle
27053         SVE constants.
27054         (aarch64_split_add_offset): New function.
27055         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
27056         them aarch64_add_offset.
27057         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
27058         and update call to aarch64_sub_sp.
27059         (aarch64_add_cfa_expression): New function.
27060         (aarch64_expand_prologue): Pass extra temporary registers to the
27061         functions above.  Handle the case in which we need to emit new
27062         DW_CFA_expressions for registers that were originally saved
27063         relative to the stack pointer, but now have to be expressed
27064         relative to the frame pointer.
27065         (aarch64_output_mi_thunk): Pass extra temporary registers to the
27066         functions above.
27067         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
27068         IP0 and IP1 values for SVE frames.
27069         (aarch64_expand_vec_series): New function.
27070         (aarch64_expand_sve_widened_duplicate): Likewise.
27071         (aarch64_expand_sve_const_vector): Likewise.
27072         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
27073         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
27074         into the register, rather than emitting a SET directly.
27075         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
27076         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
27077         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
27078         (offset_9bit_signed_scaled_p): New functions.
27079         (aarch64_replicate_bitmask_imm): New function.
27080         (aarch64_bitmask_imm): Use it.
27081         (aarch64_cannot_force_const_mem): Reject expressions involving
27082         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
27083         (aarch64_classify_index): Handle SVE indices, by requiring
27084         a plain register index with a scale that matches the element size.
27085         (aarch64_classify_address): Handle SVE addresses.  Assert that
27086         the mode of the address is VOIDmode or an integer mode.
27087         Update call to aarch64_classify_symbol.
27088         (aarch64_classify_symbolic_expression): Update call to
27089         aarch64_classify_symbol.
27090         (aarch64_const_vec_all_in_range_p): New function.
27091         (aarch64_print_vector_float_operand): Likewise.
27092         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
27093         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
27094         and the FP immediates 1.0 and 0.5.
27095         (aarch64_print_address_internal): Handle SVE addresses.
27096         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
27097         (aarch64_regno_regclass): Handle predicate registers.
27098         (aarch64_secondary_reload): Handle big-endian reloads of SVE
27099         data modes.
27100         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
27101         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
27102         (aarch64_convert_sve_vector_bits): New function.
27103         (aarch64_override_options): Use it to handle -msve-vector-bits=.
27104         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
27105         rather than an rtx.
27106         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
27107         Handle SVE vector and predicate modes.  Accept VL-based constants
27108         that need only one temporary register, and VL offsets that require
27109         no temporary registers.
27110         (aarch64_conditional_register_usage): Mark the predicate registers
27111         as fixed if SVE isn't available.
27112         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
27113         Return true for SVE vector and predicate modes.
27114         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
27115         rather than an unsigned int.  Handle SVE modes.
27116         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
27117         SVE modes.
27118         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
27119         if SVE is enabled.
27120         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
27121         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
27122         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
27123         (aarch64_sve_float_mul_immediate_p): New functions.
27124         (aarch64_sve_valid_immediate): New function.
27125         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
27126         Explicitly reject structure modes.  Check for INDEX constants.
27127         Handle PTRUE and PFALSE constants.
27128         (aarch64_check_zero_based_sve_index_immediate): New function.
27129         (aarch64_simd_imm_zero_p): Delete.
27130         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
27131         vector modes.  Accept constants in the range of CNT[BHWD].
27132         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
27133         ask for an Advanced SIMD mode.
27134         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
27135         (aarch64_simd_vector_alignment): Handle SVE predicates.
27136         (aarch64_vectorize_preferred_vector_alignment): New function.
27137         (aarch64_simd_vector_alignment_reachable): Use it instead of
27138         the vector size.
27139         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
27140         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
27141         functions.
27142         (MAX_VECT_LEN): Delete.
27143         (expand_vec_perm_d): Add a vec_flags field.
27144         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
27145         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
27146         (aarch64_evpc_ext): Don't apply a big-endian lane correction
27147         for SVE modes.
27148         (aarch64_evpc_rev): Rename to...
27149         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
27150         (aarch64_evpc_rev_global): New function.
27151         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
27152         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
27153         MAX_VECT_LEN.
27154         (aarch64_evpc_sve_tbl): New function.
27155         (aarch64_expand_vec_perm_const_1): Update after rename of
27156         aarch64_evpc_rev.  Handle SVE permutes too, trying
27157         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
27158         than aarch64_evpc_tbl.
27159         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
27160         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
27161         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
27162         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
27163         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
27164         (aarch64_expand_sve_vcond): New functions.
27165         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
27166         of aarch64_vector_mode_p.
27167         (aarch64_dwarf_poly_indeterminate_value): New function.
27168         (aarch64_compute_pressure_classes): Likewise.
27169         (aarch64_can_change_mode_class): Likewise.
27170         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
27171         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
27172         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
27173         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
27174         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
27175         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
27176         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
27177         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
27178         constraints.
27179         (Dn, Dl, Dr): Accept const as well as const_vector.
27180         (Dz): Likewise.  Compare against CONST0_RTX.
27181         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
27182         of "vector" where appropriate.
27183         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
27184         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
27185         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
27186         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
27187         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
27188         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
27189         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
27190         (v_int_equiv): Extend to SVE modes.
27191         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
27192         mode attributes.
27193         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
27194         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
27195         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
27196         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
27197         (SVE_COND_FP_CMP): New int iterators.
27198         (perm_hilo): Handle the new unpack unspecs.
27199         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
27200         attributes.
27201         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
27202         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
27203         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
27204         (aarch64_equality_operator, aarch64_constant_vector_operand)
27205         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
27206         (aarch64_sve_nonimmediate_operand): Likewise.
27207         (aarch64_sve_general_operand): Likewise.
27208         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
27209         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
27210         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
27211         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
27212         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
27213         (aarch64_sve_float_arith_immediate): Likewise.
27214         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
27215         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
27216         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
27217         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
27218         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
27219         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
27220         (aarch64_sve_float_arith_operand): Likewise.
27221         (aarch64_sve_float_arith_with_sub_operand): Likewise.
27222         (aarch64_sve_float_mul_operand): Likewise.
27223         (aarch64_sve_vec_perm_operand): Likewise.
27224         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
27225         (aarch64_mov_operand): Accept const_poly_int and const_vector.
27226         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
27227         as well as const_vector.
27228         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
27229         in file.  Use CONST0_RTX and CONSTM1_RTX.
27230         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
27231         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
27232         Use aarch64_simd_imm_zero.
27233         * config/aarch64/aarch64-sve.md: New file.
27234         * config/aarch64/aarch64.md: Include it.
27235         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
27236         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
27237         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
27238         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
27239         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
27240         (sve): New attribute.
27241         (enabled): Disable instructions with the sve attribute unless
27242         TARGET_SVE.
27243         (movqi, movhi): Pass CONST_POLY_INT operaneds through
27244         aarch64_expand_mov_immediate.
27245         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
27246         CNT[BHSD] immediates.
27247         (movti): Split CONST_POLY_INT moves into two halves.
27248         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
27249         Split additions that need a temporary here if the destination
27250         is the stack pointer.
27251         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
27252         (*add<mode>3_poly_1): New instruction.
27253         (set_clobber_cc): New expander.
27255 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27257         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
27258         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
27259         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
27260         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
27261         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
27262         Change innermode from fixed_mode_size to machine_mode.
27263         (simplify_subreg): Update call accordingly.  Handle a constant-sized
27264         subreg of a variable-length CONST_VECTOR.
27266 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27267             Alan Hayward  <alan.hayward@arm.com>
27268             David Sherwood  <david.sherwood@arm.com>
27270         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
27271         (add_offset_to_base): New function, split out from...
27272         (create_mem_ref): ...here.  When handling a scale other than 1,
27273         check first whether the address is valid without the offset.
27274         Add it into the base if so, leaving the index and scale as-is.
27276 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27278         PR c++/83778
27279         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
27280         fold_for_warn before checking if arg2 is INTEGER_CST.
27282 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
27284         * config/rs6000/predicates.md (load_multiple_operation): Delete.
27285         (store_multiple_operation): Delete.
27286         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
27287         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
27288         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
27289         guarded by TARGET_STRING.
27290         (rs6000_output_load_multiple): Delete.
27291         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
27292         OPTION_MASK_STRING / TARGET_STRING handling.
27293         (print_operand) <'N', 'O'>: Add comment that these are unused now.
27294         (const rs6000_opt_masks) <"string">: Change mask to 0.
27295         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
27296         (MASK_STRING): Delete.
27297         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
27298         parts.  Simplify.
27299         (load_multiple): Delete.
27300         (*ldmsi8): Delete.
27301         (*ldmsi7): Delete.
27302         (*ldmsi6): Delete.
27303         (*ldmsi5): Delete.
27304         (*ldmsi4): Delete.
27305         (*ldmsi3): Delete.
27306         (store_multiple): Delete.
27307         (*stmsi8): Delete.
27308         (*stmsi7): Delete.
27309         (*stmsi6): Delete.
27310         (*stmsi5): Delete.
27311         (*stmsi4): Delete.
27312         (*stmsi3): Delete.
27313         (movmemsi_8reg): Delete.
27314         (corresponding unnamed define_insn): Delete.
27315         (movmemsi_6reg): Delete.
27316         (corresponding unnamed define_insn): Delete.
27317         (movmemsi_4reg): Delete.
27318         (corresponding unnamed define_insn): Delete.
27319         (movmemsi_2reg): Delete.
27320         (corresponding unnamed define_insn): Delete.
27321         (movmemsi_1reg): Delete.
27322         (corresponding unnamed define_insn): Delete.
27323         * config/rs6000/rs6000.opt (mno-string): New.
27324         (mstring): Replace by deprecation warning stub.
27325         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
27327 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
27329         * regrename.c (regrename_do_replace): If replacing the same
27330         reg multiple times, try to reuse last created gen_raw_REG.
27332         PR debug/81155
27333         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
27334         main to workaround a bug in GDB.
27336 2018-01-12  Tom de Vries  <tom@codesourcery.com>
27338         PR target/83737
27339         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
27341 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
27343         PR rtl-optimization/80481
27344         * ira-color.c (get_cap_member): New function.
27345         (allocnos_conflict_by_live_ranges_p): Use it.
27346         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
27347         (setup_slot_coalesced_allocno_live_ranges): Ditto.
27349 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
27351         PR target/83628
27352         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
27353         (*saddl_se_1): Ditto.
27354         (*ssubsi_1): Ditto.
27355         (*ssubl_se_1): Ditto.
27357 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27359         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
27360         rather than wi::to_widest for DR_INITs.
27361         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
27362         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
27363         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
27364         INTEGER_CSTs.
27365         (vect_analyze_group_access_1): Note that here.
27367 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27369         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
27370         polynomial type sizes.
27372 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
27374         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
27375         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
27376         (gimple_add_tmp_var): Likewise.
27378 2018-01-12  Martin Liska  <mliska@suse.cz>
27380         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
27381         (gimple_alloc_sizes): Likewise.
27382         (dump_gimple_statistics): Use PRIu64 in printf format.
27383         * gimple.h: Change uint64_t to int.
27385 2018-01-12  Martin Liska  <mliska@suse.cz>
27387         * tree-core.h: Use uint64_t instead of int.
27388         * tree.c (tree_node_counts): Likewise.
27389         (tree_node_sizes): Likewise.
27390         (dump_tree_statistics): Use PRIu64 in printf format.
27392 2018-01-12  Martin Liska  <mliska@suse.cz>
27394         * Makefile.in: As qsort_chk is implemented in vec.c, add
27395         vec.o to linkage of gencfn-macros.
27396         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
27397         passing the info to record_node_allocation_statistics.
27398         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
27399         and pass the info.
27400         * ggc-common.c (struct ggc_usage): Add operator== and use
27401         it in operator< and compare function.
27402         * mem-stats.h (struct mem_usage): Likewise.
27403         * vec.c (struct vec_usage): Remove operator< and compare
27404         function. Can be simply inherited.
27406 2018-01-12  Martin Jambor  <mjambor@suse.cz>
27408         PR target/81616
27409         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
27410         * tree-ssa-math-opts.c: Include domwalk.h.
27411         (convert_mult_to_fma_1): New function.
27412         (fma_transformation_info): New type.
27413         (fma_deferring_state): Likewise.
27414         (cancel_fma_deferring): New function.
27415         (result_of_phi): Likewise.
27416         (last_fma_candidate_feeds_initial_phi): Likewise.
27417         (convert_mult_to_fma): Added deferring logic, split actual
27418         transformation to convert_mult_to_fma_1.
27419         (math_opts_dom_walker): New type.
27420         (math_opts_dom_walker::after_dom_children): New method, body moved
27421         here from pass_optimize_widening_mul::execute, added deferring logic
27422         bits.
27423         (pass_optimize_widening_mul::execute): Moved most of code to
27424         math_opts_dom_walker::after_dom_children.
27425         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
27426         * config/i386/i386.c (ix86_option_override_internal): Added
27427         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
27429 2018-01-12  Richard Biener  <rguenther@suse.de>
27431         PR debug/83157
27432         * dwarf2out.c (gen_variable_die): Do not reset old_die for
27433         inline instance vars.
27435 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
27437         PR target/81819
27438         * config/rx/rx.c (rx_is_restricted_memory_address):
27439         Handle SUBREG case.
27441 2018-01-12  Richard Biener  <rguenther@suse.de>
27443         PR tree-optimization/80846
27444         * target.def (split_reduction): New target hook.
27445         * targhooks.c (default_split_reduction): New function.
27446         * targhooks.h (default_split_reduction): Declare.
27447         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
27448         target requests first reduce vectors by combining low and high
27449         parts.
27450         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
27451         (get_vectype_for_scalar_type_and_size): Export.
27452         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
27453         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
27454         * doc/tm.texi: Regenerate.
27455         * config/i386/i386.c (ix86_split_reduction): Implement
27456         TARGET_VECTORIZE_SPLIT_REDUCTION.
27458 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27460         PR target/83368
27461         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
27462         in PIC mode except for TARGET_VXWORKS_RTP.
27463         * config/sparc/sparc.c: Include cfgrtl.h.
27464         (TARGET_INIT_PIC_REG): Define.
27465         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
27466         (sparc_pic_register_p): New predicate.
27467         (sparc_legitimate_address_p): Use it.
27468         (sparc_legitimize_pic_address): Likewise.
27469         (sparc_delegitimize_address): Likewise.
27470         (sparc_mode_dependent_address_p): Likewise.
27471         (gen_load_pcrel_sym): Remove 4th parameter.
27472         (load_got_register): Adjust call to above.  Remove obsolete stuff.
27473         (sparc_expand_prologue): Do not call load_got_register here.
27474         (sparc_flat_expand_prologue): Likewise.
27475         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
27476         (sparc_use_pseudo_pic_reg): New function.
27477         (sparc_init_pic_reg): Likewise.
27478         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27479         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
27481 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
27483         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
27484         Add item for branch_cost.
27486 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27488         PR rtl-optimization/83565
27489         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
27490         not extend the result to a larger mode for rotate operations.
27491         (num_sign_bit_copies1): Likewise.
27493 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27495         PR target/40411
27496         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
27497         -symbolic.
27498         Use values-Xc.o for -pedantic.
27499         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
27501 2018-01-12  Martin Liska  <mliska@suse.cz>
27503         PR ipa/83054
27504         * ipa-devirt.c (final_warning_record::grow_type_warnings):
27505         New function.
27506         (possible_polymorphic_call_targets): Use it.
27507         (ipa_devirt): Likewise.
27509 2018-01-12  Martin Liska  <mliska@suse.cz>
27511         * profile-count.h (enum profile_quality): Use 0 as invalid
27512         enum value of profile_quality.
27514 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
27516         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
27517         -mext-string options.
27519 2018-01-12  Richard Biener  <rguenther@suse.de>
27521         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
27522         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
27523         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
27524         Likewise.
27525         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
27527 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
27529         * configure.ac (--with-long-double-format): Add support for the
27530         configuration option to change the default long double format on
27531         PowerPC systems.
27532         * config.gcc (powerpc*-linux*-*): Likewise.
27533         * configure: Regenerate.
27534         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
27535         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
27536         used without modification.
27538 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27540         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
27541         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
27542         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
27543         MISC_BUILTIN_SPEC_BARRIER.
27544         (rs6000_init_builtins): Likewise.
27545         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
27546         enum value.
27547         (speculation_barrier): New define_insn.
27548         * doc/extend.texi: Document __builtin_speculation_barrier.
27550 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27552         PR target/83203
27553         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
27554         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
27555         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
27556         iterators.
27557         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
27558         integral modes instead of "ss" and "sd".
27559         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
27560         vectors with 32-bit and 64-bit elements.
27561         (vecdupssescalarmodesuffix): New mode attribute.
27562         (vec_dup<mode>): Use it.
27564 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
27566         PR target/83330
27567         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
27568         frame if argument is passed on stack.
27570 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27572         PR target/82682
27573         * ree.c (combine_reaching_defs): Optimize also
27574         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
27575         reg2=any_extend(exp); reg1=reg2;, formatting fix.
27577 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27579         PR middle-end/83189
27580         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
27582 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27584         PR middle-end/83718
27585         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
27586         after they are computed.
27588 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
27590         PR tree-optimization/83695
27591         * gimple-loop-linterchange.cc
27592         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
27593         reset cached scev information after interchange.
27594         (pass_linterchange::execute): Remove call to scev_reset_htab.
27596 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27598         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
27599         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
27600         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
27601         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
27602         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
27603         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
27604         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
27605         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
27606         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
27607         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
27608         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
27609         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
27610         (V_lane_reg): Likewise.
27611         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
27612         New define_expand.
27613         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
27614         (vfmal_lane_low<mode>_intrinsic,
27615         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
27616         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27617         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27618         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27619         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27620         vfmsl_lane_high<mode>_intrinsic): New define_insns.
27622 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27624         * config/arm/arm-cpus.in (fp16fml): New feature.
27625         (ALL_SIMD): Add fp16fml.
27626         (armv8.2-a): Add fp16fml as an option.
27627         (armv8.3-a): Likewise.
27628         (armv8.4-a): Add fp16fml as part of fp16.
27629         * config/arm/arm.h (TARGET_FP16FML): Define.
27630         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27631         when appropriate.
27632         * config/arm/arm-modes.def (V2HF): Define.
27633         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27634         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27635         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27636         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27637         vfmsl_low, vfmsl_high): New set of builtins.
27638         * config/arm/iterators.md (PLUSMINUS): New code iterator.
27639         (vfml_op): New code attribute.
27640         (VFMLHALVES): New int iterator.
27641         (VFML, VFMLSEL): New mode attributes.
27642         (V_reg): Define mapping for V2HF.
27643         (V_hi, V_lo): New mode attributes.
27644         (VF_constraint): Likewise.
27645         (vfml_half, vfml_half_selector): New int attributes.
27646         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27647         define_expand.
27648         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27649         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27650         New define_insn.
27651         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27652         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27653         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27654         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
27655         documentation.
27656         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27657         Document new effective target and option set.
27659 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27661         * config/arm/arm-cpus.in (armv8_4): New feature.
27662         (ARMv8_4a): New fgroup.
27663         (armv8.4-a): New arch.
27664         * config/arm/arm-tables.opt: Regenerate.
27665         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27666         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27667         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27668         Add matching rules for -march=armv8.4-a and extensions.
27669         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27671 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
27673         PR target/81821
27674         * config/rx/rx.md (BW): New mode attribute.
27675         (sync_lock_test_and_setsi): Add mode suffix to insn output.
27677 2018-01-11  Richard Biener  <rguenther@suse.de>
27679         PR tree-optimization/83435
27680         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27681         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27682         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27684 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27685             Alan Hayward  <alan.hayward@arm.com>
27686             David Sherwood  <david.sherwood@arm.com>
27688         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27689         field.
27690         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
27691         (aarch64_print_address_internal): Use it to check for a zero offset.
27693 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27694             Alan Hayward  <alan.hayward@arm.com>
27695             David Sherwood  <david.sherwood@arm.com>
27697         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27698         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27699         Return a poly_int64 rather than a HOST_WIDE_INT.
27700         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27701         rather than a HOST_WIDE_INT.
27702         * config/aarch64/aarch64.h (aarch64_frame): Protect with
27703         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
27704         hard_fp_offset, frame_size, initial_adjust, callee_offset and
27705         final_offset from HOST_WIDE_INT to poly_int64.
27706         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27707         to_constant when getting the number of units in an Advanced SIMD
27708         mode.
27709         (aarch64_builtin_vectorized_function): Check for a constant number
27710         of units.
27711         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27712         GET_MODE_SIZE.
27713         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27714         attribute instead of GET_MODE_NUNITS.
27715         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27716         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27717         GET_MODE_SIZE for fixed-size registers.
27718         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27719         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27720         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27721         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27722         (aarch64_print_operand, aarch64_print_address_internal)
27723         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27724         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27725         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27726         Handle polynomial GET_MODE_SIZE.
27727         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
27728         wider than SImode without modification.
27729         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27730         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27731         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27732         passing and returning SVE modes.
27733         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27734         rather than GEN_INT.
27735         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27736         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27737         (aarch64_allocate_and_probe_stack_space): Likewise.
27738         (aarch64_layout_frame): Cope with polynomial offsets.
27739         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27740         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
27741         polynomial offsets.
27742         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27743         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27744         poly_int64 rather than a HOST_WIDE_INT.
27745         (aarch64_get_separate_components, aarch64_process_components)
27746         (aarch64_expand_prologue, aarch64_expand_epilogue)
27747         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27748         (aarch64_anchor_offset): New function, split out from...
27749         (aarch64_legitimize_address): ...here.
27750         (aarch64_builtin_vectorization_cost): Handle polynomial
27751         TYPE_VECTOR_SUBPARTS.
27752         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27753         GET_MODE_NUNITS.
27754         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27755         number of elements from the PARALLEL rather than the mode.
27756         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27757         rather than GET_MODE_BITSIZE.
27758         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27759         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27760         (aarch64_expand_vec_perm_const_1): Handle polynomial
27761         d->perm.length () and d->perm elements.
27762         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
27763         Apply to_constant to d->perm elements.
27764         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27765         polynomial CONST_VECTOR_NUNITS.
27766         (aarch64_move_pointer): Take amount as a poly_int64 rather
27767         than an int.
27768         (aarch64_progress_pointer): Avoid temporary variable.
27769         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27770         the mode attribute instead of GET_MODE.
27772 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27773             Alan Hayward  <alan.hayward@arm.com>
27774             David Sherwood  <david.sherwood@arm.com>
27776         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27777         x exists before using it.
27778         (aarch64_add_constant_internal): Rename to...
27779         (aarch64_add_offset_1): ...this.  Replace regnum with separate
27780         src and dest rtxes.  Handle the case in which they're different,
27781         including when the offset is zero.  Replace scratchreg with an rtx.
27782         Use 2 additions if there is no spare register into which we can
27783         move a 16-bit constant.
27784         (aarch64_add_constant): Delete.
27785         (aarch64_add_offset): Replace reg with separate src and dest
27786         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
27787         Use aarch64_add_offset_1.
27788         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27789         an rtx rather than an int.  Take the delta as a poly_int64
27790         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
27791         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27792         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27793         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27794         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27795         and aarch64_add_sp.
27796         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27797         aarch64_add_constant.
27799 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27801         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27802         Use scalar_float_mode.
27804 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27806         * config/aarch64/aarch64-simd.md
27807         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27808         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27809         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27810         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27811         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27812         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27813         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27814         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27815         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27816         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27818 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27820         PR target/83514
27821         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27822         targ_options->x_arm_arch_string is non NULL.
27824 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
27826         * config/aarch64/aarch64.h
27827         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
27829 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
27831         PR target/82096
27832         * expmed.c (emit_store_flag_force): Swap if const op0
27833         and change VOIDmode to mode of op0.
27835 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27837         PR rtl-optimization/83761
27838         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27839         than bytes to mode_for_size.
27841 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27843         PR middle-end/83189
27844         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27845         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27846         profile.
27848 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27850         PR middle-end/83575
27851         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27852         when in layout mode.
27853         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27854         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27855         partition fixup.
27857 2018-01-10  Michael Collison  <michael.collison@arm.com>
27859         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27860         * config/aarch64/aarch64-option-extension.def: Add
27861         AARCH64_OPT_EXTENSION of 'fp16fml'.
27862         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27863         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27864         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27865         * config/aarch64/constraints.md (Ui7): New constraint.
27866         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27867         (VFMLA_SEL_W): Ditto.
27868         (f16quad): Ditto.
27869         (f16mac1): Ditto.
27870         (VFMLA16_LOW): New int iterator.
27871         (VFMLA16_HIGH): Ditto.
27872         (UNSPEC_FMLAL): New unspec.
27873         (UNSPEC_FMLSL): Ditto.
27874         (UNSPEC_FMLAL2): Ditto.
27875         (UNSPEC_FMLSL2): Ditto.
27876         (f16mac): New code attribute.
27877         * config/aarch64/aarch64-simd-builtins.def
27878         (aarch64_fmlal_lowv2sf): Ditto.
27879         (aarch64_fmlsl_lowv2sf): Ditto.
27880         (aarch64_fmlalq_lowv4sf): Ditto.
27881         (aarch64_fmlslq_lowv4sf): Ditto.
27882         (aarch64_fmlal_highv2sf): Ditto.
27883         (aarch64_fmlsl_highv2sf): Ditto.
27884         (aarch64_fmlalq_highv4sf): Ditto.
27885         (aarch64_fmlslq_highv4sf): Ditto.
27886         (aarch64_fmlal_lane_lowv2sf): Ditto.
27887         (aarch64_fmlsl_lane_lowv2sf): Ditto.
27888         (aarch64_fmlal_laneq_lowv2sf): Ditto.
27889         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27890         (aarch64_fmlalq_lane_lowv4sf): Ditto.
27891         (aarch64_fmlsl_lane_lowv4sf): Ditto.
27892         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27893         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27894         (aarch64_fmlal_lane_highv2sf): Ditto.
27895         (aarch64_fmlsl_lane_highv2sf): Ditto.
27896         (aarch64_fmlal_laneq_highv2sf): Ditto.
27897         (aarch64_fmlsl_laneq_highv2sf): Ditto.
27898         (aarch64_fmlalq_lane_highv4sf): Ditto.
27899         (aarch64_fmlsl_lane_highv4sf): Ditto.
27900         (aarch64_fmlalq_laneq_highv4sf): Ditto.
27901         (aarch64_fmlsl_laneq_highv4sf): Ditto.
27902         * config/aarch64/aarch64-simd.md:
27903         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27904         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27905         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27906         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27907         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27908         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27909         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27910         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27911         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27912         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27913         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27914         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27915         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27916         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27917         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27918         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27919         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27920         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27921         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27922         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27923         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27924         (vfmlsl_low_u32): Ditto.
27925         (vfmlalq_low_u32): Ditto.
27926         (vfmlslq_low_u32): Ditto.
27927         (vfmlal_high_u32): Ditto.
27928         (vfmlsl_high_u32): Ditto.
27929         (vfmlalq_high_u32): Ditto.
27930         (vfmlslq_high_u32): Ditto.
27931         (vfmlal_lane_low_u32): Ditto.
27932         (vfmlsl_lane_low_u32): Ditto.
27933         (vfmlal_laneq_low_u32): Ditto.
27934         (vfmlsl_laneq_low_u32): Ditto.
27935         (vfmlalq_lane_low_u32): Ditto.
27936         (vfmlslq_lane_low_u32): Ditto.
27937         (vfmlalq_laneq_low_u32): Ditto.
27938         (vfmlslq_laneq_low_u32): Ditto.
27939         (vfmlal_lane_high_u32): Ditto.
27940         (vfmlsl_lane_high_u32): Ditto.
27941         (vfmlal_laneq_high_u32): Ditto.
27942         (vfmlsl_laneq_high_u32): Ditto.
27943         (vfmlalq_lane_high_u32): Ditto.
27944         (vfmlslq_lane_high_u32): Ditto.
27945         (vfmlalq_laneq_high_u32): Ditto.
27946         (vfmlslq_laneq_high_u32): Ditto.
27947         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27948         (AARCH64_FL_FOR_ARCH8_4): New.
27949         (AARCH64_ISA_F16FML): New ISA flag.
27950         (TARGET_F16FML): New feature flag for fp16fml.
27951         (doc/invoke.texi): Document new fp16fml option.
27953 2018-01-10  Michael Collison  <michael.collison@arm.com>
27955         * config/aarch64/aarch64-builtins.c:
27956         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27957         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27958         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27959         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27960         (AARCH64_ISA_SHA3): New ISA flag.
27961         (TARGET_SHA3): New feature flag for sha3.
27962         * config/aarch64/iterators.md (sha512_op): New int attribute.
27963         (CRYPTO_SHA512): New int iterator.
27964         (UNSPEC_SHA512H): New unspec.
27965         (UNSPEC_SHA512H2): Ditto.
27966         (UNSPEC_SHA512SU0): Ditto.
27967         (UNSPEC_SHA512SU1): Ditto.
27968         * config/aarch64/aarch64-simd-builtins.def
27969         (aarch64_crypto_sha512hqv2di): New builtin.
27970         (aarch64_crypto_sha512h2qv2di): Ditto.
27971         (aarch64_crypto_sha512su0qv2di): Ditto.
27972         (aarch64_crypto_sha512su1qv2di): Ditto.
27973         (aarch64_eor3qv8hi): Ditto.
27974         (aarch64_rax1qv2di): Ditto.
27975         (aarch64_xarqv2di): Ditto.
27976         (aarch64_bcaxqv8hi): Ditto.
27977         * config/aarch64/aarch64-simd.md:
27978         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27979         (aarch64_crypto_sha512su0qv2di): Ditto.
27980         (aarch64_crypto_sha512su1qv2di): Ditto.
27981         (aarch64_eor3qv8hi): Ditto.
27982         (aarch64_rax1qv2di): Ditto.
27983         (aarch64_xarqv2di): Ditto.
27984         (aarch64_bcaxqv8hi): Ditto.
27985         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27986         (vsha512h2q_u64): Ditto.
27987         (vsha512su0q_u64): Ditto.
27988         (vsha512su1q_u64): Ditto.
27989         (veor3q_u16): Ditto.
27990         (vrax1q_u64): Ditto.
27991         (vxarq_u64): Ditto.
27992         (vbcaxq_u16): Ditto.
27993         * config/arm/types.md (crypto_sha512): New type attribute.
27994         (crypto_sha3): Ditto.
27995         (doc/invoke.texi): Document new sha3 option.
27997 2018-01-10  Michael Collison  <michael.collison@arm.com>
27999         * config/aarch64/aarch64-builtins.c:
28000         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
28001         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28002         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
28003         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
28004         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
28005         (AARCH64_ISA_SM4): New ISA flag.
28006         (TARGET_SM4): New feature flag for sm4.
28007         * config/aarch64/aarch64-simd-builtins.def
28008         (aarch64_sm3ss1qv4si): Ditto.
28009         (aarch64_sm3tt1aq4si): Ditto.
28010         (aarch64_sm3tt1bq4si): Ditto.
28011         (aarch64_sm3tt2aq4si): Ditto.
28012         (aarch64_sm3tt2bq4si): Ditto.
28013         (aarch64_sm3partw1qv4si): Ditto.
28014         (aarch64_sm3partw2qv4si): Ditto.
28015         (aarch64_sm4eqv4si): Ditto.
28016         (aarch64_sm4ekeyqv4si): Ditto.
28017         * config/aarch64/aarch64-simd.md:
28018         (aarch64_sm3ss1qv4si): Ditto.
28019         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
28020         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
28021         (aarch64_sm4eqv4si): Ditto.
28022         (aarch64_sm4ekeyqv4si): Ditto.
28023         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
28024         (sm3part_op): Ditto.
28025         (CRYPTO_SM3TT): Ditto.
28026         (CRYPTO_SM3PART): Ditto.
28027         (UNSPEC_SM3SS1): New unspec.
28028         (UNSPEC_SM3TT1A): Ditto.
28029         (UNSPEC_SM3TT1B): Ditto.
28030         (UNSPEC_SM3TT2A): Ditto.
28031         (UNSPEC_SM3TT2B): Ditto.
28032         (UNSPEC_SM3PARTW1): Ditto.
28033         (UNSPEC_SM3PARTW2): Ditto.
28034         (UNSPEC_SM4E): Ditto.
28035         (UNSPEC_SM4EKEY): Ditto.
28036         * config/aarch64/constraints.md (Ui2): New constraint.
28037         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
28038         * config/arm/types.md (crypto_sm3): New type attribute.
28039         (crypto_sm4): Ditto.
28040         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
28041         (vsm3tt1aq_u32): Ditto.
28042         (vsm3tt1bq_u32): Ditto.
28043         (vsm3tt2aq_u32): Ditto.
28044         (vsm3tt2bq_u32): Ditto.
28045         (vsm3partw1q_u32): Ditto.
28046         (vsm3partw2q_u32): Ditto.
28047         (vsm4eq_u32): Ditto.
28048         (vsm4ekeyq_u32): Ditto.
28049         (doc/invoke.texi): Document new sm4 option.
28051 2018-01-10  Michael Collison  <michael.collison@arm.com>
28053         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
28054         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
28055         (AARCH64_FL_FOR_ARCH8_4): New.
28056         (AARCH64_FL_V8_4): New flag.
28057         (doc/invoke.texi): Document new armv8.4-a option.
28059 2018-01-10  Michael Collison  <michael.collison@arm.com>
28061         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28062         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
28063         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
28064         * config/aarch64/aarch64-option-extension.def: Add
28065         AARCH64_OPT_EXTENSION of 'sha2'.
28066         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
28067         (crypto): Disable sha2 and aes if crypto disabled.
28068         (crypto): Enable aes and sha2 if enabled.
28069         (simd): Disable sha2 and aes if simd disabled.
28070         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
28071         New flags.
28072         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
28073         (TARGET_SHA2): New feature flag for sha2.
28074         (TARGET_AES): New feature flag for aes.
28075         * config/aarch64/aarch64-simd.md:
28076         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
28077         conditional on TARGET_AES.
28078         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
28079         (aarch64_crypto_sha1hsi): Make pattern conditional
28080         on TARGET_SHA2.
28081         (aarch64_crypto_sha1hv4si): Ditto.
28082         (aarch64_be_crypto_sha1hv4si): Ditto.
28083         (aarch64_crypto_sha1su1v4si): Ditto.
28084         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
28085         (aarch64_crypto_sha1su0v4si): Ditto.
28086         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
28087         (aarch64_crypto_sha256su0v4si): Ditto.
28088         (aarch64_crypto_sha256su1v4si): Ditto.
28089         (doc/invoke.texi): Document new aes and sha2 options.
28091 2018-01-10  Martin Sebor  <msebor@redhat.com>
28093         PR tree-optimization/83781
28094         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
28095         as string arrays.
28097 2018-01-11  Martin Sebor  <msebor@gmail.com>
28098             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28100         PR tree-optimization/83501
28101         PR tree-optimization/81703
28103         * tree-ssa-strlen.c (get_string_cst): Rename...
28104         (get_string_len): ...to this.  Handle global constants.
28105         (handle_char_store): Adjust.
28107 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
28108             Jim Wilson  <jimw@sifive.com>
28110         * config/riscv/riscv-protos.h (riscv_output_return): New.
28111         * config/riscv/riscv.c (struct machine_function): New naked_p field.
28112         (riscv_attribute_table, riscv_output_return),
28113         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
28114         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
28115         (riscv_compute_frame_info): Only compute frame->mask if not a naked
28116         function.
28117         (riscv_expand_prologue): Add early return for naked function.
28118         (riscv_expand_epilogue): Likewise.
28119         (riscv_function_ok_for_sibcall): Return false for naked function.
28120         (riscv_set_current_function): New.
28121         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
28122         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
28123         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
28124         * doc/extend.texi (RISC-V Function Attributes): New.
28126 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
28128         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
28129         check for 128-bit long double before checking TCmode.
28130         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
28131         128-bit long doubles before checking TFmode or TCmode.
28132         (FLOAT128_IBM_P): Likewise.
28134 2018-01-10  Martin Sebor  <msebor@redhat.com>
28136         PR tree-optimization/83671
28137         * builtins.c (c_strlen): Unconditionally return zero for the empty
28138         string.
28139         Use -Warray-bounds for warnings.
28140         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
28141         for non-constant array indices with COMPONENT_REF, arrays of
28142         arrays, and pointers to arrays.
28143         (gimple_fold_builtin_strlen): Determine and set length range for
28144         non-constant character arrays.
28146 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
28148         PR middle-end/81897
28149         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
28150         empty blocks.
28152 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
28154         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
28156 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
28158         PR target/83399
28159         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
28160         VECTOR_MEM_ALTIVEC_OR_VSX_P.
28161         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
28162         indexed_or_indirect_operand predicate.
28163         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
28164         (*vsx_le_perm_load_v8hi): Likewise.
28165         (*vsx_le_perm_load_v16qi): Likewise.
28166         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
28167         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
28168         (*vsx_le_perm_store_v8hi): Likewise.
28169         (*vsx_le_perm_store_v16qi): Likewise.
28170         (eight unnamed splitters): Likewise.
28172 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
28174         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
28175         * config/rs6000/emmintrin.h: Likewise.
28176         * config/rs6000/mmintrin.h: Likewise.
28177         * config/rs6000/xmmintrin.h: Likewise.
28179 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
28181         PR c++/43486
28182         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
28183         "public_flag".
28184         * tree.c (tree_nop_conversion): Return true for location wrapper
28185         nodes.
28186         (maybe_wrap_with_location): New function.
28187         (selftest::check_strip_nops): New function.
28188         (selftest::test_location_wrappers): New function.
28189         (selftest::tree_c_tests): Call it.
28190         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
28191         (maybe_wrap_with_location): New decl.
28192         (EXPR_LOCATION_WRAPPER_P): New macro.
28193         (location_wrapper_p): New inline function.
28194         (tree_strip_any_location_wrapper): New inline function.
28196 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
28198         PR target/83735
28199         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
28200         stack_realign_offset for the largest alignment of stack slot
28201         actually used.
28202         (ix86_find_max_used_stack_alignment): New function.
28203         (ix86_finalize_stack_frame_flags): Use it.  Set
28204         max_used_stack_alignment if we don't realign stack.
28205         * config/i386/i386.h (machine_function): Add
28206         max_used_stack_alignment.
28208 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
28210         * config/arm/arm.opt (-mbranch-cost): New option.
28211         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
28212         account.
28214 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
28216         PR target/83629
28217         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
28218         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
28220 2018-01-10  Richard Biener  <rguenther@suse.de>
28222         PR debug/83765
28223         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
28224         early out so it also covers the case where we have a non-NULL
28225         origin.
28227 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28229         PR tree-optimization/83753
28230         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
28231         for non-strided grouped accesses if the number of elements is 1.
28233 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28235         PR target/81616
28236         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
28237         * i386.h (TARGET_USE_GATHER): Define.
28238         * x86-tune.def (X86_TUNE_USE_GATHER): New.
28240 2018-01-10  Martin Liska  <mliska@suse.cz>
28242         PR bootstrap/82831
28243         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
28244         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
28245         partitioning.
28246         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
28247         CLEANUP_NO_PARTITIONING is not set.
28249 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
28251         * doc/rtl.texi: Remove documentation of (const ...) wrappers
28252         for vectors, as a partial revert of r254296.
28253         * rtl.h (const_vec_p): Delete.
28254         (const_vec_duplicate_p): Don't test for vector CONSTs.
28255         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
28256         * expmed.c (make_tree): Likewise.
28258         Revert:
28259         * common.md (E, F): Use CONSTANT_P instead of checking for
28260         CONST_VECTOR.
28261         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
28262         checking for CONST_VECTOR.
28264 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28266         PR middle-end/83575
28267         * predict.c (force_edge_cold): Handle in more sane way edges
28268         with no prediction.
28270 2018-01-09  Carl Love  <cel@us.ibm.com>
28272         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
28273         V4SI, V4SF types.
28274         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
28275         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
28276         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
28277         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
28278         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
28279         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
28280         * config/rs6000/rs6000-protos.h: Add extern defition for
28281         rs6000_generate_float2_double_code.
28282         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
28283         function.
28284         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
28285         (float2_v2df): Add define_expand.
28287 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
28289         PR target/83628
28290         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
28291         op_mode in the force_to_mode call.
28293 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28295         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
28296         instead of checking each element individually.
28297         (aarch64_evpc_uzp): Likewise.
28298         (aarch64_evpc_zip): Likewise.
28299         (aarch64_evpc_ext): Likewise.
28300         (aarch64_evpc_rev): Likewise.
28301         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
28302         instead of checking each element individually.  Return true without
28303         generating rtl if
28304         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
28305         whether all selected elements come from the same input, instead of
28306         checking each element individually.  Remove calls to gen_rtx_REG,
28307         start_sequence and end_sequence and instead assert that no rtl is
28308         generated.
28310 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28312         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
28313         order of HIGH and CONST checks.
28315 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
28317         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
28318         if the destination isn't an SSA_NAME.
28320 2018-01-09  Richard Biener  <rguenther@suse.de>
28322         PR tree-optimization/83668
28323         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
28324         move prologue...
28325         (canonicalize_loop_form): ... here, renamed from ...
28326         (canonicalize_loop_closed_ssa_form): ... this and amended to
28327         swap successor edges for loop exit blocks to make us use
28328         the RPO order we need for initial schedule generation.
28330 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
28332         PR tree-optimization/64811
28333         * match.pd: When optimizing comparisons with Inf, avoid
28334         introducing or losing exceptions from comparisons with NaN.
28336 2018-01-09  Martin Liska  <mliska@suse.cz>
28338         PR sanitizer/82517
28339         * asan.c (shadow_mem_size): Add gcc_assert.
28341 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
28343         Don't save registers in main().
28345         PR target/83738
28346         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
28347         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
28348         * config/avr/avr.c (avr_set_current_function): Don't error if
28349         naked, OS_task or OS_main are specified at the same time.
28350         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
28351         OS_main.
28352         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
28353         attribute.
28354         * common/config/avr/avr-common.c (avr_option_optimization_table):
28355         Switch on -mmain-is-OS_task for optimizing compilations.
28357 2018-01-09  Richard Biener  <rguenther@suse.de>
28359         PR tree-optimization/83572
28360         * graphite.c: Include cfganal.h.
28361         (graphite_transform_loops): Connect infinite loops to exit
28362         and remove fake edges at the end.
28364 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28366         * ipa-inline.c (edge_badness): Revert accidental checkin.
28368 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
28370         PR ipa/80763
28371         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
28372         symbols; not inline clones.
28374 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
28376         PR target/83507
28377         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
28378         hard registers.  Formatting fixes.
28380         PR preprocessor/83722
28381         * gcc.c (try_generate_repro): Pass
28382         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
28383         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
28384         do_report_bug.
28386 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
28387             Kito Cheng  <kito.cheng@gmail.com>
28389         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
28390         (riscv_leaf_function_p): Delete.
28391         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
28393 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28395         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
28396         function.
28397         (do_ifelse): New function.
28398         (do_isel): New function.
28399         (do_sub3): New function.
28400         (do_add3): New function.
28401         (do_load_mask_compare): New function.
28402         (do_overlap_load_compare): New function.
28403         (expand_compare_loop): New function.
28404         (expand_block_compare): Call expand_compare_loop() when appropriate.
28405         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
28406         option description.
28407         (-mblock-compare-inline-loop-limit): New option.
28409 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28411         PR target/83677
28412         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
28413         Reverse order of second and third operands in first alternative.
28414         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
28415         of first and second elements in UNSPEC_VPERMR vector.
28416         (altivec_expand_vec_perm_le): Likewise.
28418 2018-01-08  Jeff Law  <law@redhat.com>
28420         PR rtl-optimizatin/81308
28421         * tree-switch-conversion.c (cfg_altered): New file scoped static.
28422         (process_switch): If group_case_labels makes a change, then set
28423         cfg_altered.
28424         (pass_convert_switch::execute): If a switch is converted, then
28425         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
28427         PR rtl-optimization/81308
28428         * recog.c (split_all_insns): Conditionally cleanup the CFG after
28429         splitting insns.
28431 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
28433         PR target/83663 - Revert r255946
28434         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
28435         generation for cases where splatting a value is not useful.
28436         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
28437         across a vec_duplicate and a paradoxical subreg forming a vector
28438         mode to a vec_concat.
28440 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28442         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
28443         -march=armv8.3-a variants.
28444         * config/arm/t-multilib: Likewise.
28445         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
28447 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28449         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
28450         to generate rtl.
28451         (cceq_ior_compare_complement): Give it a name so I can use it, and
28452         change boolean_or_operator predicate to boolean_operator so it can
28453         be used to generate a crand.
28454         (eqne): New code iterator.
28455         (bd/bd_neg): New code_attrs.
28456         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
28457         a single define_insn.
28458         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
28459         decrement (bdnzt/bdnzf/bdzt/bdzf).
28460         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
28461         with the new names of the branch decrement patterns, and added the
28462         names of the branch decrement conditional patterns.
28464 2018-01-08  Richard Biener  <rguenther@suse.de>
28466         PR tree-optimization/83563
28467         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
28468         cache.
28470 2018-01-08  Richard Biener  <rguenther@suse.de>
28472         PR middle-end/83713
28473         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
28475 2018-01-08  Richard Biener  <rguenther@suse.de>
28477         PR tree-optimization/83685
28478         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
28479         references to abnormals.
28481 2018-01-08  Richard Biener  <rguenther@suse.de>
28483         PR lto/83719
28484         * dwarf2out.c (output_indirect_strings): Handle empty
28485         skeleton_debug_str_hash.
28486         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
28488 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28490         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
28491         (emit_store_direct): Likewise.
28492         (arc_trampoline_adjust_address): Likewise.
28493         (arc_asm_trampoline_template): New function.
28494         (arc_initialize_trampoline): Use asm_trampoline_template.
28495         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
28496         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
28497         * config/arc/arc.md (flush_icache): Delete pattern.
28499 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
28501         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
28502         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
28503         munaligned-access.
28505 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28507         PR target/83681
28508         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
28509         by not USED_FOR_TARGET.
28510         (make_pass_resolve_sw_modes): Likewise.
28512 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28514         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
28515         USED_FOR_TARGET.
28517 2018-01-08  Richard Biener  <rguenther@suse.de>
28519         PR middle-end/83580
28520         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
28522 2018-01-08  Richard Biener  <rguenther@suse.de>
28524         PR middle-end/83517
28525         * match.pd ((t * 2) / 2) -> t): Add missing :c.
28527 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
28529         PR middle-end/81897
28530         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
28531         basic blocks with a small number of successors.
28532         (convert_control_dep_chain_into_preds): Improve handling of
28533         forwarder blocks.
28534         (dump_predicates): Split apart into...
28535         (dump_pred_chain): ...here...
28536         (dump_pred_info): ...and here.
28537         (can_one_predicate_be_invalidated_p): Add debugging printfs.
28538         (can_chain_union_be_invalidated_p): Improve check for invalidation
28539         of paths.
28540         (uninit_uses_cannot_happen): Avoid unnecessary if
28541         convert_control_dep_chain_into_preds yielded nothing.
28543 2018-01-06  Martin Sebor  <msebor@redhat.com>
28545         PR tree-optimization/83640
28546         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
28547         subtracting negative offset from size.
28548         (builtin_access::overlap): Adjust offset bounds of the access to fall
28549         within the size of the object if possible.
28551 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
28553         PR rtl-optimization/83699
28554         * expmed.c (extract_bit_field_1): Restrict the vector usage of
28555         extract_bit_field_as_subreg to cases in which the extracted
28556         value is also a vector.
28558         * lra-constraints.c (process_alt_operands): Test for the equivalence
28559         substitutions when detecting a possible reload cycle.
28561 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
28563         PR debug/83480
28564         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
28565         by default if flag_selective_schedling{,2}.  Formatting fixes.
28567         PR rtl-optimization/83682
28568         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
28569         if it has non-VECTOR_MODE element mode.
28570         (vec_duplicate_p): Likewise.
28572         PR middle-end/83694
28573         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
28574         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
28576 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28578         PR target/83604
28579         * config/i386/i386-builtin.def
28580         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
28581         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
28582         Require also OPTION_MASK_ISA_AVX512F in addition to
28583         OPTION_MASK_ISA_GFNI.
28584         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
28585         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
28586         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
28587         to OPTION_MASK_ISA_GFNI.
28588         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
28589         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
28590         OPTION_MASK_ISA_AVX512BW.
28591         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
28592         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
28593         addition to OPTION_MASK_ISA_GFNI.
28594         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
28595         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
28596         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
28597         to OPTION_MASK_ISA_GFNI.
28598         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
28599         a requirement for all ISAs rather than any of them with a few
28600         exceptions.
28601         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
28602         processing.
28603         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
28604         bitmasks to be enabled with 3 exceptions, instead of requiring any
28605         enabled ISA with lots of exceptions.
28606         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
28607         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
28608         Change avx512bw in isa attribute to avx512f.
28609         * config/i386/sgxintrin.h: Add license boilerplate.
28610         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
28611         to __AVX512F__ and __AVX512VL to __AVX512VL__.
28612         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
28613         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
28614         defined.
28615         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
28616         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28617         temporarily sse2 rather than sse if not enabled already.
28619         PR target/83604
28620         * config/i386/sse.md (VI248_VLBW): Rename to ...
28621         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
28622         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28623         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28624         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28625         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28626         mode iterator instead of VI248_VLBW.
28628 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
28630         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28631         (record_modified): Skip clobbers; add debug output.
28632         (param_change_prob): Use sreal frequencies.
28634 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28636         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28637         punt for user-aligned variables.
28639 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28641         * tree-chrec.c (chrec_contains_symbols): Return true for
28642         POLY_INT_CST.
28644 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
28646         PR target/82439
28647         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28648         of (x|y) == x for BICS pattern.
28650 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28652         PR tree-optimization/83605
28653         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28654         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28655         can throw.
28657 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28659         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28660         * config/epiphany/rtems.h: New file.
28662 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28663             Uros Bizjak  <ubizjak@gmail.com>
28665         PR target/83554
28666         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28667         QIreg_operand instead of register_operand predicate.
28668         * config/i386/i386.c (ix86_rop_should_change_byte_p,
28669         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28670         comments instead of -fmitigate[-_]rop.
28672 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28674         PR bootstrap/81926
28675         * cgraphunit.c (symbol_table::compile): Switch to text_section
28676         before calling assembly_start debug hook.
28677         * run-rtl-passes.c (run_rtl_passes): Likewise.
28678         Include output.h.
28680 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28682         * tree-vrp.c (extract_range_from_binary_expr_1): Check
28683         range_int_cst_p rather than !symbolic_range_p before calling
28684         extract_range_from_multiplicative_op_1.
28686 2018-01-04  Jeff Law  <law@redhat.com>
28688         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28689         redundant test in assertion.
28691 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28693         * doc/rtl.texi: Document machine_mode wrapper classes.
28695 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28697         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28698         using tree_to_uhwi.
28700 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28702         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28703         the VEC_PERM_EXPR fold to fail.
28705 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28707         PR debug/83585
28708         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28709         to switched_sections.
28711 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28713         PR target/83680
28714         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28715         test for d.testing.
28717 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
28719         PR target/83387
28720         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28721         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28723 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28725         PR debug/83666
28726         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28727         is BLKmode and bitpos not zero or mode change is needed.
28729 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28731         PR target/83675
28732         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28733         TARGET_VIS2.
28735 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
28737         PR target/83628
28738         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28739         instead of MULT rtx.  Update all corresponding splitters.
28740         (*saddl_se): Ditto.
28741         (*ssub<modesuffix>): Ditto.
28742         (*ssubl_se): Ditto.
28743         (*cmp_sadd_di): Update split patterns.
28744         (*cmp_sadd_si): Ditto.
28745         (*cmp_sadd_sidi): Ditto.
28746         (*cmp_ssub_di): Ditto.
28747         (*cmp_ssub_si): Ditto.
28748         (*cmp_ssub_sidi): Ditto.
28749         * config/alpha/predicates.md (const23_operand): New predicate.
28750         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28751         Look for ASHIFT, not MULT inner operand.
28752         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28754 2018-01-04  Martin Liska  <mliska@suse.cz>
28756         PR gcov-profile/83669
28757         * gcov.c (output_intermediate_file): Add version to intermediate
28758         gcov file.
28759         * doc/gcov.texi: Document new field 'version' in intermediate
28760         file format. Fix location of '-k' option of gcov command.
28762 2018-01-04  Martin Liska  <mliska@suse.cz>
28764         PR ipa/82352
28765         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28767 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28769         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28771 2018-01-03  Martin Sebor  <msebor@redhat.com>
28773         PR tree-optimization/83655
28774         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28775         checking calls with invalid arguments.
28777 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28779         * tree-vect-stmts.c (vect_get_store_rhs): New function.
28780         (vectorizable_mask_load_store): Delete.
28781         (vectorizable_call): Return false for masked loads and stores.
28782         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
28783         instead of gimple_assign_rhs1.
28784         (vectorizable_load): Handle IFN_MASK_LOAD.
28785         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28787 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28789         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28790         split out from..,
28791         (vectorizable_mask_load_store): ...here.
28792         (vectorizable_load): ...and here.
28794 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28796         * tree-vect-stmts.c (vect_build_all_ones_mask)
28797         (vect_build_zero_merge_argument): New functions, split out from...
28798         (vectorizable_load): ...here.
28800 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28802         * tree-vect-stmts.c (vect_check_store_rhs): New function,
28803         split out from...
28804         (vectorizable_mask_load_store): ...here.
28805         (vectorizable_store): ...and here.
28807 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28809         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28810         split out from...
28811         (vectorizable_mask_load_store): ...here.
28813 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28815         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28816         (vect_model_store_cost): Take a vec_load_store_type instead of a
28817         vect_def_type.
28818         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28819         (vect_model_store_cost): Take a vec_load_store_type instead of a
28820         vect_def_type.
28821         (vectorizable_mask_load_store): Update accordingly.
28822         (vectorizable_store): Likewise.
28823         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28825 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28827         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28828         IFN_MASK_LOAD calls here rather than...
28829         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28831 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28832             Alan Hayward  <alan.hayward@arm.com>
28833             David Sherwood  <david.sherwood@arm.com>
28835         * expmed.c (extract_bit_field_1): For vector extracts,
28836         fall back to extract_bit_field_as_subreg if vec_extract
28837         isn't available.
28839 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28840             Alan Hayward  <alan.hayward@arm.com>
28841             David Sherwood  <david.sherwood@arm.com>
28843         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28844         they are variable or constant sized.
28845         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28846         slots for constant-sized data.
28848 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28849             Alan Hayward  <alan.hayward@arm.com>
28850             David Sherwood  <david.sherwood@arm.com>
28852         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28853         handling COND_EXPRs with boolean comparisons, try to find a better
28854         basis for the mask type than the boolean itself.
28856 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28858         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28859         is calculated and how it can be overridden.
28860         * genmodes.c (max_bitsize_mode_any_mode): New variable.
28861         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28862         if defined.
28863         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28864         if nonzero.
28866 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28867             Alan Hayward  <alan.hayward@arm.com>
28868             David Sherwood  <david.sherwood@arm.com>
28870         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28871         Remove the mode argument.
28872         (aarch64_simd_valid_immediate): Remove the mode and inverse
28873         arguments.
28874         * config/aarch64/iterators.md (bitsize): New iterator.
28875         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28876         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28877         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28878         aarch64_simd_valid_immediate.
28879         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28880         (aarch64_reg_or_bic_imm): Likewise.
28881         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28882         with an insn_type enum and msl with a modifier_type enum.
28883         Replace element_width with a scalar_mode.  Change the shift
28884         to unsigned int.  Add constructors for scalar_float_mode and
28885         scalar_int_mode elements.
28886         (aarch64_vect_float_const_representable_p): Delete.
28887         (aarch64_can_const_movi_rtx_p)
28888         (aarch64_simd_scalar_immediate_valid_for_move)
28889         (aarch64_simd_make_constant): Update call to
28890         aarch64_simd_valid_immediate.
28891         (aarch64_advsimd_valid_immediate_hs): New function.
28892         (aarch64_advsimd_valid_immediate): Likewise.
28893         (aarch64_simd_valid_immediate): Remove mode and inverse
28894         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
28895         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28896         and aarch64_float_const_representable_p on the result.
28897         (aarch64_output_simd_mov_immediate): Remove mode argument.
28898         Update call to aarch64_simd_valid_immediate and use of
28899         simd_immediate_info.
28900         (aarch64_output_scalar_simd_mov_immediate): Update call
28901         accordingly.
28903 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28904             Alan Hayward  <alan.hayward@arm.com>
28905             David Sherwood  <david.sherwood@arm.com>
28907         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28908         (mode_nunits): Likewise CONST_MODE_NUNITS.
28909         * machmode.def (ADJUST_NUNITS): Document.
28910         * genmodes.c (mode_data::need_nunits_adj): New field.
28911         (blank_mode): Update accordingly.
28912         (adj_nunits): New variable.
28913         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28914         parameter.
28915         (emit_mode_size_inline): Set need_bytesize_adj for all modes
28916         listed in adj_nunits.
28917         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28918         listed in adj_nunits.  Don't emit case statements for such modes.
28919         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28920         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
28921         nothing if adj_nunits is nonnull.
28922         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28923         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28924         (emit_mode_fbit): Update use of print_maybe_const_decl.
28925         (emit_move_size): Likewise.  Treat the array as non-const
28926         if adj_nunits.
28927         (emit_mode_adjustments): Handle adj_nunits.
28929 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28931         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28932         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28933         (VECTOR_MODES): Use it.
28934         (make_vector_modes): Take the prefix as an argument.
28936 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28937             Alan Hayward  <alan.hayward@arm.com>
28938             David Sherwood  <david.sherwood@arm.com>
28940         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28941         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28942         for MODE_VECTOR_BOOL.
28943         * machmode.def (VECTOR_BOOL_MODE): Document.
28944         * genmodes.c (VECTOR_BOOL_MODE): New macro.
28945         (make_vector_bool_mode): New function.
28946         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28947         MODE_VECTOR_BOOL.
28948         * lto-streamer-in.c (lto_input_mode_table): Likewise.
28949         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28950         Likewise.
28951         * stor-layout.c (int_mode_for_mode): Likewise.
28952         * tree.c (build_vector_type_for_mode): Likewise.
28953         * varasm.c (output_constant_pool_2): Likewise.
28954         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28955         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
28956         for MODE_VECTOR_BOOL.
28957         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28958         of mode class checks.
28959         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28960         instead of a list of mode class checks.
28961         (expand_vector_scalar_condition): Likewise.
28962         (type_for_widest_vector_mode): Handle BImode as an inner mode.
28964 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28965             Alan Hayward  <alan.hayward@arm.com>
28966             David Sherwood  <david.sherwood@arm.com>
28968         * machmode.h (mode_size): Change from unsigned short to
28969         poly_uint16_pod.
28970         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28971         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28972         or if measurement_type is not polynomial.
28973         (fixed_size_mode::includes_p): Check for constant-sized modes.
28974         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28975         return a poly_uint16 rather than an unsigned short.
28976         (emit_mode_size): Change the type of mode_size from unsigned short
28977         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
28978         (emit_mode_adjustments): Cope with polynomial vector sizes.
28979         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28980         for GET_MODE_SIZE.
28981         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28982         for GET_MODE_SIZE.
28983         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28984         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28985         * caller-save.c (setup_save_areas): Likewise.
28986         (replace_reg_with_saved_mem): Likewise.
28987         * calls.c (emit_library_call_value_1): Likewise.
28988         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28989         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28990         (gen_lowpart_for_combine): Likewise.
28991         * convert.c (convert_to_integer_1): Likewise.
28992         * cse.c (equiv_constant, cse_insn): Likewise.
28993         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28994         (cselib_subst_to_values): Likewise.
28995         * dce.c (word_dce_process_block): Likewise.
28996         * df-problems.c (df_word_lr_mark_ref): Likewise.
28997         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28998         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28999         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
29000         (rtl_for_decl_location): Likewise.
29001         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
29002         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
29003         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
29004         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
29005         (expand_expr_real_1): Likewise.
29006         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
29007         (pad_below): Likewise.
29008         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29009         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
29010         * ira.c (get_subreg_tracking_sizes): Likewise.
29011         * ira-build.c (ira_create_allocno_objects): Likewise.
29012         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
29013         (ira_sort_regnos_for_alter_reg): Likewise.
29014         * ira-costs.c (record_operand_costs): Likewise.
29015         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
29016         (resolve_simple_move): Likewise.
29017         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
29018         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
29019         (lra_constraints): Likewise.
29020         (CONST_POOL_OK_P): Reject variable-sized modes.
29021         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
29022         (add_pseudo_to_slot, lra_spill): Likewise.
29023         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29024         * optabs-query.c (get_best_extraction_insn): Likewise.
29025         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29026         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
29027         (expand_mult_highpart, valid_multiword_target_p): Likewise.
29028         * recog.c (offsettable_address_addr_space_p): Likewise.
29029         * regcprop.c (maybe_mode_change): Likewise.
29030         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
29031         * regrename.c (build_def_use): Likewise.
29032         * regstat.c (dump_reg_info): Likewise.
29033         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
29034         (find_reloads, find_reloads_subreg_address): Likewise.
29035         * reload1.c (eliminate_regs_1): Likewise.
29036         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
29037         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
29038         (simplify_binary_operation_1, simplify_subreg): Likewise.
29039         * targhooks.c (default_function_arg_padding): Likewise.
29040         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
29041         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
29042         (verify_gimple_assign_ternary): Likewise.
29043         * tree-inline.c (estimate_move_cost): Likewise.
29044         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29045         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
29046         (get_address_cost_ainc): Likewise.
29047         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
29048         (vect_supportable_dr_alignment): Likewise.
29049         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29050         (vectorizable_reduction): Likewise.
29051         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
29052         (vectorizable_operation, vectorizable_load): Likewise.
29053         * tree.c (build_same_sized_truth_vector_type): Likewise.
29054         * valtrack.c (cleanup_auto_inc_dec): Likewise.
29055         * var-tracking.c (emit_note_insn_var_location): Likewise.
29056         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
29057         (ADDR_VEC_ALIGN): Likewise.
29059 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29060             Alan Hayward  <alan.hayward@arm.com>
29061             David Sherwood  <david.sherwood@arm.com>
29063         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
29064         unsigned short.
29065         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
29066         or if measurement_type is polynomial.
29067         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
29068         * combine.c (make_extraction): Likewise.
29069         * dse.c (find_shift_sequence): Likewise.
29070         * dwarf2out.c (mem_loc_descriptor): Likewise.
29071         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
29072         (extract_bit_field, extract_low_bits): Likewise.
29073         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
29074         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
29075         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
29076         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
29077         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29078         * reload.c (find_reloads): Likewise.
29079         * reload1.c (alter_reg): Likewise.
29080         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
29081         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
29082         * tree-if-conv.c (predicate_mem_writes): Likewise.
29083         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
29084         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
29085         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
29086         * valtrack.c (dead_debug_insert_temp): Likewise.
29087         * varasm.c (mergeable_constant_section): Likewise.
29088         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
29090 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29091             Alan Hayward  <alan.hayward@arm.com>
29092             David Sherwood  <david.sherwood@arm.com>
29094         * expr.c (expand_assignment): Cope with polynomial mode sizes
29095         when assigning to a CONCAT.
29097 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29098             Alan Hayward  <alan.hayward@arm.com>
29099             David Sherwood  <david.sherwood@arm.com>
29101         * machmode.h (mode_precision): Change from unsigned short to
29102         poly_uint16_pod.
29103         (mode_to_precision): Return a poly_uint16 rather than an unsigned
29104         short.
29105         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
29106         or if measurement_type is not polynomial.
29107         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
29108         in which the mode is already known to be a scalar_int_mode.
29109         * genmodes.c (emit_mode_precision): Change the type of mode_precision
29110         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
29111         initializer.
29112         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29113         for GET_MODE_PRECISION.
29114         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29115         for GET_MODE_PRECISION.
29116         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
29117         as polynomial.
29118         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
29119         (expand_field_assignment, make_extraction): Likewise.
29120         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
29121         (get_last_value): Likewise.
29122         * convert.c (convert_to_integer_1): Likewise.
29123         * cse.c (cse_insn): Likewise.
29124         * expr.c (expand_expr_real_1): Likewise.
29125         * lra-constraints.c (simplify_operand_subreg): Likewise.
29126         * optabs-query.c (can_atomic_load_p): Likewise.
29127         * optabs.c (expand_atomic_load): Likewise.
29128         (expand_atomic_store): Likewise.
29129         * ree.c (combine_reaching_defs): Likewise.
29130         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
29131         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
29132         * tree.h (type_has_mode_precision_p): Likewise.
29133         * ubsan.c (instrument_si_overflow): Likewise.
29135 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29136             Alan Hayward  <alan.hayward@arm.com>
29137             David Sherwood  <david.sherwood@arm.com>
29139         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
29140         polynomial numbers of units.
29141         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
29142         (valid_vector_subparts_p): New function.
29143         (build_vector_type): Remove temporary shim and take the number
29144         of units as a poly_uint64 rather than an int.
29145         (build_opaque_vector_type): Take the number of units as a
29146         poly_uint64 rather than an int.
29147         * tree.c (build_vector_from_ctor): Handle polynomial
29148         TYPE_VECTOR_SUBPARTS.
29149         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
29150         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
29151         (build_vector_from_val): If the number of units is variable,
29152         use build_vec_duplicate_cst for constant operands and
29153         VEC_DUPLICATE_EXPR otherwise.
29154         (make_vector_type): Remove temporary is_constant ().
29155         (build_vector_type, build_opaque_vector_type): Take the number of
29156         units as a poly_uint64 rather than an int.
29157         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
29158         VECTOR_CST_NELTS.
29159         * cfgexpand.c (expand_debug_expr): Likewise.
29160         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
29161         (store_constructor, expand_expr_real_1): Likewise.
29162         (const_scalar_mask_from_tree): Likewise.
29163         * fold-const-call.c (fold_const_reduction): Likewise.
29164         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
29165         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
29166         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
29167         (fold_relational_const): Likewise.
29168         (native_interpret_vector): Likewise.  Change the size from an
29169         int to an unsigned int.
29170         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
29171         TYPE_VECTOR_SUBPARTS.
29172         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
29173         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
29174         duplicating a non-constant operand into a variable-length vector.
29175         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
29176         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
29177         * ipa-icf.c (sem_variable::equals): Likewise.
29178         * match.pd: Likewise.
29179         * omp-simd-clone.c (simd_clone_subparts): Likewise.
29180         * print-tree.c (print_node): Likewise.
29181         * stor-layout.c (layout_type): Likewise.
29182         * targhooks.c (default_builtin_vectorization_cost): Likewise.
29183         * tree-cfg.c (verify_gimple_comparison): Likewise.
29184         (verify_gimple_assign_binary): Likewise.
29185         (verify_gimple_assign_ternary): Likewise.
29186         (verify_gimple_assign_single): Likewise.
29187         * tree-pretty-print.c (dump_generic_node): Likewise.
29188         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29189         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
29190         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
29191         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
29192         (vect_shift_permute_load_chain): Likewise.
29193         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
29194         (expand_vector_condition, optimize_vector_constructor): Likewise.
29195         (lower_vec_perm, get_compute_type): Likewise.
29196         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29197         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
29198         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
29199         (vect_recog_mask_conversion_pattern): Likewise.
29200         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
29201         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
29202         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29203         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
29204         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
29205         (vectorizable_shift, vectorizable_operation, vectorizable_store)
29206         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
29207         (supportable_widening_operation): Likewise.
29208         (supportable_narrowing_operation): Likewise.
29209         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
29210         Likewise.
29211         * varasm.c (output_constant): Likewise.
29213 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29214             Alan Hayward  <alan.hayward@arm.com>
29215             David Sherwood  <david.sherwood@arm.com>
29217         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
29218         so that both the length == 3 and length != 3 cases set up their
29219         own permute vectors.  Add comments explaining why we know the
29220         number of elements is constant.
29221         (vect_permute_load_chain): Likewise.
29223 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29224             Alan Hayward  <alan.hayward@arm.com>
29225             David Sherwood  <david.sherwood@arm.com>
29227         * machmode.h (mode_nunits): Change from unsigned char to
29228         poly_uint16_pod.
29229         (ONLY_FIXED_SIZE_MODES): New macro.
29230         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
29231         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
29232         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
29233         New typedefs.
29234         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
29235         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
29236         or if measurement_type is not polynomial.
29237         * genmodes.c (ZERO_COEFFS): New macro.
29238         (emit_mode_nunits_inline): Make mode_nunits_inline return a
29239         poly_uint16.
29240         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
29241         Use ZERO_COEFFS when emitting initializers.
29242         * data-streamer.h (bp_pack_poly_value): New function.
29243         (bp_unpack_poly_value): Likewise.
29244         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29245         for GET_MODE_NUNITS.
29246         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29247         for GET_MODE_NUNITS.
29248         * tree.c (make_vector_type): Remove temporary shim and make
29249         the real function take the number of units as a poly_uint64
29250         rather than an int.
29251         (build_vector_type_for_mode): Handle polynomial nunits.
29252         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
29253         * emit-rtl.c (const_vec_series_p_1): Likewise.
29254         (gen_rtx_CONST_VECTOR): Likewise.
29255         * fold-const.c (test_vec_duplicate_folding): Likewise.
29256         * genrecog.c (validate_pattern): Likewise.
29257         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
29258         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29259         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
29260         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
29261         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
29262         * rtlanal.c (subreg_get_info): Likewise.
29263         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29264         (vect_grouped_load_supported): Likewise.
29265         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
29266         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
29267         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
29268         (simplify_const_unary_operation, simplify_binary_operation_1)
29269         (simplify_const_binary_operation, simplify_ternary_operation)
29270         (test_vector_ops_duplicate, test_vector_ops): Likewise.
29271         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
29272         instead of CONST_VECTOR_NUNITS.
29273         * varasm.c (output_constant_pool_2): Likewise.
29274         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
29275         explicit-encoded elements in the XVEC for variable-length vectors.
29277 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29279         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
29281 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29282             Alan Hayward  <alan.hayward@arm.com>
29283             David Sherwood  <david.sherwood@arm.com>
29285         * coretypes.h (fixed_size_mode): Declare.
29286         (fixed_size_mode_pod): New typedef.
29287         * builtins.h (target_builtins::x_apply_args_mode)
29288         (target_builtins::x_apply_result_mode): Change type to
29289         fixed_size_mode_pod.
29290         * builtins.c (apply_args_size, apply_result_size, result_vector)
29291         (expand_builtin_apply_args_1, expand_builtin_apply)
29292         (expand_builtin_return): Update accordingly.
29294 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29296         * cse.c (hash_rtx_cb): Hash only the encoded elements.
29297         * cselib.c (cselib_hash_rtx): Likewise.
29298         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
29299         CONST_VECTOR encoding.
29301 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29302             Jeff Law  <law@redhat.com>
29304         PR target/83641
29305         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
29306         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
29307         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
29308         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
29310         PR target/83641
29311         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
29312         explicitly probe *sp in a noreturn function if there were any callee
29313         register saves or frame pointer is needed.
29315 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29317         PR debug/83621
29318         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
29319         BLKmode for ternary, binary or unary expressions.
29321         PR debug/83645
29322         * var-tracking.c (delete_vta_debug_insn): New inline function.
29323         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
29324         insns from get_insns () to NULL instead of each bb separately.
29325         Use delete_vta_debug_insn.  No longer static.
29326         (vt_debug_insns_local, variable_tracking_main_1): Adjust
29327         delete_vta_debug_insns callers.
29328         * rtl.h (delete_vta_debug_insns): Declare.
29329         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
29330         instead of variable_tracking_main.
29332 2018-01-03  Martin Sebor  <msebor@redhat.com>
29334         PR tree-optimization/83603
29335         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
29336         arguments past the endof the argument list in functions declared
29337         without a prototype.
29338         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
29339         Avoid checking when arguments are null.
29341 2018-01-03  Martin Sebor  <msebor@redhat.com>
29343         PR c/83559
29344         * doc/extend.texi (attribute const): Fix a typo.
29345         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
29346         issuing -Wsuggest-attribute for void functions.
29348 2018-01-03  Martin Sebor  <msebor@redhat.com>
29350         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
29351         offset_int::from instead of wide_int::to_shwi.
29352         (maybe_diag_overlap): Remove assertion.
29353         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
29354         * gimple-ssa-sprintf.c (format_directive): Same.
29355         (parse_directive): Same.
29356         (sprintf_dom_walker::compute_format_length): Same.
29357         (try_substitute_return_value): Same.
29359 2018-01-03  Jeff Law  <law@redhat.com>
29361         PR middle-end/83654
29362         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
29363         non-constant residual for zero at runtime and avoid probing in
29364         that case.  Reorganize code for trailing problem to mirror handling
29365         of the residual.
29367 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29369         PR tree-optimization/83501
29370         * tree-ssa-strlen.c (get_string_cst): New.
29371         (handle_char_store): Call get_string_cst.
29373 2018-01-03  Martin Liska  <mliska@suse.cz>
29375         PR tree-optimization/83593
29376         * tree-ssa-strlen.c: Include tree-cfg.h.
29377         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
29378         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
29379         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
29380         to false.
29381         (strlen_dom_walker::before_dom_children): Call
29382         gimple_purge_dead_eh_edges. Dump tranformation with details
29383         dump flags.
29384         (strlen_dom_walker::before_dom_children): Update call by adding
29385         new argument cleanup_eh.
29386         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
29388 2018-01-03  Martin Liska  <mliska@suse.cz>
29390         PR ipa/83549
29391         * cif-code.def (VARIADIC_THUNK): New enum value.
29392         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
29393         thunks.
29395 2018-01-03  Jan Beulich  <jbeulich@suse.com>
29397         * sse.md (mov<mode>_internal): Tighten condition for when to use
29398         vmovdqu<ssescalarsize> for TI and OI modes.
29400 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29402         Update copyright years.
29404 2018-01-03  Martin Liska  <mliska@suse.cz>
29406         PR ipa/83594
29407         * ipa-visibility.c (function_and_variable_visibility): Skip
29408         functions with noipa attribure.
29410 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
29412         * gcc.c (process_command): Update copyright notice dates.
29413         * gcov-dump.c (print_version): Ditto.
29414         * gcov.c (print_version): Ditto.
29415         * gcov-tool.c (print_version): Ditto.
29416         * gengtype.c (create_file): Ditto.
29417         * doc/cpp.texi: Bump @copying's copyright year.
29418         * doc/cppinternals.texi: Ditto.
29419         * doc/gcc.texi: Ditto.
29420         * doc/gccint.texi: Ditto.
29421         * doc/gcov.texi: Ditto.
29422         * doc/install.texi: Ditto.
29423         * doc/invoke.texi: Ditto.
29425 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29427         * vector-builder.h (vector_builder::m_full_nelts): Change from
29428         unsigned int to poly_uint64.
29429         (vector_builder::full_nelts): Update prototype accordingly.
29430         (vector_builder::new_vector): Likewise.
29431         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
29432         (vector_builder::operator ==): Likewise.
29433         (vector_builder::finalize): Likewise.
29434         * int-vector-builder.h (int_vector_builder::int_vector_builder):
29435         Take the number of elements as a poly_uint64 rather than an
29436         unsigned int.
29437         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
29438         from unsigned int to poly_uint64.
29439         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
29440         (vec_perm_indices::new_vector): Likewise.
29441         (vec_perm_indices::length): Likewise.
29442         (vec_perm_indices::nelts_per_input): Likewise.
29443         (vec_perm_indices::input_nelts): Likewise.
29444         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
29445         number of elements per input as a poly_uint64 rather than an
29446         unsigned int.  Use the original encoding for variable-length
29447         vectors, rather than clamping each individual element.
29448         For the second and subsequent elements in each pattern,
29449         clamp the step and base before clamping their sum.
29450         (vec_perm_indices::series_p): Handle polynomial element counts.
29451         (vec_perm_indices::all_in_range_p): Likewise.
29452         (vec_perm_indices_to_tree): Likewise.
29453         (vec_perm_indices_to_rtx): Likewise.
29454         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
29455         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
29456         (tree_vector_builder::new_binary_operation): Handle polynomial
29457         element counts.  Return false if we need to know the number
29458         of elements at compile time.
29459         * fold-const.c (fold_vec_perm): Punt if the number of elements
29460         isn't known at compile time.
29462 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29464         * vec-perm-indices.h (vec_perm_builder): Change element type
29465         from HOST_WIDE_INT to poly_int64.
29466         (vec_perm_indices::element_type): Update accordingly.
29467         (vec_perm_indices::clamp): Handle polynomial element_types.
29468         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29469         (vec_perm_indices::all_in_range_p): Likewise.
29470         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
29471         than shwi trees.
29472         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
29473         polynomial vec_perm_indices element types.
29474         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
29475         * fold-const.c (fold_vec_perm): Likewise.
29476         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
29477         * tree-vect-generic.c (lower_vec_perm): Likewise.
29478         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29479         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
29480         element type to HOST_WIDE_INT.
29482 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29483             Alan Hayward  <alan.hayward@arm.com>
29484             David Sherwood  <david.sherwood@arm.com>
29486         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
29487         rather than an int.  Use plus_constant.
29488         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
29489         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
29491 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29492             Alan Hayward  <alan.hayward@arm.com>
29493             David Sherwood  <david.sherwood@arm.com>
29495         * calls.c (emit_call_1, expand_call): Change struct_value_size from
29496         a HOST_WIDE_INT to a poly_int64.
29498 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29499             Alan Hayward  <alan.hayward@arm.com>
29500             David Sherwood  <david.sherwood@arm.com>
29502         * calls.c (load_register_parameters): Cope with polynomial
29503         mode sizes.  Require a constant size for BLKmode parameters
29504         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
29505         forces a parameter to be padded at the lsb end in order to
29506         fill a complete number of words, require the parameter size
29507         to be ordered wrt UNITS_PER_WORD.
29509 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29510             Alan Hayward  <alan.hayward@arm.com>
29511             David Sherwood  <david.sherwood@arm.com>
29513         * reload1.c (spill_stack_slot_width): Change element type
29514         from unsigned int to poly_uint64_pod.
29515         (alter_reg): Treat mode sizes as polynomial.
29517 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29518             Alan Hayward  <alan.hayward@arm.com>
29519             David Sherwood  <david.sherwood@arm.com>
29521         * reload.c (complex_word_subreg_p): New function.
29522         (reload_inner_reg_of_subreg, push_reload): Use it.
29524 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29525             Alan Hayward  <alan.hayward@arm.com>
29526             David Sherwood  <david.sherwood@arm.com>
29528         * lra-constraints.c (process_alt_operands): Reject matched
29529         operands whose sizes aren't ordered.
29530         (match_reload): Refer to this check here.
29532 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29533             Alan Hayward  <alan.hayward@arm.com>
29534             David Sherwood  <david.sherwood@arm.com>
29536         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
29537         that the mode size is in the set {1, 2, 4, 8, 16}.
29539 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29540             Alan Hayward  <alan.hayward@arm.com>
29541             David Sherwood  <david.sherwood@arm.com>
29543         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
29544         Use plus_constant instead of gen_rtx_PLUS.
29546 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29547             Alan Hayward  <alan.hayward@arm.com>
29548             David Sherwood  <david.sherwood@arm.com>
29550         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
29551         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
29552         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
29553         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
29554         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
29555         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
29556         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
29557         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
29558         * config/i386/i386.c (ix86_push_rounding): ...this new function.
29559         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29560         a poly_int64.
29561         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
29562         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
29563         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
29564         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
29565         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
29566         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
29567         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
29568         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
29569         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
29570         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
29571         function.
29572         * expr.c (emit_move_resolve_push): Treat the input and result
29573         of PUSH_ROUNDING as a poly_int64.
29574         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
29575         (emit_push_insn): Likewise.
29576         * lra-eliminations.c (mark_not_eliminable): Likewise.
29577         * recog.c (push_operand): Likewise.
29578         * reload1.c (elimination_effects): Likewise.
29579         * rtlanal.c (nonzero_bits1): Likewise.
29580         * calls.c (store_one_arg): Likewise.  Require the padding to be
29581         known at compile time.
29583 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29584             Alan Hayward  <alan.hayward@arm.com>
29585             David Sherwood  <david.sherwood@arm.com>
29587         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
29588         Use plus_constant instead of gen_rtx_PLUS.
29590 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29591             Alan Hayward  <alan.hayward@arm.com>
29592             David Sherwood  <david.sherwood@arm.com>
29594         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
29595         rather than an int.
29597 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29598             Alan Hayward  <alan.hayward@arm.com>
29599             David Sherwood  <david.sherwood@arm.com>
29601         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
29602         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
29603         via stack temporaries.  Treat the mode size as polynomial too.
29605 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29606             Alan Hayward  <alan.hayward@arm.com>
29607             David Sherwood  <david.sherwood@arm.com>
29609         * expr.c (expand_expr_real_2): When handling conversions involving
29610         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
29611         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
29612         as a poly_uint64 too.
29614 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29615             Alan Hayward  <alan.hayward@arm.com>
29616             David Sherwood  <david.sherwood@arm.com>
29618         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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         * combine.c (can_change_dest_mode): Handle polynomial
29625         REGMODE_NATURAL_SIZE.
29626         * expmed.c (store_bit_field_1): Likewise.
29627         * expr.c (store_constructor): Likewise.
29628         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29629         and polynomial REGMODE_NATURAL_SIZE.
29630         (gen_lowpart_common): Likewise.
29631         * reginfo.c (record_subregs_of_mode): Likewise.
29632         * rtlanal.c (read_modify_subreg_p): Likewise.
29634 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29635             Alan Hayward  <alan.hayward@arm.com>
29636             David Sherwood  <david.sherwood@arm.com>
29638         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29639         numbers of elements.
29641 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29642             Alan Hayward  <alan.hayward@arm.com>
29643             David Sherwood  <david.sherwood@arm.com>
29645         * match.pd: Cope with polynomial numbers of vector 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         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29652         in a POINTER_PLUS_EXPR.
29654 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29655             Alan Hayward  <alan.hayward@arm.com>
29656             David Sherwood  <david.sherwood@arm.com>
29658         * omp-simd-clone.c (simd_clone_subparts): New function.
29659         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29660         (ipa_simd_modify_function_body): Likewise.
29662 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29663             Alan Hayward  <alan.hayward@arm.com>
29664             David Sherwood  <david.sherwood@arm.com>
29666         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29667         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29668         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29669         (expand_vector_condition, vector_element): Likewise.
29670         (subparts_gt): New function.
29671         (get_compute_type): Use subparts_gt.
29672         (count_type_subparts): Delete.
29673         (expand_vector_operations_1): Use subparts_gt instead of
29674         count_type_subparts.
29676 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29677             Alan Hayward  <alan.hayward@arm.com>
29678             David Sherwood  <david.sherwood@arm.com>
29680         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29681         (vect_compile_time_alias): ...this new function.  Do the calculation
29682         on poly_ints rather than trees.
29683         (vect_prune_runtime_alias_test_list): Update call accordingly.
29685 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29686             Alan Hayward  <alan.hayward@arm.com>
29687             David Sherwood  <david.sherwood@arm.com>
29689         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29690         numbers of units.
29691         (vect_schedule_slp_instance): Likewise.
29693 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29694             Alan Hayward  <alan.hayward@arm.com>
29695             David Sherwood  <david.sherwood@arm.com>
29697         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29698         constant and extern definitions for variable-length vectors.
29699         (vect_get_constant_vectors): Note that the number of units
29700         is known to be constant.
29702 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29703             Alan Hayward  <alan.hayward@arm.com>
29704             David Sherwood  <david.sherwood@arm.com>
29706         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29707         of units as polynomial.  Choose between WIDE and NARROW based
29708         on multiple_p.
29710 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29711             Alan Hayward  <alan.hayward@arm.com>
29712             David Sherwood  <david.sherwood@arm.com>
29714         * tree-vect-stmts.c (simd_clone_subparts): New function.
29715         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
29717 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29718             Alan Hayward  <alan.hayward@arm.com>
29719             David Sherwood  <david.sherwood@arm.com>
29721         * tree-vect-stmts.c (vectorizable_call): Treat the number of
29722         vectors as polynomial.  Use build_index_vector for
29723         IFN_GOMP_SIMD_LANE.
29725 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29726             Alan Hayward  <alan.hayward@arm.com>
29727             David Sherwood  <david.sherwood@arm.com>
29729         * tree-vect-stmts.c (get_load_store_type): Treat the number of
29730         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29731         for variable-length vectors.
29732         (vectorizable_mask_load_store): Treat the number of units as
29733         polynomial, asserting that it is constant if the condition has
29734         already been enforced.
29735         (vectorizable_store, vectorizable_load): Likewise.
29737 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29738             Alan Hayward  <alan.hayward@arm.com>
29739             David Sherwood  <david.sherwood@arm.com>
29741         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29742         of units as polynomial.  Punt if we can't tell at compile time
29743         which vector contains the final result.
29745 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29746             Alan Hayward  <alan.hayward@arm.com>
29747             David Sherwood  <david.sherwood@arm.com>
29749         * tree-vect-loop.c (vectorizable_induction): Treat the number
29750         of units as polynomial.  Punt on SLP inductions.  Use an integer
29751         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
29752         cast of such a series for variable-length floating-point
29753         reductions.
29755 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29756             Alan Hayward  <alan.hayward@arm.com>
29757             David Sherwood  <david.sherwood@arm.com>
29759         * tree.h (build_index_vector): Declare.
29760         * tree.c (build_index_vector): New function.
29761         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29762         of units as polynomial, forcibly converting it to a constant if
29763         vectorizable_reduction has already enforced the condition.
29764         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
29765         to create a {1,2,3,...} vector.
29766         (vectorizable_reduction): Treat the number of units as polynomial.
29767         Choose vectype_in based on the largest scalar element size rather
29768         than the smallest number of units.  Enforce the restrictions
29769         relied on above.
29771 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29772             Alan Hayward  <alan.hayward@arm.com>
29773             David Sherwood  <david.sherwood@arm.com>
29775         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29776         number of units as polynomial.
29778 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29779             Alan Hayward  <alan.hayward@arm.com>
29780             David Sherwood  <david.sherwood@arm.com>
29782         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29783         * target.def (autovectorize_vector_sizes): Return the vector sizes
29784         by pointer, using vector_sizes rather than a bitmask.
29785         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29786         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29787         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29788         Likewise.
29789         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29790         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29791         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29792         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29793         * omp-general.c (omp_max_vf): Likewise.
29794         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29795         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29796         * tree-vect-loop.c (vect_analyze_loop): Likewise.
29797         * tree-vect-slp.c (vect_slp_bb): Likewise.
29798         * doc/tm.texi: Regenerate.
29799         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29800         to a poly_uint64.
29801         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29802         the vector size as a poly_uint64 rather than an unsigned int.
29803         (current_vector_size): Change from an unsigned int to a poly_uint64.
29804         (get_vectype_for_scalar_type): Update accordingly.
29805         * tree.h (build_truth_vector_type): Take the size and number of
29806         units as a poly_uint64 rather than an unsigned int.
29807         (build_vector_type): Add a temporary overload that takes
29808         the number of units as a poly_uint64 rather than an unsigned int.
29809         * tree.c (make_vector_type): Likewise.
29810         (build_truth_vector_type): Take the number of units as a poly_uint64
29811         rather than an unsigned int.
29813 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29814             Alan Hayward  <alan.hayward@arm.com>
29815             David Sherwood  <david.sherwood@arm.com>
29817         * target.def (get_mask_mode): Take the number of units and length
29818         as poly_uint64s rather than unsigned ints.
29819         * targhooks.h (default_get_mask_mode): Update accordingly.
29820         * targhooks.c (default_get_mask_mode): Likewise.
29821         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29822         * doc/tm.texi: Regenerate.
29824 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29825             Alan Hayward  <alan.hayward@arm.com>
29826             David Sherwood  <david.sherwood@arm.com>
29828         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29829         * omp-general.c (omp_max_vf): Likewise.
29830         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29831         (expand_omp_simd): Handle polynomial safelen.
29832         * omp-low.c (omplow_simd_context): Add a default constructor.
29833         (omplow_simd_context::max_vf): Change from int to poly_uint64.
29834         (lower_rec_simd_input_clauses): Update accordingly.
29835         (lower_rec_input_clauses): Likewise.
29837 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29838             Alan Hayward  <alan.hayward@arm.com>
29839             David Sherwood  <david.sherwood@arm.com>
29841         * tree-vectorizer.h (vect_nunits_for_cost): New function.
29842         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29843         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29844         (vect_analyze_slp_cost): Likewise.
29845         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29846         (vect_model_load_cost): Likewise.
29848 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29849             Alan Hayward  <alan.hayward@arm.com>
29850             David Sherwood  <david.sherwood@arm.com>
29852         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29853         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29854         from an unsigned int * to a poly_uint64_pod *.
29855         (calculate_unrolling_factor): New function.
29856         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
29858 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29859             Alan Hayward  <alan.hayward@arm.com>
29860             David Sherwood  <david.sherwood@arm.com>
29862         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29863         from an unsigned int to a poly_uint64.
29864         (_loop_vec_info::slp_unrolling_factor): Likewise.
29865         (_loop_vec_info::vectorization_factor): Change from an int
29866         to a poly_uint64.
29867         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29868         (vect_get_num_vectors): New function.
29869         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29870         (vect_get_num_copies): Use vect_get_num_vectors.
29871         (vect_analyze_data_ref_dependences): Change max_vf from an int *
29872         to an unsigned int *.
29873         (vect_analyze_data_refs): Change min_vf from an int * to a
29874         poly_uint64 *.
29875         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29876         than an unsigned HOST_WIDE_INT.
29877         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29878         (vect_analyze_data_ref_dependence): Change max_vf from an int *
29879         to an unsigned int *.
29880         (vect_analyze_data_ref_dependences): Likewise.
29881         (vect_compute_data_ref_alignment): Handle polynomial vf.
29882         (vect_enhance_data_refs_alignment): Likewise.
29883         (vect_prune_runtime_alias_test_list): Likewise.
29884         (vect_shift_permute_load_chain): Likewise.
29885         (vect_supportable_dr_alignment): Likewise.
29886         (dependence_distance_ge_vf): Take the vectorization factor as a
29887         poly_uint64 rather than an unsigned HOST_WIDE_INT.
29888         (vect_analyze_data_refs): Change min_vf from an int * to a
29889         poly_uint64 *.
29890         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29891         vfm1 as a poly_uint64 rather than an int.  Make the same change
29892         for the returned bound_scalar.
29893         (vect_gen_vector_loop_niters): Handle polynomial vf.
29894         (vect_do_peeling): Likewise.  Update call to
29895         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29896         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29897         be constant.
29898         * tree-vect-loop.c (vect_determine_vectorization_factor)
29899         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29900         (vect_get_known_peeling_cost): Likewise.
29901         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29902         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29903         (vect_transform_loop): Likewise.  Use the lowest possible VF when
29904         updating the upper bounds of the loop.
29905         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
29906         rather than an int.
29907         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29908         polynomial unroll factors.
29909         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29910         (vect_make_slp_decision): Likewise.
29911         (vect_supported_load_permutation_p): Likewise, and polynomial
29912         vf too.
29913         (vect_analyze_slp_cost): Handle polynomial vf.
29914         (vect_slp_analyze_node_operations): Likewise.
29915         (vect_slp_analyze_bb_1): Likewise.
29916         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29917         than an unsigned HOST_WIDE_INT.
29918         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29919         (vectorizable_load): Handle polynomial vf.
29920         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29921         a poly_uint64.
29922         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29924 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29925             Alan Hayward  <alan.hayward@arm.com>
29926             David Sherwood  <david.sherwood@arm.com>
29928         * match.pd: Handle bit operations involving three constants
29929         and try to fold one pair.
29931 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29933         * tree-vect-loop-manip.c: Include gimple-fold.h.
29934         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29935         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
29936         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29937         Add a path that uses a step of VF instead of 1, but disable it
29938         for now.
29939         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29940         and niters_no_overflow parameters.  Update calls to
29941         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29942         Create a new SSA name if the latter choses to use a ste other
29943         than zero, and return it via niters_vector_mult_vf_var.
29944         * tree-vect-loop.c (vect_transform_loop): Update calls to
29945         vect_do_peeling, vect_gen_vector_loop_niters and
29946         slpeel_make_loop_iterate_ntimes.
29947         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29948         (vect_gen_vector_loop_niters): Update declarations after above changes.
29950 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
29952         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29953         128-bit round to integer instructions.
29954         (ceil<mode>2): Likewise.
29955         (btrunc<mode>2): Likewise.
29956         (round<mode>2): Likewise.
29958 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29960         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29961         unaligned VSX load/store on P8/P9.
29962         (expand_block_clear): Allow the use of unaligned VSX
29963         load/store on P8/P9.
29965 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29967         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29968         New function.
29969         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29970         swap associated with both a load and a store.
29972 2018-01-02  Andrew Waterman  <andrew@sifive.com>
29974         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29975         * config/riscv/riscv.md (clear_cache): Use it.
29977 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
29979         * web.c: Remove out-of-date comment.
29981 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29983         * expr.c (fixup_args_size_notes): Check that any existing
29984         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29985         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29986         (emit_single_push_insn): ...here.
29988 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29990         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29991         (const_vector_encoded_nelts): New function.
29992         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29993         (const_vector_int_elt, const_vector_elt): Declare.
29994         * emit-rtl.c (const_vector_int_elt_1): New function.
29995         (const_vector_elt): Likewise.
29996         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29997         of CONST_VECTOR_ELT.
29999 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30001         * expr.c: Include rtx-vector-builder.h.
30002         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
30003         directly on the tree encoding.
30004         (const_vector_from_tree): Likewise.
30005         * optabs.c: Include rtx-vector-builder.h.
30006         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
30007         sequence of "u" values.
30008         * vec-perm-indices.c: Include rtx-vector-builder.h.
30009         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
30010         directly on the vec_perm_indices encoding.
30012 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30014         * doc/rtl.texi (const_vector): Describe new encoding scheme.
30015         * Makefile.in (OBJS): Add rtx-vector-builder.o.
30016         * rtx-vector-builder.h: New file.
30017         * rtx-vector-builder.c: Likewise.
30018         * rtl.h (rtx_def::u2): Add a const_vector field.
30019         (CONST_VECTOR_NPATTERNS): New macro.
30020         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
30021         (CONST_VECTOR_DUPLICATE_P): Likewise.
30022         (CONST_VECTOR_STEPPED_P): Likewise.
30023         (CONST_VECTOR_ENCODED_ELT): Likewise.
30024         (const_vec_duplicate_p): Check for a duplicated vector encoding.
30025         (unwrap_const_vec_duplicate): Likewise.
30026         (const_vec_series_p): Check for a non-duplicated vector encoding.
30027         Say that the function only returns true for integer vectors.
30028         * emit-rtl.c: Include rtx-vector-builder.h.
30029         (gen_const_vec_duplicate_1): Delete.
30030         (gen_const_vector): Call gen_const_vec_duplicate instead of
30031         gen_const_vec_duplicate_1.
30032         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
30033         (gen_const_vec_duplicate): Use rtx_vector_builder.
30034         (gen_const_vec_series): Likewise.
30035         (gen_rtx_CONST_VECTOR): Likewise.
30036         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
30037         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30038         Build a new vector rather than modifying a CONST_VECTOR in-place.
30039         (handle_special_swappables): Update call accordingly.
30040         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
30041         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30042         Build a new vector rather than modifying a CONST_VECTOR in-place.
30043         (handle_special_swappables): Update call accordingly.
30045 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30047         * simplify-rtx.c (simplify_const_binary_operation): Use
30048         CONST_VECTOR_ELT instead of XVECEXP.
30050 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30052         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
30053         the selector elements to be different from the data elements
30054         if the selector is a VECTOR_CST.
30055         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
30056         ssizetype for the selector.
30058 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30060         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
30061         before testing each element individually.
30062         * tree-vect-generic.c (lower_vec_perm): Likewise.
30064 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30066         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
30067         * selftest-run-tests.c (selftest::run_tests): Call it.
30068         * vector-builder.h (vector_builder::operator ==): New function.
30069         (vector_builder::operator !=): Likewise.
30070         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
30071         (vec_perm_indices::all_from_input_p): New function.
30072         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30073         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
30074         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
30075         instead of reading the VECTOR_CST directly.  Detect whether both
30076         vector inputs are the same before constructing the vec_perm_indices,
30077         and update the number of inputs argument accordingly.  Use the
30078         utility functions added above.  Only construct sel2 if we need to.
30080 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30082         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
30083         the broadcast of the low byte.
30084         (expand_mult_highpart): Use an explicit encoding for the permutes.
30085         * optabs-query.c (can_mult_highpart_p): Likewise.
30086         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
30087         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30088         (vectorizable_bswap): Likewise.
30089         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
30090         explicit encoding for the power-of-2 permutes.
30091         (vect_permute_store_chain): Likewise.
30092         (vect_grouped_load_supported): Likewise.
30093         (vect_permute_load_chain): Likewise.
30095 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30097         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
30098         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
30099         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
30100         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
30101         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
30102         (vect_gen_perm_mask_any): Likewise.
30104 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30106         * int-vector-builder.h: New file.
30107         * vec-perm-indices.h: Include int-vector-builder.h.
30108         (vec_perm_indices): Redefine as an int_vector_builder.
30109         (auto_vec_perm_indices): Delete.
30110         (vec_perm_builder): Redefine as a stand-alone class.
30111         (vec_perm_indices::vec_perm_indices): New function.
30112         (vec_perm_indices::clamp): Likewise.
30113         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
30114         (vec_perm_indices::new_vector): New function.
30115         (vec_perm_indices::new_expanded_vector): Update for new
30116         vec_perm_indices class.
30117         (vec_perm_indices::rotate_inputs): New function.
30118         (vec_perm_indices::all_in_range_p): Operate directly on the
30119         encoded form, without computing elided elements.
30120         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
30121         encoding.  Update for new vec_perm_indices class.
30122         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
30123         the given vec_perm_builder.
30124         (expand_vec_perm_var): Update vec_perm_builder constructor.
30125         (expand_mult_highpart): Use vec_perm_builder instead of
30126         auto_vec_perm_indices.
30127         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
30128         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
30129         or double series encoding as appropriate.
30130         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
30131         vec_perm_indices instead of auto_vec_perm_indices.
30132         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30133         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30134         (vect_permute_store_chain): Likewise.
30135         (vect_grouped_load_supported): Likewise.
30136         (vect_permute_load_chain): Likewise.
30137         (vect_shift_permute_load_chain): Likewise.
30138         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30139         (vect_transform_slp_perm_load): Likewise.
30140         (vect_schedule_slp_instance): Likewise.
30141         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30142         (vectorizable_mask_load_store): Likewise.
30143         (vectorizable_bswap): Likewise.
30144         (vectorizable_store): Likewise.
30145         (vectorizable_load): Likewise.
30146         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
30147         vec_perm_indices instead of auto_vec_perm_indices.  Use
30148         tree_to_vec_perm_builder to read the vector from a tree.
30149         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
30150         vec_perm_builder instead of a vec_perm_indices.
30151         (have_whole_vector_shift): Use vec_perm_builder and
30152         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
30153         truncation to calc_vec_perm_mask_for_shift.
30154         (vect_create_epilog_for_reduction): Likewise.
30155         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
30156         from auto_vec_perm_indices to vec_perm_indices.
30157         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30158         instead of changing individual elements.
30159         (aarch64_vectorize_vec_perm_const): Use new_vector to install
30160         the vector in d.perm.
30161         * config/arm/arm.c (expand_vec_perm_d::perm): Change
30162         from auto_vec_perm_indices to vec_perm_indices.
30163         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
30164         instead of changing individual elements.
30165         (arm_vectorize_vec_perm_const): Use new_vector to install
30166         the vector in d.perm.
30167         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
30168         Update vec_perm_builder constructor.
30169         (rs6000_expand_interleave): Likewise.
30170         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
30171         (rs6000_expand_interleave): Likewise.
30173 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30175         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
30176         to qimode could truncate the indices.
30177         * optabs.c (expand_vec_perm_var): Likewise.
30179 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30181         * Makefile.in (OBJS): Add vec-perm-indices.o.
30182         * vec-perm-indices.h: New file.
30183         * vec-perm-indices.c: Likewise.
30184         * target.h (vec_perm_indices): Replace with a forward class
30185         declaration.
30186         (auto_vec_perm_indices): Move to vec-perm-indices.h.
30187         * optabs.h: Include vec-perm-indices.h.
30188         (expand_vec_perm): Delete.
30189         (selector_fits_mode_p, expand_vec_perm_var): Declare.
30190         (expand_vec_perm_const): Declare.
30191         * target.def (vec_perm_const_ok): Replace with...
30192         (vec_perm_const): ...this new hook.
30193         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
30194         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
30195         * doc/tm.texi: Regenerate.
30196         * optabs.def (vec_perm_const): Delete.
30197         * doc/md.texi (vec_perm_const): Likewise.
30198         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
30199         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
30200         expand_vec_perm for constant permutation vectors.  Assert that
30201         the mode of variable permutation vectors is the integer equivalent
30202         of the mode that is being permuted.
30203         * optabs-query.h (selector_fits_mode_p): Declare.
30204         * optabs-query.c: Include vec-perm-indices.h.
30205         (selector_fits_mode_p): New function.
30206         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
30207         is defined, instead of checking whether the vec_perm_const_optab
30208         exists.  Use targetm.vectorize.vec_perm_const instead of
30209         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
30210         fit in the vector mode before using a variable permute.
30211         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
30212         vec_perm_indices instead of an rtx.
30213         (expand_vec_perm): Replace with...
30214         (expand_vec_perm_const): ...this new function.  Take the selector
30215         as a vec_perm_indices rather than an rtx.  Also take the mode of
30216         the selector.  Update call to shift_amt_for_vec_perm_mask.
30217         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
30218         Use vec_perm_indices::new_expanded_vector to expand the original
30219         selector into bytes.  Check whether the indices fit in the vector
30220         mode before using a variable permute.
30221         (expand_vec_perm_var): Make global.
30222         (expand_mult_highpart): Use expand_vec_perm_const.
30223         * fold-const.c: Includes vec-perm-indices.h.
30224         * tree-ssa-forwprop.c: Likewise.
30225         * tree-vect-data-refs.c: Likewise.
30226         * tree-vect-generic.c: Likewise.
30227         * tree-vect-loop.c: Likewise.
30228         * tree-vect-slp.c: Likewise.
30229         * tree-vect-stmts.c: Likewise.
30230         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
30231         Delete.
30232         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
30233         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
30234         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
30235         (aarch64_vectorize_vec_perm_const): ...this new function.
30236         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30237         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30238         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
30239         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
30240         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30241         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30242         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
30243         into...
30244         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
30245         check for NEON modes.
30246         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
30247         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
30248         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
30249         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
30250         into...
30251         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
30252         the old VEC_PERM_CONST conditions.
30253         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
30254         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
30255         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
30256         (ia64_vectorize_vec_perm_const_ok): Merge into...
30257         (ia64_vectorize_vec_perm_const): ...this new function.
30258         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
30259         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
30260         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
30261         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
30262         * config/mips/mips.c (mips_expand_vec_perm_const)
30263         (mips_vectorize_vec_perm_const_ok): Merge into...
30264         (mips_vectorize_vec_perm_const): ...this new function.
30265         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
30266         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
30267         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
30268         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
30269         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
30270         (rs6000_expand_vec_perm_const): Delete.
30271         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
30272         Delete.
30273         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30274         (altivec_expand_vec_perm_const_le): Take each operand individually.
30275         Operate on constant selectors rather than rtxes.
30276         (altivec_expand_vec_perm_const): Likewise.  Update call to
30277         altivec_expand_vec_perm_const_le.
30278         (rs6000_expand_vec_perm_const): Delete.
30279         (rs6000_vectorize_vec_perm_const_ok): Delete.
30280         (rs6000_vectorize_vec_perm_const): New function.
30281         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30282         an element count and rtx array.
30283         (rs6000_expand_extract_even): Update call accordingly.
30284         (rs6000_expand_interleave): Likewise.
30285         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
30286         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
30287         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
30288         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
30289         (rs6000_expand_vec_perm_const): Delete.
30290         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
30291         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30292         (altivec_expand_vec_perm_const_le): Take each operand individually.
30293         Operate on constant selectors rather than rtxes.
30294         (altivec_expand_vec_perm_const): Likewise.  Update call to
30295         altivec_expand_vec_perm_const_le.
30296         (rs6000_expand_vec_perm_const): Delete.
30297         (rs6000_vectorize_vec_perm_const_ok): Delete.
30298         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
30299         reference to the SPE evmerge intructions.
30300         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
30301         an element count and rtx array.
30302         (rs6000_expand_extract_even): Update call accordingly.
30303         (rs6000_expand_interleave): Likewise.
30304         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
30305         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
30306         new function.
30307         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
30309 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30311         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
30312         vector mode and that that mode matches the mode of the data
30313         being permuted.
30314         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
30315         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
30316         directly using expand_vec_perm_1 when forcing selectors into
30317         registers.
30318         (expand_vec_perm_var): New function, split out from expand_vec_perm.
30320 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30322         * optabs-query.h (can_vec_perm_p): Delete.
30323         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
30324         * optabs-query.c (can_vec_perm_p): Split into...
30325         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
30326         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
30327         particular selector is valid.
30328         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30329         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30330         (vect_grouped_load_supported): Likewise.
30331         (vect_shift_permute_load_chain): Likewise.
30332         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
30333         (vect_transform_slp_perm_load): Likewise.
30334         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30335         (vectorizable_bswap): Likewise.
30336         (vect_gen_perm_mask_checked): Likewise.
30337         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
30338         implementations of variable permutation vectors into account
30339         when deciding which selector to use.
30340         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
30341         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
30342         with a false third argument.
30343         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
30344         to test whether the constant selector is valid and can_vec_perm_var_p
30345         to test whether a variable selector is valid.
30347 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30349         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
30350         * optabs-query.c (can_vec_perm_p): Likewise.
30351         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
30352         instead of vec_perm_indices.
30353         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
30354         (vect_gen_perm_mask_checked): Likewise,
30355         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
30356         (vect_gen_perm_mask_checked): Likewise,
30358 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30360         * optabs-query.h (qimode_for_vec_perm): Declare.
30361         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
30362         (qimode_for_vec_perm): ...this new function.
30363         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
30365 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30367         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
30368         does not have a conditional at the top.
30370 2018-01-02  Richard Biener  <rguenther@suse.de>
30372         * ipa-inline.c (big_speedup_p): Fix expression.
30374 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30376         PR target/81616
30377         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
30378         for generic 4->6.
30380 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
30382         PR target/81616
30383         Generic tuning.
30384         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
30385         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
30386         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
30387         cond_taken_branch_cost 3->4.
30389 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
30391         PR tree-optimization/83581
30392         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
30393         TODO_cleanup_cfg if any changes have been made.
30395         PR middle-end/83608
30396         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
30397         convert_modes if target mode has the right side, but different mode
30398         class.
30400         PR middle-end/83609
30401         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
30402         last argument when extracting from CONCAT.  If either from_real or
30403         from_imag is NULL, use expansion through memory.  If result is not
30404         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
30405         the parts directly to inner mode, if even that fails, use expansion
30406         through memory.
30408         PR middle-end/83623
30409         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
30410         check for bswap in mode rather than HImode and use that in expand_unop
30411         too.
30413 Copyright (C) 2018 Free Software Foundation, Inc.
30415 Copying and distribution of this file, with or without modification,
30416 are permitted in any medium without royalty provided the copyright
30417 notice and this notice are preserved.