2018-10-31 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob1f45354555a378087b9c60bfac6023b51060d564
1 2018-10-31  Richard Biener  <rguenther@suse.de>
3         PR middle-end/70359
4         PR middle-end/86270
5         * tree-outof-ssa.c (insert_backedge_copies): Restrict
6         copy generation to useful cases.  Place the copy before
7         the definition of the backedge value when possible.
9 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
11         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
12         * config/arc/arc.c (arc_active_insn): New function.
13         (check_store_cacheline_hazard): Likewise.
14         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
15         (arc_override_options): Disable delay slot scheduler for older
16         A7.
17         (arc_store_addr_hazard_p): New implementation, old one renamed to
18         ...
19         (arc_store_addr_hazard_internal_p): Renamed.
20         (arc_reorg): Don't combine into brcc instructions which are part
21         of hardware hazard solution.
22         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
23         (tune_arc700): Likewise.
24         * config/arc/arc.opt (arc7xx): New tune value.
25         * config/arc/arc700.md: Improve A7 scheduler.
27 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
29         * config/arc/arc.c (arc_override_options): Remove
30         TARGET_COMPACT_CASESI.
31         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
32         (CASE_VECTOR_MODE): Likewise.
33         (CASE_VECTOR_PC_RELATIVE): Likewise.
34         (CASE_VECTOR_SHORTEN_MODE): Likewise.
35         (CASE_VECTOR_SHORTEN_MODE1): Delete.
36         (ADDR_VEC_ALIGN): Update.
37         (ASM_OUTPUT_CASE_LABEL): Undefine.
38         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
39         (TARGET_BI_BIH): Define.
40         (DEFAULT_BRANCH_INDEX): Likewise.
41         * config/arc/arc.md (casesi): Rework to accept BI/BIH
42         instructions, remove compact_casesi use case.
43         (casesi_compact_jump): Remove.
44         (casesi_dispatch): New pattern.
45         * config/arc/arc.opt: Add mbranch-index option. Deprecate
46         compact_casesi option.
47         * doc/invoke.texi: Document mbranch-index option.
49 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
51         * config/arc/arc.c (arc_get_tp): Remove function.
52         (arc_emit_call_tls_get_addr): Likewise.
53         (arc_call_tls_get_addr): New function.
54         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
55         * config/arc/arc.md (tls_load_tp_soft): Remove.
56         (tls_gd_get_addr): Likewise.
58 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
60         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
61         (mulsi3_600_lib): Remove pattern.
62         (umulsi3_highpart_600_lib_le): Likewise.
63         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
64         (umulsidi3): Remove call to umulsidi3_600_lib.
65         (umulsidi3_600_lib): Remove pattern.
66         (peephole2): Remove peephole using the above deprecated patterns.
68 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
70         PR target/87374
71         * config/arm/arm.c (arm_option_check_internal): Disable the combined
72         use of -mslow-flash-data and -mword-relocations.
73         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
74         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
75         flag_pic.
76         * doc/invoke.texi (-mword-relocations): Mention conflict with
77         -mslow-flash-data.
78         (-mslow-flash-data): Reciprocally.
80 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
82         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
83         16-byte modes held in GP registers to use an even regno.
85         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
86         (aarch64_atomic_ldop_supported_p): Remove.
87         (aarch64_gen_atomic_ldop): Remove.
88         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
89         Fully expand LSE operations here.
90         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
91         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
92         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
93         and use ATOMIC_LDOP instead; use register_operand for the input;
94         drop the split and emit insns directly.
95         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
96         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
97         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
99         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
100         (aarch64_gen_atomic_ldop): Don't call it.
101         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
102         Use aarch64_reg_or_zero.
103         (aarch64_atomic_exchange<ALLI>): Likewise.
104         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
105         operand 0; use aarch64_reg_or_zero for input; merge ...
106         (@aarch64_atomic_swp<ALLI>): ... this and remove.
108         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
109         (aarch64_split_compare_and_swap): Use it.
110         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
111         test oldval against the proper predicate.
112         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
113         Use nonmemory_operand for expected.
114         (cas_short_expected_pred): New.
115         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
116         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
117         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
118         (aarch64_plushi_operand): New.
120         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
121         Force oldval into the rval register for TARGET_LSE; emit the compare
122         during initial expansion so that it may be deleted if unused.
123         (aarch64_gen_atomic_cas): Remove.
124         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
125         Change =&r to +r for operand 0; use match_dup for operand 2;
126         remove is_weak and mod_f operands as unused.  Drop the split
127         and merge with...
128         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
129         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
130         (@aarch64_atomic_cas<GPI>): Similarly.
132 2018-10-31  Richard Biener  <rguenther@suse.de>
134         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
135         using ABSU_EXPR.
137 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
139         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
140         * config/aarch64/aarch64.md: Include saphira.md
141         * config/aarch64/saphira.md: New file for pipeline description.
143 2018-10-30  Martin Sebor  <msebor@redhat.com>
145         PR middle-end/87041
146         * gimple-ssa-sprintf.c (format_directive): Use %G to include
147         inlining context.
148         (sprintf_dom_walker::compute_format_length):
149         Avoid setting POSUNDER4K here.
150         (get_destination_size): Handle null argument values.
151         (get_user_idx_format): New function.
152         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
153         functions, including user-defined with attribute format printf.
154         Use %G to include inlining context.
155         Set POSUNDER4K here.
157 2018-10-30  Jan Hubicka  <jh@suse.cz>
159         * params.def (lto-partitions): Bump from 32 to 128.
161 2018-10-30  Jan Hubicka  <jh@suse.cz>
163         * tree.c
164         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
165         head in file.
166         (free_lang_data_in_type): Forward declare.
167         (fld_type_variant_equal_p): New function.
168         (fld_type_variant): New function
169         (fld_incomplete_types): New hash.
170         (fld_incomplete_type_of): New function
171         (fld_simplfied-type): New function.
172         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
173         (free_lang_data): Allocate and free fld_incomplete_type; update call
174         of free_lang_data_in_decl.
176 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
178         * gcov.c (output_lines): Remove duplicate line.
180 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
182         * config/rs6000/rs6000.md (bswapdi2): Force address into register
183         if not in indexed or indirect form.
184         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
185         (bswapdi2_store): Ditto.
186         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
187         helper function.
188         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
189         Prototype for helper function.
191 2018-10-30  Martin Sebor  <msebor@redhat.com>
193         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
194         (target, pragma GCC optimize, pragma GCC target): Ditto.
196 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
198         * doc/extend.texi: Fix prototype and description of
199         __builtin_expect_with_probability.
201 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
203         * cgraph.h (clone_function_name_1): Replaced by new
204           clone_function_name_numbered that takes name as string; for
205           privatize_symbol_name_1 use only.
206           (clone_function_name): Renamed to
207           clone_function_name_numbered to be explicit about numbering.
208           (clone_function_name): New two-argument function that does
209           not number its output.
210           (clone_function_name): New three-argument function that
211           takes a number to append to its output.
212         * cgraphclones.c (duplicate_thunk_for_node):
213           (clone_function_name_1): Renamed.
214           (clone_function_name_numbered): Two new functions.
215           (clone_function_name): Improved documentation.
216           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
217         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
218         * final.c (final_scan_insn_1): Use the new clone_function_name
219           without numbering.
220         * multiple_target.c (create_dispatcher_calls): Ditto.
221           (create_target_clone): Ditto.
222         * omp-expand.c (grid_expand_target_grid_body): Ditto.
223         * omp-low.c (create_omp_child_function_name): Ditto.
224         * omp-simd-clone.c (simd_clone_create): Ditto.
225         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
226           new clone_function_name without numbering.
228 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
230         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
231         Assert that the allocation size is not zero.
233 2018-10-30  Richard Biener  <rguenther@suse.de>
235         PR tree-optimization/87800
236         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
237         non-induction or reduction PHIs.
239 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
241         * config/aarch64/falkor-tag-collision-avoidance.c
242         (execute_tag_collision_avoidance): Call df_note_add_problem.
244 2018-10-30  Martin Liska  <mliska@suse.cz>
246         * doc/extend.texi: Fix typo in documentation
247         of __builtin_expect_with_probability.
249 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
251         PR c++/87721
252         * input.c (get_substring_ranges_for_loc): Detect if
253         linemap_resolve_location gives us a NULL map, and reject
254         this case.
256 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
258         * config.gcc (xstormy16-*-elf): Set tm_d_file.
260 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
261             Martin Sebor  <msebor@redhat.com>
262             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
264         * doc/ux.texi (Quoting): New subsection, adapted from material at
265         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
266         MartinSebor and ManuelLopezIbanez.
267         (Fix-it hints): Note that fix-it hints shouldn't be marked for
268         translation.
270 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
272         PR middle-end/87469
273         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
274         max value.
276 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
278         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
280 2018-10-29  Paul Koning  <ni1d@arrl.net>
282         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
284 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
286         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
287         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
288         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
289         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
290         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
291         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
292         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
293         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
294         Change 'vector' to '__vector'.
295         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
296         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
297         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
298         _mm_avg_pu16): Likewise.  And, whitespace corrections.
300 2018-10-29  Richard Biener  <rguenther@suse.de>
302         PR tree-optimization/87785
303         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
304         internal defs.
306 2018-10-29  Olivier Hainque  <hainque@adacore.com>
308         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
310 2018-10-29  Olivier Hainque  <hainque@adacore.com>
312         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
313         ports configuration macro, defaults to "gnu".
314         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
315         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
317 2018-10-29  Olivier Hainque  <hainque@adacore.com>
319         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
321 2018-10-29  Richard Biener  <rguenther@suse.de>
323         PR tree-optimization/87790
324         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
325         (vect_make_slp_decision): Adjust.
326         (vect_slp_analyze_bb_1): Likewise.
327         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
328         edges.
330 2018-10-29  Richard Biener  <rguenther@suse.de>
332         PR tree-optimization/87785
333         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
334         and processing.
335         (vect_build_slp_tree): Likewise.
336         (vect_gather_slp_loads): New function.
337         (vect_analyze_slp_instance): Gather loads separately from the
338         SLP tree build.
340 2018-10-29  Martin Liska  <mliska@suse.cz>
342         * Makefile.in: Make dependency to json.o.
343         * doc/gcov.texi: Document new JSON format, remove
344         old intermediate format documentation.
345         * gcov.c (struct function_info): Come up with m_name and
346         m_demangled_name.
347         (function_info::function_info): Initialize it.
348         (function_info::~function_info): Release it.
349         (main): Rename flag_intermediate_format to flag_json_format.
350         (print_usage): Describe --json-format.
351         (process_args): Set flag_json_format.
352         (output_intermediate_line): Remove.
353         (output_intermediate_json_line): Likewise.
354         (get_gcov_intermediate_filename): Return new extension
355         ".gcov.json.gz".
356         (output_intermediate_file): Implement JSON emission.
357         (output_json_intermediate_file): Implement JSON emission.
358         (generate_results): Use ::get_name for function name.
359         Handle JSON output file.
360         (read_graph_file): Use ::get_name instead of cplus_demangle.
361         (read_count_file): Likewise.
362         (solve_flow_graph): Likewise.
363         (add_line_counts): Likewise.
364         (accumulate_line_counts): Use new flag_json_format.
365         (output_function_details): Use ::get_name instead of cplus_demangle.
366         (output_lines): Likewise.
367         * json.cc (test_writing_literals): Add new tests.
368         * json.h (class literal): Add new boolean constructor.
370 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
372         PR rtl-optimization/87701
373         PR rtl-optimization/87780
374         * combine.c (make_more_copies): Rewrite.
376 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
378         * doc/generic.texi (ABSU_EXPR): Document.
379         * match.pd (absu(x)*absu(x) -> x*x): Handle.
380         (absu(absu(X)) -> absu(X)): Likewise.
381         (absu(-X) -> absu(X)): Likewise.
382         (absu(X)  where X is nonnegative -> X): Likewise.
384 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
386         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
387         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
388         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
389         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
390         (s-tm-texi): Also check timestamp on d-target.def.
391         (generated_files): Add TM_D_H and d-target-hooks-def.h.
392         (build/genhooks.o): Also depend on D_TARGET_DEF.
393         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
394         variables.
395         * config/aarch64/aarch64-d.c: New file.
396         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
397         Define.
398         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
399         prototype.
400         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
401         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
402         * config/arm/arm-d.c: New file.
403         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
404         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
405         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
406         * config/arm/t-arm (arm-d.o): New rule.
407         * config/default-d.c: New file.
408         * config/glibc-d.c: New file.
409         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
410         * config/i386/i386-d.c: New file.
411         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
412         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
413         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
414         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
415         * config/i386/t-i386 (i386-d.o): New rule.
416         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
417         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
418         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
419         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
420         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
421         * config/mips/mips-d.c: New file.
422         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
423         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
424         * config/mips/t-mips (mips-d.o): New rule.
425         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
426         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
427         * config/powerpcspe/powerpcspe-d.c: New file.
428         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
429         New prototype.
430         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
431         Support GNU D by using 0 as the language type.
432         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
433         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
434         * config/riscv/riscv-d.c: New file.
435         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
436         prototype.
437         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
438         * config/riscv/t-riscv (riscv-d.o): New rule.
439         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
440         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
441         * config/rs6000/rs6000-d.c: New file.
442         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
443         prototype.
444         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
445         Support GNU D by using 0 as the language type.
446         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
447         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
448         * config/s390/s390-d.c: New file.
449         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
450         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
451         * config/s390/t-s390 (s390-d.o): New rule.
452         * config/sparc/sparc-d.c: New file.
453         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
454         prototype.
455         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
456         * config/sparc/t-sparc (sparc-d.o): New rule.
457         * config/t-glibc (glibc-d.o): New rule.
458         * configure: Regenerated.
459         * configure.ac (tm_d_file): New variable.
460         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
461         * doc/contrib.texi (Contributors): Add self for the D frontend.
462         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
463         * doc/install.texi (Configuration): Mention libphobos as an option for
464         --enable-shared.  Mention d as an option for --enable-languages.
465         (Testing): Mention check-d as a target.
466         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
467         name suffixes.  Mention d as a -x option.
468         * doc/sourcebuild.texi (Top Level): Mention libphobos.
469         * doc/standards.texi (Standards): Add section on D language.
470         * doc/tm.texi: Regenerated.
471         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
472         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
473         * dwarf2out.c (is_dlang): New function.
474         (gen_compile_unit_die): Use DW_LANG_D for D.
475         (declare_in_namespace): Return module die for D, instead of adding
476         extra declarations into the namespace.
477         (gen_namespace_die): Generate DW_TAG_module for D.
478         (gen_decl_die): Handle CONST_DECLSs for D.
479         (dwarf2out_decl): Likewise.
480         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
481         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
482         kinds of aggregates.
483         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
484         * genhooks.c: Include d/d-target.def.
486 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
488         PR target/85669
489         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
490         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
491         (STACK_DYNAMIC_OFFSET): Likewise.
493 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
495         PR target/80024
496         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
497         error message.
499 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
501         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
502         return type and other typos.
504 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
506         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
507         a shorter sequence with fewer branches.
508         (emit_final_str_compare_gpr): Ditto.
510 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
512         * config/rs6000/tmmintrin.h: New file.
513         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
515 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
517         * config/rs6000/mmintrin.h: Enable 32bit compilation.
518         * config/rs6000/xmmintrin.h: Likewise.
520 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
522         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
524 2018-10-26  Richard Biener  <rguenther@suse.de>
526         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
527         and wrapper.
528         (vect_mark_slp_stmts_relevant): Likewise.
529         (vect_detect_hybrid_slp_stmts): Likewise.
530         (vect_bb_slp_scalar_cost): Likewise.
531         (vect_remove_slp_scalar_calls): Likewise.
533 2018-10-26  Jan Hubicka  <jh@suse.cz>
535         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
536         (warn_types_mismatch): Fix walk of DECL_NAME.
537         (odr_types_equivalent_p): Fix overactive assert.
539 2018-10-26  Richard Biener  <rguenther@suse.de>
541         PR tree-optimization/87105
542         * tree-vectorizer.h (_slp_tree::refcnt): New member.
543         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
544         refcnt.
545         (vect_create_new_slp_node): Initialize refcnt to one.
546         (bst_traits): Move.
547         (scalar_stmts_set_t, bst_fail): Remove.
548         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
549         (vect_build_slp_tree): Add bst_map argument and lookup
550         already created SLP nodes.
551         (vect_print_slp_tree): Handle a SLP graph, print SLP node
552         addresses.
553         (vect_slp_rearrange_stmts): Handle a SLP graph.
554         (vect_analyze_slp_instance): Adjust and free SLP nodes from
555         the CSE map.  Fix indenting.
556         (vect_schedule_slp_instance): Add short-cut.
558 2018-10-26  Martin Liska  <mliska@suse.cz>
560         PR testsuite/86158
561         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
562         addr_expr and not with pointers.
564 2018-10-26  Jan Hubicka  <jh@suse.cz>
566         * tree.c (free_lang_data_in_type): Only check main variants.
567         * ipa-devirt.c (warn_odr): Make static.
568         (types_same_for_odr): Drop strict variant.
569         (types_odr_comparable): Likewise.
570         (odr_or_derived_type_p): Look for main variants.
571         (odr_name_hasher::equal): Cleanup comment.
572         (odr_subtypes_equivalent): Add warn and warned arguments; check main
573         variants.
574         (type_variants_equivalent_p): break out from ...
575         (odr_types_equivalent): ... here; go for main variants where needed.
576         (warn_odr): ... here; turn static.
577         (warn_types_mismatch): Compare mangled names of main variants.
578         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
579         (type_with_linkage_p): Sanity check that we look at main variant.
580         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
581         * tree.h (types_same_for_odr): Drop strict argument.
583 2018-10-26  Richard Biener  <rguenther@suse.de>
585         PR tree-optimization/87746
586         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
587         Simplify and fix WRT strided store groups with size not
588         equal to step in element count.
589         (vect_analyze_group_access_1): Dump the whole group.
591 2018-10-25  Carl Love  <cel@us.ibm.com>
593         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
594         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
595         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
596         precicion entry for each overloaded builtin.
597         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
598         VSCEDPUO): Rename overloaded name.
599         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
600         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
601         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
602         define_expand for xscmpexqp instruction.
603         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
605 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
606             Jinsong Ji <jji@us.ibm.com>
608         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
609         function with vec_sl.
610         (_mm_slli_epi32): Likewise.
611         (_mm_slli_epi64): Likewise.
612         (_mm_srai_epi16): Replace deprecated function with vec_sra.
613         (_mm_srai_epi32): Likewise.
614         (_mm_srli_epi16): Replace deprecated function with vec_sr.
615         (_mm_srli_epi32): Likewise.
616         (_mm_srli_epi64): Likewise.
617         (_mm_sll_epi16): Replace deprecated function with vec_sl.
618         (_mm_sll_epi32): Likewise.
619         (_mm_sll_epi64): Likewise.
620         (_mm_sra_epi16): Replace deprecated function with vec_sra.
621         (_mm_sra_epi32): Likewise.
622         (_mm_srl_epi16): Replace deprecated function with vec_sr.
623         (_mm_srl_epi32): Likewise.
624         (_mm_srl_epi64): Likewise.
626 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
627             Jinsong Ji <jji@us.ibm.com>
629         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
630         comparison operators with vec_cmp* for compatibility due to
631         unfortunate history; clean up formatting and use types more
632         appropriately.
633         (_mm_sll_epi32): Likewise.
634         (_mm_sll_epi64): Likewise.
635         (_mm_srl_epi16): Likewise.
636         (_mm_srl_epi32): Likewise.
637         (_mm_srl_epi64): Likewise.
639 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
640             Jinsong Ji <jji@us.ibm.com>
642         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
643         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
644         __vector __bool int.  Use vec_cmpgt in preference to deprecated
645         function vec_vcmpgtfp.
646         (_mm_max_ps): Likewise.
648 2018-10-25  Jeff Law  <law@redhat.com>
650         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
651         if returning false.
653 2018-10-25  Martin Sebor  <msebor@redhat.com>
655         * doc/extend.texi (aligned): Expand attribute description.
656         (Alignment): Rename section.  Discuss function arguments.
658 2018-10-25  Jan Hubicka  <jh@suse.cz>
660         * ipa-devirt.c (main_odr_variant): Remove.
661         (hash_odr_name, types_same_for_odr, types_odr_comparable,
662         odr_name_hasher::equal, odr_subtypes_equivalent_p):
663         Drop use of main_odr_variant.
664         (add_type_duplicate): Silence confused warnings on integer types.
665         (get_odr_type): Always look for main variant.
666         (register_odr_type): Simplify.
668 2018-10-25  Richard Biener  <rguenther@suse.de>
670         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
671         Initialize ng to silence error with release checking bootstrap.
673 2018-10-25  Richard Biener  <rguenther@suse.de>
675         * tree-if-conv.c: Include tree-ssa-sccvn.h.
676         (tree_if_conversion): Run CSE on the if-converted loop body.
678 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
680         * config/s390/constraints.md (ZL): New constraint.
681         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
682         operands.
683         * config/s390/s390.md (movdi_larl): Remove.
684         (movdi_64): Add the LARL alternative.
686 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
688         PR bootstrap/87747
689         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
690         (RTX_CODE_HWINT_P): New macro.
691         (rtx_code_size): Use RTX_CODE_HWINT_P ().
693 2018-10-25  Jan Hubicka  <jh@suse.cz>
695         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
696         is anonymous.
698 2018-10-25  Richard Biener  <rguenther@suse.de>
700         PR tree-optimization/87665
701         PR tree-optimization/87745
702         * tree-vectorizer.h (get_earlier_stmt): Remove.
703         (get_later_stmt): Pick up UID from the original non-pattern stmt.
705 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
707         * options.texi (Deprecated): Move list to Var section.
709 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
710             Jinsong Ji <jji@us.ibm.com>
712         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
713         __vector long to __vector long long.
714         (_mm_cvtpd_ps): Likewise.
715         (_mm_cvttpd_epi32): Likewise.
716         (_mm_cvtpi32_pd): Likewise.
717         (_mm_unpackhi_epi64): Likewise.
718         (_mm_unpacklo_epi64): Likewise.
720 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
722         PR rtl-optimization/87720
723         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
725 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
727         * gimple-ssa-isolate-paths.c
728         (find_implicit_erroneous_behavior): Do not change code if the
729         pass is running for warnings only.
730         (find_explicit_erroneous_behavior): Likewise.
732 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
734         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
735         Define as rs6000_mangle_decl_assembler_name.
736         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
737         long double to IEEE long double, switch the names of the long
738         double built-in functions to be <func>f128 instead of <func>l.
740 2018-10-24  Martin Sebor  <msebor@redhat.com>
742         * doc/extend.texi (nonnull): List no-argument form.  Reference
743         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
745 2018-10-24  Richard Biener  <rguenther@suse.de>
747         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
749 2018-10-24  Martin Liska  <mliska@suse.cz>
751         PR tree-optimization/84436
752         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
753         Remove.
754         (switch_conversion::contains_linear_function_p): New.
755         (switch_conversion::build_one_array): Support linear
756         transformation on input.
757         * tree-switch-conversion.h (struct switch_conversion): Add
758         contains_linear_function_p declaration.
760 2018-10-24  Richard Biener  <rguenther@suse.de>
762         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
763         if its argument is CONSTANT_CLASS_P.
765 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
767         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
768         it is wrong for forward declarations.
770 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
772         * config/s390/s390.c (s390_check_qrst_address): Add the missing
773         SYMBOL_REF_P () check.
775 2018-10-24  Richard Biener  <rguenther@suse.de>
777         PR tree-optimization/87105
778         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
779         dump classification.
780         (vect_analyze_data_ref_accesses): Handle duplicate loads and
781         stores by splitting the affected group after the fact.
782         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
783         fail the SLP build because of size constraints.
785 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
787         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
788         * configure: Regenerate.
789         * config.in: Regenerate.
790         * varasm.c (mergeable_string_section): Use readonly_data_section
791         if linker doesn't support SHF_MERGE with alignment > 8.
792         (mergeable_constant_section): Likewise.
794 2018-10-24  Richard Biener  <rguenther@suse.de>
796         PR tree-optimization/84013
797         * tree-ssa-structalias.c (struct msdi_data): New struct for
798         marshalling data to walk_stmt_load_store_ops.
799         (maybe_set_dependence_info): Refactor as callback for
800         walk_stmt_load_store_ops.
801         (compute_dependence_clique): Set restrict info on all stmt kinds.
803 2018-10-24  Martin Liska  <mliska@suse.cz>
805         * cgraph.c (cgraph_node::dump):
806         Remove reduntant dumps and make tp_first_run dump more compact.
808 2018-10-24  Richard Biener  <rguenther@suse.de>
810         PR tree-optimization/87665
811         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
812         to reflect reality.
814 2018-10-12  Jeff Law  <law@redhat.com>
816         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
817         for H8/S.
819 2018-10-23  Richard Biener  <rguenther@suse.de>
821         * tree-vrp.c (add_assert_info): Guard dump_printf with
822         dump_enabled_p.
823         * gimple-ssa-evrp-analyze.c
824         (evrp_range_analyzer::record_ranges_from_incoming_edge):
825         Use value_range::ignore_equivs_equal_p.
827 2018-10-23  Richard Biener  <rguenther@suse.de>
829         PR tree-optimization/87105
830         PR tree-optimization/87608
831         * passes.def (pass_all_early_optimizations): Add early phi-opt
832         after dce.
833         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
834         addition to debug stmts.
835         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
836         and abs replacement early.
837         * tree-cfg.c (gimple_empty_block_p): Likewise.
839 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
841         PR target/86383
842         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
843         specified to configure.
844         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
846 2018-10-23  Richard Biener  <rguenther@suse.de>
848         PR tree-optimization/87700
849         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
851 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
853         PR target/87674
854         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
855         second argument from __mmask16 to __mmask8.
856         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
857         _mm_mask_packs_epi32): Likewise.
858         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
859         Likewise.
860         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
862 2018-10-23  Richard Biener  <rguenther@suse.de>
864         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
866 2018-10-23  Richard Biener  <rguenther@suse.de>
868         PR tree-optimization/86144
869         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
870         over simd attribute.
872 2018-10-23  Richard Biener  <rguenther@suse.de>
874         PR tree-optimization/87693
875         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
876         the case we do not find the taken edge.
878 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
879             Jinsong Ji  <jji@us.ibm.com>
881         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
882         (_mm_store_pd): Use unaligned vector type for pointer cast.
883         (_mm_maskmoveu_si128): Likewise.
884         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
885         (_mm_store_ps): Use unaligned vector type for pointer cast.
887 2018-10-22  Paul Koning  <ni1d@arrl.net>
889         * symtab.c (symtab_node::increase_alignment): Correct max
890         alignment check.
892 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
894         PR tree-optimization/87633
895         * match.pd: Do not generate unordered integer comparisons.
897 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
899         PR rtl-optimization/87600
900         * combine.c: Add include of expr.h.
901         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
902         register to a pseudo.
903         (make_more_copies): New function, add a copy to a new pseudo after
904         the moves from hard registers into pseudos.
905         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
906         later.  Call make_more_copies.
908 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
910         * lra-constraints.c (process_alt_operands): New local array,
911         matching_early_clobber.  Check matching_early_clobber before
912         decrementing reject, and set matching_early_clobber after.
914 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
916         PR target/87598
917         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
918         call output_addr_const and hope for the best.
920 2018-10-22  Richard Biener  <rguenther@suse.de>
922         * gimple-ssa-evrp-analyze.c
923         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
924         smarter about what ranges to use.
925         * tree-vrp.c (add_assert_info): Dump here.
926         (register_edge_assert_for_2): Instead of here at multiple but
927         not all places.
929         * gcc.dg/tree-ssa/evrp12.c: New testcase.
930         * gcc.dg/predict-6.c: Adjust.
931         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
932         * gcc.dg/tree-ssa/vrp02.c: Likewise.
933         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
935 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
936         Richard Biener  <rguenther@suse.de>
938         * bitmap.h: Update data structure documentation, including a
939         description of bitmap views as either linked-lists or splay trees.
940         (struct bitmap_element_def): Update comments for splay tree bitmaps.
941         (struct bitmap_head_def): Likewise.
942         (bitmap_list_view, bitmap_tree_view): New prototypes.
943         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
944         tree_form fields.
945         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
946         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
947         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
948         released bitmap element here.
949         (bitmap_element_free): Remove.
950         (bitmap_elt_clear_from): Work on splay tree bitmaps.
951         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
952         this function similar ones such that linked-list bitmap implementation
953         functions are grouped.
954         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
955         and moved for grouping.
956         (bitmap_list_insert_element_after): Renamed from
957         bitmap_elt_insert_after, and moved for grouping.
958         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
959         (bitmap_tree_link_left, bitmap_tree_link_right,
960         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
961         bitmap_tree_link_element, bitmap_tree_unlink_element,
962         bitmap_tree_find_element): New functions for splay-tree bitmap
963         implementation.
964         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
965         Renamed and moved, see above entries.
966         (bitmap_tree_listify_from): New function to convert part of a splay
967         tree bitmap to a linked-list bitmap.
968         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
969         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
970         (bitmap_find_bit): Remove.
971         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
972         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
973         Handle splay tree bitmaps.
974         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
975         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
976         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
977         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
978         bitmap_intersect_compl_p, bitmap_ior_and_compl,
979         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
980         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
981         corresponding changes to use linked-list specific bitmap_element
982         manipulation functions as applicable for efficiency.
983         (bitmap_tree_to_vec): New function.
984         (debug_bitmap_elt_file): New function split out from ...
985         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
986         (bitmap_print): Likewise.
988         PR tree-optimization/63155
989         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
990         SSA edge worklists.
991         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
992         in tree-view.
994 2018-10-22  Martin Liska  <mliska@suse.cz>
996         PR tree-optimization/87686
997         Revert
998         2018-08-29  Martin Liska  <mliska@suse.cz>
1000         * tree-switch-conversion.c (switch_conversion::expand):
1001         Strenghten assumption about gswitch statements.
1003 2018-10-22  Martin Liska  <mliska@suse.cz>
1005         * ipa-icf.c (sem_item::compare_attributes): Remove.
1006         (sem_item::compare_referenced_symbol_properties): Use
1007         attribute_list_equal instead.
1008         (sem_function::equals_wpa): Likewise.
1009         * ipa-icf.h: Remove compare_attributes.
1011 2018-10-22  Richard Biener  <rguenther@suse.de>
1013         PR middle-end/87682
1014         * mem-stats.h (mem_usage::operator==): Fix pasto.
1016 2018-10-22  Richard Biener  <rguenther@suse.de>
1018         PR tree-optimization/87640
1019         * tree-vrp.c (set_value_range_with_overflow): Decompose
1020         incomplete result.
1021         (extract_range_from_binary_expr_1): Adjust.
1023 2018-10-22  Martin Jambor  <mjambor@suse.cz>
1025         * tree-eh.h (stmt_could_throw_p): Add function parameter.
1026         (stmt_can_throw_external): Likewise.
1027         (stmt_can_throw_internal): Likewise.
1028         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
1029         (lower_eh_constructs_2): Likewise.
1030         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
1031         (stmt_can_throw_external): Likewise.
1032         (stmt_can_throw_internal): Likewise.
1033         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
1034         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
1035         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
1036         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
1037         (pass_lower_eh_dispatch::execute): Pass cfun to
1038         stmt_can_throw_external.
1039         (cleanup_empty_eh): Likewise.
1040         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
1041         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
1042         stmt_can_throw_external instead of pushing it to cfun.
1043         (symbol_table::create_edge): Likewise.
1044         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
1045         stmt_can_throw_internal.
1046         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
1047         to stmt_could_throw_p.
1048         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
1049         stmt_can_throw_internal.
1050         (pass_store_merging::execute): Likewise.
1051         * gimple-ssa-strength-reduction.c
1052         (find_candidates_dom_walker::before_dom_children): Pass cfun to
1053         stmt_could_throw_p.
1054         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
1055         stmt_can_throw_internal.
1056         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
1057         to stmt_can_throw_external.
1058         (check_stmt): Pass cfun to stmt_could_throw_p.
1059         (check_stmt): Pass cfun to stmt_can_throw_external.
1060         (pass_nothrow::execute): Likewise.
1061         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
1062         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
1063         stmt_can_throw_internal.
1064         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
1065         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
1066         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
1067         * tree-complex.c (expand_complex_libcall): Pass cfun to
1068         stmt_could_throw_p and to stmt_can_throw_internal.
1069         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
1070         * tree-inline.c (copy_edges_for_bb): Likewise.
1071         (maybe_move_debug_stmts_to_successors): Likewise.
1072         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
1073         stmt_could_throw_p.
1074         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
1075         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
1076         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
1077         stmt_can_throw_internal.
1078         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1079         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
1080         stmt_could_throw_p.
1081         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
1082         stmt_can_throw_internal.
1083         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
1084         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
1085         stmt_could_throw_p.
1086         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
1087         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
1088         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1089         (convert_mult_to_fma_1): Likewise.
1090         (convert_to_divmod): Likewise.
1091         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
1092         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
1093         * tree-ssa-propagate.c
1094         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
1095         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
1096         (maybe_optimize_range_tests): Likewise.
1097         (linearize_expr_tree): Likewise.
1098         (reassociate_bb): Likewise.
1099         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1100         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
1101         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
1102         (handle_char_store): Likewise.
1103         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
1104         stmt_can_throw_internal.
1105         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
1106         stmt_could_throw_p.
1107         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
1108         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
1109         (vectorizable_simd_clone_call): Likewise.
1110         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
1111         (gimple_stringop_fixed_value): Likewise.
1113 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
1115         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
1116         literal pool references.
1117         (s390_check_qrst_address): Adapt to the new behavior of
1118         s390_loadrelative_operand_p ().
1120 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1122         PR target/72782
1123         * config/i386/sse.md (*andnot<mode>3_bcst): New.
1125 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1127         PR target/72782
1128         * config/i386/sse.md (*<code><mode>3_bcst): New.
1130 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
1132         PR target/72782
1133         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
1134         V4DI, V16SI and V8DI.
1135         (*sub<mode>3<mask_name>_bcst): New.
1136         (*add<mode>3<mask_name>_bcst): Likewise.
1138 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
1139             Jinsong Ji  <jji@us.ibm.com>
1141         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
1142         __m64 with __vector unsigned long long for compatibility.
1143         (_mm_movemask_epi8): Likewise.
1144         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
1145         (_mm_cvttps_pi32): Likewise.
1146         (_mm_cvtpi32_ps): Likewise.
1147         (_mm_cvtps_pi16): Likewise.
1148         (_mm_loadh_pi): Likewise.
1149         (_mm_storeh_pi): Likewise.
1150         (_mm_movehl_ps): Likewise.
1151         (_mm_movelh_ps): Likewise.
1152         (_mm_loadl_pi): Likewise.
1153         (_mm_storel_pi): Likewise.
1154         (_mm_movemask_ps): Likewise.
1155         (_mm_shuffle_pi16): Likewise.
1157 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1159         PR target/72782
1160         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
1161         __builtin_ia32_vfnmsubpd512_mask.
1162         (_mm512_mask_fnmsub_round_pd): Likewise.
1163         (_mm512_fnmsub_pd): Likewise.
1164         (_mm512_mask_fnmsub_pd): Likewise.
1165         (_mm512_maskz_fnmsub_round_pd): Use
1166         __builtin_ia32_vfnmsubpd512_maskz.
1167         (_mm512_maskz_fnmsub_pd): Likewise.
1168         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
1169         (_mm512_mask_fnmsub_round_ps): Likewise.
1170         (_mm512_fnmsub_ps): Likewise.
1171         (_mm512_mask_fnmsub_ps): Likewise.
1172         (_mm512_maskz_fnmsub_round_ps): Use
1173         __builtin_ia32_vfnmsubps512_maskz.
1174         (_mm512_maskz_fnmsub_ps): Likewise.
1175         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
1176         __builtin_ia32_vfnmsubpd256_mask.
1177         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
1178         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1179         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
1180         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1181         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
1182         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
1183         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
1184         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
1185         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
1186         __builtin_ia32_vfnmsubpd.
1187         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
1188         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
1189         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
1190         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
1191         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
1192         * config/i386/i386-builtin.def: Add
1193         __builtin_ia32_vfnmsubpd256_mask,
1194         __builtin_ia32_vfnmsubpd256_maskz,
1195         __builtin_ia32_vfnmsubpd128_mask,
1196         __builtin_ia32_vfnmsubpd128_maskz,
1197         __builtin_ia32_vfnmsubps256_mask,
1198         __builtin_ia32_vfnmsubps256_maskz,
1199         __builtin_ia32_vfnmsubps128_mask,
1200         __builtin_ia32_vfnmsubps128_maskz,
1201         __builtin_ia32_vfnmsubpd512_mask,
1202         __builtin_ia32_vfnmsubpd512_maskz,
1203         __builtin_ia32_vfnmsubps512_mask,
1204         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
1205         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
1206         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
1207         __builtin_ia32_vfnmsubpd256.
1208         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
1209         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
1210         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
1211         Likewise.
1212         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
1213         Likewise.
1214         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
1215         Likewise.
1216         (fmai_vmfnmsub_<mode><round_name>): Likewise.
1218 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1220         PR target/72782
1221         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
1222         __builtin_ia32_vfnmaddpd512_mask.
1223         (_mm512_mask_fnmadd_round_pd): Likewise.
1224         (_mm512_fnmadd_pd): Likewise.
1225         (_mm512_mask_fnmadd_pd): Likewise.
1226         (_mm512_maskz_fnmadd_round_pd): Use
1227         __builtin_ia32_vfnmaddpd512_maskz.
1228         (_mm512_maskz_fnmadd_pd): Likewise.
1229         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
1230         (_mm512_mask_fnmadd_round_ps): Likewise.
1231         (_mm512_fnmadd_ps): Likewise.
1232         (_mm512_mask_fnmadd_ps): Likewise.
1233         (_mm512_maskz_fnmadd_round_ps): Use
1234         __builtin_ia32_vfnmaddps512_maskz.
1235         (_mm512_maskz_fnmadd_ps): Likewise.
1236         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
1237         __builtin_ia32_vfnmaddpd256_mask.
1238         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
1239         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
1240         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
1241         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1242         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
1243         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
1244         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
1245         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
1246         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
1247         __builtin_ia32_vfnmaddpd.
1248         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
1249         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
1250         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
1251         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
1252         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
1253         * config/i386/i386-builtin.def: Add
1254         __builtin_ia32_vfnmaddpd256_mask,
1255         __builtin_ia32_vfnmaddpd256_maskz,
1256         __builtin_ia32_vfnmaddpd128_mask,
1257         __builtin_ia32_vfnmaddpd128_maskz,
1258         __builtin_ia32_vfnmaddps256_mask,
1259         __builtin_ia32_vfnmaddps256_maskz,
1260         __builtin_ia32_vfnmaddps128_mask,
1261         __builtin_ia32_vfnmaddps128_maskz,
1262         __builtin_ia32_vfnmaddpd512_mask,
1263         __builtin_ia32_vfnmaddpd512_maskz,
1264         __builtin_ia32_vfnmaddps512_mask,
1265         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1266         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1267         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1268         __builtin_ia32_vfnmaddpd256.
1269         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1270         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1271         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1272         Likewise.
1273         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1274         Likewise.
1275         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1276         Likewise.
1277         (fmai_vmfnmadd_<mode><round_name>): Likewise.
1279 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1281         PR target/72782
1282         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1283         __builtin_ia32_vfmsubpd512_mask.
1284         (_mm512_mask_fmsub_round_pd): Likewise.
1285         (_mm512_fmsub_pd): Likewise.
1286         (_mm512_mask_fmsub_pd): Likewise.
1287         (_mm512_maskz_fmsub_round_pd): Use
1288         __builtin_ia32_vfmsubpd512_maskz.
1289         (_mm512_maskz_fmsub_pd): Likewise.
1290         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1291         (_mm512_mask_fmsub_round_ps): Likewise.
1292         (_mm512_fmsub_ps): Likewise.
1293         (_mm512_mask_fmsub_ps): Likewise.
1294         (_mm512_maskz_fmsub_round_ps): Use
1295         __builtin_ia32_vfmsubps512_maskz.
1296         (_mm512_maskz_fmsub_ps): Likewise.
1297         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1298         __builtin_ia32_vfmsubpd256_mask.
1299         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1300         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1301         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1302         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1303         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1304         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1305         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1306         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1307         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1308         __builtin_ia32_vfmsubpd.
1309         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1310         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1311         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1312         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1313         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1314         * config/i386/i386-builtin.def: Add
1315         __builtin_ia32_vfmsubpd256_mask,
1316         __builtin_ia32_vfmsubpd256_maskz,
1317         __builtin_ia32_vfmsubpd128_mask,
1318         __builtin_ia32_vfmsubpd128_maskz,
1319         __builtin_ia32_vfmsubps256_mask,
1320         __builtin_ia32_vfmsubps256_maskz,
1321         __builtin_ia32_vfmsubps128_mask,
1322         __builtin_ia32_vfmsubps128_maskz,
1323         __builtin_ia32_vfmsubpd512_mask,
1324         __builtin_ia32_vfmsubpd512_maskz,
1325         __builtin_ia32_vfmsubps512_mask,
1326         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1327         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1328         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1329         __builtin_ia32_vfmsubpd256.
1330         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1331         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1332         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1333         Likewise.
1334         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1335         Likewise.
1336         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1337         Likewise.
1338         (fmai_vmfmsub_<mode><round_name>): Likewise.
1340 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1342         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1343         Remove plus.  Renamed to ...
1344         (*sub<mode>3<mask_name>_bcst): This.
1345         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1346         (*add<mode>3<mask_name>_bcst): This.
1348 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1350         PR target/72782
1351         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1353 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1355         PR target/87662
1356         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1357         (_mm_or_epi32): Likewise.
1358         (_mm256_xor_epi32): Likewise.
1359         (_mm_xor_epi32): Likewise.
1360         (_mm256_or_epi64): Likewise.
1361         (_mm_or_epi64): Likewise.
1362         (_mm256_xor_epi64): Likewise.
1363         (_mm_xor_epi64): Likewise.
1365 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
1367         PR target/72782
1368         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1370 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
1372         PR middle-end/87647
1373         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1375 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
1377         * doc/ux.texi: Move @section directly after @node.
1379 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
1381         PR middle-end/85488
1382         PR middle-end/87649
1383         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1384         depend closely nested inside of loop with ordered clause with
1385         a parameter.
1387 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
1389         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1390         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1391         * doc/ux.texi: New file.
1393 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
1395         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1396         be the first CR field allocated.
1398 2018-10-19  Richard Biener  <rguenther@suse.de>
1400         PR target/87657
1401         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1402         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1404 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1406         PR target/72782
1407         * config/i386/sse.md
1408         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1409         (*add<mode>3<mask_name>_bcst_2): Likewise.
1411 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1413         * config/i386/sse.md
1414         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1415         Replace nonimmediate_operand with register_operand.
1416         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1417         Likewise.
1418         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1419         Likewise.
1421 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
1423         PR rtl-optimization/87596
1424         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1425         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1426         for instructions in FROM..TO range.
1428 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
1430         * cfgexpand.c (expand_one_var): Use specific wording in error message
1431         for non-local frame variables.
1432         * stor-layout.c (layout_decl): Do not issue a warning for them.
1434 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
1436         * haifa-sched.c (priority): Add force_recompute parameter.
1437         (apply_replacement): Call priority () with force_recompute = true.
1438         (restore_pattern): Likewise.
1440 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1442         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
1443         HOST_BITS_PER_WIDE_INT.
1444         (test_vector_ops_duplicate): Likewise.
1446 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1448         PR target/72782
1449         * config/i386/sse.md (VF_AVX512): New.
1450         (avx512bcst): Likewise.
1451         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1452         Likewise.
1453         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1454         Likewise.
1455         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1456         Likewise.
1458 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
1460         * doc/invoke.texi (-dumpversion): Improve grammar.
1461         (-dumpfullversion): Make more consistent with -dumpversion.
1463 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
1465         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
1466         Set JUMP_LABEL to the jump insn.
1467         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
1468         Predict emitted jump and add label to jump insn.
1470 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
1472         PR tree-optimization/87562
1473         * input.c (get_substring_ranges_for_loc): Use
1474         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
1475         getting the linemap for the endpoint.  Verify that it's either
1476         in the same linemap as the start point's spelling location, or
1477         at least in the same file.
1479 2018-10-18  Richard Biener  <rguenther@suse.de>
1481         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
1482         feed width-specific load/store costs through ix86_vec_cost.
1483         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
1484         (k8_cost): Likewise.
1485         (bdver_cost): Likewise.
1486         (znver1_cost): Likewise.
1487         (btver1_cost): Likewise.
1488         (btver2_cost): Likewise.
1490 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1492         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
1493         to simplify subreg of vec_merge.
1495 2018-10-18  Richard Biener  <rguenther@suse.de>
1497         * config/i386/i386.c: Fix costing of vector FMA.
1499 2018-10-18  Richard Biener  <rguenther@suse.de>
1501         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
1502         and argument.
1503         (ix86_builtin_vectorization_cost): For vec_construct properly
1504         cost insertion into SSE regs.
1505         (...): Adjust calls to ix86_vec_cost.
1507 2018-10-18  Richard Biener  <rguenther@suse.de>
1509         PR middle-end/87087
1510         Revert
1511         2018-02-07  Richard Biener  <rguenther@suse.de>
1513         PR tree-optimization/84204
1514         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1515         this place.
1517 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1519         PR target/87537
1520         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
1521         of vec_duplicate.
1522         (test_vector_ops_duplicate): Add test for a scalar subreg of a
1523         VEC_MERGE of a VEC_DUPLICATE.
1525 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1527         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
1528         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
1529         * doc/standards.texi (C Language): Document C2X.
1530         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
1531         (rl78_option_override): Handle "GNU C2X" language name.
1533 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1535         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
1536         Document C17 as published in 2018.
1538 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
1540         PR middle-end/87623
1541         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
1542         bail out if both sides do not have the same storage order.
1544 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1546         * bitmap.c (bitmap_head::dump): New.
1547         * bitmap.h (bitmap_head): Add dump().
1548         * gimple-ssa-evrp-analyze.c
1549         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
1550         (evrp_range_analyzer::set_ssa_range_info): Same.
1551         (evrp_range_analyzer::record_ranges_from_phis): Same.
1552         (evrp_range_analyzer::record_ranges_from_stmt): Same.
1553         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
1554         * gimple-ssa-sprintf.c (get_int_range): Same.
1555         (format_integer): Same.
1556         (sprintf_dom_walker::handle_gimple_call): Same.
1557         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
1558         (ipcp_vr_lattice::top_p): Same.
1559         (ipcp_vr_lattice::bottom_p): Same.
1560         (ipcp_vr_lattice::set_to_bottom): Same.
1561         (ipa_vr_operation_and_type_effects): Same.
1562         (propagate_vr_across_jump_function): Same.
1563         (ipcp_store_vr_results): Same.
1564         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
1565         (ipa_print_node_jump_functions_for_edge): Same.
1566         (ipa_get_value_range): Same.
1567         (ipa_compute_jump_functions_for_edge): Same.
1568         (ipa_write_jump_function): Same.
1569         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
1570         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1571         Same.
1572         * vr-values.c (set_value_range_to_nonnegative): Same.
1573         (set_value_range_to_truthvalue): Same.
1574         (vr_values::get_value_range): Same.
1575         (vr_values::set_defs_to_varying): Same.
1576         (vr_values::update_value_range): Same.
1577         (symbolic_range_based_on_p): Same.
1578         (vr_values::op_with_boolean_value_range_p): Same.
1579         (vr_values::extract_range_for_var_from_comparison_expr): Same.
1580         (vr_values::extract_range_from_ssa_name): Same.
1581         (vr_values::extract_range_from_binary_expr): Same.
1582         (vr_values::extract_range_from_unary_expr): Same.
1583         (vr_values::extract_range_from_cond_expr): Same.
1584         (vr_values::extract_range_from_comparison): Same.
1585         (vr_values::check_for_binary_op_overflow): Same.
1586         (vr_values::extract_range_basic): Same.
1587         (vr_values::extract_range_from_assignment): Same.
1588         (compare_ranges): Same.
1589         (compare_range_with_value): Same.
1590         (vr_values::adjust_range_with_scev): Same.
1591         (vrp_valueize): Same.
1592         (vrp_valueize_1): Same.
1593         (vr_values::get_vr_for_comparison): Same.
1594         (vr_values::compare_name_with_value): Same.
1595         (vr_values::compare_names): Same.
1596         (vr_values::vrp_evaluate_conditional): Same.
1597         (find_case_label_ranges): Same.
1598         (vr_values::vrp_visit_switch_stmt): Same.
1599         (vr_values::extract_range_from_phi_node): Same.
1600         (vr_values::simplify_div_or_mod_using_ranges): Same.
1601         (vr_values::simplify_bit_ops_using_ranges): Same.
1602         (test_for_singularity): Same.
1603         (range_fits_type_p): Same.
1604         (vr_values::simplify_cond_using_ranges_1): Same.
1605         (vr_values::simplify_switch_using_ranges): Same.
1606         (vr_values::simplify_float_conversion_using_ranges): Same.
1607         (vr_values::two_valued_val_range_p): Same.
1608         (vr_values::add_equivalence): Move to value_range::equiv_add.
1609         * vr-values.h (vr_values::add_equivalence): Remove.
1610         (VR_INITIALIZER): Remove.
1611         * tree-vrp.c (value_range::set): New.
1612         (value_range::equiv_add): New.
1613         (value_range::value_range): New.
1614         (value_range::deep_copy): New.
1615         (value_range::check): New.
1616         (value_range::equal_p): New.
1617         (value_range::ignore_equivs_equal_p): New.
1618         (value_range::operator==): New.
1619         (value_range::operator!=): New.
1620         (value_range::symbolic_p): New.
1621         (value_range::numeric_p): New.
1622         (value_range::set_undefined): New.
1623         (value_range::set_varying): New.
1624         (value_range::may_contain_p): New.
1625         (value_range::equiv_clear): New.
1626         (value_range::singleton_p): New.
1627         (value_range::intersect): New.
1628         (value_range::dump): New.
1629         (value_range::set_and_canonicalize): New.
1630         (set_value_range): Adjust for value_range API.
1631         (set_value_range_to_undefined): Same.
1632         (set_value_range_to_varying): Same.
1633         (set_and_canonicalize_value_range): Same.
1634         (set_value_range_to_nonnull): Same.
1635         (set_value_range_to_null): Same.
1636         (range_is_null): Same.
1637         (range_is_nonnull): Same.
1638         (range_int_cst_p): Same.
1639         (range_int_cst_singleton_p): Same.
1640         (symbolic_range_p): Same.
1641         (range_includes_zero_p): Same.
1642         (value_range_constant_singleton): Same.
1643         (vrp_set_zero_nonzero_bits): Same.
1644         (ranges_from_anti_range): Same.
1645         (extract_range_into_wide_ints): Same.
1646         (extract_range_from_multiplicative_op): Same.
1647         (set_value_range_with_overflow): Same.
1648         (extract_range_from_binary_expr_1): Same.
1649         (extract_range_from_unary_expr): Same.
1650         (dump_value_range): Same.
1651         (debug_value_range): Same.
1652         (vrp_prop::check_array_ref): Same.
1653         (vrp_prop::check_mem_ref): Same.
1654         (vrp_prop::vrp_initialize): Same.
1655         (vrp_prop::visit_stmt): Same.
1656         (intersect_ranges): Same.
1657         (vrp_prop::visit_phi): Same.
1658         (vrp_prop::vrp_finalize): Same.
1659         (determine_value_range_1): Same.
1660         (determine_value_range): Same.
1661         (vrp_intersect_ranges_1): Rename to...
1662         (vrp_intersect_1): this.
1663         (vrp_intersect_ranges): Rename to...
1664         (value_range::intersect_helper): ...this.
1665         (vrp_meet_1): Rename to...
1666         (value_range::union_helper): ...this.
1667         (vrp_meet): Rename to...
1668         (value_range::union_): ...this.
1669         (copy_value_range): Remove.
1670         * tree-vrp.h (struct value_range): Rewrite into a proper class.
1671         (value_range::vrtype): New.
1672         (value_range::type): New.
1673         (value_range::equiv): New.
1674         (value_range::min): New.
1675         (value_range::max): New.
1676         (value_range::varying_p): New.
1677         (value_range::undefined_p): New.
1678         (value_range::null_p): New.
1679         (value_range::equiv_add): New.
1680         (copy_value_range): Remove.
1682 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
1684         * Makefile.in (SELFTEST_TARGETS): New.
1685         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
1686         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
1687         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
1688         c/Make-lang.in.
1689         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
1690         (selftest-c++-gdb, selftest-c++-valgrind): Move to
1691         cp/Make-lang.in.
1692         * configure: Regenerate.
1693         * configure.ac (selftest_languages): New.
1695 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1697         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
1698         overflow wraps argument.
1699         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
1700         wide_int_range_multiplicative_op.
1701         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
1702         overflow wraps argument.
1703         (wide_int_range_multiplicative_op): Same.
1704         (wide_int_range_lshift): Same.
1705         (wide_int_range_div): Same.
1706         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
1707         (wide_int_range_lshift): Same.
1708         (wide_int_range_div): Same.
1710 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1712         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
1713         use sign as argument.
1714         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
1715         wide_int_range_shift_undefined_p.
1717 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1719         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
1720         Rename to...
1721         (@despeculate_copy<ALLI_TI:mode>): ... This.
1722         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
1723         switch statement.
1725 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1727         * config.gcc: Obsolete *-*-solaris2.10*.
1728         * doc/install.texi (Specific, *-*-solaris2*): Document it.
1730 2018-10-12  Jeff Law  <law@redhat.com>
1732         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
1733         reg + sym +- const_int addressing modes.
1735 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1737         * common.opt (fdiagnostics-minimum-margin-width=): New option.
1738         * diagnostic-show-locus.c (layout::layout): Apply the minimum
1739         margin width.
1740         (layout::start_annotation_line): Only print up to 3 of the
1741         margin character, to avoid touching the left-hand side.
1742         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
1743         minimum margin width, as set by test_diagnostic_context's ctor.
1744         (selftest::test_fixit_insert_containing_newline): Likewise.
1745         (selftest::test_fixit_insert_containing_newline_2): Likewise.
1746         (selftest::test_line_numbers_multiline_range): Clear
1747         dc.min_margin_width.
1748         * diagnostic.c (diagnostic_initialize): Initialize
1749         min_margin_width.
1750         * diagnostic.h (struct diagnostic_context): Add field
1751         "min_margin_width".
1752         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
1753         * opts.c (common_handle_option): Handle
1754         OPT_fdiagnostics_minimum_margin_width_.
1755         * selftest-diagnostic.c
1756         (selftest::test_diagnostic_context::test_diagnostic_context):
1757         Initialize min_margin_width to 6.
1758         * toplev.c (general_init): Initialize global_dc->min_margin_width.
1760 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1762         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
1763         Fix usage of "error_at_rich_loc" in the comment.
1765 2018-10-15  Renlin Li  <renlin.li@arm.com>
1767         PR target/87563
1768         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
1769         if-conversioned loop when it contains ifn with types not
1770         supported by backend.
1771         * internal-fn.c (expand_direct_optab_fn): Add an assert.
1772         (direct_internal_fn_supported_p): New helper function.
1773         * internal-fn.h (direct_internal_fn_supported_p): Declare.
1775 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
1777         PR target/87572
1778         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1779         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1780         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1782 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
1784         PR tree-optimization/87022
1785         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
1786         bits in dist vector rather than the first one.
1788 2018-10-15  Richard Biener  <rguenther@suse.de>
1790         PR middle-end/87610
1791         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
1792         (visit_loadstore): When a used restrict tag escaped verify that
1793         the points-to solution of "other" pointers do not include
1794         escaped.
1795         (compute_dependence_clique): If a used restrict tag escaped
1796         communicated that down to visit_loadstore.
1798 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
1800         * config/s390/s390.c (s390_expand_vec_init): Force vector element
1801         into reg if it isn't a general operand.
1803 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1805         PR target/87599
1806         * config/i386/sse.md (*vec_dupv2di): Add register source to
1807         movddup.
1809 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1811         PR target/87572
1812         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1813         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
1814         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1815         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1817 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
1819         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
1820         (notice_args_size): Set it in the current trace if no insn that can
1821         throw internally has been seen yet.
1822         (connect_traces): When connecting args_size between traces, allow the
1823         incoming values not to match if there is an insn setting it before the
1824         first insn that can throw internally; in that case, force the creation
1825         of a CFI note on this latter insn.
1827 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
1829         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
1830         the base class, because using the injected-class-name was not clearly
1831         specified until DR 176.
1833 2018-10-12  Paul Koning  <ni1d@arrl.net>
1835         * config/pdp11/pdp11.md (doloop_end): New expander.
1836         (doloop_end_insn): renamed from "doloop_end".
1837         (addqi3): New pattern.
1838         (subqi3): New pattern.
1839         * config/pdp11/predicates.md (incdec_operand): New predicate.
1841 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
1843         PR middle-end/81376
1844         * real.c (format_helper::can_represent_integral_type_p): New function
1845         * real.h (format_helper::can_represent_integral_type_p): Ditto.
1846         * match.pd: New pattern.
1848 2018-10-12  Alexandre Oliva <oliva@adacore.com>
1850         * configure.ac: Introduce --enable-large-address-aware
1851         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
1852         * doc/install.texi: Document it.
1853         * configure, config.in: Rebuilt.
1854         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
1855         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
1856         (LINK_SPEC): Insert it.
1857         * config/i386/mingw-264.h: Likewise.
1859         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
1861 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
1863         PR rtl-optimization/87600
1864         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
1866 2018-10-12  Paul Koning  <ni1d@arrl.net>
1868         * doc/md.texi (doloop_end): Document that the pattern code may
1869         need to check operand mode.
1871 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
1873         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
1874         to zero-extend between int and floating-point registers.
1875         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
1876         ldp into floating-point registers.  Add type and arch attributes.
1877         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
1878         Use f_loads for type attribute.
1880 2018-10-11  Martin Sebor  <msebor@redhat.com>
1882         * doc/extend.texi (attribute packed): Correct typos.
1884 2018-10-11  Martin Sebor  <msebor@redhat.com>
1886         * doc/extend.texi (attribute flatten): Mention interaction with
1887         noinline.
1889 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1891         PR target/87156
1892         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
1893         Set new_decl virtual flag to zero.
1895 2018-10-11  Martin Sebor  <msebor@redhat.com>
1897         PR middle-end/87593
1898         * doc/extend.texi (attribute format_arg): Discuss using multiple
1899         attributes on a single function.
1901 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
1903         PR tree-optimization/86829
1904         * match.pd (sin (atan (x))): New simplification rules.
1905         (cos (atan (x))): Likewise.
1906         * real.c (build_sinatan_real): New function.
1907         * real.h (build_sinatan_real): Prototype.
1909 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
1911         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
1912         function.
1913         (fold_mergeeo_helper): New helper function.
1914         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
1915         intrinsics.  Correct some whitespace indentation issues.
1917 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
1919         PR target/87511
1920         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
1921         Use HOST_WIDE_INT_1U for shift.
1923 2018-10-11  Doug Rupp  <rupp@adacore.com>
1924             Olivier Hainque  <hainque@adacore.com>
1926         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
1927         Pass --relax to the linker for RTPs.
1928         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
1930 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
1931             Jan Hubicka  <jh@suse.cz>
1932             Martin Jambor  <mjambor@suse.cz>
1934         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
1935         the same elements are repeated rather than printing all of them.
1936         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
1937         "repeated" elements.
1938         * read-rtl-function.c (test_loading_repeat): New function.
1939         (read_rtl_function_c_tests): Call test_loading_repeat.
1940         * rtl-tests.c (test_dumping_repeat): New function.
1941         (rtl_tests_c_tests): Call test_dumping_repeat.
1943 2018-10-11  Richard Biener  <rguenther@suse.de>
1945         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
1946         bdver?_cost): Unify to ...
1947         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
1948         * config/i386/i386.c (processor_cost_table): Adjust.
1950 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
1952         PR middle-end/87574
1953         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
1954         the thunk when expanding to GIMPLE.
1956 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1958         * varasm.c (mergeable_string_section): Don't try to move zero-length
1959         strings to the merge section.
1961 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
1963         PR target/87573
1964         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
1966 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
1968         PR target/87550
1969         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
1970         to special_args set.
1972 2018-10-10  Richard Biener  <rguenther@suse.de>
1974         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
1975         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
1976         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
1977         and recursing and pattern terminating the recursion on SSE
1978         vector width using ix86_expand_reduc.
1979         (reduc_sminmax_scal_<mode>): Split into part reducing to half
1980         width and recursing and SSE2 vector variant doing the final
1981         reduction with ix86_expand_reduc.
1982         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
1983         with terminating the recursion at AVX level, splitting that
1984         to SSE there.
1986 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
1988         * genmatch.c (error_cb): Rename to...
1989         (diagnostic_cb): ...this, converting int params to enums.
1990         (fatal_at): Update for renaming.
1991         (warning_at): Likewise.
1992         (main): Likewise.
1993         * input.c (selftest::ebcdic_execution_charset::apply):
1994         Update for renaming of...
1995         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
1996         to...
1997         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
1998         converting level and reason to enums.
1999         (class selftest::lexer_error_sink): Rename to...
2000         (class selftest::lexer_test_options): ...this, renaming field
2001         "m_errors" to "m_diagnostics".
2002         (selftest::lexer_test_options::apply): Update for renaming of...
2003         (selftest::lexer_test_options::on_error): ...this, renaming to...
2004         (selftest::lexer_test_options::on_diagnostic): ...this
2005         converting level and reason to enums.
2006         (selftest::test_lexer_string_locations_raw_string_unterminated):
2007         Update for renamings.
2008         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
2009         "reason".
2011 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
2013         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
2014         * config/rs6000/pmmintrin.h: New file.
2016 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
2018         PR tree-optimization/86659
2019         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
2021 2018-10-09  Richard Biener  <rguenther@suse.de>
2023         PR tree-optimization/63155
2024         * tree-ssa-structalias.c: Include tree-ssa.h.
2025         (get_constraint_for_ssa_var): For undefs return nothing_id.
2026         (find_func_aliases): Cleanup PHI handling.
2028 2018-10-09  Richard Biener  <rguenther@suse.de>
2030         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
2031         replacements.
2033 2018-10-09  Martin Liska  <mliska@suse.cz>
2035         * asan.c (asan_emit_stack_protection): If a stack variable
2036         is located in a same file as current function, then emit
2037         line info into variable definition string.
2039 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2041         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
2042         information.
2044 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2046         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
2047         on the thunk.
2049 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
2051         PR tree-optimization/86659
2052         * gimple-match.h (struct gimple_match_op): Add reverse field.
2053         (gimple_match_op::set_op): New overloaded method.
2054         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
2055         the REF_REVERSE_STORAGE_ORDER flag on the value.
2056         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
2057         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
2058         
2059 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
2061         PR middle-end/63155
2062         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2063         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2065 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
2067         PR target/87517
2068         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2069         Defined with __builtin_ia32_vfmaddsubpd512_mask.
2071 2018-10-08  Richard Biener  <rguenther@suse.de>
2073         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
2074         cost the same as AVX128 ones.
2076 2018-10-08  Paul Koning  <ni1d@arrl.net>
2078         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
2079         (expand_block_move): New function.
2080         * config/pdp11/pdp11.c (output_block_move): Remove.
2081         (expand_block_move): New function.
2082         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
2083         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
2084         (*movmemhi1): Remove.
2086 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
2088         * config/s390/2827.md: Increase latencies for some FP instructions.
2090 2018-10-08  Richard Biener  <rguenther@suse.de>
2092         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2093         Open a dump scope.
2094         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
2095         * tree-vectorizer.h (dump_stmt_cost): Adjust.
2096         (add_stmt_cost): Dump return value of the hook.
2098 2018-10-08  Richard Biener  <rguenther@suse.de>
2100         PR tree-optimization/63155
2101         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2102         (ssa_propagation_engine::ssa_propagate): Remove redundant
2103         bitmap bit clearing.
2105 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
2107         PR rtl-optimization/86939
2108         PR rtl-optimization/87479
2109         * ira.h (non_conflicting_reg_copy_p): New prototype.
2110         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
2111         (make_hard_regno_dead): Don't add conflicts for register
2112         ignore_reg_for_conflicts.
2113         (make_object_dead): Likewise.
2114         (non_conflicting_reg_copy_p): New function.
2115         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
2116         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
2117         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
2118         (make_hard_regno_dead): Don't add conflicts for register
2119         ignore_reg_for_conflicts.  Remove special conflict handling of
2120         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
2121         check_pic_pseudo_p and update callers.
2122         (mark_pseudo_dead): Don't add conflicts for register
2123         ignore_reg_for_conflicts.
2124         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
2126 2018-10-05  Andrew Waterman  <andrew@sifive.com>
2127             Jim Wilson  <jimw@sifive.com>
2129         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
2130         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
2131         new pattern using HONOR_SNANS that emits one extra instruction.
2133 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2135         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
2136         patterns): Merge SI and DI patterns to a GPR pattern.
2137         (unnamed define_insn and define_split for record form of that): Merge
2138         to a single define_insn_and_split pattern.
2140 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
2142         PR c++/56856
2143         * input.c (expand_location_to_spelling_point): Add param "aspect"
2144         and use rather than hardcoding LOCATION_ASPECT_CARET.
2145         (get_substring_ranges_for_loc): Handle the case of a single token
2146         within a macro expansion.
2147         * input.h (expand_location_to_spelling_point): Add "aspect" param,
2148         defaulting to LOCATION_ASPECT_CARET.
2150 2018-10-05  Paul Koning  <ni1d@arrl.net>
2152         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
2153         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
2154         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2155         (pdp11_guard_type): New function.
2157 2018-10-05  Paul Koning  <ni1d@arrl.net>
2159         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
2160         * config/pdp11/pdp11.opt (mfloat32): Remove.
2161         (mfloat64): Remove.
2162         * doc/invoke.texi (pdp11 -mfloat32): Remove:
2163         (pdp11 -mfloat64): Remove.
2165 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2167         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
2168         (*cmp<mode>_cc_i387): Ditto.
2169         (*cmpu<mode>_cc_i387): Ditto.
2170         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2171         * config/i386/i386.c (ix86_expand_fp_compare): Remove
2172         "scratch" argument.
2173         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
2174         Emit x86_sahf_1 pattern.
2175         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
2176         (ix86_expand_carry_flag_compare): Ditto.
2178 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2180         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
2181         to reg_or_0_operand.  Add "C" constraint.
2182         (*cmpxf_cc_i387): Ditto.
2183         (*cmp<mode>_i387): Change operand 2 predicate
2184         to nonimm_or_0_operand.  Add "C" constraint.
2185         (*cmp<mode>_cc_i387): Ditto.
2186         (*cmp<mode>_0_i387): Remove insn pattern.
2187         (*cmp<mode>_0_cc_i387): Ditto.
2189 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
2191         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
2192         * config/i386/predicates.md (nonimm_or_0_operand): Rename
2193         from vector_move_operand.  Update all uses.
2195 2018-10-05  Martin Sebor  <msebor@redhat.com>
2197         PR tree-optimization/87490
2198         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
2199         consistently.
2201 2018-10-05  Richard Biener  <rguenther@suse.de>
2203         PR tree-optimization/63155
2204         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2205         vertical space in dumpfiles.
2206         * tree-ssa-propagate.h
2207         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2208         * tree-ssa-propagate.c (cfg_blocks_back): New global.
2209         (ssa_edge_worklist_back): Likewise.
2210         (curr_order): Likewise.
2211         (cfg_blocks_get): Remove abstraction.
2212         (cfg_blocks_add): Likewise.
2213         (cfg_blocks_empty_p): Likewise.
2214         (add_ssa_edge): Add to current or next worklist based on
2215         RPO index.
2216         (add_control_edge): Likewise.
2217         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2218         into ...
2219         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
2220         iteration from CFG and SSA edge worklist so we process
2221         everything in RPO order, prioritizing forward progress
2222         over iteration.
2223         (ssa_prop_init): Allocate new worklists, do not dump
2224         immediate uses.
2225         (ssa_prop_fini): Free new worklists.
2227 2018-10-05  Richard Biener  <rguenther@suse.de>
2229         * tree-core.h (tree_block::abstract_flag): Remove.
2230         (tree_block::block_num): Make full 32bits.
2231         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
2232         * tree.h (BLOCK_ABSTRACT): Remove.
2233         * dwarf2out.c (gen_lexical_block_die): Remove dead code
2234         resulting from BLOCK_ABSTRACT being always false.
2235         (gen_inlined_subroutine_die): Likewise.
2236         (gen_block_die): Likewise.
2237         * tree.c (block_ultimate_origin): Likewise.
2238         * tree-pretty-print.c (dump_block_node): Remove code dealing
2239         with BLOCK_ABSTRACT.
2240         * tree-ssa-live.c (dump_scope_block): Likewise.
2241         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
2242         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
2244 2018-10-05   Richard Biener  <rguenther@suse.de>
2246         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
2247         is asked for initialize mode to the component mode of the
2248         vector type.
2250 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
2252         PR target/87522
2253         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2254         assembler for -mavx.
2255         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2257 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2259         PR target/87509
2260         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2261         RS6000_BTM_DFP.
2262         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2263         to be DImode.  When using mffscrn, force the operand to a register.
2265 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
2267         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2268         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2269         X87MODEF mode iterator.
2270         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2271         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2272         X87MODEF mode iterator.
2274 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
2276         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2277         -Wno-prio-ctor-dtor.
2279 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2281         * Makefile.in (OBJS): Add opt-problem.o.
2282         * dump-context.h: Include "selftest.h.
2283         (selftest::temp_dump_context): New forward decl.
2284         (class dump_context): Make friend of class
2285         selftest::temp_dump_context.
2286         (dump_context::dump_loc_immediate): New decl.
2287         (class dump_pretty_printer): Move here from dumpfile.c.
2288         (class temp_dump_context): Move to namespace selftest.
2289         (temp_dump_context::temp_dump_context): Add param
2290         "forcibly_enable_dumping".
2291         (selftest::verify_dumped_text):
2292         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2293         (selftest::verify_item):
2294         (ASSERT_IS_TEXT): Move here from dumpfile.c.
2295         (ASSERT_IS_TREE): Likewise.
2296         (ASSERT_IS_GIMPLE): Likewise.
2297         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2298         to...
2299         (dump_context::dump_loc_immediate): ...this new function.
2300         (class dump_pretty_printer): Move to dump-context.h.
2301         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2302         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2303         (temp_dump_context::temp_dump_context): Move to "selftest"
2304         namespace.  Add param "forcibly_enable_dumping", and use it to
2305         conditionalize the use of m_pp;
2306         (selftest::verify_dumped_text): Make non-static.
2307         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2308         (selftest::verify_item): Make non-static.
2309         (ASSERT_IS_TEXT): Move to dump-context.h.
2310         (ASSERT_IS_TREE): Likewise.
2311         (ASSERT_IS_GIMPLE): Likewise.
2312         (selftest::test_capture_of_dump_calls): Pass "true" for new
2313         param of temp_dump_context.
2314         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2315         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
2316         TDF_COMPARE_DEBUG.
2317         * opt-problem.cc: New file.
2318         * opt-problem.h: New file.
2319         * optinfo-emit-json.cc
2320         (selftest::test_building_json_from_dump_calls): Pass "true" for
2321         new param of temp_dump_context.
2322         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2323         (optinfo::emit_for_opt_problem): New function.
2324         (optinfo::emit): Clarity which emit_item is used.
2325         * optinfo.h (optinfo::get_dump_location): New accessor.
2326         (optinfo::emit_for_opt_problem): New decl.
2327         (optinfo::emit): Make const.
2328         * selftest-run-tests.c (selftest::run_tests): Call
2329         selftest::opt_problem_cc_tests.
2330         * selftest.h (selftest::opt_problem_cc_tests): New decl.
2331         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2332         bool to opt_result, converting fprintf messages to
2333         opt_result::failure_at calls.  Add "stmt" param for use by the
2334         failure_at calls.
2335         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2336         (runtime_alias_check_p): Convert return type from bool to
2337         opt_result, converting dump_printf calls to
2338         opt_result::failure_at, using the statement DDR_A for their
2339         location.
2340         (find_data_references_in_stmt): Convert return type from bool to
2341         opt_result, converting "return false" to opt_result::failure_at
2342         with a new message.
2343         * tree-data-ref.h: Include "opt-problem.h".
2344         (dr_analyze_innermost): Convert return type from bool to opt_result,
2345         and add a const gimple * param.
2346         (find_data_references_in_stmt): Convert return type from bool to
2347         opt_result.
2348         (runtime_alias_check_p): Likewise.
2349         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2350         dr_analyze_innermost.
2351         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2352         Convert return type from bool to opt_result, adding a message for
2353         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2354         (vect_analyze_data_ref_dependence): Convert return type from bool
2355         to opt_result.  Change sense of return type from "false"
2356         effectively meaning "no problems" to "false" meaning a problem,
2357         so that "return false" becomes "return opt_result::success".
2358         Convert "return true" calls to opt_result::failure_at, using
2359         the location of statement A rather than vect_location.
2360         (vect_analyze_data_ref_dependences): Convert return type from bool
2361         to opt_result.
2362         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2363         calls to opt_result::failure_at, using the stmt location rather
2364         than vect_location.
2365         (vect_verify_datarefs_alignment): Convert return type from bool
2366         to opt_result.
2367         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
2368         into multiple more-tightly-scoped copies.
2369         (vect_analyze_data_refs_alignment): Convert return type from bool
2370         to opt_result.
2371         (vect_analyze_data_ref_accesses): Likewise, converting a
2372         "return false" to a "return opt_result::failure_at", adding a
2373         new message.
2374         (vect_prune_runtime_alias_test_list): Convert return type from
2375         bool to opt_result, converting dump_printf_loc to
2376         opt_result::failure_at.  Add a %G to show the pertinent statement,
2377         and use the stmt's location rather than vect_location.
2378         (vect_find_stmt_data_reference): Convert return type from
2379         bool to opt_result, converting dump_printf_loc to
2380         opt_result::failure_at, using stmt's location.
2381         (vect_analyze_data_refs):  Convert return type from bool to
2382         opt_result.  Convert "return false" to "return
2383         opt_result::failure_at", adding messages as needed.
2384         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2385         type from bool to opt_result.
2386         (vect_determine_vf_for_stmt): Likewise.
2387         (vect_determine_vectorization_factor): Likewise, converting
2388         dump_printf_loc to opt_result::failure_at, using location of phi
2389         rather than vect_location.
2390         (vect_analyze_loop_form_1): Convert return type from bool to
2391         opt_result, converting dump_printf_loc calls, retaining the use of
2392         vect_location.
2393         (vect_analyze_loop_form): Convert return type from loop_vec_info
2394         to opt_loop_vec_info.
2395         (vect_analyze_loop_operations): Convert return type from bool to
2396         opt_result, converting dump_printf_loc calls, using the location
2397         of phi/stmt rather than vect_location where available.  Convert
2398         various "return false" to "return opt_result::failure_at" with
2399         "unsupported phi" messages.
2400         (vect_get_datarefs_in_loop): Convert return type from bool to
2401         opt_result.  Add a message for the
2402         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2403         (vect_analyze_loop_2): Convert return type from bool to
2404         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
2405         each "goto again;", adding new messages where needed.
2406         Add "unsupported grouped {store|load}" messages.
2407         (vect_analyze_loop): Convert return type from loop_vec_info to
2408         opt_loop_vec_info.
2409         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2410         bool to opt_result.
2411         * tree-vect-stmts.c (process_use): Likewise, converting
2412         dump_printf_loc call and using stmt location, rather than
2413         vect_location.
2414         (vect_mark_stmts_to_be_vectorized): Likeise.
2415         (vect_analyze_stmt): Likewise, adding a %G.
2416         (vect_get_vector_types_for_stmt): Convert return type from bool to
2417         opt_result, converting dump_printf_loc calls and using stmt
2418         location, rather than vect_location.
2419         (vect_get_mask_type_for_stmt): Convert return type from tree to
2420         opt_tree, converting dump_printf_loc calls and using stmt location.
2421         * tree-vectorizer.c: Include "opt-problem.h.
2422         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2423         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
2424         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
2425         enabled, use it to report at the top level "couldn't vectorize
2426         loop" followed by the problem.
2427         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2428         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2429         to opt_result.
2430         (vect_analyze_stmt): Likewise.
2431         (vect_get_vector_types_for_stmt): Likewise.
2432         (tree vect_get_mask_type_for_stmt): Likewise.
2433         (vect_analyze_data_ref_dependences): Likewise.
2434         (vect_enhance_data_refs_alignment): Likewise.
2435         (vect_analyze_data_refs_alignment): Likewise.
2436         (vect_verify_datarefs_alignment): Likewise.
2437         (vect_analyze_data_ref_accesses): Likewise.
2438         (vect_prune_runtime_alias_test_list): Likewise.
2439         (vect_find_stmt_data_reference): Likewise.
2440         (vect_analyze_data_refs): Likewise.
2441         (vect_analyze_loop): Convert return type from loop_vec_info to
2442         opt_loop_vec_info.
2443         (vect_analyze_loop_form): Likewise.
2444         (vect_analyze_slp): Convert return type from bool to opt_result.
2446 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2448         * doc/invoke.texi (-fopt-info): Document new "internals"
2449         sub-option.
2450         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
2451         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
2452         MSG_ALL_KINDS.
2453         (optinfo_verbosity_options): Add "internals".
2454         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
2455         (dump_context::apply_dump_filter_p): New member function.
2456         (dump_context::dump_loc): Use apply_dump_filter_p rather than
2457         explicitly masking the dump_kind.
2458         (dump_context::begin_scope): Increment the scope depth first.  Use
2459         apply_dump_filter_p rather than explicitly masking the dump_kind.
2460         (dump_context::emit_item): Use apply_dump_filter_p rather than
2461         explicitly masking the dump_kind.
2462         (dump_dec): Likewise.
2463         (dump_hex): Likewise.
2464         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
2465         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
2466         (opt_info_switch_p): Update handling of default
2467         MSG_OPTIMIZED_LOCATIONS to cope with default of
2468         MSG_PRIORITY_USER_FACING.
2469         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
2470         masking the dump_kind.
2471         (selftest::test_capture_of_dump_calls): Update test_dump_context
2472         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
2473         than MSG_ALL.  Generalize scope test to be run at all four
2474         combinations of with/without MSG_PRIORITY_USER_FACING and
2475         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
2476         for each of the two values.
2477         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
2478         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
2479         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
2480         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
2481         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
2482         with MSG_PRIORITY_*.
2483         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
2484         dump messages as MSG_PRIORITY_USER_FACING.
2485         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
2486         about the interaction with MSG_PRIORITY_*.
2488 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2490         * varasm.c (output_constant): Add new parameter merge_strings.
2491         Make strings properly zero terminated in merge string sections.
2492         (mergeable_string_section): Don't fail if the last char is non-zero.
2493         (assemble_variable_contents): Handle merge string sections.
2494         (assemble_variable): Likewise.
2495         (assemble_constant_contents): Likewise.
2496         (output_constant_def_contents): Likewise.
2497         (output_constructor_array_range,
2498         output_constructor_regular_field): Adjust call to output_constant.
2499         (output_object_block): Adjust call to assemble_constant_contents
2500         and assemble_variable_contents.
2502 2018-10-04  Martin Liska  <mliska@suse.cz>
2504         PR c/87483
2505         * cgraphunit.c (process_function_and_variable_attributes):
2506         Warn about a function with alias attribute and a body.
2508 2018-10-04  Martin Liska  <mliska@suse.cz>
2510         PR ipa/82625
2511         * multiple_target.c (redirect_to_specific_clone): New function.
2512         (ipa_target_clone): Use it.
2513         * tree-inline.c: Fix comment.
2515 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2517         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
2518         fields.
2519         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
2520         (gcc::dump_manager::register_pass): New member function, adapted
2521         from loop body in gcc::pass_manager::register_pass, adding a
2522         call to update_dfi_for_opt_info.
2523         (gcc::dump_manager::opt_info_enable_passes): Store the
2524         -fopt-info options into the new fields.  Move the loop
2525         bodies into...
2526         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
2527         function.
2528         * dumpfile.h (struct opt_pass): New forward decl.
2529         (gcc::dump_manager::register_pass): New decl.
2530         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
2531         (class gcc::dump_manager): Add fields "m_optgroup_flags",
2532         "m_optinfo_flags", and "m_optinfo_filename".
2533         * passes.c (gcc::pass_manager::register_pass): Move all of the
2534         dump-handling code to gcc::dump_manager::register_pass.
2536 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
2538         PR rtl-optimization/87466
2539         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
2540         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
2541         * doc/tm.texi: Regenerate.
2542         * ira-lives.c (process_bb_node_lives): Use the new target hook.
2543         * lra-lives.c (process_bb_lives): Likewise.
2544         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
2545         Define.
2547 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
2549         * params.c (add_params): Fix initialization.
2551 2018-10-04  Martin Liska  <mliska@suse.cz>
2553         PR gcov-profile/84107
2554         * tree-profile.c (init_ic_make_global_vars):
2555         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
2556         Come up with new ic_tuple* variables.  Emit
2557         __gcov_indirect_call{,_topn} variables.
2558         (gimple_gen_ic_profiler): Access the variable
2559         and emit gimple.
2560         (gimple_gen_ic_func_profiler): Access
2561         __gcov_indirect_call.callee field.
2562         (gimple_init_gcov_profiler): Use ptr_type_node.
2563         * value-prof.c (gimple_ic): Use ptr_type_node.
2565 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2567         PR tree-optimization/85787
2568         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
2569         into this function and add support for detecting multiple phis.
2570         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
2572 2018-10-04  Martin Liska  <mliska@suse.cz>
2574         PR ipa/87491
2575         * ipa-inline.c (inline_to_all_callers_1):
2576         Call ultimate_alias_target for node being inlined.
2578 2018-10-03  Jeff Law  <law@redhat.com>
2580         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
2581         target's wchar_t.
2582         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
2583         * tree.h (get_typenode_from_name): Prototype.
2585 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2587         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
2588         Change operand 2 predicate to nonimmediate_operand.
2589         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2591 2018-10-03  Martin Sebor  <msebor@redhat.com>
2592             Jeff Law  <law@redhat.com>
2594         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
2595         initialize it.
2596         (get_string_length): Detect unterminated arrays.
2597         (format_string): Same.
2598         (format_directive): Warn about unterminated arrays.
2599         (handle_gimple_call): Mark statements with no_warning as needed.
2601 2018-10-03  Jim Wilson  <jimw@sifive.com>
2603         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
2604         also define __riscv_abi_rve.  Delete trailing white space.
2606 2018-10-03  Paul Koning  <ni1d@arrl.net>
2608         Enable LRA register allocator for PDP11.
2609         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
2610         (R): Likewise.
2611         (D): Likewise.
2612         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
2613         * config/pdp11/pdp11.opt (-mlra): New option.
2614         * doc/invoke.texi (PDP-11 Options): Document -mlra.
2616 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2618         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
2619         (*<absneg:code>extend<mode>xf2): Ditto.
2621 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2623         PR tree-optimization/87415
2624         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
2625         precision fields.
2627 2018-10-02  Jeff Law  <law@redhat.com>
2629         * gimple-fold.c (get_range_strlen): Only set *nonstr when
2630         an unterminated string is discovered.  Bubble up range
2631         even for unterminated strings.
2632         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
2633         indicates the string was not terminated via NONSTR.
2635 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2637         * tree-vrp.c (extract_range_from_unary_expr): Special case all
2638         pointer conversions.
2639         Do not do anything special for anti-ranges.
2641 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
2643         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
2644         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
2646 2018-10-03  Martin Liska  <mliska@suse.cz>
2648         PR gcov-profile/86109
2649         * coverage.c (coverage_begin_function): Do not
2650         mark lambdas as artificial.
2651         * tree-core.h (struct GTY): Remove tm_clone_flag
2652         and introduce new lambda_function.
2653         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2655 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2657         PR target/87474
2658         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
2659         P8_VECTOR and VSX are enabled.
2661 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
2663         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
2664         0x3907 as CPU model number.
2666 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
2668         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
2669         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
2670         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
2671         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
2672         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
2673         * config/s390/s390.md: Likewise. Rename also the cpu attribute
2674         value from arch12 to z14.
2676 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
2678         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
2679         (isinfxf2): Ditto.
2680         (isinf<mode>2): Ditto.
2682 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
2684         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
2685         before emitting fxam.  Perform calculations in XFmode.
2687 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
2689         * match.pd (((X /[ex] A) +- B) * A): New transformation.
2691 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
2693         PR middle-end/87319
2694         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
2695         * tree.c (signed_or_unsigned_type_for): Handle complex.
2697 2018-10-02  Jeff Law  <law@redhat.com>
2699         * gimple-fold.c (get_range_strlen): Remove dead code.
2701 2018-10-02  Martin Sebor  <msebor@redhat.com>
2702             Jeff Law  <law@redhat.com>
2704         * builtins.c (unterminated_array): Add new arguments.
2705         If argument is not terminated, bubble up size and exact
2706         state to callers.
2707         (expand_builtin_strnlen): Detect, avoid expanding
2708         and diagnose unterminated arrays.
2709         (c_strlen): Fill in offset of start of unterminated strings.
2710         * builtins.h (unterminated_array): Update prototype.
2712 2018-10-02  Richard Biener  <rguenther@suse.de>
2714         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
2715         of haddv4df, first reduce to SSE width and exploit the fact
2716         that we only need element zero with the reduction result.
2717         (reduc_plus_scal_v2df): Likewise.
2719 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2721         * dojump.h (do_jump): Delete.
2722         (do_jump_1): Likewise.
2723         (split_comparison): Move around.
2724         * dojump.c (do_jump): Make static.
2725         (do_jump_1): Likewise.
2726         (jumpifnot): Move around.
2727         (jumpifnot_1): Likewise.
2728         (jumpif): Likewise.
2729         (jumpif_1): Likewise.
2730         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
2732 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2734         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
2735         insns in the delay slot and add_insn_after for the jump insn.
2737 2018-10-02  Richard Biener  <rguenther@suse.de>
2739         * tree-inline.c (expand_call_inline): Use the location of
2740         the callee declaration for the inline-entry marker.
2741         * final.c (notice_source_line): Remove special-casing of
2742         NOTE_INSN_INLINE_ENTRY.
2744 2018-10-01  Carl Love  <cel@us.ibm.com>
2746         PR 69431
2747         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
2748         (__builtin_mtfsb0): New.
2749         (__builtin_mtfsb1): New.
2750         ( __builtin_set_fpscr_rn): New.
2751         (__builtin_set_fpscr_drn): New.
2752         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
2753         (rs6000_expand_set_fpscr_rn_builtin): Add.
2754         (rs6000_expand_set_fpscr_drn_builtin): Add.
2755         (rs6000_expand_builtin): Add case statement entries for
2756         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
2757         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
2758         RS6000_BUILTIN_MFFSL.
2759         (rs6000_init_builtins): Add ftype initialization and def_builtin
2760         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
2761         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
2762         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
2763         rs6000_mffscdrn): Add define_insn.
2764         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
2765         * doc/extend.texi: Add documentation for the builtins.
2767 2018-10-01  Richard Biener  <rguenther@suse.de>
2769         PR tree-optimization/87465
2770         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
2771         causing branch miscounts.
2773 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2775         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
2776         aarch64_option_default_param):  New.
2777         (params.h): Include.
2778         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
2779         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
2780         stack-clash protection validation code.
2782 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2784         * params.c (validate_param): New.
2785         (add_params): Use it.
2786         (set_param_value): Refactor param validation into validate_param.
2787         (diagnostic.h): Include.
2788         * diagnostic.h (diagnostic_ready_p): New.
2790 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2792         * params.c (set_param_value):
2793         Add index of parameter being validated.
2794         * common/common-target.def (option_validate_param): New.
2795         * common/common-targhooks.h (default_option_validate_param): New.
2796         * common/common-targhooks.c (default_option_validate_param): New.
2797         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
2798         * doc/tm.texi: Regenerate.
2800 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2802         PR target/86486
2803         * config/aarch64/aarch64.c (aarch64_override_options_internal):
2804         Add validation for stack-clash parameters and set defaults.
2806 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2808         PR target/86486
2809         * configure.ac: Add stack-clash-protection-guard-size.
2810         * doc/install.texi: Document it.
2811         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
2812         * params.def: Update comment for guard-size.
2813         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
2814         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
2815         * configure: Regenerate.
2817 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2819         PR target/86486
2820         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
2821         STACK_DYNAMIC_OFFSET): New.
2822         * config/aarch64/aarch64.c (aarch64_layout_frame):
2823         Update outgoing args size.
2824         (aarch64_stack_clash_protection_alloca_probe_range,
2825         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2827 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2829         PR target/86486
2830         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
2831         probe ranges.
2832         * target.def (stack_clash_protection_alloca_probe_range): New.
2833         (stack_clash_protection_final_dynamic_probe): Remove.
2834         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
2835         (default_stack_clash_protection_final_dynamic_probe): Remove.
2836         * targhooks.c: Likewise.
2837         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2838         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
2839         * doc/tm.texi: Regenerate.
2841 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2843         PR target/86486
2844         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
2845         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
2846         aarch64_clamp_to_uimm12_shift): New.
2847         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
2848         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
2850 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2852         PR target/86486
2853         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
2855 2018-10-01  Jeff Law  <law@redhat.com>
2856             Richard Sandiford <richard.sandiford@linaro.org>
2857             Tamar Christina  <tamar.christina@arm.com>
2859         PR target/86486
2860         * config/aarch64/aarch64.md
2861         (probe_stack_range): Add k (SP) constraint.
2862         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
2863         STACK_CLASH_MAX_UNROLL_PAGES): New.
2864         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
2865         stack probes for stack clash.
2866         (aarch64_allocate_and_probe_stack_space): New.
2867         (aarch64_expand_prologue): Use it.
2868         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
2869         (aarch64_sub_sp): Add emit_move_imm optional param.
2871 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
2873         PR tree-optimization/87261
2874         * match.pd: Remove trailing whitespace.
2875         Add (x & y) | ~(x | y) -> ~(x ^ y),
2876         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
2878 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2880         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
2881         constraints.
2882         (ashlsi3_insn): Update instruction constraints.
2883         (ashrsi3_insn): Likewise.
2884         (rotrsi3): Likewise.
2885         (add_shift): Likewise.
2886         * config/arc/constraints.md (Csz): New 32 bit constraint. It
2887         avoids placing in the limm field small constants which, otherwise,
2888         could end into a small instruction.
2890 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2892         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
2893         destination register is not odd-even.
2894         (umaddsidi4_split): Likewise.
2896 2018-10-01  Richard Biener  <rguenther@suse.de>
2898         * tree-inline.c (expand_call_inline): Store origin of fn
2899         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
2900         * tree.c (block_ultimate_origin): Simplify and do some
2901         checking.
2903 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
2905         * config/i386/mmx.md (EMMS): New int iterator.
2906         (emms): New int attribute.
2907         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
2908         EMMS int iterator.  Explicitly declare clobbers.
2909         (mmx_emms): Remove expander.
2910         (mmx_femms): Ditto.
2911         * config/i386/predicates.md (emms_operation): Remove predicate.
2912         (vzeroall_pattern): New predicate.
2913         (vzeroupper_pattern): Rename from vzeroupper_operation.
2914         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
2915         vzeroupper_pattern and vzeroall_pattern predicates.
2917 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
2919         PR rtl-optimization/86939
2920         * ira-lives.c (make_hard_regno_born): Rename from this...
2921         (make_hard_regno_live): ... to this.  Remove update to conflict
2922         information.  Update function comment.
2923         (make_hard_regno_dead): Add conflict information update.  Update
2924         function comment.
2925         (make_object_born): Rename from this...
2926         (make_object_live): ... to this.  Remove update to conflict information.
2927         Update function comment.
2928         (make_object_dead):  Add conflict information update.  Update function
2929         comment.
2930         (mark_pseudo_regno_live): Call make_object_live.
2931         (mark_pseudo_regno_subword_live): Likewise.
2932         (mark_hard_reg_dead): Update function comment.
2933         (mark_hard_reg_live): Call make_hard_regno_live.
2934         (process_bb_node_lives): Likewise.
2935         * lra-lives.c (make_hard_regno_born): Rename from this...
2936         (make_hard_regno_live): ... to this.  Remove update to conflict
2937         information.  Remove now uneeded check_pic_pseudo_p argument.
2938         Update function comment.
2939         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
2940         to conflict information.  Update function comment.
2941         (mark_pseudo_live): Remove update to conflict information.  Update
2942         function comment.
2943         (mark_pseudo_dead): Add conflict information update.
2944         (mark_regno_live): Call make_hard_regno_live.
2945         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
2946         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
2948 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
2950         PR target/87370
2951         * config/i386/i386.c (construct_container): Use TImode for
2952         BLKmode values in 2 integer registers.
2954 2018-09-29  Jeff Law  <law@redhat.com>
2956         * builtins.c (unterminated_array): Pass in c_strlen_data * to
2957         c_strlen rather than just a tree *.
2958         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
2959         Update recursive calls appropriately.  If caller did not provide a
2960         suitable data pointer, create a local one.  When a non-terminated
2961         string is discovered, bubble up information about the string via the
2962         c_strlen_data object.
2963         * builtins.h (c_strlen): Update prototype.
2964         (c_strlen_data): New structure.
2965         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
2966         For a type 2 call, if c_strlen indicates a non-terminated string
2967         use the length of the non-terminated string.
2968         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
2970 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
2972         PR target/87467
2973         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2974         __m512d type for __A argument rather than __m512.
2976 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
2978         * match.pd (simple_comparison): Don't optimize if either operand is
2979         a function pointer when target needs function pointer canonicalization.
2981 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2983         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
2984         power5 .. power9 to remove indirection.
2985         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
2986         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
2987         ASM_CPU_476_SPEC): Delete.
2988         (ASM_CPU_SPEC): Adjust.
2989         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
2990         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
2992 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2994         * config.in: Delete HAVE_AS_DCI.
2995         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
2996         * config/rs6000/rs6000.h: Ditto.
2997         * configure.ac: Delete HAVE_AS_DCI.
2998         * configure: Regenerate.
3000 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
3002         * config.in (HAVE_AS_LWSYNC): Delete.
3003         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3004         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
3005         do it as a .long .
3006         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
3007         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
3008         as a .long .
3009         * configure.ac: Delete HAVE_AS_LWSYNC.
3010         * configure: Regenerate.
3012 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3013             Pierre-Marie de Rodat  <derodat@adacore.com>
3015         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
3016         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
3017         (cgraph_node::create_thunk): Add indirect_offset parameter.
3018         (thunk_adjust): Likewise.
3019         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
3020         and initialize the corresponding field with it.
3021         (cgraph_node::dump): Dump indirect_offset field.
3022         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
3023         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
3024         (thunk_adjust): Add indirect_offset parameter and deal with it.
3025         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
3026         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
3027         or if the thunk is external or local.  Fix formatting.  Do not chain
3028         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
3029         if any, in the GIMPLE representation.
3030         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
3031         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
3032         (input_node): Read indirect_offset field.
3033         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
3034         call to thunk_adjust.
3035         * tree-nested.c (struct nesting_info): Add thunk_p field.
3036         (create_nesting_tree): Set it.
3037         (convert_all_function_calls): Copy static chain from targets to thunks.
3038         (finalize_nesting_tree_1): Return early for thunks.
3039         (unnest_nesting_tree_1): Do not finalize thunks.
3040         (gimplify_all_functions): Do not gimplify thunks.
3042 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3044         * opt-suggestions.c (option_proposer::build_option_suggestions):
3045         Release "option_values".
3047 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
3049         * coverage.c (get_coverage_counts): Convert problem-reporting dump
3050         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
3051         * dumpfile.c (kind_as_string): New function.
3052         (dump_loc): Rather than a hardcoded prefix of "note: ", use
3053         kind_as_string to vary the prefix based on dump_kind.
3054         (selftest::test_capture_of_dump_calls): Update for above.
3056 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3058         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
3059         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
3061 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
3063         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
3064         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
3065         INVALID_REGNUM instead of FPSR_REG.
3066         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
3067         * config/i386/i386.md: Update comment of FP compares.
3068         (fldenv): Do not clobber FPSR_REG.
3070 2018-09-28  Richard Biener  <rguenther@suse.de>
3072         * tree.h (BLOCK_ORIGIN): New.
3073         * omp-expand.c (grid_expand_target_grid_body): Assign
3074         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
3075         * tree-inline.c (remap_block): Likewise.
3076         * auto-profile.c (get_function_decl_from_block): Simplify
3077         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
3078         * langhooks.c (lhd_print_error_function): Likewise.
3079         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
3080         Likewise.
3081         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3082         * tree.c (block_nonartificial_location): Likewise.
3083         (block_ultimate_origin): Likewise.
3084         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
3085         no longer needed LTO case.
3087 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
3088             Jan Hubicka  <jh@suse.cz>
3089             Martin Jambor  <mjambor@suse.cz>
3091         * simplify-rtx.c (simplify_merge_mask): New function.
3092         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
3093         same masks are used in op1 or op2.
3094         (test_vec_merge): New function.
3095         (test_vector_ops): Call test_vec_merge.
3097 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
3099         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
3100         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
3101         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
3102         (cypress_costs): Set it.
3103         (supersparc_costs): Likewise.
3104         (hypersparc_costs): Likewise.
3105         (leon_cost): Likewise.
3106         (leon3_costs): Likewise.
3107         (sparclet_costs): Likewise.
3108         (ultrasparc_costs): Likewise.
3109         (ultrasparc_costs): Likewise.
3110         (niagara_costs): Likewise.
3111         (niagara2_costs): Likewise.
3112         (niagara3_costs): Likewise.
3113         (niagara4_costs): Likewise.
3114         (niagara7_costs): Likewise.
3115         (m8_costs): Likewise.
3116         (TARGET_CAN_FOLLOW_JUMP): Define.
3117         (pass_work_around_errata::gate): Minor tweak.
3118         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
3119         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
3120         Automaitcally clear MASK_FSMULD mask for V7 processors.
3121         (sparc_can_follow_jump): New static function.
3122         (output_ubranch): Deal with CROSSING_JUMP_P.
3123         (sparc_use_sched_lookahead): Rewrite using switch statement.
3124         (sparc_issue_rate): Reorder.
3125         (sparc_branch_cost): New function.
3127 2018-09-27  Martin Sebor  <msebor@redhat.com>
3129         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
3130         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
3131         (int_fits_type_p): Same.
3133 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
3135         * config/i386/i386.md (FPCR_REG): Remove.
3136         (UNSPEC_FLDCW): Remove.
3137         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
3138         (x86_fldcw_1): Remove insn pattern.
3139         (fnstenv): Do not clobber FPCR_REG.
3140         (fldenv): Ditto.
3141         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
3142         (CALL_USED_REGISTERS): Ditto.
3143         (REG_ALLOC_ORDER): Ditto.
3144         (REG_CLASS_CONTENTS): Ditto.
3145         (HI_REGISTER_NAMES): Ditto.
3146         (ADDITIONAL_REGISTER_NAMES): Use defines instead
3147         of numerical constants.
3148         * config/i386/i386.c (regclass_map): Remove fpsr register.
3149         (dbx_register_map): Ditto.
3150         (dbx64_register_map): Ditto.
3151         (svr4_dbx_register_map): Ditto.
3152         (print_reg): Do not handle FPCR_REG.
3154 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
3156         PR target/87149
3157         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
3158         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
3159         Delete, always treat as true.
3160         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
3161         Ditto.  Simplify remaining code.
3162         * config/powerpcspe/powerpcspe.h: Ditto.
3163         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
3164         Simplify remaining code.
3165         (rs6000_expand_builtin): Ditto.
3166         * config/rs6000/rs6000.h: Ditto.
3167         * configure.ac: Ditto.
3168         * configure: Regenerate.
3170 2018-09-27  Martin Liska  <mliska@suse.cz>
3172         * coverage.c (get_coverage_counts): Revert the formatting
3173         of missing profile opt info.
3175 2018-09-27  Richard Biener  <rguenther@suse.de>
3177         PR debug/37801
3178         PR debug/87440
3179         * dwarf2out.c (set_block_origin_self): Do not mark outermost
3180         block as we do not output that.
3181         (gen_inlined_subroutine_die): Elide the originally outermost
3182         block, matching what we do for concrete instances.
3183         (decls_for_scope): Add parameter specifying whether to recurse
3184         to subblocks.
3186 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
3187             Tom de Vries  <tom@codesourcery.com>
3189         PR 82089
3191         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
3192         STORE_FLAG_VALUE == 1.
3194 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
3196         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
3197         constant definitions.
3198         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
3199         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
3200         ("speculation_barrier"): New expander definition.
3202 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
3204         PR gcov-profile/86957
3205         * common.opt: New warning option -Wmissing-profile.
3206         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
3207         * doc/invoke.texi: Document -Wmissing-profile.
3209 2018-09-26  Jim Wilson  <jimw@sifive.com>
3211         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
3212         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
3214 2018-09-26  Martin Sebor  <msebor@redhat.com>
3216         * tree.c (zerop): Change return type to bool.
3217         (integer_zerop, integer_onep, integer_each_onep): Same.
3218         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3219         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3220         (real_onep, real_minus_onep, chain_index): Same.
3221         (print_type_hash_statistics, type_list_equal): Same.
3222         * tree.h (zerop): Same.
3223         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
3224         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
3225         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
3226         (real_onep, real_minus_onep, chain_index): Same.
3227         (print_type_hash_statistics, type_list_equal): Same.
3229 2018-09-26  Jim Wilson  <jimw@sifive.com>
3231         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
3233 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
3235         PR target/87414
3236         * config/i386/i386.c: Include debug.h and dwarf2out.h.
3237         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
3238         call.
3240 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
3242         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
3244 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3246         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
3247         and FP_SECOND_SSE_REGS.
3248         (REG_CLASS_NAMES): Ditto.
3249         (REG_CLASS_CONTENTS): Ditto.
3250         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
3251         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
3252         (ix86_preferred_output_reload_class): Ditto.
3253         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
3254         clobber constraint to "=&f".
3255         (fix_truncdi_i387): Ditto.
3256         (lrintxfdi2): Ditto.
3257         (fistdi2_<rounding>): Ditto.
3258         (fpremxf4_i387): Change "=u" constraint to "=f".
3259         (fprem1xf4_i387): Ditto.
3260         (sincosxf3): Ditto.
3261         (fptanxf4_i387): Ditto.
3262         (fxtractxf3_i387): Ditto.
3263         (fscalexf4_i387): Ditto.
3264         (atan2xf3): Change "u" constraint to "f".
3265         (fyl2xxf3_i387): Ditto.
3266         (fyl2xp1xf3_i387): Ditto.
3268 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3270         PR target/87439
3271         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3272         for removed I387_MASK_PM entity.
3275 2018-09-26  Jeff Law  <law@redhat.com>
3276         Revert
3277         2018-09-26  Alexey Neyman  <stilor@att.net>
3279         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3280         headers are no longer pulled in by <isl/val.h>.
3282 2018-09-26  Richard Biener  <rguenther@suse.de>
3284         PR debug/87443
3285         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3286         or concrete instance DIE to the tree.  Create abstract origin
3287         attributes also for concrete instances.
3289 2018-09-26  Alexey Neyman  <stilor@att.net>
3291         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3292         headers are no longer pulled in by <isl/val.h>.
3294 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
3296         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3297         Use new helper functions.
3298         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3299         Use new helper functions.
3300         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3301         aarch_mm_needs_release): New declarations.
3302         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3303         aarch_mm_needs_release): New.
3305 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
3307         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3308         (arm32_output_mi_thunk): Deal with long calls.
3310 2018-09-26  Richard Biener  <rguenther@suse.de>
3312         PR debug/87428
3313         PR debug/87362
3314         * tree-inline.c (expand_call_inline): When the location
3315         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3316         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3317         the inserted BLOCK to make inlined_function_outer_scope_p
3318         recognize it.
3319         * dwarf2out.c (add_call_src_coords_attributes): Do not add
3320         coords for reserved locations.
3322 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3324         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3325         (*call_indirect_nonlocal_sysv<mode>): Ditto.
3326         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3327         (*sibcall_nonlocal_sysv<mode>): Ditto.
3328         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3329         (<bd>_<mode>): Ditto.
3330         (<bd>tf_<mode>): Ditto.
3332 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3334         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3335         control string as a list of templates instead of as C code.
3336         (*altivec_movti): Ditto.
3337         * config/rs6000/darwin.md (movdf_low_di): Ditto.
3339 2018-09-25  Jim Wilson  <jimw@sifive.com>
3341         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3342         when target symbol is weak.
3344 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3346         PR c/87387
3347         * builtins.c (unterminated_array): Simplify.
3348         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
3349         where pointer arithmetic is safe.
3351 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3353         PR target/86987
3354         * config/rs6000/altivec.md (altivec_vspltb): Use
3355         const_0_to_15_operand instead of u5bit_cint_operand.
3356         (*altivec_vspltb_internal): Ditto.
3357         (altivec_vspltb_direct): Ditto.
3358         (altivec_vsplth): Use const_0_to_7_operand instead of
3359         u5bit_cint_operand.
3360         (*altivec_vsplth_internal): Ditto.
3361         (altivec_vsplth_direct): Ditto.
3362         (altivec_vspltw): Use const_0_to_3_operand instead of
3363         u5bit_cint_operand.
3364         (*altivec_vspltw_internal): Ditto.
3365         (altivec_vspltw_direct): Ditto.
3366         (altivec_vspltsf): Ditto.
3367         (*altivec_vspltsf_internal): Ditto.
3368         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3369         various splats with the proper size immediate.  Reorder the various
3370         cases by ascending size of immediate, and put all such together.
3372 2018-09-25  Richard Biener  <rguenther@suse.de>
3374         PR debug/83941
3375         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3376         GC-ification.
3377         (maybe_create_die_with_external_ref): Do not create
3378         DW_TAG_imported_unit here.
3379         (add_abstract_origin_attribute): Handle external BLOCK refs.
3380         (dwarf2out_abstract_function): Simplify LTO case.
3381         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3382         rather than using maybe_create_die_with_external_ref.
3384 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
3386         PR target/71278
3387         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3388         (frndintxf2_mask_pm_i387): Ditto.
3389         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3390         Enable for !flag_trapping_math.
3391         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3392         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3393         Change operand 1 predicate to nonimmediate_operand.
3394         (attr "i387_cw"): Remove mask_pm.
3395         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3396         (enum ix86_entity): Remove I387_MASK_PM.
3397         * config/i386/i386.c (ix86_i387_mode_needed): Do not
3398         handle I387_MASK_PM.
3399         (ix86_mode_needed): Ditto.
3400         (ix86_mode_after): Ditto.
3401         (ix86_mode_entry): Ditto.
3402         (ix86_mode_exit): Ditto.
3403         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3405 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
3407         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3408         to_update_switch_stmts to vNULL instead of calling create on them
3409         immediately.
3411 2018-09-25  Richard Biener  <rguenther@suse.de>
3413         PR tree-optimization/87402
3414         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3415         (visit_phi): Re-instantiate handling of supposed to be VARYING
3416         but non-VARYING backedge value.
3418 2018-09-25  Richard Biener  <rguenther@suse.de>
3420         PR debug/83941
3421         * dwarf2out.c (struct sym_off_pair): New.
3422         (external_die_map): New global.
3423         (lookup_decl_die): When in LTO create DIEs lazily from the
3424         external_die_map.
3425         (lookup_block_die): New function, create DIEs lazily in LTO.
3426         (equate_block_to_die): New function.
3427         (dwarf2out_die_ref_for_decl): During WPA get the association
3428         from the external DIE map.
3429         (dwarf2out_register_external_die): Record mapping into the
3430         external DIE map.
3431         (maybe_create_die_with_external_ref): New function split out from
3432         DIE generation part of old dwarf2out_register_external_die.
3433         (add_abstract_origin_attribute): Do not return the DIE.  When
3434         in LTO reference externals directly.
3435         (dwarf2out_abstract_function): When in LTO ignore calls for
3436         decls with external DIEs (already present abstract instances).
3437         (gen_call_site_die): Adjust.
3438         (add_high_low_attributes): Likewise.
3439         (gen_lexical_block_die): Likewise.
3440         (gen_inlined_subroutine_die): Likewie.
3441         (gen_block_die): Likewise.
3442         (dwarf2out_inline_entry): Likewise.
3443         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
3444         DIEs.
3446 2018-09-25  Martin Liska  <mliska@suse.cz>
3448         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
3449         integers and not by a float value.
3451 2018-09-25  Martin Liska  <mliska@suse.cz>
3453         PR fortran/87394
3454         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
3455         instead of NULL.
3456         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
3457         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
3458         * godump.c (go_format_type): Remove extra parenthesis.
3460 2018-09-25  Martin Liska  <mliska@suse.cz>
3462         * alias.c (set_dest_equal_p): Remove unused function.
3463         * config/i386/i386.c (def_builtin_pure2): Likewise.
3464         * diagnostic-show-locus.c (class layout): Remove
3465         unused field.
3466         (layout::layout): Likewise here.
3467         * dump-context.h (class temp_dump_context): Likewise.
3468         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
3469         (add_AT_loclistsptr): Likewise.
3470         (add_AT_offset): Likewise.
3471         (get_AT_hi_pc): Likewise.
3472         (is_comdat_die): Likewise.
3473         (type_is_enum): Likewise.
3474         (ceiling): Likewise.
3475         (add_AT_vms_delta): Likewise.
3476         (is_class_die): Likewise.
3477         * edit-context.c (class line_event): Remove unused field.
3478         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
3479         unused function.
3480         * ipa-cp.c (ipa_get_vr_lat): Likewise.
3481         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
3482         (ok_for_base_p_nonstrict): Likewise.
3483         * tree-chrec.c (is_not_constant_evolution): Likewise.
3484         (chrec_fold_poly_cst): Likewise.
3485         * tree-if-conv.c (has_pred_critical_p): Likewise.
3486         * tree-ssa-coalesce.c (print_exprs): Likewise.
3487         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
3488         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
3489         * tree-vrp.c (value_ranges_intersect_p): Likewise.
3490         (value_range_nonnegative_p): Likewise.
3492 2018-09-25  Martin Liska  <mliska@suse.cz>
3494         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3495         Do not handle "GNU Pascal".
3496         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3497         Likewise.
3498         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
3499         from documentation. Likewise.
3500         * dbxout.c (dbxout_range_type): Likewise.
3501         * doc/cpp.texi: Likewise.
3502         * doc/extend.texi: Likewise.
3503         * doc/frontends.texi: Likewise.
3504         * doc/invoke.texi: Remove Pascal entry.
3505         * tree.def (CLEANUP_POINT_EXPR): Likewise.
3506         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
3508 2018-09-25  Martin Liska  <mliska@suse.cz>
3510         PR middle-end/86078
3511         * doc/invoke.texi: Document all parameters and remove default
3512         of the parameters.
3514 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3516         PR bootstrap/87417
3517         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
3518         contains HOST_WIDE_INTs when computing its size.
3520 2018-09-24  Jim Wilson  <jimw@sifive.com>
3522         PR target/87391
3523         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
3524         not TARGET_RVE.
3525         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
3527 2018-09-24  Andrew Pinski  <apinski@marvell.com>
3529         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3530         access prev before checking it for NULLness in the
3531         AARCH64_FUSE_CMP_BRANCH case.
3533 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
3535         PR target/82699
3536         * config/i386/i386.c (rest_of_insert_endbranch): Set
3537         endbr_queued_at_entrance to true and don't insert ENDBR if
3538         x86_function_profiler will be called.
3539         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3540         is true.
3541         * config/i386/i386.h (machine_function): Add
3542         endbr_queued_at_entrance.
3544 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3546         * genattrtab.c (mk_attr_alt): Use alternative_mask.
3547         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
3548         types.
3549         (check_attr_test): Use alternative_mask.
3550         (get_attr_value): Likewise.
3551         (compute_alternative_mask): Use alternative_mask and XWINT.
3552         (make_alternative_compare): Use alternative_mask.
3553         (attr_alt_subset_p): Use XWINT.
3554         (attr_alt_subset_of_compl_p): Likewise.
3555         (attr_alt_intersection): Use alternative_mask and XWINT.
3556         (attr_alt_union): Likewise.
3557         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
3558         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
3559         (simplify_test_exp): Use alternative_mask and XWINT.
3560         (write_test_expr): Use alternative_mask and XWINT, adjust bit
3561         number calculation to support 64 bits.  Generate code that
3562         checks 64-bit masks.
3563         (main): Use alternative_mask.
3564         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
3566 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3568         PR target/80080
3569         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
3570         RETURN+USE when returning via %r14.
3572 2018-09-24  Martin Liska  <mliska@suse.cz>
3574         * gcov.c (output_lines): Print colorization legend
3575         for both flag_use_colors and flag_use_hotness_colors.
3576         Reword the help.
3578 2018-09-24  Martin Liska  <mliska@suse.cz>
3580         * coverage.c (get_coverage_counts): Use warning_at
3581         with current_function_decl location. Use %qD in warning
3582         message.
3584 2018-09-24  Martin Liska  <mliska@suse.cz>
3586         * memory-block.h (memory_block_pool::release): Annotate with
3587         valgrind that the memory is not accessible.
3589 2018-09-24  Martin Liska  <mliska@suse.cz>
3591         PR sanitizer/85774
3592         * asan.c: Make asan_handled_variables extern.
3593         * asan.h: Likewise.
3594         * cfgexpand.c (expand_stack_vars): Make sure
3595         a representative is unpoison if another
3596         variable in the partition is handled by
3597         use-after-scope sanitization.
3599 2018-09-24  Richard Biener  <rguenther@suse.de>
3601         PR tree-optimization/63155
3602         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
3603         the worklist when the edge of the respective argument isn't
3604         executable.
3606 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3608         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
3609         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
3610         (MASK_CLASS_P): Update for rename.
3611         (MAYBE_MASK_CLASS_P): Ditto.
3612         (REG_CLASS_NAMES): Update.
3613         (REG_CLASS_CONTENT): Update.
3614         * config/i386/i386.c (regclass_map): Update for MASK_REG
3615         and ALL_MASK_REGS rename.
3616         * config/i386/constraints.md (Yk): Update for rename.
3617         (k): Ditto.
3619 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3621         * config/i386/i386.h (enum reg_class): Remove
3622         EVEX_SSE_REGS and MOD4_SSE_REGS.
3623         (REG_CLASS_NAMES): Update.
3624         (REG_CLASS_CONTENT): Update.
3625         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
3626         registers as ALL_SSE_REGS.
3627         (ix86_additional_allocno_class_p): Remove.
3628         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
3629         (ix86_register_priority): Lower priority of EVEX SSE registers.
3630         Use IN_RANGE macro where appropriate.
3631         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
3632         AVX-5124VNNIW checks.
3633         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
3634         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
3635         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
3636         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
3637         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
3638         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
3639         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
3640         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
3641         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
3642         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
3643         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
3644         * config/i386/constraints.md (Yh): Remove.
3646 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3648         * config/i386/i386.c (regclass_map): Declare integer REX registers
3649         as GENERAL_REGS.
3651 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
3653         * doc/service.texi (Service): Switch the fsf.org link to https.
3655 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3657         PR target/86798
3658         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3659         Define to speculation_safe_value_not_needed.
3661 2018-09-21  Florian Weimer  <fweimer@redhat.com>
3663         PR middle-end/81035
3664         * doc/extend.texi (Common Function Attributes): Mention that
3665         noreturn suppresses tail call optimization.
3667 2018-09-21  Jeff Law  <law@redhat.com>
3669         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
3670         vr_values::cleanup_edges_and_switches.
3671         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
3672         vr_values class.
3673         (identify_jump_threads): Remove EDGE_IGNORE handling.
3674         (execute_vrp): Move handling of to_remove_edges and
3675         to_update_switch_stmts into vr_values class member functions.
3676         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
3677         (to_update_switch_stmts): Likewise.
3678         * vr-values.c: Include cfghooks.h.
3679         (vr_values::vr_values): Initialize to_remove_edges and
3680         to_update_switch_stmts.
3681         (vr_values::~vr_values): Verify to_remove_edges and
3682         to_update_switch_stmts are empty.
3683         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
3684         (vr_values::cleanup_edges_and_switches): New member function.
3685         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
3686         function.  Add new data members.
3688 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
3690         PR tree-optimization/87309
3691         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
3692         calls with pflags and alt_flags.
3693         (selftest::test_capture_of_dump_calls): Add test of interaction of
3694         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
3696 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3698         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
3700 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3702         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
3704 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3706         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
3707         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
3708         (VXWORKS_LIBS_RTP): Minor reordering.
3710 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3712         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
3713         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
3715 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3717         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
3718         (PTRDIFF_TYPE): Likewise.
3720 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3722         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
3723         triplet, similar to support for VxWorks7.
3724         * config/vxworks-dummy.h: Provide a default definition
3725         of TARGET_VXWORKS64 to 0.
3727 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3729         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
3730         * config/vxworks-dummy.h: here.
3732 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3734         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
3736 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
3737             Bo Zhou  <zbo.zhou@hisilicon.com>
3739         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
3740         * config/aarch64/aarch64-tune.md: Regenerated.
3741         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
3742         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
3743         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
3745 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
3746             Julian Brown  <julian@codesourcery.com>
3748         * builtins.c (get_builtin_sync_mem): Handle address spaces.
3750 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3752         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
3753         if the call takes a static chain.
3755 2018-09-21  Martin Liska  <mliska@suse.cz>
3757         * auto-profile.c (autofdo_source_profile::read): Do not
3758         set sum_all.
3759         (read_profile): Do not add working sets.
3760         (read_autofdo_file): Remove sum_all.
3761         (afdo_callsite_hot_enough_for_early_inline): Remove const
3762         qualifier.
3763         * coverage.c (struct counts_entry): Remove gcov_summary.
3764         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
3765         do not support GCOV_TAG_PROGRAM_SUMMARY.
3766         (get_coverage_counts): Remove summary and expected
3767         arguments.
3768         * coverage.h (get_coverage_counts): Likewise.
3769         * doc/gcov-dump.texi: Remove -w option.
3770         * gcov-dump.c (dump_working_sets): Remove.
3771         (main): Do not support '-w' option.
3772         (print_usage): Likewise.
3773         (tag_summary): Likewise.
3774         * gcov-io.c (gcov_write_summary): Do not dump
3775         histogram.
3776         (gcov_read_summary): Likewise.
3777         (gcov_histo_index): Remove.
3778         (gcov_histogram_merge): Likewise.
3779         (compute_working_sets): Likewise.
3780         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
3781         it not obsolete.
3782         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
3783         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
3784         (GCOV_HISTOGRAM_SIZE): Remove.
3785         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
3786         (struct gcov_summary): Simplify rapidly just
3787         to runs and sum_max fields.
3788         (gcov_histo_index): Remove.
3789         (NUM_GCOV_WORKING_SETS): Likewise.
3790         (compute_working_sets): Likewise.
3791         * gcov-tool.c (print_overlap_usage_message): Remove
3792         trailing empty line.
3793         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
3794         (output_lines): Remove program related line.
3795         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
3796         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
3797         histogram.
3798         (input_profile_summary): Do not read it.
3799         (merge_profile_summaries): And do not merge it.
3800         (input_symtab): Do not call removed function.
3801         * modulo-sched.c (sms_schedule): Do not print sum_max.
3802         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
3803         removed when histogram method was invented.
3804         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
3805         mode.
3806         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
3807         GCOV coding style.
3808         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
3809         and dump selected value.
3810         * profile.c (add_working_set): Remove.
3811         (get_working_sets): Likewise.
3812         (find_working_set): Likewise.
3813         (get_exec_counts): Do not work with working sets.
3814         (read_profile_edge_counts): Do not inform as sum_max is removed.
3815         (compute_branch_probabilities): Likewise.
3816         (compute_value_histograms): Remove argument for call of
3817         get_coverage_counts.
3818         * profile.h: Do not make gcov_summary const.
3820 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
3822         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
3824 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3826         PR tree-optimization/86990
3827         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
3828         Check that the entire merged store group is made of constants only for
3829         overlapping stores.
3831 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
3833         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
3834         (VTABLE_VERIFICATION_SPEC): Likewise.
3835         (SANITIZER_EARLY_SPEC): Likewise.
3836         (SANITIZER_SPEC): Likewise.
3837         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
3838         * doc/invoke.texi (Link Options): Document -r.
3840 2018-09-20  Richard Biener <rguenther@suse.de>
3842         PR middle-end/87054
3843         * gimplify.c (gimplify_expr): Retain alignment of
3844         addressable lvalue in dereference.
3846 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
3848         PR bootstrap/87013
3849         * configure.ac: Check for .loc is_stmt support.
3850         * configure, config.in: Rebuilt.
3851         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
3852         if not supported.
3854 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3856         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
3857         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
3858         -misel=no.
3860 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3862         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3863         VECTOR_OTHER.
3864         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
3865         case VECTOR_OTHER.
3867 2018-09-20  Marek Polacek  <polacek@redhat.com>
3869         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
3871 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3873         PR tree-optimization/87288
3874         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
3875         into account when determining PEELING_FOR_NITERS.
3877 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3879         PR tree-optimization/86877
3880         * tree-vect-loop.c (vect_analyze_loop_2): Call
3881         vect_verify_datarefs_alignment.
3883 2018-09-19  Marek Polacek  <polacek@redhat.com>
3885         * doc/invoke.texi: Document -Wclass-conversion.
3887 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
3889         * config/pa/pa.c (pa_adjust_priority): Delete.
3890         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
3892         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
3893         (atomic_storehi): Likewise.
3894         (atomic_storesi): Likewise.
3895         (atomic_loaddi): Restore compare and swap exchange loop code.
3897 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
3899         PR rtl-optimization/86902
3900         * combine.c (try_combine): When changing the CC mode used, don't change
3901         an unrelated mode in other_insn to that new CC mode.
3903 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
3905         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
3906         with %T in place of calls to dump_generic_expr.
3907         (prune_runtime_alias_test_list): Likewise.
3908         (create_runtime_alias_checks): Likewise.
3909         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
3910         (vect_analyze_data_ref_dependence): Likewise.
3911         (vect_slp_analyze_data_ref_dependence): Likewise.
3912         (vect_record_base_alignment): Likewise.  Use %G in place of call
3913         to dump_gimple_stmt.
3914         (vect_compute_data_ref_alignment): Likewise.
3915         (verify_data_ref_alignment): Likewise.
3916         (vect_find_same_alignment_drs): Likewise.
3917         (vect_analyze_group_access_1): Likewise.
3918         (vect_analyze_data_ref_accesses): Likewise.
3919         (dependence_distance_ge_vf): Likewise.
3920         (dump_lower_bound): Likewise.
3921         (vect_prune_runtime_alias_test_list): Likewise.
3922         (vect_find_stmt_data_reference): Likewise.
3923         (vect_analyze_data_refs): Likewise.
3924         (vect_create_addr_base_for_vector_ref): Likewise.
3925         (vect_create_data_ref_ptr): Likewise.
3926         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
3927         (vect_can_advance_ivs_p): Likewise.
3928         (vect_update_ivs_after_vectorizer): Likewise.
3929         (vect_gen_prolog_loop_niters): Likewise.
3930         (vect_prepare_for_masked_peels): Likewise.
3931         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3932         (vect_determine_vectorization_factor): Likewise.
3933         (vect_is_simple_iv_evolution): Likewise.
3934         (vect_analyze_scalar_cycles_1): Likewise.
3935         (vect_analyze_loop_operations): Likewise.
3936         (report_vect_op): Likewise.
3937         (vect_is_slp_reduction): Likewise.
3938         (check_reduction_path): Likewise.
3939         (vect_is_simple_reduction): Likewise.
3940         (vect_create_epilog_for_reduction): Likewise.
3941         (vect_finalize_reduction:): Likewise.
3942         (vectorizable_induction): Likewise.
3943         (vect_transform_loop_stmt): Likewise.
3944         (vect_transform_loop): Likewise.
3945         (optimize_mask_stores): Likewise.
3946         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
3947         (vect_split_statement): Likewise.
3948         (vect_recog_over_widening_pattern): Likewise.
3949         (vect_recog_average_pattern): Likewise.
3950         (vect_determine_min_output_precision_1): Likewise.
3951         (vect_determine_precisions_from_range): Likewise.
3952         (vect_determine_precisions_from_users): Likewise.
3953         (vect_mark_pattern_stmts): Likewise.
3954         (vect_pattern_recog_1): Likewise.
3955         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3956         (vect_record_max_nunits): Likewise.
3957         (vect_build_slp_tree_1): Likewise.
3958         (vect_build_slp_tree_2): Likewise.
3959         (vect_print_slp_tree): Likewise.
3960         (vect_analyze_slp_instance): Likewise.
3961         (vect_detect_hybrid_slp_stmts): Likewise.
3962         (vect_detect_hybrid_slp_1): Likewise.
3963         (vect_slp_analyze_operations): Likewise.
3964         (vect_slp_analyze_bb_1): Likewise.
3965         (vect_transform_slp_perm_load): Likewise.
3966         (vect_schedule_slp_instance): Likewise.
3967         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3968         (vect_mark_stmts_to_be_vectorized): Likewise.
3969         (vect_init_vector_1): Likewise.
3970         (vect_get_vec_def_for_operand): Likewise.
3971         (vect_finish_stmt_generation_1): Likewise.
3972         (vect_check_load_store_mask): Likewise.
3973         (vectorizable_call): Likewise.
3974         (vectorizable_conversion): Likewise.
3975         (vectorizable_operation): Likewise.
3976         (vectorizable_load): Likewise.
3977         (vect_analyze_stmt): Likewise.
3978         (vect_is_simple_use): Likewise.
3979         (vect_get_vector_types_for_stmt): Likewise.
3980         (vect_get_mask_type_for_stmt): Likewise.
3981         * tree-vectorizer.c (increase_alignment): Likewise.
3983 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
3985         * doc/rtl.texi: Adjust vec_select description.
3986         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
3987         non-constant selectors.
3989 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
3991         * config/aarch64/aarch64-protos.h
3992         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
3993         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
3994         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
3995         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
3996         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
3997         (AARCH64_FL_PROFILE): Move index so flags are ordered.
3998         (AARCH64_ISA_RCPC8_4): New flag.
3999         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
4000         to aarch64_offset_9bit_signed_unscaled_p.
4001         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
4002         and use stlur.
4003         * config/aarch64/constraints.md (Ust): New constraint.
4004         * config/aarch64/predicates.md.
4005         (aarch64_9bit_offset_memory_operand): New predicate.
4006         (aarch64_rcpc_memory_operand): New predicate.
4008 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
4010         PR rtl-optimization/87361
4011         * rtlanal.c (nonzero_bits1): Revert accidental change.
4013 2018-09-19  Richard Biener  <rguenther@suse.de>
4015         PR tree-optimization/87349
4016         PR tree-optimization/87342
4017         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
4019 2018-09-18  Marek Polacek  <polacek@redhat.com>
4021         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
4022         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
4024 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4026         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
4028 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
4030         PR rtl-optimization/86882
4031         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
4033 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
4035         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
4036         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
4038 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
4040         PR other/87353
4041         * doc/invoke.texi (Link Options): Fix formatting and grammar.
4043 2018-09-18  Richard Biener  <rguenther@suse.de>
4045         PR middle-end/63155
4046         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
4047         (compute_samebase_partition_bases): Likewise.
4048         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
4049         (gimple_can_coalesce_p): Simplify.
4051 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
4053         Handle a library implementation of ffs calling __builtin_ffs.
4054         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
4055         (mmix_init_libfuncs): New function: make __builtin_ffs expand
4056         to __ffsdi2.
4058 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4060         * diagnostic-show-locus.c (class layout_range): Add field
4061         "m_original_idx".
4062         (layout_range::layout_range): Add "original_idx" param and use it
4063         to initialize new field.
4064         (make_range): Use 0 for original_idx.
4065         (layout::layout): Pass in index to calls to
4066         maybe_add_location_range.
4067         (layout::maybe_add_location_range): Add param "original_idx" and
4068         pass it on to layout_range.
4069         (layout::print_any_labels): Pass on range->m_original_idx to
4070         get_text call.
4071         (gcc_rich_location::add_location_if_nearby): Use 0 for
4072         original_idx.
4073         * gcc-rich-location.h (text_range_label::get_text): Update for new
4074         param.
4075         (range_label_for_type_mismatch::get_text): Likewise.
4077 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4079         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
4081 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
4083         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4084         format_string_diagnostic_t.
4085         (fmtwarn_n): Likewise.
4086         * substring-locations.c
4087         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
4088         (format_warning_n_va): Convert to...
4089         (format_string_diagnostic_t::emit_warning_n_va): ...this.
4090         (format_warning_va): Convert to...
4091         (format_string_diagnostic_t::emit_warning_va): ...this.
4092         (format_warning_at_substring): Convert to...
4093         (format_string_diagnostic_t::emit_warning): ...this.
4094         (format_warning_at_substring_n): Convert to...
4095         (format_string_diagnostic_t::emit_warning_n): ...this.
4096         * substring-locations.h (class format_string_diagnostic_t): New
4097         class.
4098         (format_warning_va): Convert to
4099         format_string_diagnostic_t::emit_warning_va.
4100         (format_warning_n_va): Convert to
4101         format_string_diagnostic_t::emit_warning_n_va.
4102         (format_warning_at_substring): Convert to
4103         format_string_diagnostic_t::emit_warning.
4104         (format_warning_at_substring_n): Convert to
4105         format_string_diagnostic_t::emit_warning_n.
4107 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
4108             Bernd Schmidt <bernds_cb1@t-online.de>
4110         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
4111         SImode args.
4113 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
4115         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
4116         operand 0 predicate to nonimmediate operand.
4117         (rint<mode>2_frndint): Remove insn pattern.
4118         (rint<mode>2): Change operand 1 predicate to general_operand.
4119         Extend operand 1 to XFmode and generate rintxf2 insn.
4120         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
4121         Do not use X87MODEF mode macro.
4122         (frndintxf2_<rounding>_i387): Rename from
4123         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
4124         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
4125         to XFmode and generate significandxf3 insn.
4127 2018-09-17  Richard Biener  <rguenther@suse.de>
4129         PR tree-optimization/87328
4130         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
4131         visiting unexecutable backedges when not iterating.
4132         (do_rpo_vn): Mark all edges not executable even when not
4133         iterating.
4135 2018-09-17  Martin Jambor  <mjambor@suse.cz>
4137         PR c/63886
4138         * doc/invoke.texi (Warning Options): Likewise.
4140 2018-09-17  Richard Biener  <rguenther@suse.de>
4142         PR tree-optimization/87301
4143         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
4144         clean EH info from leftover copy assignments.
4146 2018-09-17  Martin Liska  <mliska@suse.cz>
4148         PR gcov-profile/85871
4149         * gcov.c (output_intermediate_file): Fix out of bounds
4150         access.
4152 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
4154         * config/arc/arc.c: Object attributes for core4 not reflected
4155         correctly.
4156         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
4157         core3).
4159 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
4161         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
4163 2018-09-17  Martin Liska  <mliska@suse.cz>
4165         * doc/gcov.texi: Document new option --use-hotness-colors.
4166         * gcov.c (struct source_info): Declare new field.
4167         (source_info::source_info): Set default for maximum_count.
4168         (print_usage): Add new -q option.
4169         (process_args): Process it.
4170         (accumulate_line_info): Save src->maximum_count.
4171         (output_line_beginning): Make color line number if
4172         flag_use_hotness_colors is set.
4173         (output_line_details): Pass default argument value.
4174         (output_lines): Pass src->maximum_count.
4176 2018-09-17  Martin Liska  <mliska@suse.cz>
4178         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
4179         Use processor_names table.
4180         * config/i386/i386.c (ix86_default_align): Use
4181         processor_cost_table for alignment values.
4182         (ix86_option_override_internal): Use processor_names.
4183         (ix86_function_specific_print): Likewise.
4184         * config/i386/i386.h (struct processor_costs):
4185         Add alignment values.
4186         (struct ptt): Remove and replace with const char *.
4187         * config/i386/x86-tune-costs.h (struct processor_costs):
4188         Declare default alignments for all costs.
4190 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
4192         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
4193         symbolics or VR_VARYING ranges for ABS_EXPR.
4194         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
4195         when range will wrap.
4197 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
4199         PR middle-end/86864
4200         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
4201         before and after a JUMP_TABLE_DATA.
4203 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
4205         PR middle-end/87188
4206         * dojump.c (do_compare_and_jump): Canonicalize function pointers
4207         when one operand is a function pointer.  Use POINTER_TYPE_P and
4208         FUNC_OR_METHOD_TYPE_P.
4209         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
4210         * fold-const.c (build_range_check): Likewise.
4211         * match.pd (simple_comparison): Likewise.
4213 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
4215         PR c/82967
4216         * spellcheck.c (get_edit_distance_cutoff): New function.
4217         (selftest::test_edit_distance_unit_test_oneway): Rename to...
4218         (selftest::test_get_edit_distance_one_way): ...this.
4219         (selftest::test_get_edit_distance_unit): Rename to...
4220         (selftest::test_get_edit_distance_both_ways): ...this.
4221         (selftest::test_edit_distances): Move tests to this new function,
4222         and test some more pairs of strings.  Update for above renaming.
4223         (selftest::get_old_cutoff): New function.
4224         (selftest::test_get_edit_distance_cutoff): New function.
4225         (selftest::assert_suggested_for): New function.
4226         (ASSERT_SUGGESTED_FOR): New macro.
4227         (selftest::assert_not_suggested_for): New function.
4228         (ASSERT_NOT_SUGGESTED_FOR): New macro.
4229         (selftest::test_suggestions): New function.
4230         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
4231         tests to selftest::test_edit_distances and call it.  Add calls to
4232         selftest::test_get_edit_distance_cutoff and
4233         selftest::test_suggestions.
4234         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
4235         (best_match::consider): Replace hard-coded cutoff calculation with
4236         a call to...
4237         (best_match::get_cutoff): New declaration.
4238         (best_match::get_best_meaningful_candidate): Likewise.
4240 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4242         * builtins.c (fold_builtin_strlen): Remove TODO comment.
4244 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4246         revert:
4247         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4249         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4250         terminated string literal.
4252 2018-09-14  Martin Sebor  <msebor@redhat.com>
4254         * builtins.c (unterminated_array): Handle ARRAY_REF.
4255         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
4256         * builtins.h (unterminated_array): Declare extern.
4257         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
4258         arrays.
4259         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4260         calls.
4262 2018-09-14  Martin Sebor  <msebor@redhat.com>
4263             Jeff Law  <law@redhat.com>
4265         * builtins.c (unterminated_array): New.
4266         (expand_builtin_strcpy): Adjust.
4267         (expand_builtin_strcpy_args): Detect unterminated arrays.
4268         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
4269         unterminated arrays.
4270         * gimple-fold.h (get_maxval_strlen): Add argument.
4271         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4273         * gimple-fold.c (get_range_strlen): Add argument.
4274         (get_maxval_strlen): Adjust.
4275         * gimple-fold.h (get_range_strlen): Add argument.
4277 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
4279         * config/i386/movdirintrin.h: Fix copyright year.
4281 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
4283         * reg-stack.c: Include regs.h.
4284         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4285         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4286         FIRST_STACK_REG, not DFmode.
4287         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4288         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
4289         (change stack): Default register mode to the reg_raw_mode of
4290         FIRST_STACK_REG, not DFmode.
4291         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4292         (*swapxf): Rename from swapxf.
4294 2018-09-14  Carl Love  <cel@us.ibm.com>
4296         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4297         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4299 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
4301         PR target/87224
4302         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4303         alternatives.
4305 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
4307         PR target/85628
4308         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4310 2018-09-14  Jason Merrill  <jason@redhat.com>
4312         Fix --enable-gather-detailed-mem-stats.
4313         * hash-table.c (hash_table_usage): Change from variable to function.
4314         * hash-table.h: Adjust.
4315         * Makefile.in: Add missing dependencies on hash-table.h.
4317 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4319         PR tree-optimization/87259
4320         PR lto/87283
4321         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4322         execute_cse_reciprocals_1 has tried transforming.
4324 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
4326         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4327         VR_VARYING for PLUS/MINUS_EXPR.
4329 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4331         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4332         formatting.
4334 2018-09-14  Richard Biener  <rguenther@suse.de>
4336         PR middle-end/63155
4337         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4338         bits for the merged partition.
4340 2018-09-13  Martin Sebor  <msebor@redhat.com>
4341             Bernd Edlinger  <bernd.edlinger@hotmail.de>
4343         * builtins.h (c_srlen): Add argument.
4344         * builtins.c (warn_string_no_nul): New function.
4345         (c_strlen): Add argument and use it.  Update recursive calls.
4346         Pass DECL argument to string_constant to get info on non
4347         terminated strings.  Update *NONSTR as needed.
4348         (fold_builtin_strlen): Add argument to calls to c_strlen.
4349         Warn for unterminated arrays.
4350         (warn_string_no_null): Add prototype.
4351         * expr.c (string_constant): Update arguments.  Update recursive
4352         calls appropriately.  Detect missing NUL terminator and outermost
4353         declaration its missing in.
4354         Improve checks for arrays with nonzero lower bound or elements
4355         that are not a single byte.  Simplify offset computation.
4356         Simplify checks for non-NUL terminated strings.
4357         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4358         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4360 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4362         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4363         correctly.
4364         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
4365         argument.  Fix range checks.
4366         * fold-const.h (c_getstr): Adjust protoype.
4367         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4368         string is constant but contains no NUL byte.
4370         * expr.c (string_constant): Adjust function comment.
4371         Remove bogus check for zero termination.
4373         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4375         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4376         (get_constant_size): Don't make STRING_CSTs larger than they are.
4377         (check_string_literal): New check function for STRING_CSTs.
4378         (output_constant): Use it.
4380 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
4382         PR target/86812
4383         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4385 2018-09-13  Richard Biener  <rguenther@suse.de>
4387         PR tree-optimization/87263
4388         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4389         (struct unwind_state): Add max_rpo field.
4390         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4391         Compute max_rpo, the max RPO number a block can be backwards reached
4392         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
4393         separating it from the iterating mode.
4395 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4397         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4398         (rfs_decision): New scheduling decision.
4400 2018-09-13  Richard Biener  <rguenther@suse.de>
4402         PR bootstrap/87134
4403         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4404         (vn_nary_op_insert_pieces_predicated): Do not write useless
4405         valid_dominated_by_p entry outside of the allocated storage.
4407 2018-09-13  Omar Sandoval  <osandov@osandov.com>
4408             Tom de Vries  <tdevries@suse.de>
4410         PR debug/86985
4411         * dwarf2out.c (is_c): New function.
4412         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4414 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
4416         PR target/85628
4417         * config/aarch64/aarch64.md (*aarch64_bfxil):
4418         Define.
4419         * config/aarch64/constraints.md (Ulc): Define.
4420         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4421         Define.
4422         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4423         New function.
4425 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4427         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4428         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4429         aarch64_layout_frame call.
4430         (aarch64_expand_epilogue): Likewise.
4431         (aarch64_initial_elimination_offset): Likewise.
4432         (aarch64_get_separate_components): Likewise.
4433         (aarch64_use_return_insn_p): Likewise.
4434         (aarch64_layout_frame): Remove unneeded check.
4436 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4438         * configure.ac: Only append
4439         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
4440         gcc_config_arguments if it was never reconfigured or last reconfigure
4441         was with different arguments.
4442         * configure: Regenerated.
4444 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4445             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4447         PR middle-end/87290
4448         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
4449         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
4451 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4453         PR tree-optimization/87287
4454         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
4455         X % C == 0 to X % (unsigned) C == 0 optimization to ...
4456         * match.pd (X % C == 0): ... here.  New optimization.
4458 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4460         PR middle-end/82853
4461         * expr.h (maybe_optimize_mod_cmp): Declare.
4462         * expr.c (mod_inv): New function.
4463         (maybe_optimize_mod_cmp): New function.
4464         (do_store_flag): Use it.
4465         * cfgexpand.c (expand_gimple_cond): Likewise.
4467 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
4468             Julian Brown  <julian@codesourcery.com>
4470         PR middle-end/86336
4471         * gimplify.c (gimplify_scan_omp_clauses): Set
4472         target_firstprivatize_array_bases in OpenACC parallel and kernels
4473         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
4474         OpenACC data regions.
4476 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
4478         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
4479         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
4481 2018-09-12  Richard Biener  <rguenther@suse.de>
4483         PR tree-optimization/87280
4484         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
4485         edge but unreachable target.
4486         (do_rpo_vn): For conservatively handling a PHI only mark
4487         the backedge executable but not the block reachable.
4489 2018-09-12  Richard Biener  <rguenther@suse.de>
4491         PR tree-optimization/87266
4492         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
4493         visited blocks.
4495 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
4497         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
4498         constants.
4499         ("trunc<BFP:mode><DFP_ALL:mode>2")
4500         ("trunc<DFP_ALL:mode><BFP:mode>2")
4501         ("extend<BFP:mode><DFP_ALL:mode>2")
4502         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
4503         according to the target operand type.
4505 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4506             Andreas Krebbel  <krebbel@linux.ibm.com>
4508         PR tree-optimization/86844
4509         * gimple-ssa-store-merging.c
4510         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
4511         there are any overlapping stores in between them, make sure they are
4512         also coalesced or we give up completely.
4514 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4516         PR middle-end/87248
4517         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
4518         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
4520 2018-09-12  Tom de Vries  <tdevries@suse.de>
4522         * common.opt (gdescribe-dies): Add option.
4523         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
4524         attribute for artifical and nameless decls.
4525         (dwarf2out_register_external_die): Add description attribute to
4526         external reference die.
4527         (add_desc_attribute): New functions.
4528         (gen_subprogram_die): Add description attribute to
4529         DW_TAG_call_site_parameter.
4530         * tree-pretty-print.c (print_generic_expr_to_str): New function.
4531         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
4532         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
4533         -gno-describe-dies.
4534         (@item -gdescribe-dies): Add.
4536 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4538         * tree-vrp.c (vrp_shift_undefined_p): Remove.
4539         (extract_range_from_binary_expr_1: Call
4540         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
4541         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
4542         depend on sign.
4544 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4546         * gimple-ssa-warn-alloca.c
4547         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
4548         field for ALLOCA_BOUND_*_LARGE.
4550 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4552         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
4554 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4556         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
4557         for clobbers.  Remove obsolete comment.
4559 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4561         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
4562         mpxchk, mpxld and mpxst types.
4563         (define_attr length_immediate): Remove all processing of mpx types.
4564         (define_attr prefix_0f): Ditto.
4565         (define_attr memory): Ditto.
4567 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4569         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
4570         (log<mode>2): Change operand 1 predicate to general_operand.
4571         Extend operand 1 to XFmode and generate logxf3 insn.
4572         (log10<mode>2): Change operand 1 predicate to general_operand.
4573         Extend operand 1 to XFmode and generate log10xf3 insn.
4574         (log2<mode>2): Change operand 1 predicate to general_operand.
4575         Extend operand 1 to XFmode and generate log2xf3 insn.
4576         (fyl2xp1_extend<mode>xf3_i387): Remove.
4577         (log1p<mode>2): Change operand 1 predicate to general_operand.
4578         Extend operand 1 to XFmode and generate log1pxf3 insn.
4579         (fxtract_extend<mode>xf3_i387): Remove.
4580         (logb<mode>2): Change operand 1 predicate to general_operand.
4581         Extend operand 1 to XFmode and generate logbxf3 insn.
4582         (ilogb<mode>2): Change operand 1 predicate to general_operand.
4583         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
4584         (significand<mode>2): Change operand 1 predicate to general_operand.
4585         Extend operand 1 to XFmode and generate significandxf3 insn.
4587 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4589         * gcc.c (perror_with_name, pfatal_with_name): Delete.
4590         (load_specs): Use fatal_error.
4591         (DELETE_IF_ORDINARY, process_command): Use error.
4592         (execute, run_attempt): Use fatal_error.
4594 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
4596         * diagnostic-core.h (sorry_at): New prototype.
4597         * diagnostic.c (sorry_at): New function.
4599 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
4601         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
4602         by zero as VR_UNDEFINED.
4604 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4606         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
4607         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
4608         (<sincos>mode2): New expander.
4609         (sincos_extend<mode>xf3_i387): Remove insn pattern.
4610         (sincos -> sin, cos splitters): Remove splitter patterns.
4611         (sincos<mode>3): Change operand 2 predicate to general_operand.
4612         Extend operand 2 to XFmode and generate sincosxf3 insn.
4613         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
4614         Change operand 3 predicate to const1_operand.
4615         (fptan_extend<mode>xf4_i387): Remove insn pattern.
4616         (tanxf2): Update operands in the call to fptanxf4_i387.
4617         (tan<mode>2): Change operand 1 predicate to general_operand.
4618         Extend operand 1 to XFmode and generate tanxf3 insn.
4619         (atan2xf3): Rename from *fpatanxf3_i387.
4620         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
4621         (atan2xf3): Remove expander.
4622         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
4623         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
4624         (atan<mode>2): Change operand 1 predicate to general_operand.
4625         Extend operand 1 to XFmode and generate atanxf3 insn.
4627 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4629         * config/i386/i386.md (x87/SSE constant load splitter): Use
4630         memory_operand instead of nonimmediate_operand for input operand
4631         predicate.
4633 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
4635         * config/i386/i386.md (float partial SSE register stall splitter): Move
4636         splitter near its instruction pattern.
4637         (float_extend partial SSE register stall splitter): Ditto.
4638         (float_truncate partial SSE register stall splitter): Ditto.
4640 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
4642         PR target/86794
4643         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4644         to speculation_safe_value_not_needed.
4646         PR target/85666
4647         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
4648         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
4649         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
4650         leaf_function_p, instead use has_hard_reg_initial_val.
4652 2018-09-09  Nathan Sidwell  <nathan@acm.org>
4654         * gcc.h (pfatal_with_name): Don't declare here.
4655         * gcc.c (pfatal_with_name): Make static.
4657 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4659         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
4660         earlyclobber.
4662 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
4664         PR rtl-optimization/85458
4665         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
4666         priority hook to reduce the priority of EXPR.
4668 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
4670         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
4671         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
4672         with X87_ENABLE_FLOAT.  Remove preparation code.
4673         (*float<SWI48:mode><MODEF:mode>2): Rename from
4674         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
4675         math using "enabled" attribute.
4676         (*floatdi<MODEF:mode>2_i387): Rename from
4677         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
4678         enable for 32bit targets only.
4679         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
4680         splitter.
4681         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
4682         as operand 1 predicate.  Rewrite as define_insn_and_split.
4683         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
4685 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
4687         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
4688         to fallthru to FLOAT case.
4690 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4692         PR target/86731
4693         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
4694         around folding of vec_sl to handle out of range shift values.
4696 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
4698         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
4699         Update callers to gen_fix_trunc<mode>_i387_fisttp
4700         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
4701         nonimmediate_operand.
4702         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
4703         and corresponding splitters.
4704         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
4705         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
4706         (fix_truncdi_i387_with_temp): Remove insn pattern
4707         and corresponding splitters.
4708         (fix_trunc<mode>_i387): Change operand 0 predicate to
4709         nonimmediate_operand.
4710         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
4711         and corresponding splitters.
4712         (*fistdi2_1): Remove.
4713         (fistdi2): Ditto.
4714         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
4715         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
4716         (*fist<mode>2_1): Remove.
4717         (fist<mode>2): Ditto.
4718         (fist<mode>2_with_temp): Remove insn pattern and corresponding
4719         splitters.
4720         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
4721         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
4722         (fistdi2_<rounding>): Change operand 0 predicate to
4723         nonimmediate_operand.
4724         (fistdi2_<rounding>_with_temp): Remove insn pattern
4725         and corresponding splitters.
4726         (fist<mode>2_<rounding>): Change operand 0 predicate to
4727         nonimmediate_operand.
4728         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
4729         and corresponding splitters.
4731         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
4733 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4735         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
4736         the init value.
4738 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4740         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4741         early gimple folding of vec_splat().
4742         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
4743         * gimple-fold.h: Add an extern define for tree_vec_extract().
4745 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4747         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
4748         wrappers around TREE_TYPE comparisons.
4750 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4752         PR target/80080
4753         * config/s390/predicates.md: Add nonsym_memory_operand.
4754         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
4755         contains a SYMBOL_REF, load it into an intermediate pseudo.
4756         (s390_emit_compare_and_swap): Legitimize operand.
4757         * config/s390/s390.md: Use the new nonsym_memory_operand
4758         with UNSPECV_CAS patterns.
4760 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4762         PR target/80080
4763         * config/s390/s390-passes.def: New file.
4764         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
4765         declaration.
4766         (make_pass_s390_early_mach): Add declaration.
4767         * config/s390/s390.c (make_pass_s390_early_mach):
4768         (s390_option_override): Remove dynamic registration.
4769         * config/s390/t-s390: Add s390-passes.def.
4771 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4773         * config/s390/s390.c (s390_decompose_constant_pool_ref):
4774         Remove UNSPEC_LTREL_BASE check.
4775         (annotate_constant_pool_refs): Likewise.
4776         (find_constant_pool_ref): Likewise.
4777         (find_ltrel_base): Removed.
4778         (replace_ltrel_base): Removed.
4779         (s390_mainpool_finish): Remove replace_ltrel_base call.
4780         (s390_chunkify_start): Remove pending LTREL_BASE logic.
4781         (s390_chunkify_finish): Remove replace_ltrel_base call.
4782         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
4784 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
4786         PR target/86779
4787         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4788         to speculation_safe_value_not_needed.
4790 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
4791             Bernd Schmidt  <bernds_cb1@t-online.de>
4793         * config/nvptx/nvptx-opts.h: New file.
4794         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
4795         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
4796         (ASM_SPEC): Define.
4797         (TARGET_SM35): New macro.
4798         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
4799         correct predicate.
4800         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
4801         values.
4802         (misa=): New option.
4803         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
4805 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
4807         * config/i386/i386.md (truncdfsf2): Remove expander.
4808         (truncdfsf2_with_temp): Ditto.
4809         (truncxf<mode>2): Ditto.
4810         (*truncdfsf_fast_mixed): Remove insn pattern.
4811         (*truncdfsf_fast_i387): Ditto.
4812         (*truncdfsf_mixed): Ditto.
4813         (*truncdfsf_i387): Ditto.
4814         (*truncdfsf2_i387_1): Ditto.
4815         (*truncxfsf2_mixed): Ditto.
4816         (*truncxfdf2_mixed): Ditto.
4817         (*truncxf<mode>2_i387_noop): Ditto. Update callers
4818         to call gen_truncxf<mode>2 instead.
4819         (*truncxf<mode>2_i387): Remove.
4820         (reg->reg splitters): Remove splitter pattern.
4821         (reg->mem splitters): Ditto.
4823         (truncdfsf2): New insn pattern.
4824         (truncxf<mode>2): Ditto.
4826 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4828         * tree-ssa-math-opts.c (is_mult_by): New function.
4829         (is_square_of): Use the above.
4830         (optimize_recip_sqrt): New function.
4831         (pass_cse_reciprocals::execute): Use the above.
4833 2018-09-05  Richard Biener  <rguenther@suse.de>
4835         PR bootstrap/87134
4836         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
4837         to zero-init the emplaced vec.
4839 2018-09-05  Martin Liska  <mliska@suse.cz>
4841         PR tree-optimization/87205
4842         * tree-switch-conversion.c (pass_lower_switch::execute):
4843         Group cases for switch statements.
4845 2018-09-05  Richard Biener  <rguenther@suse.de>
4847         PR tree-optimization/87217
4848         * tree-ssa-sccvn.c (vuse_valueize): New.
4849         (vn_reference_lookup_pieces): Use it.
4850         (vn_reference_lookup): Likewise.
4852 2018-09-05  Nathan Sidwell  <nathan@acm.org>
4854         PR c++/87137
4855         * stor-layout.c (place_field): Scan forwards to check last
4856         bitfield when ms_bitfield_placement is in effect.
4858 2018-09-05  Richard Biener  <rguenther@suse.de>
4860         PR bootstrap/87225
4861         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
4862         return.
4864 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
4865             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
4867         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
4868         * config.gcc (extra_objs): Build it.
4869         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
4870         Likewise.
4871         * config/aarch64/aarch64-passes.def
4872         (pass_tag_collision_avoidance): New pass.
4873         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
4874         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
4875         (aarch64_classify_address): Remove static qualifier.
4876         (aarch64_address_info, aarch64_address_type): Move to...
4877         * config/aarch64/aarch64-protos.h: ... here.
4878         (make_pass_tag_collision_avoidance): New function.
4879         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
4880         New tuning flag.
4882 2018-09-05  Martin Liska  <mliska@suse.cz>
4884         * doc/gcov.texi: Update documentation of humar
4885         readable mode.
4886         * gcov.c (format_count): Print one decimal place, it provides
4887         more fine number of situations like '1G' vs. '1.4G'.
4889 2018-09-05  Martin Liska  <mliska@suse.cz>
4891         PR target/87164
4892         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
4893         * optc-gen.awk: Allow 'Var' for Deprecated options in order
4894         to generate a MASK value.
4896 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
4898         PR debug/86593
4899         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
4900         if frame pointer isn't used.
4901         (compute_frame_pointer_to_fb_displacement): Likewise.
4903 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
4905         PR target/87198
4906         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
4907         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
4908         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
4909         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
4910         and OPTION_MASK_ISA_XSAVEC_UNSET.
4912 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
4914         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
4915         XOR operations in NAND case.
4917 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4919         * wide-int-range.cc (wide_int_range_convert): New.
4920         * wide-int-range.h (wide_int_range_convert): New.
4921         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
4922         code into wide_int_range_convert.
4923         (extract_range_into_wide_ints): Do not munge anti range constants
4924         into the entire domain.  Just return the range back.
4926 2018-09-04  Martin Liska  <mliska@suse.cz>
4928         * genmatch.c (output_line_directive): Add new argument
4929         fnargs.
4930         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
4932 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4934         * doc/invoke.texi (Option Summary): Add whitespace.
4936         * doc/invoke.texi (Option Summary): Add -Waligned-new.
4938 2018-09-04  Richard Biener  <rguenther@suse.de>
4940         PR tree-optimization/87211
4941         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
4942         backedge value we're supposed to treat as VARYING also number
4943         the PHI to VARYING in case it got a different value-number already.
4945 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4947         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
4948         (extract_range_from_binary_expr_1): Do not call
4949         vrp_can_optimize_bit_op.
4950         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
4951         static.
4952         (wide_int_range_get_mask_and_bounds): New.
4953         (wide_int_range_optimize_bit_op): New.
4954         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
4955         (wide_int_range_bit_and): Same.
4956         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
4957         (wide_int_range_optimize_bit_op): New.
4958         (wide_int_range_get_mask_and_bounds): New.
4960 2018-09-04  Richard Biener  <rguenther@suse.de>
4962         PR tree-optimization/87176
4963         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
4964         variable.  When value-numbering a virtual PHI node make sure
4965         to not value-number to the backedge value.
4967 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4969         * doc/extend.texi (Long Long, Hex Floats): Document support for
4970         long long and hex floats in more recent versions of ISO C++.
4972 2018-09-03  Richard Biener  <rguenther@suse.de>
4974         PR tree-optimization/87177
4975         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
4976         cleanup.
4978 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4980         * bb-reorder.c (edge_order): Convert to C-qsort-style
4981         tri-state comparator.
4982         (reorder_basic_blocks_simple): Change std::stable_sort to
4983         gcc_stablesort.
4985 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4987         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
4988         tri-state comparator.
4989         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
4991 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4993         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
4994         (mergesort): ... here as maximum count for using netsort.
4995         (gcc_qsort): Set nlim to 3 if stable sort is requested.
4996         (gcc_stablesort): New.
4997         * system.h (gcc_stablesort): Declare.
4999 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
5001         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
5002         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
5003         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
5005 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
5007         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
5008         lxsdx and stxsdx alternatives.
5009         (*mov<mode>_hardfloat64): Ditto.
5010         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
5012 2018-09-03  Richard Biener  <rguenther@suse.de>
5014         PR tree-optimization/87200
5015         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
5016         simplify result.
5018 2018-09-03  Martin Liska  <mliska@suse.cz>
5020         PR tree-optimization/87201
5021         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5022         Fix parenthesis in an expression.
5024 2018-09-03  Richard Biener  <rguenther@suse.de>
5026         PR tree-optimization/87197
5027         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
5028         visited.  CSE the VN_INFO hashtable lookup.
5030         PR tree-optimization/87169
5031         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
5032         iterating make sure there's no extra backedges from irreducible
5033         regions feeding the header.  Mark the destination block
5034         executable.
5036 2018-09-03  Martin Liska  <mliska@suse.cz>
5038         PR driver/83193
5039         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
5040         * common/common-targhooks.c (default_get_valid_option_values):
5041         New function.
5042         * common/common-targhooks.h (default_get_valid_option_values):
5043         Likewise.
5044         * common/config/i386/i386-common.c: Move processor_target_table
5045         from i386.c.
5046         (ix86_get_valid_option_values): New function.
5047         (TARGET_GET_VALID_OPTION_VALUES): New macro.
5048         * config/i386/i386.c (struct ptt): Move to i386-common.c.
5049         (PTA_*): Move all defined masks into i386-common.c.
5050         (ix86_function_specific_restore): Use new processor_cost_table.
5051         * config/i386/i386.h (struct ptt): Moved from i386.c.
5052         (struct pta): Likewise.
5053         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
5054         * doc/tm.texi.in: Likewise.
5055         * opt-suggestions.c (option_proposer::suggest_option):
5056         Pass prefix to build_option_suggestions.
5057         (option_proposer::get_completions): Likewise.
5058         (option_proposer::build_option_suggestions): Use the new target
5059         hook.
5060         * opts.c (struct option_help_tuple): New struct.
5061         (print_filtered_help): Use the new target hook.
5063 2018-09-03  Martin Liska  <mliska@suse.cz>
5065         PR middle-end/59521
5066         * predict.c (set_even_probabilities): Add likely_edges
5067         argument and handle cases where we have precisely one
5068         likely edge.
5069         (combine_predictions_for_bb): Catch also likely_edges.
5070         (tree_predict_by_opcode): Handle gswitch statements.
5071         * tree-cfg.h (find_case_label_for_value): New declaration.
5072         (find_taken_edge_switch_expr): Likewise.
5073         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
5074         Find pivot in decision tree based on probabily, not by number of
5075         nodes.
5077 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
5079         * doc/standards.texi (Standards): Update Objective-C reference.
5081 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5083         * doc/install.texi (Prerequisites): Update link for MPC.
5085 2018-09-01  Michael Matz  <matz@suse.de>
5087         PR tree-optimization/87074
5088         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
5089         PHIs for outer-loop uses.
5091 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5093         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
5094         * doc/invoke.texi (C Dialect Options): Ditto.
5096 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
5098         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
5100 2018-08-31  Richard Biener  <rguenther@suse.de>
5102         PR tree-optimization/87168
5103         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
5104         (rpo_elim::eliminate_avail): When OP was not visited it must
5105         be available.
5107 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
5109         * tree-vrp.c (copy_value_range): Convert param "from" from
5110         "value_range *" to "const value_range *".
5111         (range_is_null): Likewise for param "vr".
5112         (range_int_cst_p): Likewise.
5113         (range_int_cst_singleton_p): Likewise.
5114         (symbolic_range_p): Likewise.
5115         (value_ranges_intersect_p): Likewise for both params.
5116         (value_range_nonnegative_p): Likewise for param "vr".
5117         (value_range_constant_singleton): Likewise.
5118         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
5119         (extract_range_into_wide_ints): Likewise for param "vr".
5120         (extract_range_from_multiplicative_op): Likewise for params "vr0"
5121         and "vr1".
5122         (vrp_can_optimize_bit_op): Likewise.
5123         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
5124         "vr1_".
5125         (extract_range_from_unary_expr): Likewise.
5126         (debug_value_range): Likewise for param "vr".
5127         (value_range::dump): Add "const" qualifier.
5128         (vrp_prop::check_array_ref): Convert local "vr" from
5129         "value_range *" to "const value_range *".
5130         (vrp_prop::check_mem_ref): Likewise.
5131         (vrp_prop::visit_stmt): Likewise for local "old_vr".
5132         (vrp_intersect_ranges_1): Likewise for param "vr_1".
5133         (vrp_intersect_ranges): Likewise.
5134         (simplify_stmt_for_jump_threading): Likewise for local "vr".
5135         (vrp_prop::vrp_finalize): Likewise.
5136         * tree-vrp.h (value_range::dump): Add "const" qualifier.
5137         (vrp_intersect_ranges): Add "const" qualifier to params as above.
5138         (extract_range_from_unary_expr): Likewise.
5139         (value_range_constant_singleton): Likewise.
5140         (symbolic_range_p): Likewise.
5141         (copy_value_range): Likewise.
5142         (extract_range_from_binary_expr_1): Likewise.
5143         (range_int_cst_p): Likewise.
5144         (vrp_set_zero_nonzero_bits): Likewise.
5145         (range_int_cst_singleton_p): Likewise.
5147 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
5149         * config/aarch64/arm_neon.h (vabsd_s64): New.
5150         (vnegd_s64): Likewise.
5152 2018-08-31  Martin Jambor  <mjambor@suse.cz>
5154         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
5156 2018-08-31  Martin Liska  <mliska@suse.cz>
5158         * ipa-icf.c (sem_item::add_type): Use
5159         sem_item::m_type_hash_cache.
5160         * ipa-icf.h: Move the cache from sem_item_optimizer
5161         to sem_item.
5163 2018-08-31  Nathan Sidwell  <nathan@acm.org>
5165         * doc/extend.texi (Backwards Compatibility): Remove implicit
5166         extern C leeway of () being (...).
5168 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5170         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
5172 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
5174         PR target/86684
5175         PR target/87149
5176         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
5178 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
5180         PR middle-end/87138
5181         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
5182         gen_int_mode.  Formatting fixes.
5184 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
5186         * target.def (custom_function_descriptors): Improve documentation.
5187         * doc/tm.texi.in (Trampolines): Expand discussion of function
5188         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
5189         beginning of the section.
5190         * doc/tm.texi: Regenerated.
5192 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
5194         * cfg.h (class auto_edge_flag): Spell out the template-id of the
5195         base class in the initializer list.  This is a workaround for
5196         building with older GCC.
5197         (class auto_bb_flag): Likewise.
5199 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
5201         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
5202         (altivec_vcmpequ<VI_char>_p): Remove star.
5203         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
5204         vector load modes.
5205         (expand_strncmp_vec_sequence): New function.
5206         (emit_final_str_compare_vec): New function.
5207         (expand_strn_compare): Add support for vector strncmp.
5208         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
5209         length specification to bytes.
5210         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
5211         (vcmpnezb_p): New pattern.
5212         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
5213         for option -mstring-compare-inline-limit.
5215 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
5217         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
5218         (PTA_SKYLAKE): Add PTA_AES.
5219         (PTA_GOLDMONT): Likewise.
5221 2018-08-29  Jan Hubicka  <jh@suse.cz>
5223         PR lto/86517
5224         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
5225         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
5227 2018-08-29  Jan Hubicka  <jh@suse.cz>
5229         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
5230         TYPE_STUB_DECL.
5231         (hash_tree): Do not visit TYPE_STUB_DECL.
5232         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
5233         stream TYPE_STUB_DECL.
5234         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
5235         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
5236         after free_lang_data.
5237         (type_in_anonymous_namespace_p): Likewise.
5239 2018-08-29  Jan Hubicka  <jh@suse.cz>
5241         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5242         comment that it has to be even number.
5243         (class sreal): Change m_sig type to int32_t.
5244         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5245         int64_t for temporary calculations.
5246         (sreal_verify_basics): Drop one bit from minimum and maximum.
5248 2018-08-30  Richard Biener  <rguenther@suse.de>
5250         PR tree-optimization/87147
5251         * tree-ssa-sccvn.c (SSA_VISITED): New function.
5252         (visit_phi): When the degenerate result is from the backedge and
5253         we didn't visit its definition yet drop to VARYING.
5254         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
5256 2018-08-29  Jan Hubicka  <jh@suse.cz>
5258         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5259         DECL_VINDEX.
5260         (hash_tree): Likewise.
5262 2018-08-29  Jan Hubicka  <jh@suse.cz>
5264         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5265         and TYPE_NEXT_REF_TO.
5267 2018-08-29  Jan Hubicka  <jh@suse.cz>
5269         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5270         comment that it has to be even number.
5271         (class sreal): Change m_sig type to int32_t.
5272         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5273         int64_t for temporary calculations.
5274         (sreal_verify_basics): Drop one bit from minimum and maximum.
5276 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
5278         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5280 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
5282         PR middle-end/86995
5283         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5284         if to_add is negative.
5286 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5288         PR middle-end/87053
5289         * builtins.c (c_strlen): Improve range checks.
5291 2018-08-29  Martin Sebor  <msebor@redhat.com>
5292             Jeff Law  <law@redhat.com>
5294         PR tree-optimization/86714
5295         PR tree-optimization/86711
5296         * builtins.c (c_strlen): Add arguments to call to string_constant.
5297         * expr.c (string_constant): Add argument.  Detect missing nul
5298         terminator and outermost declaration it's missing in.
5299         * expr.h (string_constant): Add argument.
5300         * fold-const.c (read_from_constant_string): Add arguments to call to
5301         string_constant.
5302         (c_getstr): Likewise.
5303         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5304         to string_constant.
5305         * tree-ssa-strlen.c (get_stridx): Likewise.
5307 2018-08-29  Jan Hubicka  <jh@suse.cz>
5309         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5310         Do not stream DECL_VINDEX.
5311         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5312         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5313         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5315 2018-08-29  Richard Biener  <rguenther@suse.de>
5317         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5318         virtual operands that are not default defs to honor region
5319         boundaries.
5320         (rpo_vn_valueize): Remove ineffective code here.
5322 2018-08-29  Richard Biener  <rguenther@suse.de>
5324         PR tree-optimization/87132
5325         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5326         when skipping defs reachable over backedges.
5328 2018-08-29  Richard Biener  <rguenther@suse.de>
5330         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5331         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5332         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5333         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5334         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5335         (vn_reference_lookup_3): Remove use of const_parms.
5336         (free_rpo_vn): Do not free const_parms.
5337         (do_rpo_vn): Do not call init_const_parms.
5338         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5339         SSA_NAME_POINTS_TO_READONLY_MEMORY.
5340         (call_may_clobber_ref_p_1): Likewise.
5342 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
5344         PR other/86726
5345         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5346         (-O): Ditto.
5347         (-ftree-scev-cprop): Document.
5349 2018-08-29  Jan Hubicka  <jh@suse.cz>
5351         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5352         parameters.
5353         (sreal constructor): Update.
5354         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5355         sreal:operator/): Update.
5357 2018-08-29  Martin Liska  <mliska@suse.cz>
5359         * tree-switch-conversion.c (switch_conversion::expand):
5360         Strenghten assumption about gswitch statements.
5362 2018-08-29  Richard Biener  <rguenther@suse.de>
5364         PR tree-optimization/87117
5365         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5366         re-value-number released SSA VDEFs.
5368 2018-08-29  Richard Biener  <rguenther@suse.de>
5370         PR tree-optimization/87126
5371         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5373 2018-08-28  Jim Wilson  <jimw@sifive.com>
5375         * config/riscv/pic.md: Rewrite.
5376         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5377         invalid address.
5378         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5379         (SOFTF, default_load, softload, softstore): New.
5381 2018-08-28  Jeff Law  <law@redhat.com>
5383         * fold-const.c (fold_binary_loc): Remove recently added assert.
5385 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
5387         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5388         to OP parmeter of generated function.
5390 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
5392         PR tree-optimization/87009
5393         * match.pd: Add boolean optimizations.
5395 2018-08-28  Martin Sebor  <msebor@redhat.com>
5397         PR middle-end/86631
5398         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5399         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5400         (pass_walloca::gate): Use it.
5401         (alloca_call_type): Same.
5402         (pass_walloca::execute): Same.
5403         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5405 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
5407         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5408         GCC_VERSION for usage of "__gcc_dump_printf__" format from
5409         >= 3005 to >= 9000.
5411 2018-08-28  Richard Biener  <rguenther@suse.de>
5413         PR tree-optimization/87124
5414         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5415         constants before looking up avail.
5417 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
5419         PR middle-end/87099
5420         * calls.c (maybe_warn_nonstring_arg): Punt early if
5421         warn_stringop_overflow is zero.  Don't call get_range_strlen
5422         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5423         Swap comparison operands to have constants on rhs.  Only use
5424         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
5425         increment lenrng[0].
5427 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
5429         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5430         use of tree_to_shwi.  Remove duplicated test for the size being
5431         a whole number of bytes.
5433 2018-08-28  Richard Biener  <rguenther@suse.de>
5435         PR tree-optimization/87117
5436         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5437         Handle removed stmt without LHS (GIMPLE_NOP).
5439 2018-08-28  Richard Biener  <rguenther@suse.de>
5441         PR tree-optimization/87117
5442         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
5443         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
5445 2018-08-28  Richard Biener  <rguenther@suse.de>
5447         PR tree-optimization/87117
5448         * tree-ssa-pre.c (compute_avail): Do not make expressions
5449         with predicated values available.
5450         (get_expr_value_id): Assert we do not run into predicated value
5451         expressions.
5453 2018-08-28  Richard Biener  <rguenther@suse.de>
5455         PR tree-optimization/87117
5456         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
5457         get virtual operands.
5458         (get_expr_operands): Handle STRING_CST like other decls.
5460 2018-08-28  Martin Liska  <mliska@suse.cz>
5462         * tree.h: Update documentation of fndecl_built_in_p
5463         functions.
5466 2018-08-27  Jeff Law  <law@redhat.com>
5467         PR tree-optimization/87110
5468         * tree-ssa-dse.c (compute_trims): Handle non-constant
5469         TYPE_SIZE_UNIT.
5471 2018-08-27  Martin Sebor  <msebor@redhat.com>
5473         PR tree-optimization/86914
5474         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
5476 2018-08-27  Martin Sebor  <msebor@redhat.com>
5478         PR tree-optimization/87112
5479         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
5480         the type of the bound argument.
5482 2018-08-27  Jeff Law  <law@redhat.com>
5484         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
5485         type does not have a TYPE_SIZE_UNIT.
5487 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
5489         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
5490         with include of backend.h.
5492 2018-08-27  Richard Biener  <rguenther@suse.de>
5494         PR tree-optimization/86927
5495         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
5496         use const cond reduction code.
5498 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
5500         PR tree-optimization/85758
5501         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
5503 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5505         PR c++/87091
5506         * diagnostic-show-locus.c (class layout_range): Update for
5507         conversion of show_caret_p to a tri-state.
5508         (layout_range::layout_range): Likewise.
5509         (make_range): Likewise.
5510         (layout::maybe_add_location_range): Likewise.
5511         (layout::should_print_annotation_line_p): Don't show annotation
5512         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
5513         (layout::get_state_at_point): Update for conversion of
5514         show_caret_p to a tri-state.  Bail out early for
5515         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
5516         underlining or source colorization.
5517         (gcc_rich_location::add_location_if_nearby): Update for conversion
5518         of show_caret_p to a tri-state.
5519         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
5520         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
5521         Likewise.
5522         (selftest::test_one_liner_labels): Likewise.
5523         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
5524         conversion of show_caret_p to a tri-state.
5525         * pretty-print.c (text_info::set_location): Likewise.
5526         * pretty-print.h (text_info::set_location): Likewise.
5527         * substring-locations.c (format_warning_n_va): Likewise.
5528         * tree-diagnostic.c (default_tree_printer): Likewise.
5529         * tree-pretty-print.c (newline_and_indent): Likewise.
5531 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5533         PR c++/87091
5534         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
5535         line above for line-insertion fix-it hints.
5536         (selftest::test_fixit_insert_containing_newline): Update the
5537         expected results, and add a test with line-numbering enabled.
5539 2018-08-27  Martin Liska  <mliska@suse.cz>
5541         PR sanitizer/86962
5542         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
5543         params with DECL_HAS_VALUE_EXPR_P.
5545 2018-08-27  Martin Liska  <mliska@suse.cz>
5547         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
5548         selected expansion strategy.
5550 2018-08-27  Martin Liska  <mliska@suse.cz>
5552         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
5553         * builtins.c (is_builtin_fn): Likewise.
5554         * attribs.c (diag_attr_exclusions): Use new function
5555         fndecl_built_in_p and remove check for FUNCTION_DECL if
5556         possible.
5557         (builtin_mathfn_code): Likewise.
5558         (fold_builtin_expect): Likewise.
5559         (fold_call_expr): Likewise.
5560         (fold_builtin_call_array): Likewise.
5561         (fold_call_stmt): Likewise.
5562         (set_builtin_user_assembler_name): Likewise.
5563         (is_simple_builtin): Likewise.
5564         * calls.c (gimple_alloca_call_p): Likewise.
5565         (maybe_warn_nonstring_arg): Likewise.
5566         * cfgexpand.c (expand_call_stmt): Likewise.
5567         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
5568         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
5569         (cgraph_node::verify_node): Likewise.
5570         * cgraphclones.c (build_function_decl_skip_args): Likewise.
5571         (cgraph_node::create_clone): Likewise.
5572         * config/arm/arm.c (arm_insert_attributes): Likewise.
5573         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
5574         * dse.c (scan_insn): Likewise.
5575         * expr.c (expand_expr_real_1): Likewise.
5576         * fold-const.c (operand_equal_p): Likewise.
5577         (fold_binary_loc): Likewise.
5578         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
5579         * gimple-low.c (lower_stmt): Likewise.
5580         * gimple-pretty-print.c (dump_gimple_call): Likewise.
5581         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
5582         Likewise.
5583         * gimple.c (gimple_build_call_from_tree): Likewise.
5584         (gimple_call_builtin_p): Likewise.
5585         (gimple_call_combined_fn): Likewise.
5586         * gimplify.c (gimplify_call_expr): Likewise.
5587         (gimple_boolify): Likewise.
5588         (gimplify_modify_expr): Likewise.
5589         (gimplify_addr_expr): Likewise.
5590         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
5591         * ipa-cp.c (determine_versionability): Likewise.
5592         * ipa-fnsummary.c (compute_fn_summary): Likewise.
5593         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
5594         * ipa-split.c (visit_bb): Likewise.
5595         (split_function): Likewise.
5596         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5597         * lto-cgraph.c (input_node): Likewise.
5598         * lto-streamer-out.c (write_symbol): Likewise.
5599         * omp-low.c (setjmp_or_longjmp_p): Likewise.
5600         (lower_omp_1): Likewise.
5601         * predict.c (strip_predict_hints): Likewise.
5602         * print-tree.c (print_node): Likewise.
5603         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
5604         * trans-mem.c (is_tm_irrevocable): Likewise.
5605         (is_tm_load): Likewise.
5606         (is_tm_simple_load): Likewise.
5607         (is_tm_store): Likewise.
5608         (is_tm_simple_store): Likewise.
5609         (is_tm_abort): Likewise.
5610         (tm_region_init_1): Likewise.
5611         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
5612         * tree-cfg.c (verify_gimple_call): Likewise.
5613         (move_stmt_r): Likewise.
5614         (stmt_can_terminate_bb_p): Likewise.
5615         * tree-eh.c (lower_eh_constructs_2): Likewise.
5616         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
5617         * tree-inline.c (remap_gimple_stmt): Likewise.
5618         (copy_bb): Likewise.
5619         (estimate_num_insns): Likewise.
5620         (fold_marked_statements): Likewise.
5621         * tree-sra.c (scan_function): Likewise.
5622         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
5623         (optimize_stack_restore): Likewise.
5624         (pass_fold_builtins::execute): Likewise.
5625         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
5626         (mark_all_reaching_defs_necessary_1): Likewise.
5627         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
5628         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5629         (pass_forwprop::execute): Likewise.
5630         * tree-ssa-loop-im.c (stmt_cost): Likewise.
5631         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
5632         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
5633         * tree-ssa-strlen.c (get_string_length): Likewise.
5634         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
5635         (find_func_aliases_for_call): Likewise.
5636         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
5637         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
5638         * tree-tailcall.c (find_tail_calls): Likewise.
5639         * tree.c (need_assembler_name_p): Likewise.
5640         (free_lang_data_in_decl): Likewise.
5641         (get_call_combined_fn): Likewise.
5642         * ubsan.c (is_ubsan_builtin_p): Likewise.
5643         * varasm.c (incorporeal_function_p): Likewise.
5644         * tree.h (DECL_BUILT_IN): Remove and replace with
5645         fndecl_built_in_p.
5646         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
5647         (fndecl_built_in_p): New.
5649 2018-08-27  Martin Liska  <mliska@suse.cz>
5651         PR tree-optimization/86847
5652         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
5653         Dump also subtree probability.
5654         (switch_decision_tree::do_jump_if_equal): New function.
5655         (switch_decision_tree::emit_case_nodes): Handle special
5656         situations in balanced tree that can be emitted much simpler.
5657         Fix calculation of probabilities that happen in tree expansion.
5658         * tree-switch-conversion.h (struct cluster): Add
5659         is_single_value_p.
5660         (struct simple_cluster): Likewise.
5661         (struct case_tree_node): Add new function has_child.
5662         (do_jump_if_equal): New.
5664 2018-08-27  Martin Liska  <mliska@suse.cz>
5666         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
5667         Add new argument to bit_test_cluster constructor.
5668         (bit_test_cluster::emit): Set bits really number of values
5669         handlel by a test.
5670         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
5671         probability argument.
5672         * tree-switch-conversion.h (struct bit_test_cluster):
5673         Add m_handles_entire_switch.
5675 2018-08-27  Martin Liska  <mliska@suse.cz>
5677         PR tree-optimization/86702
5678         * tree-switch-conversion.c (jump_table_cluster::emit):
5679         Make probabilities even for values in jump table
5680         according to number of cases handled.
5681         (switch_decision_tree::compute_cases_per_edge): Pass
5682         argument to reset_out_edges_aux function.
5683         (switch_decision_tree::analyze_switch_statement): Likewise.
5684         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
5685         Make it static.
5687 2018-08-27  Martin Liska  <mliska@suse.cz>
5689         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
5690         cfun argument explicitly.
5691         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
5692         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
5693         function gimple_switch_default_bb.
5694         (convert_switch_statements):
5695         (expand_builtins):
5696         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
5697         * stmt.c (label_to_block_fn): Use label_to_block and pass
5698         cfun argument explicitly and use gimple_switch_label_bb.
5699         (expand_case): Likewise.
5700         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
5701         cfun argument explicitly. Likewise.
5702         (make_edges_bb): Likewise.
5703         (make_cond_expr_edges): Likewise.
5704         (get_cases_for_edge): Likewise.
5705         (make_gimple_switch_edges): Likewise.
5706         (label_to_block_fn): Likewise.
5707         (label_to_block): Likewise.
5708         (make_goto_expr_edges): Likewise.
5709         (make_gimple_asm_edges): Likewise.
5710         (main_block_label): Likewise.
5711         (group_case_labels_stmt): Likewise.
5712         (find_taken_edge_computed_goto): Likewise.
5713         (find_taken_edge_switch_expr): Likewise.
5714         (gimple_verify_flow_info): Likewise.
5715         (gimple_redirect_edge_and_branch): Likewise.
5716         (gimple_switch_label_bb): New function.
5717         (gimple_switch_default_bb): Likewise.
5718         (gimple_switch_edge): Likewise.
5719         (gimple_switch_default_edge): Likewise.
5720         * tree-cfg.h (label_to_block_fn): Remove and replace ...
5721         (label_to_block): ... with this.
5722         (gimple_switch_label_bb): New.
5723         (gimple_switch_default_bb): Likewise.
5724         (gimple_switch_edge): Likewise.
5725         (gimple_switch_default_edge): Likewise.
5726         * tree-cfgcleanup.c (convert_single_case_switch): Use
5727         new gimple functions and pass new argument to label_to_block.
5728         (cleanup_control_flow_bb):
5729         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
5730         cfun argument explicitly.
5731         (make_eh_edges): Likewise.
5732         (redirect_eh_dispatch_edge): Likewise.
5733         (lower_resx): Likewise.
5734         (lower_eh_dispatch): Likewise.
5735         (maybe_remove_unreachable_handlers): Likewise.
5736         (unsplit_eh): Likewise.
5737         (cleanup_empty_eh): Likewise.
5738         (verify_eh_edges): Likewise.
5739         (verify_eh_dispatch_edge): Likewise.
5740         * tree-ssa-dom.c (record_edge_info): Likewise.
5741         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
5742         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
5743         (thread_through_normal_block): Likewise.
5744         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
5745         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
5746         * tree-switch-conversion.c (switch_conversion::collect): Use new
5747         gimple functions.
5748         (switch_conversion::check_final_bb): Likewise.
5749         (switch_conversion::gather_default_values): Pass new argument
5750         to label_to_block.
5751         (switch_conversion::build_constructors): Likewise.
5752         (switch_decision_tree::compute_cases_per_edge): Use new
5753         gimple_switch_edge function.
5754         (switch_decision_tree::analyze_switch_statement): Pass new argument
5755         to label_to_block.
5756         (switch_decision_tree::try_switch_expansion): Use
5757         gimple_switch_default_edge.
5758         * tree-vrp.c (find_switch_asserts): Pass new argument
5759         to label_to_block.
5760         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
5761         (vr_values::simplify_switch_using_ranges): Likewise.
5763 2018-08-27  Richard Biener  <rguenther@suse.de>
5765         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
5766         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
5768         * tree-ssa-sccvn.h (struct vn_pval): New structure.
5769         (struct vn_nary_op_s): Add unwind_to member.  Add
5770         predicated_values flag and put result into a union together
5771         with a linked list of vn_pval.
5772         (struct vn_ssa_aux): Add name member to make maintaining
5773         a map of SSA name to vn_ssa_aux possible.  Remove no longer
5774         needed info, dfsnum, low, visited, on_sccstack, use_processed
5775         and range_info_anti_range_p members.
5776         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
5777         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
5778         New functions.
5779         (vn_valueize): New global.
5780         (vn_context_bb): Likewise.
5781         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
5782         VN_INFO_PTR_INFO): Remove.
5783         * tree-ssa-sccvn.c: ... (rewrite)
5784         (pass_fre::execute): For -O2+ initialize loops and run
5785         RPO VN in optimistic mode (iterating).  For -O1 and -Og
5786         run RPO VN in non-optimistic mode.
5787         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
5788         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
5789         * doc/invoke.texi (sccvn-max-scc-size): Remove.
5790         (rpo-vn-max-loop-depth): Document.
5791         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
5792         when valuezing the VUSE signals we walked out of the region.
5793         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
5794         (phi_translate): Set VN context block to use for availability
5795         lookup.
5796         (compute_avail): Likewise.
5797         (pre_valueize): New function.
5798         (pass_pre::execute): Adjust to the RPO VN API.
5800         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
5801         (propagate_constants_for_unrolling): Remove.
5802         (tree_unroll_loops_completely): Perform value-numbering
5803         on the unrolled bodies loop parent.
5805 2018-08-27  Richard Biener  <rguenther@suse.de>
5807         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
5808         for partial antic compute.
5810 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
5812         PR rtl-optimization/87065
5813         * combine.c (simplify_if_then_else): Formatting fix.
5814         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
5815         check.
5816         (known_cond): Don't return const_true_rtx for vector modes.  Use
5817         CONST0_RTX instead of const0_rtx.  Formatting fixes.
5819 2018-08-27  Martin Liska  <mliska@suse.cz>
5821         PR gcov-profile/87069
5822         * gcov.c (process_file): Record files already processed
5823         and warn about a file being processed multiple times.
5825 2018-08-27  Martin Liska  <mliska@suse.cz>
5827         PR driver/83193
5828         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5829         Set default values for x_aarch64_*_string strings.
5830         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
5831         prefix.  For -mabi do not print '=ABI' in help and use
5832         <option_value> format for -msve-vector-bits and -moverride
5833         options.
5835 2018-08-26  Jeff Law  <law@redhat.com>
5837         * config/mips/frame-header-opt.c: Include "backend.h" rather than
5838         "cfg.h"
5840 2018-08-26  Marek Polacek  <polacek@redhat.com>
5842         PR c++/87029, Implement -Wredundant-move.
5843         * doc/invoke.texi: Document -Wredundant-move.
5845 2018-08-25  Martin Sebor  <msebor@redhat.com>
5847         PR tree-optimization/87059
5848         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
5849         to the same type as the other.
5850         * fold-const.c (fold_binary_loc): Assert expectation.
5852 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
5854         * config/darwin.c (machopic_legitimize_pic_address): Clean up
5855         extraneous parentheses, dead code section and formatting.
5857 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
5859         PR c++/87091
5860         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
5861         wide enough for jumps in the line-numbering to be visible.
5862         (layout::print_gap_in_line_numbering): New member function.
5863         (layout::calculate_line_spans): When using line numbering, merge
5864         line spans that are only 1 line apart.
5865         (diagnostic_show_locus): When printing line numbers, show gaps in
5866         line numbering directly, rather than printing headers.
5867         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
5868         line-numbering with multiple line spans.
5869         (selftest::test_fixit_insert_containing_newline_2): Add test of
5870         line-numbering, in which the spans are close enough to be merged.
5872 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5874         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
5875         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
5876         * tree-vrp.c (range_is_nonnull): Remove.
5877         (range_includes_zero_p): Accept value_range instead of min/max.
5878         (extract_range_from_binary_expr_1): Do not early bail on
5879         POINTER_PLUS_EXPR.
5880         Use range_includes_zero_p instead of range_is_nonnull.
5881         (extract_range_from_unary_expr): Use range_includes_zero_p instead
5882         of range_is_nonnull.
5883         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
5884         special case VR_ANTI_RANGE.
5885         (vrp_finalize): Same.
5886         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
5887         instead of min/max.
5888         (range_is_nonnull): Remove.
5889         * vr-values.c (vrp_stmt_computes_nonzero): Use
5890         range_includes_zero_p instead of range_is_nonnull.
5891         (extract_range_basic): Pass value_range to range_includes_zero_p
5892         instead of range_is_nonnull.
5894 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
5896         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
5897         * emit-rtl.h (rtl_data): Remove return_bnd.
5898         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
5899         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
5900         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
5901         (POINTER_BOUNDS_MODE): Remove definition.
5902         (make_pointer_bounds_mode): Remove.
5903         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
5904         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
5905         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
5906         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
5907         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
5908         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
5909         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
5911         * config/i386/i386-modes.def (BND32, BND64): Remove.
5912         * config/i386/i386.c (dbx_register_map): Remove bound registers.
5913         (dbx64_register_map): Ditto.
5914         (svr4_dbx_register_map): Ditto.
5915         (indirect_thunk_bnd_needed): Remove.
5916         (indirect_thunks_bnd_used): Ditto.
5917         (indirect_return_bnd_needed): Ditto.
5918         (indirect_return_via_cx_bnd): Ditto.
5919         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
5920         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
5921         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
5922         (output_indirect_thunk_function): Remove handling of
5923         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
5924         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
5925         (ix86_save_reg): Remove handling of crtl->return_bnd.
5926         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
5927         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
5928         and UNSPEC_BNDLX_ADDR.
5929         (ix86_output_indirect_branch_via_reg): Remove handling of
5930         indirect_thunk_prefix_bnd.
5931         (ix86_output_indirect_branch_via_push): Ditto.
5932         (ix86_output_function_return): Ditto.
5933         (ix86_output_indirect_function_return): Ditto.
5934         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
5935         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
5936         (CALL_USED_REGISTERS): Ditto.
5937         (REG_ALLOC_ORDER): Update for removal of bound registers.
5938         (HI_REGISTER_NAMES): Ditto.
5939         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
5940         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
5941         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
5942         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
5943         (FIRST_PSEUDO_REG): Update.
5944         (BND): Remove mode iterator.
5945         * config/i386/predicates.md (bnd_mem_operator): Remove.
5947 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5949         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
5950         vectors.
5952 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5954         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
5955         the case in which the permute needs only a single element and
5956         repeats for every vector of the result.  Extend that case to
5957         handle variable-length vectors.
5958         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5960 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
5962         PR debug/79342
5963         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
5964         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
5966 2018-08-24  Richard Biener  <rguenther@suse.de>
5968         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
5969         bb_flags_allocated members.
5970         (auto_flag): New RAII class for allocating flags.
5971         (auto_edge_flag): New RAII class for allocating edge flags.
5972         (auto_bb_flag): New RAII class for allocating bb flags.
5973         * cfgloop.c (verify_loop_structure): Allocate temporary edge
5974         flag dynamically.
5975         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
5976         in favor of temporarily allocated BB flag.
5977         * hsa-brig.c: Re-order includes.
5978         * hsa-dump.c: Likewise.
5979         * hsa-regalloc.c: Likewise.
5980         * print-rtl.c: Likewise.
5981         * profile-count.c: Likewise.
5983 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
5985         PR target/86989
5986         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
5987         the TOC register.
5989 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5991         PR 87073/bootstrap
5992         * wide-int-range.cc (wide_int_range_div): Do not ignore result
5993         from wide_int_range_multiplicative_op.
5995 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5997         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
5998         "permutaion".
6000 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
6002         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
6003         to 'expanded'.
6005 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
6007         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
6008         full GENERIC expression used for replacement.
6010 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
6012         * tree-vrp.c (abs_extent_range): Remove.
6013         (extract_range_into_wide_ints): Pass wide ints by reference.
6014         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
6015         Pass wide ints by reference in all calls to
6016         extract_range_into_wide_ints.
6017         * wide-int-range.cc (wide_int_range_div): New.
6018         * wide-int-range.h (wide_int_range_div): New.
6019         (wide_int_range_includes_zero_p): New.
6020         (wide_int_range_zero_p): New.
6022 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
6024         * config/aarch64/aarch64.md (arches): New enum.
6025         (arch): New enum attr.
6026         (arch_enabled): New attr.
6027         (enabled): Now uses arch_enabled only.
6028         (simd, sve, fp16): Removed attribute.
6029         (fp): Attr now defined in terms of 'arch'.
6030         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
6031         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
6032         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
6033         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
6034         attributes into 'arch'.
6035         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
6036         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
6037         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
6038         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
6039         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
6040         'simd' attribute into 'arch'.
6041         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
6042         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
6043         Convert use of 'fp' attribute to 'arch'.
6044         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
6045         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
6046         into 'arch'.
6047         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
6048         (different modes) Merge 'fp' and 'simd' into 'arch'.
6049         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
6050         'simd' into 'arch'.
6052 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
6054         PR rtl-optimization/87026
6055         * expmed.c (canonicalize_comparison): If we can no longer create
6056         pseudoregisters, don't.
6058 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
6060         PR target/86951
6061         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
6062         prototype.
6063         * config/arm/arm.c (speculation_barrier_libfunc): New static
6064         variable.
6065         (arm_init_libfuncs): Initialize it.
6066         (arm_emit_speculation_barrier): New function.
6067         * config/arm/arm.md (speculation_barrier): Call
6068         arm_emit_speculation_barrier for architectures that do not have
6069         DSB or ISB.
6070         (speculation_barrier_insn): Only match on Armv7 or later.
6072 2018-08-23  Richard Biener  <rguenther@suse.de>
6074         PR middle-end/87024
6075         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
6076         calls.
6078 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6080         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
6081         of single-vector TBLs.
6082         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
6083         one input is given.
6085 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
6087         PR target/85910
6088         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
6089         aarch64_evpc_tbl guard.
6091 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6093         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6094         behaviour.
6096 2018-08-22  Martin Sebor  <msebor@redhat.com>
6098         PR middle-end/87052
6099         * tree-pretty-print.c (pretty_print_string): Add argument.
6100         (dump_generic_node): Call to pretty_print_string with string size.
6102 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
6104         PR rtl-optimization/86771
6105         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
6106         of two SETs into those two SETs, one to be placed at i2, if that SETs
6107         destination is modified between i2 and i3.
6109 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6111         PR tree-optimization/86725
6112         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
6113         function.
6114         (vect_analyze_scalar_cycles_1): Check it.
6116 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6118         PR tree-optimization/86725
6119         * tree-vect-loop.c (vect_is_simple_reduction): When treating
6120         an outer loop phi as a double reduction, make sure that the
6121         single user of the phi result is an inner loop phi.
6123 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6125         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
6126         grouped stores with gaps to a strided group.
6128 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
6130         * tree-vect-stmts.c (get_group_load_store_type)
6131         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
6132         first statement in a group.
6134 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6136         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
6137         the sequence used in gcc/gcc.c.
6139 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6141         PR other/704
6142         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
6143         building it.
6145 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6147         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
6148         Darwin10-specific unwinder-shim.
6149         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
6150         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
6151         New to cater for Darwin10 Rosetta.
6153 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6155         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
6156         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
6158 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
6160         PR bootstrap/81033
6161         PR target/81733
6162         PR target/52795
6163         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
6164         (dwarf2out_switch_text_section): Generate a local label for the
6165         second function sub-section and apply it as the second FDE start
6166         label.
6167         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
6168         second sub-section start.
6170 2018-08-22  Richard Biener  <rguenther@suse.de>
6172         PR tree-optimization/86988
6173         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
6175 2018-08-22  Richard Biener  <rguenther@suse.de>
6177         PR tree-optimization/86945
6178         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
6180 2018-08-22  Alexandre Oliva <oliva@adacore.com>
6182         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
6183         a comment about how uses of r2 for .sdata2 come about.
6185 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
6187         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
6189 2018-08-21  Marek Polacek  <polacek@redhat.com>
6191         PR c++/86981, Implement -Wpessimizing-move.
6192         * doc/invoke.texi: Document -Wpessimizing-move.
6194 2018-08-21  Jan Hubicka  <jh@suse.cz>
6196         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
6197         * tree.h (is_redundant_typedef): Remove.
6198         * dwarf2out.c (is_redundant_typedef): Turn into static function.
6200 2018-08-21  Jan Hubicka  <jh@suse.cz>
6202         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
6203         when possible.
6205 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
6207         * expmed.c (extract_low_bits): Reject invalid subregs early.
6209 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6211         PR middle-end/86121
6212         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
6213         behaviour.
6215 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6217         * config/vxworks.h: Guard vxworks_asm_out_constructor and
6218         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
6219         * config/vxworks.c: Likewise.
6221 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6223         * config/vxworks.c: Set targetm.have_ctors_dtors
6224         if HAVE_INITFINI_ARRAY_SUPPORT.
6225         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
6226         if HAVE_INITFINI_ARRAY_SUPPORT.
6228 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
6230         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
6231         default search path for VxWorks < 7.
6233 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6235         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
6236         (get_format_string): Refer to c_getstr.
6238 2018-08-21  Tom de Vries  <tdevries@suse.de>
6240         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6241         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6242         (debuginfo_early_stop): Declare.
6243         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
6244         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
6245         (debuginfo_early_stop): New function.
6246         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
6247         and debuginfo_early_stop.
6248         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
6249         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
6250         (general_init): Call debuginfo_early_init.
6251         (finalize): Call debuginfo_fini.
6252         (do_compile): Call debuginfo_init.
6253         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
6254         -fdump-early-debug.
6255         (@item -fdump-debug, @item -fdump-earlydebug): Add.
6257 2018-08-21  Tom de Vries  <tdevries@suse.de>
6259         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6260         flag_dump_noaddr and flag_dump_unnumbered.
6262 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
6264         * wide-int-range.cc (wide_int_range_abs): New.
6265         (wide_int_range_order_set): Rename from wide_int_range_min_max.
6266         * wide-int-range.h (wide_int_range_abs): New.
6267         (wide_int_range_min_max): New.
6268         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6269         case to call wide_int_range_abs.
6270         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6271         (extract_range_from_abs_expr): Delete.
6273 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
6275         PR target/87033
6276         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6277         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6278         for indexed loads.
6280 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6281             Jeff Law <law@redhat.com>
6283         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6284         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6286 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
6288         PR other/84889
6289         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6290         (decl_attributes): Likewise.
6291         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6292         instance.
6293         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6294         * diagnostic-core.h (class auto_diagnostic_group): New class.
6295         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6296         (diagnostic_report_diagnostic): Handle the first diagnostics within
6297         a group.
6298         (emit_diagnostic): Add auto_diagnostic_group instance.
6299         (inform): Likewise.
6300         (inform_n): Likewise.
6301         (warning): Likewise.
6302         (warning_at): Likewise.
6303         (warning_n): Likewise.
6304         (pedwarn): Likewise.
6305         (permerror): Likewise.
6306         (error): Likewise.
6307         (error_n): Likewise.
6308         (error_at): Likewise.
6309         (sorry): Likewise.
6310         (fatal_error): Likewise.
6311         (internal_error): Likewise.
6312         (internal_error_no_backtrace): Likewise.
6313         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6314         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6315         * diagnostic.h (struct diagnostic_context): Add fields
6316         "diagnostic_group_nesting_depth",
6317         "diagnostic_group_emission_count", "begin_group_cb",
6318         "end_group_cb".
6319         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6320         Add auto_diagnostic_group instance(s).
6321         (find_explicit_erroneous_behavior): Likewise.
6322         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6323         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6324         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6325         (gimplify_va_arg_expr): Likewise.
6326         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6327         (HSA_SORRY_AT): Likewise.
6328         * ipa-devirt.c (compare_virtual_tables): Likewise.
6329         (warn_odr): Likewise.
6330         * multiple_target.c (expand_target_clones): Likewise.
6331         * opts-common.c (cmdline_handle_error): Likewise.
6332         * reginfo.c (globalize_reg): Likewise.
6333         * substring-locations.c (format_warning_n_va): Likewise.
6334         * tree-inline.c (expand_call_inline): Likewise.
6335         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6336         * tree-ssa-loop-niter.c
6337         (do_warn_aggressive_loop_optimizations): Likewise.
6338         * tree-ssa-uninit.c (warn_uninit): Likewise.
6339         * tree.c (warn_deprecated_use): Likewise.
6341 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
6343         PR target/87014
6344         * config/i386/i386.md (eh_return): Always update EH return
6345         address in word_mode.
6347 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
6349         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6350         TARGET_SPLIT_COMPLEX_ARG is defined.
6352 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6354         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6356 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6358         PR target/86984
6359         * expr.c (expand_assignment): Assert that bitpos is positive.
6360         (store_field): Likewise
6361         (expand_expr_real_1): Make sure that bitpos is positive.
6362         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6363         integer overflow.
6365 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6367         * Makefile.in (CPP_ID_DATA_H): Delete.
6368         (CPP_INTERNAL_H): Don't add it.
6369         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6370         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6372 2018-08-20  Richard Biener  <rguenther@suse.de>
6374         PR tree-optimization/78655
6375         * tree-vrp.c (extract_range_from_binary_expr_1): Make
6376         pointer + offset nonnull if either operand is nonnull work.
6378 2018-08-20  Tom de Vries  <tdevries@suse.de>
6380         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6381         unless the referenced die describes the added property using
6382         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
6383         Otherwise, add a DW_AT_location to the referenced die.
6385 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
6387         PR target/86994
6388         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6389         register_operand when calling ix86_set_reg_reg_cost.
6390         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6391         Set *total to 0 for operands that satisfy x86_64_immediate_operand
6392         predicate and to 1 otherwise.
6394 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
6396         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6397         emit a diagnostic that it is not supported and reset the option.
6398         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6399         supported and consume the option.  (ASM_FINAL_SPEC): New.
6401 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6403         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6404         a sentence.
6406 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
6408         C-SKY port: Documentation
6410         * doc/extend.texi (C-SKY Function Attributes): New section.
6411         * doc/invoke.texi (Option Summary): Add C-SKY options.
6412         (C-SKY Options): New section.
6413         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6415 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6416             Huibin Wang  <huibin_wang@c-sky.com>
6417             Sandra Loosemore  <sandra@codesourcery.com>
6418             Chung-Lin Tang  <cltang@codesourcery.com>
6420         C-SKY port: Backend implementation
6422         * config/csky/*: New.
6423         * common/config/csky/*: New.
6425 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6426             Huibin Wang  <huibin_wang@c-sky.com>
6427             Sandra Loosemore  <sandra@codesourcery.com>
6428             Chung-Lin Tang  <cltang@codesourcery.com>
6429             Andrew Jenner  <andrew@codesourcery.com>
6431         C-SKY port: Configury
6433         * config.gcc (csky-*-*): New.
6434         * configure.ac: Add csky to targets for dwarf2 debug_line support.
6435         * configure: Regenerated.
6437 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
6439         * dump-context.h: Include "dumpfile.h".
6440         (dump_context::dump_printf_va): Convert final param from va_list
6441         to va_list *.  Convert from ATTRIBUTE_PRINTF to
6442         ATTRIBUTE_GCC_DUMP_PRINTF.
6443         (dump_context::dump_printf_loc_va): Likewise.
6444         * dumpfile.c: Include "stringpool.h".
6445         (make_item_for_dump_printf_va): Delete.
6446         (make_item_for_dump_printf): Delete.
6447         (class dump_pretty_printer): New class.
6448         (dump_pretty_printer::dump_pretty_printer): New ctor.
6449         (dump_pretty_printer::emit_items): New member function.
6450         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
6451         function.
6452         (dump_pretty_printer::emit_item): New member function.
6453         (dump_pretty_printer::stash_item): New member function.
6454         (dump_pretty_printer::format_decoder_cb): New member function.
6455         (dump_pretty_printer::decode_format): New member function.
6456         (dump_context::dump_printf_va): Reimplement in terms of
6457         dump_pretty_printer.
6458         (dump_context::dump_printf_loc_va): Convert final param from va_list
6459         to va_list *.
6460         (dump_context::begin_scope): Reimplement call to
6461         make_item_for_dump_printf.
6462         (dump_printf): Update for change to dump_printf_va.
6463         (dump_printf_loc): Likewise.
6464         (selftest::test_capture_of_dump_calls): Convert "stmt" from
6465         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
6466         with %T, %E, and %G.
6467         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
6468         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
6469         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
6470         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
6471         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
6472         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
6473         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
6474         within a dump_printf_loc call to "%wu".
6475         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
6476         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
6477         missing space after "=".
6478         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
6479         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
6480         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
6481         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
6482         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
6483         duplicate "vectorized" from message.
6485 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
6487         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
6488         polyNxK_t element's TYPE_STRING_FLAG.
6490 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6492         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
6493         (they were unnamed before).  Fix comments.
6495 2018-08-17  Nathan Sidwell  <nathan@acm.org>
6497         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
6499 2018-08-17  Richard Biener  <rguenther@suse.de>
6501         PR tree-optimization/86841
6502         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6504 2018-08-17  Martin Liska  <mliska@suse.cz>
6506         * common.opt: Remove Warn, Init and Report for options with
6507         Ignore/Deprecated flag. Warning is done automatically for
6508         Deprecated flags.
6509         * config/i386/i386.opt: Likewise.
6510         * config/ia64/ia64.opt: Likewise.
6511         * config/rs6000/rs6000.opt: Likewise.
6512         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6513         Remove usage of flag_check_pointer_bounds.
6514         * lto-wrapper.c (merge_and_complain): Do not handle
6515         OPT_fcheck_pointer_bounds.
6516         (append_compiler_options): Likewise.
6517         * opt-functions.awk: Do not handle Deprecated.
6518         * optc-gen.awk: Check that Var, Report and Init are not
6519         used for an option with Ignore/Deprecated flag.
6520         * opts-common.c (decode_cmdline_option): Do not report
6521         CL_ERR_DEPRECATED.
6522         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
6523         options.
6524         * opts.h (struct cl_option): Remove cl_deprecated flag.
6525         (CL_ERR_DEPRECATED): Remove error enum value.
6527 2018-08-17  Richard Biener  <rguenther@suse.de>
6529         PR middle-end/86505
6530         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
6531         across a va-arg-pack using call adjust its return value accordingly.
6533 2018-08-16  Martin Sebor  <msebor@redhat.com>
6535         PR tree-optimization/86853
6536         * gimple-ssa-sprintf.c (struct format_result): Rename member.
6537         (struct fmtresult): Add member and initialize it in ctors.
6538         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
6539         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
6540         (format_directive): Set POSUNDER4K when MAYFAIL is set.
6541         (parse_directive): Handle %C same as %c.
6542         (sprintf_dom_walker::compute_format_length): Adjust.
6543         (is_call_safe): Adjust.
6545 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6547         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
6548         for determining how to count the elements.
6549         * builtins.h (c_strlen): Adjust prototype.
6550         * expr.c (string_constant): Add new parameter mem_size.
6551         Set *mem_size appropriately.
6552         * expr.h (string_constant): Adjust protoype.
6553         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
6554         * gimple-fold.h (get_range_strlen): Adjust prototype.
6555         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
6556         (format_string): Call get_string_length with eltsize.
6558 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6560         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
6561         to emit the span, rather than setting it as the prefix.
6563 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6565         * diagnostic-show-locus.c (layout::start_annotation_line): Add
6566         "margin_char" parameter, defaulting to space.  Use it in place
6567         of pp_space for the initial part of the margin.
6568         (layout::print_leading_fixits): Use '+' when filling the margin
6569         of line-insertion fix-it hints.
6571 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6573         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
6574         Delete.
6576 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6578         * config/rs6000/altivec.md: Don't set length attribute to the default
6579         value.
6580         * config/rs6000/darwin.md: Ditto.
6581         * config/rs6000/dfp.md: Ditto.
6582         * config/rs6000/htm.md: Ditto.
6583         * config/rs6000/rs6000.md: Ditto.
6584         * config/rs6000/sync.md: Ditto.
6585         * config/rs6000/vsx.md: Ditto.
6587 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6589         * config/rs6000/altivec.md: Don't set length attribute to the default
6590         value, for branch instructions.
6591         * config/rs6000/darwin.md: Ditto.
6592         * config/rs6000/rs6000.md: Ditto.
6594 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6596         * config/rs6000/rs6000.md (length): Always define as const_int 4.
6597         (unnamed conditional branch define_insn): Set length to 4 or 8
6598         depending on offset.
6599         (<bd>_<mode>): Similar, for alternative 0.
6600         (<bd>tf_<mode>): Ditto.
6602 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6604         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
6606 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
6608         * doc/rtl.texi: Replace old RTX class names with new names.
6611 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
6613         * expmed.h (canonicalize_comparison): New declaration.
6614         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
6615         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
6616         * optabs.c (prepare_cmp_insn): Likewise.
6617         * rtl.h (unsigned_condition_p): New function which checks if a
6618         comparison operator is unsigned.
6620 2018-08-16  Nathan Sidwell  <nathan@acm.org>
6622         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
6623         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
6625 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6627         PR target/84711
6628         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
6629         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
6630         (mov<mov>): ..this and enable unconditionally.
6632 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6634         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
6636 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
6638         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
6639         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
6640         "Common" with "Target".
6642 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
6644         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
6645         * doc/invoke.texi (mmitigate-rop): Remove.
6646         * config/i386/i386.c: Do not include "regrename.h".
6647         (ix86_rop_should_change_byte_p, reg_encoded_number)
6648         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
6649         Remove.
6650         (ix86_reorg): Remove call to ix86_mitigate_rop.
6651         * config/i386/i386.md (attr "modrm_class"): Remove.
6652         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
6653         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
6654         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
6656 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
6658         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
6659         allow folding of mergeh() and mergel() for the float and double types.
6660         (fold_mergehl_helper): Rework to handle building a permute tree
6661         for float vectors.
6663 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
6665         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
6666         for TARGET_SSE.
6668 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
6670         * common.opt (fdiagnostics-show-labels): New option.
6671         * diagnostic-show-locus.c (class layout_range): Add field
6672         "m_label".
6673         (class layout): Add field "m_show_labels_p".
6674         (layout_range::layout_range): Add param "label" and use it to
6675         initialize m_label.
6676         (make_range): Pass in NULL for new "label" param of layout_range's
6677         ctor.
6678         (layout::layout): Initialize m_show_labels_p.
6679         (layout::maybe_add_location_range): Pass in loc_range->m_label
6680         when constructing layout_range instances.
6681         (struct line_label): New struct.
6682         (layout::print_any_labels): New member function.
6683         (layout::print_line): Call it if label-printing is enabled.
6684         (selftest::test_one_liner_labels): New test.
6685         (selftest::test_diagnostic_show_locus_one_liner): Call it.
6686         * diagnostic.c (diagnostic_initialize): Initialize
6687         context->show_labels_p.
6688         * diagnostic.h (struct diagnostic_context): Add field
6689         "show_labels_p".
6690         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6691         -fno-diagnostics-show-labels.
6692         * dwarf2out.c (gen_producer_string): Add
6693         OPT_fdiagnostics_show_labels to the ignored options.
6694         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
6695         param.
6696         (gcc_rich_location::maybe_add_expr): Likewise.
6697         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
6698         label" param, defaulting to NULL.
6699         (gcc_rich_location::add_expr): Add "label" param.
6700         (gcc_rich_location::maybe_add_expr): Likewise.
6701         (class text_range_label): New class.
6702         (class range_label_for_type_mismatch): New class.
6703         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
6704         of format_warning_va.
6705         (fmtwarn_n): Likewise for new params of format_warning_n_va.
6706         * lto-wrapper.c (merge_and_complain): Add
6707         OPT_fdiagnostics_show_labels to the "pick one setting" options.
6708         (append_compiler_options): Likewise to the dropped options.
6709         (append_diag_options): Likewise to the passed-on options.
6710         * opts.c (common_handle_option): Handle the new option.
6711         * selftest-diagnostic.c
6712         (test_diagnostic_context::test_diagnostic_context): Enable
6713         show_labels_p.
6714         * substring-locations.c: Include "gcc-rich-location.h".
6715         (format_warning_n_va): Add "fmt_label" and "param_label" params
6716         and use them as appropriate.
6717         (format_warning_va): Add "fmt_label" and "param_label" params,
6718         passing them on to format_warning_n_va.
6719         (format_warning_at_substring): Likewise.
6720         (format_warning_at_substring_n): Likewise.
6721         * substring-locations.h (format_warning_va): Add "fmt_label" and
6722         "param_label" params.
6723         (format_warning_n_va): Likewise.
6724         (format_warning_at_substring): Likewise.
6725         (format_warning_at_substring_n): Likewise.
6726         * toplev.c (general_init): Initialize global_dc->show_labels_p.
6728 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
6730         PR testsuite/86519
6731         * builtins.c (expand_builtin_memcmp): Do not expand the call
6732         when overflow is detected.
6734 2018-08-15  Martin Sebor  <msebor@redhat.com>
6736         PR tree-optimization/71625
6737         * config/aarch64/aarch64-builtins.c
6738         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
6740 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
6742         * config/s390/s390.c (s390_reorg): Remove loop.
6744 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6746         * config/darwin.c
6747          (darwin_function_switched_text_sections): Delete.
6748         * gcc/config/darwin.h
6749          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
6751 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6753         PR target/81685
6754         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
6755         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
6756         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
6758 2018-08-15  Martin Liska  <mliska@suse.cz>
6760         PR tree-optimization/86925
6761         * predict.c (expr_expected_value_1): When taking
6762         later predictor, assign also probability.
6763         Use fold_build2_initializer_loc in order to fold
6764         the expression in -frounding-math.
6766 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
6768         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
6769         patterns.
6770         (expand_vec_perm_1): Try the new method.
6772 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6774         PR target/86547
6775         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
6776         Check whether lra_live_max_point is 0 before dividing.
6778 2018-08-14  Martin Sebor  <msebor@redhat.com>
6780         PR tree-optimization/86650
6781         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
6782         (vrp_prop::check_mem_ref): Same.
6784 2018-08-13  Liu Hao <lh_mouse@126.com>
6786         * pretty-print.c (eat_esc_sequence): Swap the foreground and
6787         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
6788         and clear it thereafter, as it only works for DBCS.
6790 2018-08-13  Liu Hao <lh_mouse@126.com>
6792         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
6793         handle returned by _get_osf_handle().
6795 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6797         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
6798         for folding vec_perm.
6800 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6802         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
6803         Add support for gimple-folding of vec_pack() and vec_unpack()
6804         intrinsics.
6806 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
6808         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
6809         vec_xst variants to the list.
6810         (rs6000_gimple_fold_builtin): Add support for folding unaligned
6811         vector loads and stores.
6813 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
6815         * config.gcc (rs6000-ibm-aix4.x): Delete.
6816         (rs6000-ibm-aix5.1): Delete.
6817         (rs6000-ibm-aix5.2): Delete.
6818         (rs6000-ibm-aix5.3): Delete.
6819         * config/rs6000/aix43.h: Delete.
6820         * config/rs6000/aix51.h: Delete.
6821         * config/rs6000/aix52.h: Delete.
6822         * config/rs6000/t-aix43: Delete.
6824 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
6826         * config/s390/s390.c (s390_decompose_constant_pool_ref):
6827         New function.
6828         (s390_decompose_address): Factor out constant pool ref
6829         decomposition.
6831 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6833         * config/nds32/nds32-predicates.c
6834         (nds32_can_use_bclr_p): Change return type as bool.
6835         (nds32_can_use_bset_p): Ditto.
6836         (nds32_can_use_btgl_p): Ditto.
6837         (nds32_can_use_bitci_p): Ditto.
6838         * config/nds32/nds32-protos.h
6839         (nds32_can_use_bclr_p): Change declaration.
6840         (nds32_can_use_bset_p): Ditto.
6841         (nds32_can_use_btgl_p): Ditto.
6842         (nds32_can_use_bitci_p): Ditto.
6844 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6846         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
6847         Support -msched-prolog-epilog option.
6848         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
6850 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6852         * common/config/nds32/nds32-common.c
6853         (nds32_option_optimization_table): Enalbe -malways-align.
6855 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6857         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
6858         extra_headers.
6859         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
6860         OPT_misr_secure_ case.
6861         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
6862         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
6863         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
6864         secure attribute.
6865         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
6866         (nds32_isr_info): New field security_level.
6867         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
6868         * config/nds32/nds32.md (return_internal): Consider critical attribute.
6869         * config/nds32/nds32.opt (misr-secure): New option.
6870         * config/nds32/nds32_init.inc: New file.
6871         * config/nds32/nds32_isr.h: New file.
6873 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
6875         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
6876         Update comment for atomic instructions.
6877         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
6878         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
6879         Remove.
6880         (atomic_loaddi): Revise fence expansion to only emit fence prior to
6881         load for __ATOMIC_SEQ_CST model.
6882         (atomic_loaddi_1): Remove float register target.
6883         (atomic_storedi): Handle CONST_INT values.
6884         (atomic_storedi_1): Remove float register source.  Add special case
6885         for zero value.
6886         (memory_barrier): New expander and insn.
6888 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
6890         PR tree-optimization/86835
6891         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
6892         new_stmt after def_gsi, make sure to insert new_square_stmt after
6893         that stmt, not 2 stmts before it.
6895 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
6897         PR target/82418
6898         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
6899         instead of SWI48.
6901 2018-08-10  Martin Liska  <mliska@suse.cz>
6903         PR target/83610
6904         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
6905         function type.
6906         * builtins.c (expand_builtin_expect_with_probability):
6907         New function.
6908         (expand_builtin_expect_with_probability): New function.
6909         (build_builtin_expect_predicate): Add new argumnet probability
6910         for BUILT_IN_EXPECT_WITH_PROBABILITY.
6911         (fold_builtin_expect):
6912         (fold_builtin_2):
6913         (fold_builtin_3):
6914         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
6915         * builtins.h (fold_builtin_expect): Set new argument.
6916         * doc/extend.texi: Document __builtin_expect_with_probability.
6917         * doc/invoke.texi: Likewise.
6918         * gimple-fold.c (gimple_fold_call): Pass new argument.
6919         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
6920         also BUILT_IN_EXPECT_WITH_PROBABILITY.
6921         * predict.c (get_predictor_value): New function.
6922         (expr_expected_value): Add new argument probability. Assume
6923         that predictor and probability are always non-null.
6924         (expr_expected_value_1): Likewise.  For __builtin_expect and
6925         __builtin_expect_with_probability set probability.  Handle
6926         combination in binary expressions.
6927         (tree_predict_by_opcode): Simplify code by simply calling
6928         get_predictor_value.
6929         (pass_strip_predict_hints::execute): Add handling of
6930         BUILT_IN_EXPECT_WITH_PROBABILITY.
6931         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
6932         new predictor.
6933         * tree.h (DECL_BUILT_IN_P): New function.
6935 2018-08-10  Martin Liska  <mliska@suse.cz>
6937         PR tree-optimization/85799
6938         * passes.def: Add argument for pass_strip_predict_hints.
6939         * predict.c (class pass_strip_predict_hints): Add new argument
6940         early_p.
6941         (strip_predictor_early): New function.
6942         (pass_strip_predict_hints::execute): Call the function to
6943         strip predictors.
6944         (strip_predict_hints): New function.
6945         * predict.def: Fix comment.
6947 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
6949         * Makefile.in: Clarify which tm.texi to copy over to assert the
6950         right to grant a GFDL license for all.
6952 2018-08-09  Jeff Law  <law@redhat.com>
6954         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
6955         unused variable.
6957 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
6959         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
6960         prototype.
6962 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6964         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
6965         reductions for variable-length vectors.
6967 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
6969         PR other/84889
6970         * common.opt (fdiagnostics-show-line-numbers): New option.
6971         * diagnostic-show-locus.c (class layout): Add fields
6972         "m_show_line_numbers_p" and "m_linenum_width";
6973         (num_digits): New function.
6974         (test_num_digits): New function.
6975         (layout::layout): Initialize new fields.  Update m_x_offset
6976         logic to handle any left margin.
6977         (layout::print_source_line): Print line number when requested.
6978         (layout::start_annotation_line): New member function.
6979         (layout::print_annotation_line): Call it.
6980         (layout::print_leading_fixits): Likewise.
6981         (layout::print_trailing_fixits): Likewise.  Update calls to
6982         move_to_column for new parameter.
6983         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
6984         it to potentially call start_annotation_line.
6985         (layout::show_ruler): Call start_annotation_line.
6986         (selftest::test_line_numbers_multiline_range): New selftest.
6987         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
6988         and selftest::test_line_numbers_multiline_range.
6989         * diagnostic.c (diagnostic_initialize): Initialize
6990         show_line_numbers_p.
6991         * diagnostic.h (struct diagnostic_context): Add field
6992         "show_line_numbers_p".
6993         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6994         -fno-diagnostics-show-line-numbers.
6995         * dwarf2out.c (gen_producer_string): Add
6996         OPT_fdiagnostics_show_line_numbers to the ignored options.
6997         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
6998         one setting" options.
6999         (append_compiler_options): Likewise to the dropped options.
7000         (append_diag_options): Likewise to the passed-on options.
7001         * opts.c (common_handle_option): Handle the new option.
7002         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
7004 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7006         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
7007         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
7008         third argument of type "const signed char" to descriptions of
7009         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
7010         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
7011         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
7012         __builtin_bcdsub_ov functions.
7014 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7016         PR tree-optimization/86858
7017         * tree-vect-loop.c (vect_is_simple_reduction): Restore
7018         flow_bb_inside_loop_p calls.
7020 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
7022         PR tree-optimization/86871
7023         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
7024         instead of gimple_assign_lhs.
7026 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
7028         PR target/86887
7029         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
7030         register constraint to operand 0.
7031         (add<mode>3_carryinC): Likewise.
7032         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
7034 2018-08-09  Martin Liska  <mliska@suse.cz>
7036         PR c/86895
7037         * common.opt: Remove extra line.
7039 2018-08-09  Martin Liska  <mliska@suse.cz>
7041         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
7042         at the end of a line, make first letter capital and end up
7043         a sentence with a dot.
7044         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
7045         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
7046         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
7047         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
7048         (PARAM_MAX_ISL_OPERATIONS): Likewise.
7049         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
7050         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
7051         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
7052         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
7053         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
7054         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
7055         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
7056         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
7057         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
7058         (PARAM_TREE_REASSOC_WIDTH): Likewise.
7059         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
7060         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
7061         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
7063 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
7065         PR target/84332
7066         * config/s390/s390.c (s390_option_override_internal): Reduce the
7067         stack-clash-protection-probe-interval param if it would be too big
7068         for z900.
7070 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
7072         PR target/46179
7073         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
7074         * config/m68k/m68k.c (handle_move_double): Don't call
7075         m68k_final_prescan_insn.
7076         (m68k_adjust_decorated_operand): Renamed from
7077         m68k_final_prescan_insn, remove first and third operand and
7078         simplify.
7079         (print_operand): Call it.
7080         (print_operand_address): Call it.
7082 2018-08-08  Nathan Sidwell  <nathan@acm.org>
7084         * diagnostic.c (diagnostic_report_current_module): Use
7085         linemap_included_from & linemap_included_from_linemap.
7087 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
7089         * config/aarch64/aarch64-cores.def: Add phecda core.
7090         * config/aarch64/aarch64-tune.md: Regenerate.
7091         * doc/invoke.texi: Add phecda core.
7093 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7095         PR target/85295
7096         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
7097         definitions.
7098         * config/s390/s390.md ("movti"): Add more alternatives for
7099         constant to GPR copies.
7101 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
7103         * config/s390/s390.c: Fix whitespace damage throughout the file.
7104         * config/s390/s390.h: Likewise.
7105         * config/s390/tpf.h: Likewise.
7107 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7109         * config/s390/s390.c (s390_loadrelative_operand_p):
7110         Remove TARGET_CPU_ZARCH usages.
7111         (s390_rtx_costs): Likewise.
7112         (s390_legitimate_constant_p): Likewise.
7113         (s390_cannot_force_const_mem): Likewise.
7114         (legitimate_reload_constant_p): Likewise.
7115         (s390_preferred_reload_class): Likewise.
7116         (legitimize_pic_address): Likewise.
7117         (legitimize_tls_address): Likewise.
7118         (s390_split_branches): Removed.
7119         (s390_add_execute): Removed.
7120         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
7121         (s390_mainpool_start): Likewise.
7122         (s390_mainpool_finish): Likewise.
7123         (s390_mainpool_cancel): Removed.
7124         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
7125         (s390_chunkify_cancel): Likewise.
7126         (s390_return_addr_rtx): Likewise.
7127         (s390_register_info): Remove split_branches_pending_p uages.
7128         (s390_optimize_register_info): Likewise.
7129         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
7130         split_branches_pending_p usages.
7131         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
7132         (s390_load_got): Likewise.
7133         (s390_expand_split_stack_prologue): Likewise.
7134         (output_asm_nops): Likewise.
7135         (s390_function_profiler): Likewise.
7136         (s390_emit_call): Likewise.
7137         (s390_conditional_register_usage): Likewise.
7138         (s390_optimize_prologue): Likewise.
7139         (s390_reorg): Remove TARGET_CPU_ZARCH and
7140         split_branches_pending_p usages.
7141         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
7142         usages.
7143         (s390_output_indirect_thunk_function): Likewise.
7144         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
7145         (TARGET_CPU_ZARCH_P): Removed.
7146         (struct machine_function): Remove split_branches_pending_p.
7147         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
7149 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
7151         * common/config/s390/s390-common.c (processor_flags_table):
7152         Remove flags.
7153         * config.gcc: Remove with_arch/with_tune support.
7154         * config/s390/2064.md: Remove cpu attribute comparisons.
7155         * config/s390/driver-native.c (s390_host_detect_local_cpu):
7156         Remove MTN.
7157         * config/s390/linux.h (ASM_SPEC):
7158         Remove -march support.
7159         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
7160         Use a table to get an arch level.
7161         * config/s390/s390-opts.h (enum processor_type):
7162         Remove enum values.
7163         * config/s390/s390.c
7164         (processor_table): Remove entries, add arch_level values.
7165         (s390_issue_rate): Remove cases.
7166         (s390_option_override): Adjust
7167         s390_option_override_internal() call.
7168         (s390_option_override_internal): Remove deprecation warning.
7169         (s390_valid_target_attribute_tree): Adjust
7170         s390_option_override_internal() call.
7171         * config/s390/s390.h (struct s390_processor):
7172         Share with s390-c.c, add arch_level field.
7173         * config/s390/s390.md:
7174         Remove occurrences in cpu attribute.
7175         * config/s390/s390.opt: Remove -march/-mtune support.
7176         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
7177         * doc/invoke.texi: Remove deprecation warning.
7179 2018-08-08  Luis Machado  <luis.machado@linaro.org>
7181         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
7182         global.
7183         (qdf24xx_tunings): Set vector cost structure to
7184         qdf24xx_vector_cost.
7186         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
7187         <register_sextend>: Set to 3.
7189 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
7191         PR target/86838
7192         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
7193         * config/aarch64/aarch64-simd.md
7194         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
7195         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
7196         (aarch64_frecpx<mode>): ...this new pattern.
7197         * config/aarch64/aarch64-simd-builtins.def: Remove comment
7198         about aarch64_frecp<FRECP:frecp_suffix><mode>.
7200 2018-08-07  Martin Liska  <mliska@suse.cz>
7202         PR middle-end/83023
7203         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
7204         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
7205         * predict.def (PRED_MALLOC_NONNULL): New predictor.
7206         * doc/extend.texi: Document that malloc attribute adds
7207         hit to compiler.
7209 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
7211         PR target/86785
7212         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7213         Define to speculation_safe_value_not_needed.
7215 2018-08-06  Jeff Law  <law@redhat.com>
7217         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
7218         the vr_values instance to cprop_into_stmt.
7219         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
7220         (cprop_operand): Also query EVRP to determine if OP is a constant.
7222 2018-08-06  Nathan Sidwell  <nathan@acm.org>
7224         * diagnostic.c (diagnostic_report_current_module): Reroll
7225         included-at loop.  Translate text.
7227 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
7229         * function-tests.c (selftest::test_expansion_to_rtl): Call
7230         free_after_compilation.
7232 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7234         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
7236 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
7238         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
7239         loops with memory block operations from getting unrolled.
7241 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
7243         PR target/86807
7244         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7245         Define to speculation_safe_value_not_needed.
7247 2018-08-06  Jeff Law  <law@redhat.com>
7249         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
7250         assert.
7252 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7254         PR target/86662
7255         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
7256         with all enabled __intN types.
7258         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7260 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7262         * alias.c (record_set): Check for clobber high.
7263         * cfgexpand.c (expand_gimple_stmt): Likewise.
7264         * combine-stack-adj.c (single_set_for_csa): Likewise.
7265         * combine.c (find_single_use_1): Likewise.
7266         (set_nonzero_bits_and_sign_copies): Likewise.
7267         (get_combine_src_dest): Likewise.
7268         (is_parallel_of_n_reg_sets): Likewise.
7269         (try_combine): Likewise.
7270         (record_dead_and_set_regs_1): Likewise.
7271         (reg_dead_at_p_1): Likewise.
7272         (reg_dead_at_p): Likewise.
7273         * dce.c (deletable_insn_p): Likewise.
7274         (mark_nonreg_stores_1): Likewise.
7275         (mark_nonreg_stores_2): Likewise.
7276         * df-scan.c (df_find_hard_reg_defs): Likewise.
7277         (df_uses_record): Likewise.
7278         (df_get_call_refs): Likewise.
7279         * dwarf2out.c (mem_loc_descriptor): Likewise.
7280         * haifa-sched.c (haifa_classify_rtx): Likewise.
7281         * ira-build.c (create_insn_allocnos): Likewise.
7282         * ira-costs.c (scan_one_insn): Likewise.
7283         * ira.c (equiv_init_movable_p): Likewise.
7284         (rtx_moveable_p): Likewise.
7285         (interesting_dest_for_shprep): Likewise.
7286         * jump.c (mark_jump_label_1): Likewise.
7287         * postreload-gcse.c (record_opr_changes): Likewise.
7288         * postreload.c (reload_cse_simplify): Likewise.
7289         (struct reg_use): Add source expr.
7290         (reload_combine): Check for clobber high.
7291         (reload_combine_note_use): Likewise.
7292         (reload_cse_move2add): Likewise.
7293         (move2add_note_store): Likewise.
7294         * print-rtl.c (print_pattern): Likewise.
7295         * recog.c (decode_asm_operands): Likewise.
7296         (store_data_bypass_p): Likewise.
7297         (if_test_bypass_p): Likewise.
7298         * regcprop.c (kill_clobbered_value): Likewise.
7299         (kill_set_value): Likewise.
7300         * reginfo.c (reg_scan_mark_refs): Likewise.
7301         * reload1.c (maybe_fix_stack_asms): Likewise.
7302         (eliminate_regs_1): Likewise.
7303         (elimination_effects): Likewise.
7304         (mark_not_eliminable): Likewise.
7305         (scan_paradoxical_subregs): Likewise.
7306         (forget_old_reloads_1): Likewise.
7307         * reorg.c (find_end_label): Likewise.
7308         (try_merge_delay_insns): Likewise.
7309         (redundant_insn): Likewise.
7310         (own_thread_p): Likewise.
7311         (fill_simple_delay_slots): Likewise.
7312         (fill_slots_from_thread): Likewise.
7313         (dbr_schedule): Likewise.
7314         * resource.c (update_live_status): Likewise.
7315         (mark_referenced_resources): Likewise.
7316         (mark_set_resources): Likewise.
7317         * rtl.c (copy_rtx): Likewise.
7318         * rtlanal.c (reg_referenced_p): Likewise.
7319         (single_set_2): Likewise.
7320         (noop_move_p): Likewise.
7321         (note_stores): Likewise.
7322         * sched-deps.c (sched_analyze_reg): Likewise.
7323         (sched_analyze_insn): Likewise.
7325 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7327         * cse.c (invalidate_reg): New function extracted from...
7328         (invalidate): ...here.
7329         (canonicalize_insn): Check for clobber high.
7330         (invalidate_from_clobbers): invalidate clobber highs.
7331         (invalidate_from_sets_and_clobbers): Likewise.
7332         (count_reg_usage): Check for clobber high.
7333         (insn_live_p): Likewise.
7334         * cselib.c (cselib_expand_value_rtx_1):Likewise.
7335         (cselib_invalidate_regno): Check for clobber in setter.
7336         (cselib_invalidate_rtx): Pass through setter.
7337         (cselib_invalidate_rtx_note_stores):
7338         (cselib_process_insn): Check for clobber high.
7339         * cselib.h (cselib_invalidate_rtx): Add operand.
7341 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7343         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7344         (mark_not_eliminable): Likewise.
7345         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7346         * lra-lives.c (process_bb_lives): Check for clobber high.
7347         * lra.c (new_insn_reg): Remember clobber highs.
7348         (collect_non_operand_hard_regs): Check for clobber high.
7349         (lra_set_insn_recog_data): Likewise.
7350         (add_regs_to_insn_regno_info): Likewise.
7351         (lra_update_insn_regno_info): Likewise.
7353 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7355         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7356         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7358 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7360         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7361         (copy_insn_1): Likewise.
7362         (gen_hard_reg_clobber_high): New gen function.
7363         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7364         * genemit.c (gen_exp): Likewise.
7365         (gen_emit_seq): Pass through info.
7366         (gen_insn): Check for CLOBBER_HIGH.
7367         (gen_expand): Pass through info.
7368         (gen_split): Likewise.
7369         (output_add_clobbers): Likewise.
7370         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7371         (remove_clobbers): Likewise.
7372         * rtl.h (gen_hard_reg_clobber_high): New declaration.
7374 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7376         * doc/rtl.texi (clobber_high): Add.
7377         (parallel): Add in clobber high
7378         * rtl.c (rtl_check_failed_code3): Add function.
7379         * rtl.def (CLOBBER_HIGH): Add expression.
7380         * rtl.h (RTL_CHECKC3): Add macro.
7381         (rtl_check_failed_code3): Add declaration.
7382         (XC3EXP): Add macro.
7384 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
7386         PR target/86386
7387         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7388         cfun->machine->max_used_stack_alignment if needed.
7390 2018-08-04  Martin Sebor  <msebor@redhat.com>
7392         PR tree-optimization/86571
7393         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7394         NaN output to 4.
7396 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
7398         PR target/86799
7399         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7400         Define.
7402 2018-08-03  Jeff Law  <law@redhat.com>
7404         PR target/86795
7405         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7406         Define to speculation_safe_value_not_needed.
7408 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
7410         * doc/gcov.texi (-x): Remove duplicate "to".
7411         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7412         (-Wif-not-aligned): Remove duplicate "is".
7413         (-flto): Remove duplicate "the".
7414         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7415         duplicate "v5.00.b".
7416         (MSP430 Options): Remove duplicate "and" from the description
7417         of "-mgprel-sec=regexp".
7418         (x86 Options): Remove duplicate copies of "vmldLog102" and
7419         vmlsLog104 from description of "-mveclibabi=type".
7421 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
7423         * internal-fn.h (first_commutative_argument): Declare.
7424         * internal-fn.c (first_commutative_argument): New function.
7425         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7426         restrictions for pattern statements.  Use first_commutative_argument
7427         to look for commutative operands in calls to internal functions.
7429 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
7431         * Makefile.in (wide-int-range.o): New.
7432         * tree-vrp.c: Move all the wide_int_* functions to...
7433         * wide-int-range.cc: ...here.
7434         * tree-vrp.h: Move all the wide_int_* prototypes to...
7435         * wide-int-range.h: ...here.
7437 2018-08-03  Tom de Vries  <tdevries@suse.de>
7439         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
7440         UI_NONE.
7441         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
7442         * except.c (output_function_exception_table): Do early exit if
7443         targetm_common.except_unwind_info (&global_options) == UI_NONE.
7445 2018-08-03  Martin Liska  <mliska@suse.cz>
7447         * predict.c (dump_prediction): Change to 2 digits
7448         in fraction part.
7450 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7452         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
7453         neon_dup_q to...
7454         (falkor_am_1_gtov_gtov): ... a new insn reservation.
7456 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
7458         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
7459         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
7460         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
7461         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
7462         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
7464 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7466         * diagnostic-show-locus.c (diagnostic_show_locus): Use
7467         pp_take_prefix when saving the existing prefix.
7468         * diagnostic.c (diagnostic_append_note): Likewise.
7469         * langhooks.c (lhd_print_error_function): Likewise.
7470         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
7471         param's type.  Free the existing prefix.
7472         (pp_take_prefix): New function.
7473         (pretty_printer::pretty_printer): Drop the prefix parameter.
7474         Rename the length parameter to match the comment.
7475         (pretty_printer::~pretty_printer): Free the prefix.
7476         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
7477         parameter.
7478         (struct pretty_printer): Drop the "const" from "prefix" field's
7479         type and clarify memory management.
7480         (pp_set_prefix): Drop the "const" from the 2nd param.
7481         (pp_take_prefix): New decl.
7483 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
7485         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
7486         for word_mode_ok here instead of passing as argument.
7487         (expand_block_compare): Change select_block_compare_mode() call.
7488         (expand_strncmp_gpr_sequence): New function.
7489         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
7491 2018-08-02  Jeff Law  <law@redhat.com>
7493         PR target/86790
7494         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7495         Define to speculation_safe_value_not_needed.
7497         PR target/86784
7498         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7499         Define to speculation_safe_value_not_needed.
7501 2018-08-02  Tom de Vries  <tdevries@suse.de>
7503         PR target/86660
7504         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
7505         function.  Return UI_TARGET unconditionally.
7506         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
7507         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
7509 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7511         * genemit.c (print_overload_test): Fix typo.
7513 2018-08-02  Richard Biener  <rguenther@suse.de>
7515         PR tree-optimization/86816
7516         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
7517         which checks for value availability before querying it.
7518         (gvn_uses_equal): Use it.
7519         (same_succ_hash): Likewise.
7520         (gimple_equal_p): Likewise.
7522 2018-08-02  Nick Clifton  <nickc@redhat.com>
7524         PR target/86813
7525         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7526         Define to speculation_safe_value_not_needed.
7528         PR target/86810
7529         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7530         Define to speculation_safe_value_not_needed.
7532         PR target/86810
7533         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7534         Define to speculation_safe_value_not_needed.
7536         PR target/86803
7537         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7538         Define to speculation_safe_value_not_needed.
7540         PR target/86797
7541         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7542         Define to speculation_safe_value_not_needed.
7544         PR target/86791
7545         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7546         Define to speculation_safe_value_not_needed.
7548         PR target/86789
7549         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7550         Define to speculation_safe_value_not_needed.
7552         PR target/86787
7553         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7554         Define to speculation_safe_value_not_needed.
7556         PR target/86782
7557         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7558         speculation_safe_value_not_needed.
7560         PR target/86781
7561         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
7562         to speculation_safe_value_not_needed.
7564 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7566         * doc/md.texi: Expand the documentation of instruction names
7567         to mention port-local uses.  Document '@' in pattern names.
7568         * read-md.h (overloaded_instance, overloaded_name): New structs.
7569         (mapping): Declare.
7570         (md_reader::handle_overloaded_name): New member function.
7571         (md_reader::get_overloads): Likewise.
7572         (md_reader::m_first_overload): New member variable.
7573         (md_reader::m_next_overload_ptr): Likewise.
7574         (md_reader::m_overloads_htab): Likewise.
7575         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
7576         m_next_overload_ptr and m_overloads_htab.
7577         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
7578         (get_mode_token, get_code_token, get_int_token): New functions.
7579         (map_attr_string): Add an optional argument that passes back
7580         the associated iterator.
7581         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
7582         (md_reader::handle_overloaded_name, add_overload_instance): New
7583         functions.
7584         (apply_iterators): Handle '@' names.  Report an error if '@'
7585         is used without iterators.
7586         (initialize_iterators): Initialize the new iterator_group fields.
7587         * genopinit.c (handle_overloaded_code_for)
7588         (handle_overloaded_gen): New functions.
7589         (main): Use them to print declarations of maybe_code_for_* and
7590         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
7591         * genemit.c (print_overload_arguments, print_overload_test)
7592         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
7593         (main): Use it to print definitions of maybe_code_for_* and
7594         maybe_gen_* functions.
7595         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
7596         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
7597         instead of explicit mode checks.
7598         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
7599         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
7600         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
7601         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
7602         (aarch64_expand_compare_and_swap): Likewise
7603         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
7604         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
7605         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
7606         (aarch64_constant_pool_reload_icode): Delete.
7607         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
7608         instead of aarch64_constant_pool_reload_icode.  Use
7609         code_for_aarch64_reload_mov instead of explicit mode checks.
7610         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
7611         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
7612         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
7613         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
7614         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
7615         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
7616         (aarch64_atomic_load_op_code): Delete.
7617         (aarch64_emit_atomic_load_op): Likewise.
7618         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
7619         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
7620         instead of aarch64_emit_atomic_load_op.
7621         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
7622         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
7623         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
7624         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
7625         character before the pattern name.
7626         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
7627         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
7628         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
7629         (aarch64_frecps<mode>): Likewise.
7630         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
7631         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
7632         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
7633         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
7634         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
7636 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7638         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
7639         Allow HFmode constants if TARGET_FP_F16INST.
7641 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
7643         PR target/86014
7644         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
7645         No longer check last store for clobber of address register.
7647 2018-08-02  Martin Liska  <mliska@suse.cz>
7649         PR gcov-profile/86817
7650         * gcov.c (process_all_functions): New function.
7651         (main): Call it.
7652         (process_file): Move functions processing to
7653         process_all_functions.
7655 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7657         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
7658         "const" to the "gimple *" and "rtx_insn *" parameters.
7659         * dumpfile.h (dump_user_location_t::dump_user_location_t):
7660         Likewise.
7661         (dump_location_t::dump_location_t): Likewise.
7663 2018-08-01  Martin Sebor  <msebor@redhat.com>
7665         PR tree-optimization/86650
7666         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
7667         rather than a "gcall *".  Directly pass the data of interest
7668         to percent_K_format, rather than building a temporary CALL_EXPR
7669         to hold it.
7670         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
7671         (gimple_fold_builtin_strncat): Adjust.
7672         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
7673         gcall* argument with gimple*.
7674         * gimple-ssa-warn-restrict.c (check_call): Same.
7675         (wrestrict_dom_walker::before_dom_children): Same.
7676         (builtin_access::builtin_access): Same.
7677         (check_bounds_or_overlap): Same
7678         (maybe_diag_overlap): Same.
7679         (maybe_diag_offset_bounds): Same.
7680         * tree-diagnostic.c (default_tree_printer): Move usage of
7681         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
7682         to this callsite.
7683         * tree-pretty-print.c (percent_K_format): Add argument.
7684         * tree-pretty-print.h: Add argument.
7685         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
7686         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
7687         (maybe_diag_stxncpy_trunc): Same.
7688         (handle_builtin_stxncpy): Same.
7689         (handle_builtin_strcat): Same.
7691 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7693         * match.pd: Optimise pointer range checks.
7695 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7697         PR tree-optimization/86758
7698         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
7699         to remove pattern statements.
7701 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7703         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
7704         result of dfs_enumerate_from when constructing stmt_vec_infos,
7705         instead of additionally calling get_loop_body.
7707 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7709         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
7710         parameter.
7711         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
7712         When creating an iv, assert that the step is not known to be zero.
7713         (vect_setup_realignment): Update call accordingly.
7714         * tree-vect-stmts.c (vectorizable_store): Likewise.
7715         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
7717 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7719         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
7720         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
7721         (vectorizable_reduction): Likewise.
7722         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7723         (vect_detect_hybrid_slp_stmts): Likewise.
7724         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7726 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
7728         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
7729         (wide_int_set_zero_nonzero_bits): ...this.
7730         (zero_nonzero_bits_from_vr): Rename to...
7731         (vrp_set_zero_nonzero_bits): ...this.
7732         (extract_range_from_multiplicative_op_1): Abstract wide int
7733         code...
7734         (wide_int_range_multiplicative_op): ...here.
7735         (extract_range_from_binary_expr_1): Extract wide int binary
7736         operations into their own functions.
7737         (wide_int_range_lshift): New.
7738         (wide_int_range_can_optimize_bit_op): New.
7739         (wide_int_range_shift_undefined_p): New.
7740         (wide_int_range_bit_xor): New.
7741         (wide_int_range_bit_ior): New.
7742         (wide_int_range_bit_and): New.
7743         (wide_int_range_trunc_mod): New.
7744         (extract_range_into_wide_ints): New.
7745         (vrp_shift_undefined_p): New.
7746         (extract_range_from_multiplicative_op): New.
7747         (vrp_can_optimize_bit_op): New.
7748         * tree-vrp.h (value_range::dump): New.
7749         (wide_int_range_multiplicative_op): New.
7750         (wide_int_range_lshift):New.
7751         (wide_int_range_shift_undefined_p): New.
7752         (wide_int_range_bit_xor): New.
7753         (wide_int_range_bit_ior): New.
7754         (wide_int_range_bit_and): New.
7755         (wide_int_range_trunc_mod): New.
7756         (zero_nonzero_bits_from_bounds): Rename to...
7757         (wide_int_set_zero_nonzero_bits): ...this.
7758         (zero_nonzero_bits_from_vr): Rename to...
7759         (vrp_set_zero_nonzero_bits): ...this.
7760         (range_easy_mask_min_max): Rename to...
7761         (wide_int_range_can_optimize_bit_op): this.
7762         * vr-values.c (simplify_bit_ops_using_ranges): Rename
7763         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
7765 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7767         * tree-vectorizer.h (vect_orig_stmt): New function.
7768         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
7769         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
7770         (vect_create_epilog_for_reduction): Likewise.
7771         (vectorizable_live_operation): Likewise.
7772         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7773         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
7774         * tree-vect-stmts.c (vectorizable_call): Likewise.
7775         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
7777 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7779         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
7780         argument.
7781         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
7782         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
7783         (vect_transform_loop): Likewise.
7784         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7786 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7788         * tree-vectorizer.h (vect_schedule_slp): Return void.
7789         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7790         (vect_schedule_slp): Likewise.
7792 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7794         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
7795         argument.
7796         (vect_transform_loop): Update calls accordingly.  Schedule SLP
7797         instances before the main loop, if any exist.
7799 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7801         PR tree-optimization/86749
7802         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
7803         If the lhs is used in a COND_EXPR, check that it is being used
7804         as the "then" or "else" value.
7806 2018-08-01  Tom de Vries  <tdevries@suse.de>
7808         PR target/86800
7809         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7810         speculation_safe_value_not_needed.
7812 2018-08-01  Richard Biener  <rguenther@suse.de>
7814         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
7815         as base and offset.
7817 2018-08-01  Martin Liska  <mliska@suse.cz>
7819         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
7820         format how successful transformation is dumped.
7821         (gimple_mod_pow2_value_transform): Likewise.
7822         (gimple_mod_subtract_transform): Likewise.
7823         (gimple_stringops_transform): Likewise.
7825 2018-08-01  Martin Liska  <mliska@suse.cz>
7827         PR value-prof/35543
7828         * value-prof.c (interesting_stringop_to_profile_p):
7829         Simplify the code and add BUILT_IN_MEMMOVE.
7830         (gimple_stringops_transform): Likewise.
7832 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
7834         * config/aarch64/aarch64-simd.md
7835         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
7836         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
7837         use GPI iterator instead of SI mode.
7839 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7841         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
7842         rs6000_speculation_barrier.
7843         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
7844         new barrier pattern name.
7846 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7848         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
7849         (speculation_barrier): New insn.
7851 2018-08-01  Richard Biener  <rguenther@suse.de>
7853         PR bootstrap/86724
7854         * graphite.h: Include isl/id.h and isl/space.h to allow build
7855         with ISL 0.20.
7857 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
7859         PR target/86651
7860         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7861         mode for COFF targets.
7862         * defaults.h (TARGET_COFF): Define.
7863         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
7864         TARGET_COFF): Define.
7865         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
7866         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
7867         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
7869 2018-07-31  Alexandre Oliva <oliva@adacore.com>
7871         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
7872         * gimple-streamer-out.c (output_bb): Save it.
7873         * lto-streamer-in.c (input_struct_function_base): Restore
7874         instance discriminator if available.  Create map on demand.
7875         * lto-streamer-out.c (output_struct_function_base): Save it if
7876         available.
7877         * final.c (decl_to_instance_map): Document LTO strategy.
7879 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
7880             Olivier Hainque  <hainque@adacore.com>
7882         * debug.h (decl_to_instance_map_t): New type.
7883         (decl_to_instance_map): Declare.
7884         (maybe_create_decl_to_instance_map): New inline function.
7885         * final.c (bb_discriminator, last_bb_discriminator): New statics,
7886         to track basic block discriminators.
7887         (final_start_function_1): Initialize them.
7888         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
7889         bb_discriminator.
7890         (decl_to_instance_map): New variable.
7891         (map_decl_to_instance, maybe_set_discriminator): New functions.
7892         (notice_source_line): Set discriminator.
7894 2018-07-31  Ian Lance Taylor  <iant@golang.org>
7896         * targhooks.c (default_have_speculation_safe_value): Add
7897         ATTRIBUTE_UNUSED.
7899 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7901         * dump-context.h: Include "pretty-print.h".
7902         (dump_context::refresh_dumps_are_enabled): New decl.
7903         (dump_context::emit_item): New decl.
7904         (class dump_context): Add fields "m_test_pp" and
7905         "m_test_pp_flags".
7906         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7907         (temp_dump_context::get_dumped_text): New decl.
7908         (class temp_dump_context): Add field "m_pp".
7909         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
7910         (dump_context::refresh_dumps_are_enabled): ...and add a test for
7911         m_test_pp.
7912         (set_dump_file): Update for above change.
7913         (set_alt_dump_file): Likewise.
7914         (dump_loc): New overload, taking a pretty_printer *.
7915         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
7916         to any test pretty-printer.
7917         (make_item_for_dump_gimple_stmt): New function, adapted from
7918         optinfo::add_gimple_stmt.
7919         (dump_context::dump_gimple_stmt): Call it, and use the result,
7920         eliminating the direct usage of dump_file and alt_dump_file in
7921         favor of indirectly using them via emit_item.
7922         (make_item_for_dump_gimple_expr): New function, adapted from
7923         optinfo::add_gimple_expr.
7924         (dump_context::dump_gimple_expr): Call it, and use the result,
7925         eliminating the direct usage of dump_file and alt_dump_file in
7926         favor of indirectly using them via emit_item.
7927         (make_item_for_dump_generic_expr): New function, adapted from
7928         optinfo::add_tree.
7929         (dump_context::dump_generic_expr): Call it, and use the result,
7930         eliminating the direct usage of dump_file and alt_dump_file in
7931         favor of indirectly using them via emit_item.
7932         (make_item_for_dump_printf_va): New function, adapted from
7933         optinfo::add_printf_va.
7934         (make_item_for_dump_printf): New function.
7935         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
7936         and use the result, eliminating the direct usage of dump_file and
7937         alt_dump_file in favor of indirectly using them via emit_item.
7938         (make_item_for_dump_dec): New function.
7939         (dump_context::dump_dec): Call it, and use the result,
7940         eliminating the direct usage of dump_file and alt_dump_file in
7941         favor of indirectly using them via emit_item.
7942         (make_item_for_dump_symtab_node): New function, adapted from
7943         optinfo::add_symtab_node.
7944         (dump_context::dump_symtab_node): Call it, and use the result,
7945         eliminating the direct usage of dump_file and alt_dump_file in
7946         favor of indirectly using them via emit_item.
7947         (dump_context::begin_scope): Reimplement, avoiding direct usage
7948         of dump_file and alt_dump_file in favor of indirectly using them
7949         via emit_item.
7950         (dump_context::emit_item): New member function.
7951         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7952         Set up test pretty-printer on the underlying context.  Call
7953         refresh_dumps_are_enabled.
7954         (temp_dump_context::~temp_dump_context): Call
7955         refresh_dumps_are_enabled.
7956         (temp_dump_context::get_dumped_text): New member function.
7957         (selftest::verify_dumped_text): New function.
7958         (ASSERT_DUMPED_TEXT_EQ): New macro.
7959         (selftest::test_capture_of_dump_calls): Run all tests twice, with
7960         and then without optinfo enabled.  Add uses of
7961         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
7962         * dumpfile.h: Update comment for the dump_* API.
7963         * optinfo-emit-json.cc
7964         (selftest::test_building_json_from_dump_calls): Update for new
7965         param for temp_dump_context ctor.
7966         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
7967         and "m_owned" field.
7968         (optinfo_item::~optinfo_item): Likewise.
7969         (optinfo::add_item): New member function.
7970         (optinfo::emit): Update comment.
7971         (optinfo::add_string): Delete.
7972         (optinfo::add_printf): Delete.
7973         (optinfo::add_printf_va): Delete.
7974         (optinfo::add_gimple_stmt): Delete.
7975         (optinfo::add_gimple_expr): Delete.
7976         (optinfo::add_tree): Delete.
7977         (optinfo::add_symtab_node): Delete.
7978         (optinfo::add_dec): Delete.
7979         * optinfo.h (class dump_context): New forward decl.
7980         (optinfo::add_item): New decl.
7981         (optinfo::add_string): Delete.
7982         (optinfo::add_printf): Delete.
7983         (optinfo::add_printf_va): Delete.
7984         (optinfo::add_gimple_stmt): Delete.
7985         (optinfo::add_gimple_expr): Delete.
7986         (optinfo::add_tree): Delete.
7987         (optinfo::add_symtab_node): Delete.
7988         (optinfo::add_dec): Delete.
7989         (optinfo::add_poly_int): Delete.
7990         (optinfo_item::optinfo_item): Remove "owned" param.
7991         (class optinfo_item): Remove field "m_owned".
7993 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7995         PR middle-end/86705
7996         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
7997         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
7998         requested variable alignment.
7999         (expand_one_ssa_partition): Likewise.
8000         (expand_one_var): Likewise.
8002 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8004         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8005         to speculation_safe_value_not_needed.
8007 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8009         * targhooks.h (speculation_safe_value_not_needed): New prototype.
8010         * targhooks.c (speculation_safe_value_not_needed): New function.
8011         * target.def (have_speculation_safe_value): Update documentation.
8012         * doc/tm.texi: Regenerated.
8014 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8016         * config/aarch64/iterators.md (ALLI_TI): New iterator.
8017         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
8018         expand.
8019         (despeculate_copy<ALLI:mode>_insn): New insn.
8020         (despeculate_copyti_insn): New insn.
8021         (despeculate_simple<ALLI:mode>): New insn
8022         (despeculate_simpleti): New insn.
8023         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
8024         function.
8025         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
8026         aarch64_speculation_safe_value.
8027         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
8029 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8031         * config/aarch64/aarch64-speculation.cc: New file.
8032         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
8033         before pass_reorder_blocks.
8034         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
8035         prototype.
8036         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
8037         X14 and X15 when tracking speculation.
8038         * config/aarch64/aarch64.md (register name constants): Add
8039         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
8040         (unspec): Add UNSPEC_SPECULATION_TRACKER.
8041         (speculation_barrier): New insn attribute.
8042         (cmp<mode>): Allow SP in comparisons.
8043         (speculation_tracker): New insn.
8044         (speculation_barrier): Add speculation_barrier attribute.
8045         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
8046         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
8047         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
8049 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8051         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
8052         aarch64_track_speculation is true.
8053         (tb<optab><mode>1): Likewise.
8054         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
8055         generate CB[N]Z when tracking speculation.
8056         (aarch64_split_compare_and_swap): Likewise.
8057         (aarch64_split_atomic_op): Likewise.
8059 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8061         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
8063 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8065         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
8066         (speculation_barrier): New insn.
8068 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8070         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
8071         * config/arm/arm.md (speculation_barrier): New expand.
8072         (speculation_barrier_insn): New pattern.
8074 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
8076         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
8077         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
8078         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
8079         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
8080         list.
8081         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
8082         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
8083         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
8084         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
8085         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
8086         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
8087         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
8088         * builtins.c (expand_speculation_safe_value): New function.
8089         (expand_builtin): Call it.
8090         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
8091         * doc/extend.texi: Document __builtin_speculation_safe_value.
8092         * doc/md.texi: Document "speculation_barrier" pattern.
8093         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
8094         TARGET_HAVE_SPECULATION_SAFE_VALUE.
8095         * doc/tm.texi: Regenerated.
8096         * target.def (have_speculation_safe_value, speculation_safe_value): New
8097         hooks.
8098         * targhooks.c (default_have_speculation_safe_value): New function.
8099         (default_speculation_safe_value): New function.
8100         * targhooks.h (default_have_speculation_safe_value): Add prototype.
8101         (default_speculation_safe_value): Add prototype.
8103 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
8105         * dump-context.h (dump_context::dump_loc): New decl.
8106         * dumpfile.c (dump_context::dump_loc): New member function.
8107         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
8108         and dump_gimple_stmt.
8109         (dump_context::dump_gimple_expr_loc): Likewise, using
8110         dump_gimple_expr.
8111         (dump_context::dump_generic_expr_loc): Likewise, using
8112         dump_generic_expr.
8113         (dump_context::dump_printf_loc_va): Likewise, using
8114         dump_printf_va.
8115         (dump_context::begin_scope): Explicitly using the global function
8116         "dump_loc", rather than the member function.
8118 2018-07-31  Martin Sebor  <msebor@redhat.com>
8120         PR tree-optimization/86741
8121         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
8123 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
8125         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
8126         depend on whether prefetch instructions will be emitted or not.
8127         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
8128         will be emitted or not.
8129         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
8130         (TARGET_SETMEM_PFD): New macros.
8132 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8134         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
8135         (NULL_STMT_VEC_INFO): Delete.
8136         (stmt_vec_info::operator*): Likewise.
8137         (stmt_vec_info::operator gimple *): Likewise.
8138         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
8139         of NULL_STMT_VEC_INFO.
8140         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8141         (vect_reassociating_reduction_p): Likewise.
8142         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8143         (vectorizable_store): Likewise.
8144         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
8145         (vec_info::free_stmt_vec_infos): Likewise.
8147 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8149         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
8150         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
8152 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8154         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
8155         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8156         (vec_info::free_stmt_vec_info): New private member functions.
8157         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
8158         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
8159         * tree-parloops.c (gather_scalar_reductions): Remove calls to
8160         set_stmt_vec_info_vec and free_stmt_vec_infos.
8161         * tree-vect-loop.c (_loop_vec_info): Remove call to
8162         set_stmt_vec_info_vec.
8163         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
8164         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
8165         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
8166         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
8167         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
8168         assignments in {vec_info::,}new_stmt_vec_info that are redundant
8169         with the clearing in the xcalloc.
8170         (stmt_vec_info_vec): Delete.
8171         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
8172         (vectorize_loops): Likewise.
8173         (vec_info::~vec_info): Remove argument from call to
8174         free_stmt_vec_infos.
8175         (vec_info::add_stmt): Remove vinfo argument from call to
8176         new_stmt_vec_info.
8178 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8180         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
8181         rather than a gimple stmt.
8182         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
8183         information for pattern statements when passed the original
8184         statement; instead wait to be passed the pattern statement itself.
8185         Don't call set_vinfo_for_stmt here.
8186         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
8187         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
8188         stmt_vec_infos here.
8189         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
8190         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
8191         stmt_vec_infos entry.
8193 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8195         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
8196         * tree-vectorizer.c (vec_info::replace_stmt): New function.
8197         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
8198         * tree-vect-stmts.c (vectorizable_call): Likewise.
8199         (vectorizable_simd_clone_call): Likewise.
8201 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8203         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
8204         * tree-vectorizer.c (vec_info::remove_stmt): New function.
8205         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
8206         * tree-vect-loop.c (vect_transform_loop): Likewise.
8207         * tree-vect-slp.c (vect_schedule_slp): Likewise.
8208         * tree-vect-stmts.c (vect_remove_stores): Likewise.
8210 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8212         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
8213         (vect_dr_stmt): Delete.
8214         * tree-vectorizer.c (vec_info::lookup_dr): New function.
8215         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
8216         of DR_VECT_AUX.
8217         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
8218         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
8219         (vect_verify_datarefs_alignment, vect_peeling_supportable)
8220         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8221         (vect_analyze_data_refs): Likewise.
8222         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
8223         argument.
8224         (vect_find_same_alignment_drs): Likewise.
8225         (vect_slp_analyze_node_dependences): Update calls accordingly.
8226         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8227         instead of DR_VECT_AUX.
8228         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
8229         of a vector data references.  Use vec_info::lookup_dr instead of
8230         DR_VECT_AUX.
8231         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
8232         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
8233         instead of DR_VECT_AUX.
8235 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8237         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
8238         dr_vec_info.
8239         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8240         accordingly.
8241         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8242         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8243         (vect_gen_prolog_loop_niters): Likewise.
8245 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8247         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
8248         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
8249         (vect_known_alignment_in_bytes, vect_dr_behavior)
8250         (vect_get_scalar_dr_size): Take references as dr_vec_infos
8251         instead of data_references.  Update calls to other routines for
8252         which the same change has been made.
8253         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
8254         dr_vec_infos instead of stmt_vec_infos.
8255         (vect_analyze_data_ref_dependence): Update call accordingly.
8256         (vect_slp_analyze_data_ref_dependence)
8257         (vect_record_base_alignments): Use DR_VECT_AUX.
8258         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8259         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8260         (vector_alignment_reachable_p, vect_get_data_access_cost)
8261         (vect_peeling_supportable, vect_analyze_group_access_1)
8262         (vect_analyze_group_access, vect_analyze_data_ref_access)
8263         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8264         (vect_compile_time_alias, vect_small_gap_p)
8265         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8266         (vect_supportable_dr_alignment): Take references as dr_vec_infos
8267         instead of data_references.  Update calls to other routines for
8268         which the same change has been made.
8269         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8270         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8271         (vect_slp_analyze_and_verify_node_alignment)
8272         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8273         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8274         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
8275         above changes.
8276         (_vect_peel_info::dr): Replace with...
8277         (_vect_peel_info::dr_info): ...this new field.
8278         (vect_peeling_hash_get_most_frequent)
8279         (vect_peeling_hash_choose_best_peeling): Update accordingly.
8280         (vect_peeling_hash_get_lowest_cost):
8281         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
8282         routines for which the same change has been made.
8283         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
8284         data_reference.
8285         * tree-vect-loop-manip.c (get_misalign_in_elems)
8286         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
8287         above changes.
8288         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8289         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8290         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8291         (get_group_load_store_type, get_negative_load_store_type)
8292         (vect_get_data_ptr_increment, vectorizable_store)
8293         (vectorizable_load): Likewise.
8294         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8295         Update calls to other routines for which the same change has been made.
8297 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8299         * tree-vectorizer.h (vec_info::move_dr): New member function.
8300         (dataref_aux): Rename to...
8301         (dr_vec_info): ...this and add "dr" and "stmt" fields.
8302         (_stmt_vec_info::dr_aux): Update accordingly.
8303         (_stmt_vec_info::data_ref_info): Delete.
8304         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8305         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8306         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8307         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8308         of data_ref.
8309         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
8310         (STMT_VINFO_DR_INFO): New macro.
8311         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8312         (set_dr_misalignment): Update after rename of dataref_aux.
8313         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
8314         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8315         initialization of STMT_VINFO_DATA_REF.
8316         * tree-vectorizer.c (vec_info::move_dr): New function.
8317         * tree-vect-patterns.c (vect_recog_bool_pattern)
8318         (vect_recog_mask_conversion_pattern)
8319         (vect_recog_gather_scatter_pattern): Use it.
8320         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8321         the "dr" and "stmt" fields of dr_vec_info instead of
8322         STMT_VINFO_DATA_REF.
8324 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8326         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8327         (is_pattern_stmt_p): Use it.
8328         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8329         on pattern statements.
8331 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8333         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8334         original stmt as a stmt_vec_info rather than a gimple stmt.
8335         (vect_pattern_recog_1): Take the statement directly as a
8336         stmt_vec_info, rather than via a gimple_stmt_iterator.
8337         Update call to vect_mark_pattern_stmts.
8338         (vect_pattern_recog): Update calls accordingly.
8340 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8342         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8343         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8344         a vect_def_type for the first argument.
8345         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8346         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
8347         operand if it isn't defined by a vectorized statement.
8348         (vect_build_gather_load_calls): Remove the mask_dt argument and
8349         update calls to vect_get_vec_def_for_stmt_copy.
8350         (vectorizable_bswap): Likewise the dt argument.
8351         (vectorizable_call): Update calls to vectorizable_bswap and
8352         vect_get_vec_def_for_stmt_copy.
8353         (vectorizable_simd_clone_call, vectorizable_assignment)
8354         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8355         (vectorizable_comparison): Update calls to
8356         vect_get_vec_def_for_stmt_copy.
8357         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
8358         vect_is_simple_use.
8359         (vect_get_loop_based_defs): Remove dt argument and update call
8360         to vect_get_vec_def_for_stmt_copy.
8361         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8362         and vect_get_vec_def_for_stmt_copy.
8363         (vectorizable_load): Update calls to vect_build_gather_load_calls
8364         and vect_get_vec_def_for_stmt_copy.
8365         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8366         (vectorizable_reduction, vectorizable_live_operation): Update calls
8367         to vect_get_vec_def_for_stmt_copy.
8369 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8371         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8372         and gimple stmt arguments with a stmt_vec_info.
8373         (vect_record_base_alignments): Update calls accordingly.
8374         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8375         and gimple stmt arguments with a stmt_vec_info.
8376         (vect_build_slp_tree_1): Remove vinfo argument and update call
8377         to vect_record_max_nunits.
8378         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8379         and vect_record_max_nunits.
8381 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8383         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8384         file and take a stmt_vec_info instead of a gimple stmt.
8385         (supportable_widening_operation, vect_finish_replace_stmt)
8386         (vect_finish_stmt_generation, vect_get_store_rhs)
8387         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8388         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8389         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8390         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8391         (vect_create_data_ref_ptr, bump_vector_ptr)
8392         (vect_permute_store_chain, vect_setup_realignment)
8393         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8394         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8395         (vectorizable_reduction, vectorizable_induction)
8396         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8397         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8398         than gimple stmts as arguments.
8399         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8400         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8401         (can_group_stmts_p, vect_check_gather_scatter)
8402         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8403         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8404         (vect_permute_load_chain, vect_shift_permute_load_chain)
8405         (vect_transform_grouped_load)
8406         (vect_record_grouped_load_vectors): Likewise.
8407         * tree-vect-loop.c (vect_fixup_reduc_chain)
8408         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8409         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8410         (vectorizable_reduction, vectorizable_induction)
8411         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8412         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8413         (vect_get_load_store_mask): Likewise.
8414         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8415         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8416         * tree-vect-stmts.c (vect_mark_relevant)
8417         (is_simple_and_all_uses_invariant)
8418         (exist_non_indexing_operands_for_use_p, process_use)
8419         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8420         (vect_get_vec_def_for_operand, vect_get_vec_defs)
8421         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8422         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8423         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8424         (get_negative_load_store_type, get_load_store_type)
8425         (vect_check_load_store_mask, vect_check_store_rhs)
8426         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8427         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8428         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8429         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8430         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8431         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8432         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8433         (supportable_widening_operation): Likewise.
8435 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8437         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8438         a stmt_vec_info instead of a gcall.
8439         (vect_check_gather_scatter): Update call accordingly.
8440         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
8441         of a gphi.
8442         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
8443         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
8444         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
8445         instead of a gimple stmt.
8446         (vect_transform_loop): Update calls accordingly.
8447         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
8448         stmt_vec_infos instead of gimple stmts.
8449         (vect_analyze_slp_instance): Update use accordingly.
8450         * tree-vect-stmts.c (read_vector_array, write_vector_array)
8451         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
8452         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
8453         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
8454         (vect_gen_widened_results_half, vect_get_loop_based_defs)
8455         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
8456         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
8457         down to subroutines.
8459 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8461         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
8462         of the worklist from a vector of gimple stmts to a vector of
8463         stmt_vec_infos.
8464         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8465         (vect_mark_stmts_to_be_vectorized): Likewise
8467 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8469         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
8470         statement before passing it to vect_analyze_stmt.
8471         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
8472         the chain of phi vector definitions.  Track the exit phi via its
8473         stmt_vec_info.
8474         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
8475         STMT_VINFO_REDUC_DEF.
8476         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
8477         stmt_vec_infos to handle the statement chains.
8478         (vect_get_slp_defs): Record the first statement in the node
8479         using a stmt_vec_info.
8480         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
8481         statements here and pass their stmt_vec_info down to subroutines.
8482         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
8483         down to vect_finish_stmt_generation.
8484         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
8485         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
8486         stmt_vec_infos to subroutines.
8487         (vect_remove_stores): Use stmt_vec_infos to handle the statement
8488         chains.
8490 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8492         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
8493         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
8494         (vect_permute_store_chain, vect_setup_realignment)
8495         (vect_permute_load_chain, vect_shift_permute_load_chain)
8496         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
8497         stmts internally, and when passing values to other vectorizer routines.
8498         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8499         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
8500         (vect_analyze_loop_operations, get_initial_def_for_reduction)
8501         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
8502         (vectorizable_reduction, vectorizable_induction)
8503         (vectorizable_live_operation, vect_transform_loop_stmt)
8504         (vect_transform_loop): Likewise.
8505         * tree-vect-patterns.c (vect_reassociating_reduction_p)
8506         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
8507         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
8508         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8509         (vect_slp_analyze_node_operations_1): Likewise.
8510         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8511         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
8512         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
8513         (vect_finish_stmt_generation_1, get_group_load_store_type)
8514         (get_load_store_type, vect_build_gather_load_calls)
8515         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8516         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8517         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8518         (vectorizable_store, vectorizable_load, vectorizable_condition)
8519         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
8520         (supportable_widening_operation): Likewise.
8521         (vect_get_vector_types_for_stmt): Likewise.
8522         * tree-vectorizer.h (vect_dr_behavior): Likewise.
8524 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8526         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8527         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
8528         (vect_permute_store_chain, vect_permute_load_chain)
8529         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
8530         repeated stmt_vec_info lookups.
8531         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8532         (vect_update_ivs_after_vectorizer): Likewise.
8533         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
8534         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8535         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
8536         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8537         (vect_bb_slp_scalar_cost): Likewise.
8538         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
8540 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8542         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
8543         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
8544         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
8545         to get gassigns and gcalls, rather than operating on generc gimple
8546         stmts.
8547         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
8548         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
8549         (vectorizable_load, vect_analyze_stmt): Likewise.
8550         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
8552 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8554         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
8555         return stmt_vec_infos rather than gimple stmts.  Do not accept
8556         null arguments.
8557         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8558         of a gimple stmt.
8559         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8560         Update use of get_later_stmt.
8561         (vect_get_constant_vectors): Update call accordingly.
8562         (vect_schedule_slp_instance): Likewise
8563         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
8564         (vect_slp_analyze_instance_dependence): Likewise.
8565         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
8567 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8569         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
8570         (stmt_info_for_cost::stmt_info): ...this new field.
8571         (add_stmt_costs): Update accordingly.
8572         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8573         (vect_get_known_peeling_cost): Likewise.
8574         (vect_estimate_min_profitable_iters): Likewise.
8575         * tree-vect-stmts.c (record_stmt_cost): Likewise.
8577 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8579         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
8580         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8581         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8582         accordingly.
8583         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
8585 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8587         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
8588         a gimple stmt to a stmt_vec_info.
8589         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8591 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8593         * tree-vectorizer.h (vec_info::grouped_stores): Change from
8594         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8595         (_loop_vec_info::reduction_chains): Likewise.
8596         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
8597         accordingly.
8598         * tree-vect-slp.c (vect_analyze_slp): Likewise.
8600 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8602         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
8603         a gimple stmt to a stmt_vec_info.
8604         (_stmt_vec_info::next_element): Likewise.
8605         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
8606         (vect_slp_analyze_and_verify_node_alignment)
8607         (vect_analyze_group_access_1, vect_analyze_group_access)
8608         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
8609         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
8610         (vect_supportable_dr_alignment): Update accordingly.
8611         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
8612         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
8613         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
8614         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8615         * tree-vect-slp.c (vect_build_slp_tree_1)
8616         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
8617         (vect_split_slp_store_group, vect_analyze_slp_instance)
8618         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
8619         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
8620         (get_group_load_store_type, get_load_store_type)
8621         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
8622         (vect_transform_stmt, vect_remove_stores): Likewise.
8624 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8626         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
8627         than a gimple stmt.
8628         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8629         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
8630         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
8631         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
8632         (vector_alignment_reachable_p, vect_get_data_access_cost)
8633         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
8634         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
8635         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8636         (vect_analyze_group_access_1, vect_analyze_group_access)
8637         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
8638         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
8639         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
8640         result of vect_dr_stmt and use the stmt_vec_info instead of
8641         the associated gimple stmt.
8642         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8643         (vect_gen_prolog_loop_niters): Likewise.
8644         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8646 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8648         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
8649         to a vec<stmt_vec_info>.
8650         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
8651         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
8652         vec<stmt_vec_info>.
8653         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
8654         to a vec<stmt_vec_info>.
8655         (bst_traits::value_type, bst_traits::value_type): Likewise.
8656         (bst_traits::hash): Update accordingly.
8657         (vect_get_and_check_slp_defs): Change the stmts parameter from
8658         a vec<gimple *> to a vec<stmt_vec_info>.
8659         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
8660         (vect_build_slp_tree): Likewise.
8661         (vect_build_slp_tree_2): Likewise.  Update uses of
8662         SLP_TREE_SCALAR_STMTS.
8663         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
8664         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
8665         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
8666         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
8667         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
8668         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
8669         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
8670         (vect_get_constant_vectors, vect_get_slp_defs)
8671         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
8672         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
8673         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
8674         instead of gimple stmts.
8675         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
8676         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
8677         (vect_slp_analyze_instance_dependence): Update uses of
8678         SLP_TREE_SCALAR_STMTS.
8679         (vect_slp_analyze_and_verify_node_alignment): Likewise.
8680         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
8681         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
8682         (get_initial_defs_for_reduction): Likewise.
8683         (vect_create_epilog_for_reduction): Likewise.
8684         (vectorize_fold_left_reduction): Likewise.
8685         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
8686         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
8687         (can_vectorize_live_stmts): Likewise.
8689 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8691         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
8692         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8693         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
8694         than gimple stmts.
8695         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
8696         of a gimple stmt.
8697         (gather_scalar_reductions): Update after above interface changes.
8698         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
8699         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
8700         than gimple stmts.
8701         (vect_force_simple_reduction): Likewise.
8702         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
8703         LOOP_VINFO_REDUCTIONS.
8704         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8706 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8708         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
8709         a gimple stmt to a stmt_vec_info.
8710         * tree-vect-loop.c (vect_active_double_reduction_p)
8711         (vect_force_simple_reduction, vectorizable_reduction): Update
8712         accordingly.
8714 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8716         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
8717         vec<gimple *> to a vec<stmt_vec_info>.
8718         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
8719         the reduction_phis argument from a vec<gimple *> to a
8720         vec<stmt_vec_info>.
8721         (vectorizable_reduction): Likewise the phis local variable that
8722         is passed to vect_create_epilog_for_reduction.  Update for new type
8723         of SLP_TREE_VEC_STMTS.
8724         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
8725         (vectorizable_live_operation): Likewise.
8726         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
8727         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
8729 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8731         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
8732         a gimple stmt to a stmt_vec_info.
8733         (vectorizable_condition, vectorizable_live_operation)
8734         (vectorizable_reduction, vectorizable_induction): Pass back the
8735         vectorized statement as a stmt_vec_info.
8736         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
8737         use of STMT_VINFO_VEC_STMT.
8738         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
8739         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
8740         as stmt_vec_infos rather than gimple stmts.
8741         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
8742         to a stmt_vec_info.
8743         (vectorizable_live_operation): Likewise.
8744         (vectorizable_reduction, vectorizable_induction): Likewise,
8745         updating use of STMT_VINFO_VEC_STMT.
8746         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
8747         of STMT_VINFO_VEC_STMT.
8748         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
8749         (vectorizable_simd_clone_call, vectorizable_conversion)
8750         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8751         (vectorizable_store, vectorizable_load, vectorizable_condition)
8752         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
8753         from a gimple stmt to a stmt_vec_info.
8754         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
8755         pointer to a stmt_vec_info to the vectorizable_* routines.
8757 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8759         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
8760         a gimple stmt to a stmt_vec_info.
8761         (is_pattern_stmt_p): Update accordingly.
8762         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
8763         (vect_record_grouped_load_vectors): Likewise.
8764         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
8765         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
8766         (vect_model_reduction_cost): Likewise.
8767         (vect_create_epilog_for_reduction): Likewise.
8768         (vectorizable_reduction, vectorizable_induction): Likewise.
8769         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8770         Return the stmt_vec_info for the pattern statement.
8771         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
8772         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
8773         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8774         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
8775         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
8776         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
8777         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
8778         (free_stmt_vec_info, vect_is_simple_use): Likewise.
8780 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8782         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
8783         (vect_finish_stmt_generation): Likewise.
8784         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8785         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
8786         (vect_build_gather_load_calls): Use the return value of the above
8787         functions instead of a separate call to vinfo_for_stmt.  Use narrow
8788         scopes for the input gimple stmt and wider scopes for the associated
8789         stmt_vec_info.  Use vec_info::lookup_def when setting these
8790         stmt_vec_infos from an SSA_NAME definition.
8791         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8792         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8793         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8794         (vectorizable_store, vectorizable_load, vectorizable_condition)
8795         (vectorizable_comparison): Likewise.
8796         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
8797         (vectorizable_reduction): Likewise.
8799 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8801         * tree-vectorizer.h (vect_is_simple_use): Add an optional
8802         stmt_vec_info * parameter before the optional gimple **.
8803         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8804         (process_use, vect_get_vec_def_for_operand_1): Update callers.
8805         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
8806         * tree-vect-loop.c (vectorizable_reduction): Likewise.
8807         (vectorizable_live_operation): Likewise.
8808         * tree-vect-patterns.c (type_conversion_p): Likewise.
8809         (vect_look_through_possible_promotion): Likewise.
8810         (vect_recog_rotate_pattern): Likewise.
8811         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8813 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8815         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
8816         a typedef to a wrapper class.
8817         (NULL_STMT_VEC_INFO): New macro.
8818         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
8819         (stmt_vec_info::operator*): New function.
8820         (stmt_vec_info::operator gimple *): Likewise.
8821         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
8822         (add_stmt_costs): Likewise.
8823         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
8824         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8825         (vect_get_known_peeling_cost): Likewise.
8826         (vect_estimate_min_profitable_iters): Likewise.
8827         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8828         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
8829         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8830         (vectorizable_store, free_stmt_vec_infos): Likewise.
8831         (new_stmt_vec_info): Change return type of xcalloc to
8832         _stmt_vec_info *.
8834 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8836         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
8837         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
8838         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
8839         a single_imm_use-based sequence.
8840         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
8842 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8844         * tree-vectorizer.h (vec_info::lookup_def): Declare.
8845         * tree-vectorizer.c (vec_info::lookup_def): New function.
8846         * tree-vect-patterns.c (vect_get_internal_def): Use it.
8847         (vect_widened_op_tree): Likewise.
8848         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8849         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
8850         (vectorizable_reduction): Likewise.
8851         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
8852         of a gimple *.
8853         (vect_is_slp_reduction): Update calls accordingly.  Use
8854         vec_info::lookup_def.
8855         (vect_is_simple_reduction): Likewise
8856         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
8858 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8860         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
8861         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
8862         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
8863         of vinfo_for_stmt.
8864         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
8865         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
8866         (vect_update_vf_for_slp, vect_analyze_loop_operations)
8867         (vect_is_slp_reduction, vectorizable_induction)
8868         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
8869         * tree-vect-patterns.c (vect_init_pattern_stmt):
8870         (vect_determine_min_output_precision_1, vect_determine_precisions)
8871         (vect_pattern_recog): Likewise.
8872         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
8873         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
8874         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
8875         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8876         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
8877         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
8878         info field from a loop to a loop_vec_info.
8880 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8882         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
8883         (vec_info::add_stmt): Declare.
8884         * tree-vectorizer.c (vec_info::add_stmt): New function.
8885         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
8886         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
8887         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8888         (vectorizable_induction): Likewise.
8889         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
8890         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8891         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
8892         (vectorizable_load): Likewise.
8893         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8894         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
8895         (vect_recog_gather_scatter_pattern): Likewise.
8896         (append_pattern_def_seq): Likewise.  Remove a check that is
8897         performed by add_stmt itself.
8899 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8901         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
8902         which make_ssa_name was called with new_stmt before new_stmt
8903         had been created.
8905 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8907         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
8908         split out from...
8909         (vect_is_slp_reduction): ...here...
8910         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
8911         that are already known to be false.
8913 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8915         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
8916         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
8917         STMT_VINFO_NUM_SLP_USES when it's true.
8918         (vect_free_slp_instance): Add a final_p parameter and pass it to
8919         vect_free_slp_tree.
8920         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
8921         (vect_analyze_slp_instance): Likewise.
8922         (vect_slp_analyze_operations): Likewise.
8923         (vect_slp_analyze_bb_1): Likewise.
8924         * tree-vectorizer.c (vec_info): Likewise.
8925         * tree-vect-loop.c (vect_transform_loop): Likewise.
8927 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8929         * tree-vect-loop.c (vectorizable_reduction): Assert that the
8930         function is not called for second and subsequent members of
8931         a reduction group.
8933 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8935         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
8936         cases for nested loops from here to ...
8937         (vect_create_epilog_for_reduction): ...here.  Only call
8938         vect_is_simple_use for inner-loop reductions.
8940 2018-07-31  Martin Liska  <mliska@suse.cz>
8942         PR gcov-profile/85338
8943         PR gcov-profile/85350
8944         PR gcov-profile/85372
8945         * profile.c (struct location_triplet): New.
8946         (struct location_triplet_hash): Likewise.
8947         (output_location): Do not output a BB that
8948         is already recorded for a line.
8949         (branch_prob): Use streamed_locations.
8951 2018-07-31  Martin Liska  <mliska@suse.cz>
8953         PR gcov-profile/85370
8954         * coverage.c (coverage_begin_function): Do not mark target
8955         clones as artificial functions.
8957 2018-07-31  Martin Liska  <mliska@suse.cz>
8959         PR gcov-profile/83813
8960         PR gcov-profile/84758
8961         PR gcov-profile/85217
8962         PR gcov-profile/85332
8963         * profile.c (branch_prob): Do not record GOTO expressions
8964         for GIMPLE statements which locations are already streamed.
8966 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8968         * gcc.c (handle_spec_function): Accept a soft_matched_part
8969         argument, as do_spec_1.  Pass it down to ...
8970         (eval_spec_function): Accept a soft_matched_part argument,
8971         and pass it down to ...
8972         (do_spec_2): Accept a soft_matched_part argument, and pass
8973         it down to do_spec_1.
8974         (do_spec_1): Pass soft_matched_part to handle_spec_function.
8975         (handle_braces): Update call to handle_spec_function.
8976         (driver::set_up_specs): Update calls to do_spec_2.
8977         (compare_debug_dump_opt_spec_function): Likewise.
8978         (compare_debug_self_opt_spec_function): Likewise.
8980 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8982         * common.opt (nolibc): New option.
8983         * doc/invoke.texi (Link Options): Document it.
8984         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
8985         * config/alpha/linux.h: Likewise.
8986         * config/arc/elf.h: Likewise.
8987         * config/arm/uclinux-elf.h: Likewise.
8988         * config/arm/unknown-elf.h: Likewise.
8989         * config/avr/avrlibc.h: Likewise.
8990         * config/bfin/bfin.h: Likewise.
8991         * config/bfin/linux.h: Likewise.
8992         * config/bfin/uclinux.h: Likewise.
8993         * config/darwin.h: Likewise.
8994         * config/darwin10.h: Likewise.
8995         * config/darwin12.h: Likewise.
8996         * config/gnu-user.h: Likewise.
8997         * config/lm32/uclinux-elf.h: Likewise.
8998         * config/pa/pa-hpux11.h: Likewise.
8999         * config/pa/pa64-hpux.h: Likewise.
9000         * config/sparc/sparc.h: Likewise.
9002 2018-07-31  Olivier Hainque  <hainque@adacore.com>
9004         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
9005         undefined variables.
9007 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9009         PR target/86640
9010         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
9011         instead of GEN_INT.
9013 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9015         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
9016         terminated string literal.
9018 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
9020         PR rtl-optimization/85160
9021         * combine.c (is_just_move): New function.
9022         (try_combine): Allow combining two instructions into two if neither of
9023         the original instructions was a move.
9025 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
9027         PR target/86673
9028         * doc/extend.texi (Global Register Variables): Discourage use of type
9029         qualifiers.
9030         (Local Register Variables): Likewise.
9032 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
9034         PR tree-optimization/86506
9035         * hwint.h (ceil_log2): Resync with hwint.c implementation.
9037 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
9039         PR target/86547
9040         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
9041         hard_regno, make sure no insn between `from` and `to` clobbers it.
9043 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
9044             Tom de Vries  <tdevries@suse.de>
9046         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
9047         (PTX_DEFAULT_RUNTIME_DIM): ... this.
9048         (nvptx_goacc_validate_dims): Set default worker and gang dims to
9049         PTX_DEFAULT_RUNTIME_DIM.
9050         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
9052 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
9054         * config/pa/pa.c (pa_output_addr_vec): Align address table.
9055         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
9056         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
9058 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
9060         * config/rs6000/constraints.md (wG constraint): Delete, no longer
9061         used.
9062         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
9063         predicate to reflect toc fusion has been deleted.
9064         (toc_fusion_mem_raw): Delete, no longer used.
9065         (toc_fusion_mem_wrapped): Likewise.
9066         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
9067         fusion mask bit.
9068         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
9069         Delete, no longer used.
9070         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
9071         meant to be used for toc fusion.
9072         (rs6000_debug_print_mode): Delete toc fusion debugging.
9073         (rs6000_debug_reg_global): Likewise.
9074         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
9075         fusion and secondary reload support that were never used.
9076         (rs6000_option_override_internal): Delete TOC fusion, that was only
9077         partially defined, and it did not work unless you also used the
9078         -mcmodel= switch.
9079         (rs6000_legitimate_address_p): Delete TOC fusion support.
9080         (rs6000_opt_masks): Likewise.
9081         (fusion_wrap_memory_address): Delete function, no longer used.
9082         (fusion_split_address); Delete TOC fusion support.
9083         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
9084         longer used with toc fusion being deleted.
9085         (TARGET_TOC_FUSION_FP): Likewise.
9086         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
9087         UNSPEC.
9088         (toc fusion spliter): Delete TOC fusion support.
9089         (toc_fusionload_<mode>): Likewise.
9090         (toc_fusionload_di): Likewise.
9091         (fusion_gpr_load_<mode>): Delete generator function, this insn no
9092         longer needs to be named.  Rename predicate to delete TOC fusion.
9093         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9094         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9095         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
9096         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
9097         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
9099 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9101         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
9102         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
9103         __int128 in built-in function prototypes.
9104         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
9105         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
9107 2018-07-27  Martin Sebor  <msebor@redhat.com>
9109         PR tree-optimization/86696
9110         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
9111         types, including enums.
9112         (handle_char_store): Be prepared for the above function to fail.
9114 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
9116         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
9117         when optimization level is lower than 2 or optimize for size.
9119 2018-07-26  Martin Sebor  <msebor@redhat.com>
9121         PR tree-optimization/86043
9122         PR tree-optimization/86042
9123         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
9124         (get_string_cst_length): Rename...
9125         (get_min_string_length): ...to this.  Add argument.
9126         (handle_char_store): Extend to handle multi-character stores by
9127         MEM_REF.
9128         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
9129         * tree.h (initializer_zerop): Add argument.
9131 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
9133         PR middle-end/86660
9134         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
9135         declare target to variables if they have always,{to,from,tofrom} map
9136         kinds.
9138 2018-07-26  Martin Liska  <mliska@suse.cz>
9140         PR lto/86548
9141         * lto-wrapper.c: Add linker_output as prefix
9142         for ltrans_output_file.
9144 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
9146         PR rtl-optimization/85805
9147         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
9148         value for hard registers if that was written in the same mode.
9150 2018-07-26  Martin Liska  <mliska@suse.cz>
9152         PR gcov-profile/86536
9153         * gcov.c (format_gcov): Use printf format %.*f directly
9154         and do not handle special values.
9156 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9158         * common/config/arc/arc-common.c (arc_option_optimization_table):
9159         Update default optimizations for size.
9161 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9163         * config/arc/arc.md (movsf_insn): Add short instruction selection.
9164         * config/arc/constraints.md (CfZ): New constraint.
9165         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
9166         (subsf3_fpu): Likewise.
9167         (cmpsf_fpu): Likewise.
9168         (cmpsf_fpu_uneq): Likewise.
9170 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9172         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
9173         accesses as well.
9174         (arc_is_uncached_mem_p): uncached applies to both the variable and
9175         the pointer.
9177 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
9179         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
9180         register names.
9182 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
9184         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
9185         field "m_scopes" from vec to auto_vec.
9187 2018-07-25  Martin Liska  <mliska@suse.cz>
9189         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
9190         return type.
9192 2018-07-25  Richard Biener  <rguenther@suse.de>
9194         PR debug/86654
9195         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
9196         special wrt context_die late.
9197         (gen_subprogram_die): Re-use DIEs in local scope.
9199 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
9201         PR tree-optimization/86644
9202         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
9204 2018-07-25  Martin Liska  <mliska@suse.cz>
9206         PR middle-end/86645
9207         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
9208         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
9210 2018-07-25  Martin Liska  <mliska@suse.cz>
9212         PR sanitizer/79635
9213         * params.def: Explain ASan abbreviation and provide
9214         a documentation link.
9216 2018-07-24  Martin Sebor  <msebor@redhat.com>
9218         PR tree-optimization/86622
9219         PR tree-optimization/86532
9220         * builtins.h (string_length): Declare.
9221         * builtins.c (c_strlen): Correct handling of non-constant offsets.
9222         (check_access): Be prepared for non-constant length ranges.
9223         (string_length): Make extern.
9224         * expr.c (string_constant): Only handle the minor non-constant
9225         array index.  Use string_constant to compute the length of
9226         a generic string constant.
9228 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
9230         PR tree-optimization/86618
9231         * tree-vect-stmts.c (vectorizable_call): Don't take the address
9232         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
9234 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
9236         PR tree-optimization/86636
9237         * json.cc (json::object::set): Fix comment.  Add assertions.
9238         (json::array::append): Move here from json.h.  Add comment and an
9239         assertion.
9240         (json::string::string): Likewise.
9241         * json.h (json::array::append): Move to json.cc.
9242         (json::string::string): Likewise.
9243         * optinfo-emit-json.cc
9244         (optrecord_json_writer::impl_location_to_json): Assert that we
9245         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
9246         wrapper around it.  Expand the location once, rather than three
9247         times.
9248         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
9249         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
9250         wrappers.
9251         (optrecord_json_writer::optinfo_to_json): Likewise, in four
9252         places.  Fix some overlong lines.
9254 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
9256         * config/aarch64/aarch64-simd.md
9257         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
9258         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9259         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9260         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9261         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9262         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9263         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9264         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9265         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9267 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
9269         PR middle-end/86627
9270         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9271         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
9272         and abs_d == d, do the power of two handling if profitable.
9274 2018-07-24  Richard Biener  <rguenther@suse.de>
9276         * match.pd: Add BIT_FIELD_REF canonicalizations.
9278 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9280         PR c/86617
9281         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9283 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9285         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9286         terminated STRING_CST object.
9288 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9290         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9292 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9294         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9295         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9296         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9297         the elements into a register.
9298         (rs6000_split_v4si_init_di_reg): Delete.
9299         (rs6000_split_v4si_init): Delete.
9300         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9301         (vsx_init_v4si): Rewrite as a define_expand.
9303 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9305         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9306         zero_extend argument from memory): New.
9308 2018-07-22  Martin Sebor  <msebor@redhat.com>
9310         PR bootstrap/86621
9311         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9312         diagnosing calls with unknown arguments unless -Walloca-larger-than
9313         is restricted to less than PTRDIFF_MAX bytes.
9315 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
9317         * doc/gcov.texi (Invoking Gcov): Editorial changes.
9319 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9321         * pretty-print.c (text_info::set_location): Remove redundant
9322         "line_table" parameter from call to rich_location::set_range.
9324 2018-07-20  Martin Sebor  <msebor@redhat.com>
9326         PR middle-end/82063
9327         * builtins.c (expand_builtin_alloca): Adjust.
9328         * calls.c (alloc_max_size): Simplify.
9329         * cgraphunit.c (cgraph_node::expand): Adjust.
9330         * common.opt (larger_than_size, warn_frame_larger_than): Remove
9331         variables.
9332         (frame_larger_than_size): Same.
9333         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9334         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9335         Initialize.
9336         * doc/invoke.texi (GCC Command Options): Document option arguments.
9337         Explain byte-size arguments and suffixes.
9338         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9339         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9340         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9341         * doc/options.texi (UInteger): Expand.
9342         (Host_Wide_Int, ByteSize): Document new properties.
9343         * final.c (final_start_function_1): Include sizes in an error message.
9344         * function.c (frame_offset_overflow): Same.
9345         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9346         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9347         Diagnose unbounded alloca calls only for limits of less than
9348         PTRDIFF_MAX.
9349         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
9350         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
9351         for alloca(0).
9352         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
9353         only for limits of less than PTRDIFF_MAX.
9354         * langhooks-def.h (lhd_handle_option): Change function argument
9355         to HOST_WIDE_INT.
9356         * langhooks.c (lhd_handle_option): Same.
9357         * langhooks.h (handle_option): Same.
9358         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9359         ByteSize flags.
9360         (var_type, var_type_struct): Same.
9361         (var_set): Handle ByteSize flag.
9362         * optc-gen.awk: Add comments to output to ease debugging.  Make
9363         use of HOST_WIDE_INT where appropriate.
9364         * opts-gen-save.awk:  Use %lx to format unsigned long.
9365         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9366         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9367         arguments.  Parse bytes-size suffixes.
9368         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9369         (enum_value_to_arg): Same.
9370         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
9371         (handle_option): Adjust.
9372         (generate_option): Change function argument to HOST_WIDE_INT.
9373         (cmdline_handle_error): Adjust.
9374         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9375         (set_option): Change function argument to HOST_WIDE_INT.
9376         (option_enabled): Handle cl_host_wide_int.
9377         (get_option_state): Handle CLVC_SIZE.
9378         (control_warning_option): Same.
9379         * opts.c (common_handle_option): Change function argument to
9380         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
9381         OPT_Wvla_larger_than_.
9382         * opts.h (enum cl_var_type): Add an enumerator.
9383         * stor-layout.c (layout_decl): Print a more meaningful warning.
9384         * toplev.c (output_stack_usage): Adjust.
9386 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
9388         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9389         call to inline_expand_builtin_string_cmp.
9390         (expand_builtin_strcmp): Likewise.
9391         (expand_builtin_strncmp): Likewise.
9392         (inline_string_cmp): Delete the last parameter, change char_type_node
9393         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9394         two operands.
9395         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9396         the inlining expansion on target where the type of the call has same or
9397         narrower precision than unsigned char.
9399 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9401         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9402         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9403         * common.opt (fsave-optimization-record): New option.
9404         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9405         * doc/invoke.texi (-fsave-optimization-record): New option.
9406         * dumpfile.c: Include "optinfo-emit-json.h".
9407         (struct kv_pair): Move to coretypes.h.
9408         (optgroup_options): Make non-static.
9409         (dump_context::end_scope): Call
9410         optimization_records_maybe_pop_dump_scope.
9411         * dumpfile.h (optgroup_options): New decl.
9412         * json.cc: New file.
9413         * json.h: New file.
9414         * optinfo-emit-json.cc: New file.
9415         * optinfo-emit-json.h: New file.
9416         * optinfo.cc: Include "optinfo-emit-json.h".
9417         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9418         (optinfo_enabled_p): Check optimization_records_enabled_p.
9419         (optinfo_wants_inlining_info_p): Likewise.
9420         * optinfo.h: Update comment.
9421         * profile-count.c (profile_quality_as_string): New function.
9422         * profile-count.h (profile_quality_as_string): New decl.
9423         (profile_count::quality): New accessor.
9424         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9425         and optinfo_emit_json_cc_tests.
9426         * selftest.h (selftest::json_cc_tests): New decl.
9427         (selftest::optinfo_emit_json_cc_tests): New decl.
9428         * toplev.c: Include "optinfo-emit-json.h".
9429         (compile_file): Call optimization_records_finish.
9430         (do_compile): Call optimization_records_start.
9431         * tree-ssa-live.c: Include optinfo.h.
9432         (remove_unused_scope_block_p): Retain inlining information if
9433         optinfo_wants_inlining_info_p returns true.
9435 2018-07-20  Richard Biener  <rguenther@suse.de>
9437         PR debug/86585
9438         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9439         to cover -flto-partition=none.
9441 2018-07-20  Martin Liska  <mliska@suse.cz>
9443         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
9444         (get_decl_source_range): Remove unused function.
9446 2018-07-20  Richard Biener  <rguenther@suse.de>
9448         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
9449         (struct vn_phi_s): Likewise.
9450         (struct vn_reference_s): Likewise.
9451         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
9452         for searching the slot of an entry known to be in the hash itself.
9453         (vn_phi_hasher::equal): Likewise.
9454         (vn_reference_hasher::equal): Likewise.
9455         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
9456         globals.
9457         (optimistic_info, current_info): Remove, keeping only valid_info.
9458         (vn_reference_lookup_1): Remove fallback lookup.
9459         (vn_reference_lookup_2): Likewise.
9460         (vn_nary_op_lookup_1): Likewise.
9461         (vn_phi_lookup): Likewise.
9462         (vn_nary_build_or_lookup_1): Make sure to not chain the built
9463         hash element.
9464         (vn_reference_insert): Adjust, chain the inserted hash element
9465         at last_inserted_ref.
9466         (vn_reference_insert_pieces): Likewise.
9467         (visit_reference_op_call): Likewise.
9468         (vn_nary_op_insert_into): Chain the inserted hash element at
9469         last_inserted_nary.
9470         (vn_nary_op_insert_pieces): Adjust.
9471         (vn_nary_op_insert): Likewise.
9472         (vn_nary_op_insert_stmt): Likewise.
9473         (vn_phi_insert): Adjust, chain the inserted hash element at
9474         last_inserted_phi.
9475         (process_scc): Remove clearing and copying the optimistic
9476         table.  Instead remove elements inserted during an optimistic
9477         iteration from the single table we maintain.
9478         (init_scc_vn): Adjust.
9479         (free_scc_vn): Likewise.
9480         (sccvn_dom_walker::record_cond): Likewise.
9481         (sccvn_dom_walker::after_dom_children): Likewise.
9483 2018-07-19  Martin Sebor  <msebor@redhat.com>
9485         PR tree-optimization/84047
9486         PR tree-optimization/83776
9487         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
9488         (check_array_bounds): Call it.
9490 2018-07-19  Martin Sebor  <msebor@redhat.com>
9492         * align.h (align_flags): Use member initialization.
9494 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
9496         * Makefile.in (OBJS): Add optinfo.o.
9497         * coretypes.h (class symtab_node): New forward decl.
9498         (struct cgraph_node): New forward decl.
9499         (class varpool_node): New forward decl.
9500         * dump-context.h: New file.
9501         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
9502         "tree-pass.h".
9503         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
9504         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
9505         (set_alt_dump_file): Likewise.
9506         (dump_context::~dump_context): New dtor.
9507         (dump_gimple_stmt): Move implementation to...
9508         (dump_context::dump_gimple_stmt): ...this new member function.
9509         Add the stmt to any pending optinfo, creating one if need be.
9510         (dump_gimple_stmt_loc): Move implementation to...
9511         (dump_context::dump_gimple_stmt_loc): ...this new member function.
9512         Start a new optinfo and add the stmt to it.
9513         (dump_gimple_expr): Move implementation to...
9514         (dump_context::dump_gimple_expr): ...this new member function.
9515         Add the stmt to any pending optinfo, creating one if need be.
9516         (dump_gimple_expr_loc): Move implementation to...
9517         (dump_context::dump_gimple_expr_loc): ...this new member function.
9518         Start a new optinfo and add the stmt to it.
9519         (dump_generic_expr): Move implementation to...
9520         (dump_context::dump_generic_expr): ...this new member function.
9521         Add the tree to any pending optinfo, creating one if need be.
9522         (dump_generic_expr_loc): Move implementation to...
9523         (dump_context::dump_generic_expr_loc): ...this new member
9524         function.  Add the tree to any pending optinfo, creating one if
9525         need be.
9526         (dump_printf): Move implementation to...
9527         (dump_context::dump_printf_va): ...this new member function.  Add
9528         the text to any pending optinfo, creating one if need be.
9529         (dump_printf_loc): Move implementation to...
9530         (dump_context::dump_printf_loc_va): ...this new member function.
9531         Start a new optinfo and add the stmt to it.
9532         (dump_dec): Move implementation to...
9533         (dump_context::dump_dec): ...this new member function.  Add the
9534         value to any pending optinfo, creating one if need be.
9535         (dump_context::dump_symtab_node): New member function.
9536         (dump_context::get_scope_depth): New member function.
9537         (dump_context::begin_scope): New member function.
9538         (dump_context::end_scope): New member function.
9539         (dump_context::ensure_pending_optinfo): New member function.
9540         (dump_context::begin_next_optinfo): New member function.
9541         (dump_context::end_any_optinfo): New member function.
9542         (dump_context::s_current): New global.
9543         (dump_context::s_default): New global.
9544         (dump_scope_depth): Delete global.
9545         (dumpfile_ensure_any_optinfo_are_flushed): New function.
9546         (dump_symtab_node): New function.
9547         (get_dump_scope_depth): Reimplement in terms of dump_context.
9548         (dump_begin_scope): Likewise.
9549         (dump_end_scope): Likewise.
9550         (selftest::temp_dump_context::temp_dump_context): New ctor.
9551         (selftest::temp_dump_context::~temp_dump_context): New dtor.
9552         (selftest::verify_item): New function.
9553         (ASSERT_IS_TEXT): New macro.
9554         (ASSERT_IS_TREE): New macro.
9555         (ASSERT_IS_GIMPLE): New macro.
9556         (selftest::test_capture_of_dump_calls): New test.
9557         (selftest::dumpfile_c_tests): Call it.
9558         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
9559         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
9560         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
9561         descriptive comment.
9562         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
9563         (dump_node, dump_bb): Move these unrelated decls.
9564         (class dump_manager): Add leading comment.
9565         * optinfo.cc: New file.
9566         * optinfo.h: New file.
9568 2018-07-19  Michael Collison  <michael.collison@arm.com>
9569             Richard Henderson <rth@redhat.com>
9571         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
9572         (subti): Handle op1 zero.
9573         (subvti4, usub4ti4): New.
9574         (*sub<GPI>3_compare1_imm): New.
9575         (sub<GPI>3_carryinCV): New.
9576         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
9577         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
9579 2018-07-19  Michael Collison  <michael.collison@arm.com>
9580             Richard Henderson <rth@redhat.com>
9582         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
9583         (addti3): Create simpler code if low part is already known to be 0.
9584         (addvti4, uaddvti4): New.
9585         (*add<GPI>3_compareC_cconly_imm): New.
9586         (*add<GPI>3_compareC_cconly): New.
9587         (*add<GPI>3_compareC_imm): New.
9588         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
9589         handle constants within this pattern..
9590         (*add<GPI>3_compareV_cconly_imm): New.
9591         (*add<GPI>3_compareV_cconly): New.
9592         (*add<GPI>3_compareV_imm): New.
9593         (add<GPI>3_compareV): New.
9594         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
9595         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
9596         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
9597         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
9598         with 'comparison' operator.
9599         (*add<GPI>3_compareV_cconly_imm): Ditto.
9600         (*add<GPI>3_compareV_cconly): Ditto.
9601         (*add<GPI>3_compareV_imm): Ditto.
9602         (add<GPI>3_compareV): Ditto.
9603         (add<mode>3_carryinC): Ditto.
9604         (*add<mode>3_carryinC_zero): Ditto.
9605         (*add<mode>3_carryinC): Ditto.
9606         (add<mode>3_carryinV): Ditto.
9607         (*add<mode>3_carryinV_zero): Ditto.
9608         (*add<mode>3_carryinV): Ditto.
9610 2018-07-19  Michael Collison  <michael.collison@arm.com>
9611             Richard Henderson <rth@redhat.com>
9613         * config/aarch64/aarch64-modes.def (CC_V): New.
9614         * config/aarch64/aarch64-protos.h
9615         (aarch64_addti_scratch_regs): Declare
9616         (aarch64_subvti_scratch_regs): Declare.
9617         (aarch64_expand_subvti): Declare.
9618         (aarch64_gen_unlikely_cbranch): Declare
9619         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
9620         for signed overflow using CC_Vmode.
9621         (aarch64_get_condition_code_1): Handle CC_Vmode.
9622         (aarch64_gen_unlikely_cbranch): New function.
9623         (aarch64_addti_scratch_regs): New function.
9624         (aarch64_subvti_scratch_regs): New function.
9625         (aarch64_expand_subvti): New function.
9627 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9629         * config/aarch64/aarch64-option-extensions.def: New entry for profile
9630         extension.
9631         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
9632         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
9633         extension.
9635 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9637         PR target/83009
9638         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9639         address check not strict.
9641 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9643         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
9644         Umq with Umn.
9645         (store_pair_lanes<mode>): Likewise.
9646         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
9647         enum value 'ADDR_QUERY_LDP_STP_N'.
9648         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
9649         (aarch64_print_address_internal): Add declaration.
9650         (aarch64_print_ldpstp_address): Remove.
9651         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
9652         (aarch64_print_operand): Change printing of 'y'.
9653         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
9654         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
9655         'true' rather than '1'.
9656         * config/aarch64/constraints.md (Uml): Likewise.
9657         (Uml): Rename to Umn.
9658         (Umq): Remove.
9660 2018-07-19  Richard Biener  <rguenther@suse.de>
9662         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
9663         a trailing array.
9664         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
9665         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
9666         (vn_reference_hasher): Likewise.
9667         (struct vn_tables_s): Remove obstack and alloc-pool members.
9668         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
9669         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
9670         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
9671         (vn_reference_insert_pieces): Likewise.
9672         (alloc_vn_nary_op_noinit): Adjust.
9673         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
9674         (vn_phi_eq): Adjust.
9675         (shared_lookup_phiargs): Remove.
9676         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
9677         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
9678         (visit_reference_op_call): Likewise.
9679         (copy_nary, copy_phi, copy_reference): Remove.
9680         (process_scc): Rewind the obstack when iterating.  Do not
9681         copy the elements to valid_info but just move them from one
9682         hashtable to the other.
9683         (allocate_vn_table): Adjust.
9684         (free_vn_table): Likewise.
9685         (init_scc_vn): Likewise.
9686         (free_scc_vn): Likewise.
9688 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
9690         PR target/86560
9691         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
9692         indirect_return as function type attribute.
9693         (ix86_attribute_table): Change indirect_return to function
9694         type attribute.
9695         * doc/extend.texi: Update indirect_return attribute.
9697 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
9699         * wide-int.h (widest2_int): New.
9700         * gimple-fold.c (arith_overflowed_p): Use it.
9701         * tree.h (widest2_int_cst): New.
9702         * tree-vrp.c (wide_int_binop_overflow): Rename from
9703         vrp_int_const_binop.
9704         Rewrite to work on trees.
9705         (extract_range_from_multiplicative_op_1): Abstract code to...
9706         (wide_int_range_min_max): ...here.
9707         (wide_int_range_cross_product): ...and here.
9708         (extract_range_from_binary_expr_1): Abstract overflow code to...
9709         (wide_int_range_mult_wrapping): ...here.
9710         * tree-vrp.h (wide_int_range_cross_product): New.
9711         (wide_int_range_mult_wrapping): New.
9713 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
9714             Julia Koval  <julia.koval@intel.com>
9716         * config/i386/x86-tune-costs.h (skylake_memcpy,
9717         skylake_memset): Replace rep_prefix with unrolling for size 512.
9719 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
9721         PR middle-end/86544
9722         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
9723         comparision with EQ_EXPR in last stmt.
9725 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9727         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
9728         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
9729         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
9730         previously known as "PowerPC AltiVec Built-in Functions".  Move
9731         some material to new subsubsections "PowerPC AltiVec Built-in
9732         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
9733         ISA 2.07".
9734         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
9735         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
9736         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
9737         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
9739 2018-07-18  Richard Biener  <rguenther@suse.de>
9741         PR tree-optimization/86557
9742         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
9743         EXACT_DIV_EXPR.
9745 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
9747         * config/s390/s390.c (s390_function_profiler): Generate CFI.
9749 2018-07-17  Jeff Law  <law@redhat.com>
9751         * config/arm/arm.c (get_label_padding): Update for recent
9752         changes to label_to_alignment.
9754         PR tree-optimization/86010
9755         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
9757         * config/mips/mips.c (vr4130_align_insns): Update for recent
9758         changes to label_to_alignment.
9760         * config/frv/frv.c (frv_label_align): Update for recent changes
9761         to label_to_alignment.
9763         * config/nios2/nios2.c (nios2_label_align): Update for recent
9764         changes which dropped ALIGN_LABELS_LOG.
9766 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
9768         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
9769         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
9771 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
9773         * config/arc/arc.c (arc_label_align): Use align_labels instead of
9774         deprecated align_labels_log.
9776 2018-07-17  Richard Biener  <rguenther@suse.de>
9778         PR lto/86456
9779         * dwarf2out.c (init_sections_and_labels): Always generate
9780         a debug_line_str_section for early LTO debug.
9781         (dwarf2out_finish): Reset debug_line_str_hash output early.
9782         Bump counter for extra dwarf5 .debug_loc labels to not conflict
9783         with fat LTO part.
9784         (dwarf2out_early_finish): Output debug_line_str.
9786 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9788         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
9789         index register on z196 or later.
9791 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9793         * config/s390/s390.c (s390_default_align): Set default function
9794         alignment to 16.
9795         (s390_override_options_after_change): Call s390_default align.
9796         (s390_option_override_internal): Call s390_default_align.
9797         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
9799 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9801         PR middle-end/86542
9802         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
9803         to _looptemp_ clauses, other than the first two.
9805 2018-07-17  Martin Liska  <mliska@suse.cz>
9807         * opts.c: Do not enable OPT_falign_* for -Os.
9809 2018-07-17  Martin Liska  <mliska@suse.cz>
9811         * align.h (MAX_CODE_ALIGN): New.
9812         (MAX_CODE_ALIGN_VALUE): New.
9813         * common/config/i386/i386-common.c (ix86_handle_option):
9814         (MAX_CODE_ALIGN): Moved to align.h.
9815         * final.c (MAX_CODE_ALIGN): Likewise.
9816         * opts.c (parse_and_check_align_values):
9817         (MAX_CODE_ALIGN): Likewise.
9818         (MAX_CODE_ALIGN_VALUE): Likewise.
9820 2018-07-17  Martin Liska  <mliska@suse.cz>
9822         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
9823         in order to fulfil coding style.
9824         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
9825         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9826         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9827         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
9828         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
9829         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
9830         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
9831         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
9832         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9834 2018-07-17  Martin Liska  <mliska@suse.cz>
9836         * align.h: New file.
9837         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
9838         directly.
9839         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
9840         align_flags of label_to_alignment.
9841         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
9842         align_flags class.
9843         * config/m68k/m68k.c: Do not use removed align_labels_value and
9844         align_loops_value.
9845         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
9846         (LOOP_ALIGN): Likewise.
9847         (LABEL_ALIGN): Likewise.
9848         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9849         Remove not used macro.
9850         (rs6000_loop_align): Change return type to align_flags.
9851         (rs6000_loop_align_max_skip): Remove.
9852         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
9853         Change return type to align_flags.
9854         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9855         Remove not used macro.
9856         (rs6000_loop_align):  Change return type to align_flags.
9857         (rs6000_loop_align_max_skip): Remove.
9858         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
9859         * config/rx/rx-protos.h (rx_align_for_label): Make it
9860         static function.
9861         * config/rx/rx.c (rx_align_for_label): Change return type
9862         to align_flags.
9863         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
9864         macro definitions.
9865         into align_flags class.
9866         (LABEL_ALIGN): Likewise.
9867         (LOOP_ALIGN): Likewise.
9868         * config/s390/s390.c (s390_label_align): Use align_flags
9869         class member.
9870         (s390_asm_output_function_label): Likewise.
9871         * config/sh/sh.c (sh_override_options_after_change):
9872         Use align_flags class directly without macros.
9873         (find_barrier): Likewise.
9874         (barrier_align): Likewise.
9875         (sh_loop_align): Likewise.
9876         * config/spu/spu.c (spu_option_override):
9877         Use align_flags_tuple::get_value instead of removed macros.
9878         (spu_sched_init): Likewise.
9879         * config/spu/spu.h (GTY): Likewise.
9880         * config/visium/visium.c (visium_option_override):
9881         Set "8" as default secondary alignment.
9882         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
9883         in order to guarantee secondary alignment of 8.
9884         * coretypes.h: Include align.h header file.
9885         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
9886         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
9887         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
9888         * doc/tm.texi.in: Likewise.
9889         * final.c (struct label_alignment): Remove not used structure.
9890         (LABEL_ALIGN): Change type to align_flags.
9891         (LOOP_ALIGN): Likewise.
9892         (JUMP_ALIGN): Likewise.
9893         (default_loop_align_max_skip): Remove.
9894         (default_label_align_max_skip): Likewise.
9895         (default_jump_align_max_skip): Likewise.
9896         (default_label_align_after_barrier_max_skip):
9897         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
9898         (LABEL_TO_MAX_SKIP): Remove.
9899         (label_to_alignment): Return align_flags type instead of integer.
9900         (label_to_max_skip): Remove.
9901         (align_fuzz): Use align_flags type.
9902         (compute_alignments): Use align_flags type and use align_flags::max
9903         to combine multiple alignments.
9904         (grow_label_align): Grow vec instead of C array.
9905         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
9906         (shorten_branches):  Use align_flags type and use align_flags::max
9907         to combine multiple alignments.
9908         (final_scan_insn_1): Remove usage of secondary alignment that comes
9909         from label alignment, but instead use proper secondary alignment
9910         which is computed in grow_label_align.
9911         * flags.h (struct align_flags_tuple): Move to align.h.
9912         (struct align_flags): Likewise.
9913         (state_align_loops): Rename to align_loops.
9914         (state_align_jumps): Rename to align_jumps.
9915         (state_align_labels): Rename to align_labels.
9916         (state_align_functions): Rename to align_functions.
9917         (align_loops_log): Remove.
9918         (align_jumps_log): Remove.
9919         (align_labels_log): Remove.
9920         (align_functions_log): Remove.
9921         (align_loops_max_skip): Remove.
9922         (align_jumps_max_skip): Remove.
9923         (align_labels_max_skip): Remove.
9924         (align_functions_max_skip): Remove.
9925         (align_loops_value): Remove.
9926         (align_jumps_value): Remove.
9927         (align_labels_value): Remove.
9928         (align_functions_value): Remove.
9929         * output.h (label_to_alignment): Change return type to align_flags.
9930         (label_to_max_skip): Remove.
9931         * target.def: Remove loop_align_max_skip, label_align_max_skip,
9932         jump_align_max_skip macros.
9933         * targhooks.h (default_loop_align_max_skip): Remove.
9934         (default_label_align_max_skip): Likewise.
9935         (default_jump_align_max_skip): Likewise.
9936         (default_label_align_after_barrier_max_skip): Remove.
9937         * toplev.c (read_log_maxskip): Use ::normalize function.
9938         (parse_N_M): Remove not used argument and also call ::normalize.
9939         (parse_alignment_opts): Do not pass unused arguments.
9940         * varasm.c (assemble_start_function): Use directly align_functions
9941         instead of removed macros.
9942         * system.h: Do not poison removed macros.
9944 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9946         PR middle-end/86539
9947         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
9948         and cond temporaries don't have reference type if iterator has
9949         pointer type.  For init use &for_pre_body instead of pre_p if
9950         for_pre_body is non-empty.
9952 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9954         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
9955         double-double modes to SFmode directly directly.
9956         (trunc<mode>sf2_fprs): Delete.
9958 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9960         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
9961         for conversions between IFmode and the decimal floating point modes.
9962         (init_float128_ieee): Use the correct names for conversions between
9963         KFmode and the decimal floating point modes.
9965 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9967         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
9968         for the conversions between TDmode and IFmode.
9969         (init_float128_ieee): Use more correct names for the conversions
9970         between TDmode and KFmode.
9972 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
9974         PR tree-optimization/86526
9975         * builtins.c (expand_builtin_memcmp): Formatting fixes.
9976         (inline_expand_builtin_string_cmp): Likewise.
9977         (inline_string_cmp): Likewise.  Use c_readstr instead of
9978         builtin_memcpy_read_str.  Add unit_mode temporary.
9980 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9982         PR middle-end/86528
9983         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
9984         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
9986 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9988         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
9989         Alphabetize prototypes of built-in functions, separating out
9990         built-in functions that are listed in this section but should be
9991         described elsewhere.
9993 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
9995         PR target/86511
9996         * expmed.c (emit_store_flag): Do not emit setcc followed by a
9997         conditional move when trapping comparison was split to a
9998         non-trapping one (and vice versa).
10000 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10002         * config/s390/s390.c (s390_function_profiler): Generate nops
10003         instead of profiler call sequences.
10004         * config/s390/s390.opt: Add the new option.
10006 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10008         * config/s390/s390.c (s390_function_profiler): Generate
10009         __mcount_loc section.
10010         * config/s390/s390.opt: Add the new option.
10012 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
10014         * common.opt: Add the new warning.
10015         * config/s390/s390.c (s390_function_profiler): Emit "brasl
10016         %r0,__fentry__" when -mfentry is specified.
10017         (s390_option_override_internal): Disallow -mfentry for 31-bit
10018         CPUs.
10019         * config/s390/s390.opt: Add the new option.
10021 2018-07-16  Richard Biener  <rguenther@suse.de>
10023         PR lto/86523
10024         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
10025         for function-local FUNCTION_DECL and RESULT_DECL immediately.
10027 2018-07-16  Martin Liska  <mliska@suse.cz>
10029         PR ipa/86529
10030         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
10031         to ::get_create.
10033 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10035         * config/arc/arcHS.md: Update ARCHS scheduling rules.
10037 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
10039         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
10040         for ARCHS4x.
10041         * config/arc/arc-cpus.def (hs4x): New cpu.
10042         (hs4xd): Likewise.
10043         * config/arc/arc-tables.opt: Regenerate.
10044         * config/arc/arc.c (arc_sched_issue_rate): New function.
10045         (TARGET_SCHED_ISSUE_RATE): Define.
10046         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
10047         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
10048         fpu_cvt.
10049         (attr tune): Add ARCHS4x tune values.
10050         (attr tune_dspmpy): Define.
10051         (*tst): Correct instruction type.
10052         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
10053         * config/arc/arcHS4x.md: New file.
10054         * config/arc/fpu.md: Update instruction type attributes.
10055         * config/arc/t-multilib: Regenerate.
10057 2018-07-16  Tom de Vries  <tdevries@suse.de>
10059         PR debug/86455
10060         * var-tracking.c (vt_initialize): Fix pre_dec handling.
10062 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
10064         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
10065         early clobber.
10067 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
10069         PR tree-optimization/86514
10070         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
10071         conversion to a boolean type from a type with greater precision.
10073 2018-07-16  Tom de Vries  <tdevries@suse.de>
10075         * var-tracking.c (vt_initialize): Print adjusted insn slim if
10076         dump_flags request TDF_SLIM.
10078 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
10080         * fold-const.c (int_const_binop_1): Abstract...
10081         (wide_int_binop): ...wide int code here.
10082         (poly_int_binop): ...poly int code here.
10083         Abstract the rest of int_const_binop_1 into int_const_binop.
10084         * fold-const.h (wide_int_binop): New.
10085         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
10086         Remove useless PLUS/MINUS_EXPR case.
10087         (zero_nonzero_bits_from_vr): Move wide int code...
10088         (zero_nonzero_bits_from_bounds): ...here.
10089         (extract_range_from_binary_expr_1): Move mask optimization code...
10090         (range_easy_mask_min_max): ...here.
10091         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
10092         (range_easy_mask_min_max): New.
10094 2018-07-15  Jeff Law  <law@redhat.com>
10096         PR target/85993
10097         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
10098         block.
10100 2018-07-14  Jim Wilson  <jimw@sifive.com>
10102         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
10104 2018-07-14  Paul Koning  <ni1d@arrl.net>
10106         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
10108 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
10110         * lto-streamer-out.c (copy_function_or_variable): Dump info about
10111         copying section.
10113 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
10114             Steve Munroe  <munroesj52@gmail.com>
10116         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
10117         (_mm_andnot_si128): Likewise.
10118         (_mm_or_si128): Likewise.
10119         (_mm_xor_si128): Likewise.
10121 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
10123         PR middle-end/78809
10124         * builtins.c (expand_builtin_memcmp): Inline the calls first
10125         when result_eq is false.
10126         (expand_builtin_strcmp): Inline the calls first.
10127         (expand_builtin_strncmp): Likewise.
10128         (inline_string_cmp): New routine. Expand a string compare
10129         call by using a sequence of char comparison.
10130         (inline_expand_builtin_string_cmp): New routine. Inline expansion
10131         a call to str(n)cmp/memcmp.
10132         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
10133         New option.
10134         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
10136 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
10138         * config/arm/driver-arm.c: Include arm-native.h.
10139         (host_detect_local_cpu): Use auto-generated data tables.
10140         (vendors, arm_cpu_table): Delete.  Move part information to ...
10141         * config/arm/arm-cpus.in: ... here.
10142         * config/arm/parsecpu.awk (gen_native): New function.
10143         (vendor, part): New CPU fields.
10144         (END): Add support for building the native CPU detection tables.
10145         * config/arm/t-arm (arm-native.h): Add build rule.
10146         (driver-arm.o): Add dependency on arm-native.h.
10148 2018-07-13  Richard Biener  <rguenther@suse.de>
10150         PR middle-end/85974
10151         * match.pd (addr1 - addr2): Allow either of the operand to
10152         have a conversion.
10154 2018-07-13  Tom de Vries  <tdevries@suse.de>
10156         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
10157         in remap_ssa_name.
10159 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
10161         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
10162         arrays instead of numbered variables.
10164 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
10166         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
10167         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
10169 2018-07-13  Richard Biener  <rguenther@suse.de>
10171         PR debug/86452
10172         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
10173         instead of get_context_die.
10175 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
10176             Richard Biener  <rguenther@suse.de>
10178         PR middle-end/86489
10179         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
10180         that the loop latch destination where phi is defined.
10182 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
10184         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
10185         (riscv_expand_epilogue): Add assertion to check interrupt mode.
10186         (riscv_set_current_function): Extract getting interrupt type to new
10187         function.
10188         (riscv_get_interrupt_type): New function.
10189         (riscv_merge_decl_attributes): New function, checking interrupt type is
10190         same.
10191         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
10193 2018-07-12  Paul Koning  <ni1d@arrl.net>
10195         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
10196         directive.
10198 2018-07-12  Paul Koning  <ni1d@arrl.net>
10200         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
10201         zero reference, add doloop_end instead.
10202         * doc/md.texi (decrement_and_branch_until_zero): Remove.
10203         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
10204         detail for doloop_end.
10206 2018-07-12  Martin Sebor  <msebor@redhat.com>
10208         PR c/86453
10209         * attribs.c (decl_attributes): Reject conflicting attributes before
10210         calling attribute handlers.
10212 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10214         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
10215          parameter.
10216         (gcc::dump_manager::get_dump_file_name): likewise.
10217         (dump_begin): Likewise.
10218         * dumpfile.h (dump_begin): Update prototype.
10219         (gcc::dump_manager::get_dump_file_name,
10220         gcc::dump_manager::get_dump_file_name): Update prototype.
10222 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10224         * internal-fn.h (vectorizable_internal_fn_p): New function.
10225         * tree-vect-slp.c (compatible_calls_p): Likewise.
10226         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
10227         to internal functions.
10228         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
10230 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10232         * fold-const.h (inverse_conditions_p): Declare.
10233         * fold-const.c (inverse_conditions_p): New function.
10234         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
10235         that test the inverse condition of a conditional internal function.
10236         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
10237         * internal-fn.c (internal_fn_mask_index): Handle conditional
10238         internal functions.
10239         (vectorized_internal_fn_supported_p): New function.
10240         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
10241         (any_pred_load_store): Replace with...
10242         (need_to_predicate): ...this new variable.
10243         (redundant_ssa_names): New variable.
10244         (ifcvt_can_use_mask_load_store): Move initial checks to...
10245         (ifcvt_can_predicate): ...this new function.  Handle tree codes
10246         for which a conditional internal function exists.
10247         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
10248         instead of ifcvt_can_use_mask_load_store.  Update after variable
10249         name change.
10250         (predicate_load_or_store): New function, split out from
10251         predicate_mem_writes.
10252         (check_redundant_cond_expr): New function.
10253         (value_available_p): Likewise.
10254         (predicate_rhs_code): Likewise.
10255         (predicate_mem_writes): Rename to...
10256         (predicate_statements): ...this.  Use predicate_load_or_store
10257         and predicate_rhs_code.
10258         (combine_blocks, tree_if_conversion): Update after above name changes.
10259         (ifcvt_local_dce): Handle redundant_ssa_names.
10260         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10261         general conditional functions.
10262         * tree-vect-stmts.c (vectorizable_call): Likewise.
10264 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10265             Alan Hayward  <alan.hayward@arm.com>
10266             David Sherwood  <david.sherwood@arm.com>
10268         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10269         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10270         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10271         plus and minus and convert them into IFN_COND_FMA-based sequences.
10272         (convert_mult_to_fma): Handle conditional plus and minus.
10274 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10276         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10277         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10278         (cond_fnms_optab): New optabs.
10279         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10280         internal functions.
10281         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10282         * internal-fn.h (get_conditional_internal_fn): Declare.
10283         (get_unconditional_internal_fn): Likewise.
10284         * internal-fn.c (cond_ternary_direct): New macro.
10285         (expand_cond_ternary_optab_fn): Likewise.
10286         (direct_cond_ternary_optab_supported_p): Likewise.
10287         (FOR_EACH_COND_FN_PAIR): Likewise.
10288         (get_conditional_internal_fn): New function.
10289         (get_unconditional_internal_fn): Likewise.
10290         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10291         (gimple_match_op::gimple_match_op): Add a new overload for 5
10292         operands.
10293         (gimple_match_op::set_op): Likewise.
10294         (gimple_resimplify5): Declare.
10295         * genmatch.c (decision_tree::gen): Generate simplifications for
10296         5 operands.
10297         * gimple-match-head.c (gimple_simplify): Define an overload for
10298         5 operands.  Handle calls with 5 arguments in the top-level overload.
10299         (convert_conditional_op): Handle conversions from unconditional
10300         internal functions to conditional ones.
10301         (gimple_resimplify5): New function.
10302         (build_call_internal): Pass a fifth operand.
10303         (maybe_push_res_to_seq): Likewise.
10304         (try_conditional_simplification): Try converting conditional
10305         internal functions to unconditional internal functions.
10306         Handle 3-operand unconditional forms.
10307         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10308         Define ternary equivalents of the current rules for binary conditional
10309         internal functions.
10310         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10311         ternary operations.
10312         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10313         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10314         (optab): Handle them.
10315         (SVE_COND_FP_TERNARY): New int iterator.
10316         (sve_fmla_op, sve_fmad_op): New int attributes.
10317         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10318         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10319         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10321 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10323         * target.def (preferred_else_value): New target hook.
10324         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10325         * doc/tm.texi: Regenerate.
10326         * targhooks.h (default_preferred_else_value): Declare.
10327         * targhooks.c (default_preferred_else_value): New function.
10328         * internal-fn.h (conditional_internal_fn_code): Declare.
10329         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10330         (get_conditional_internal_fn): Use it.
10331         (conditional_internal_fn_code): New function.
10332         * gimple-match.h (gimple_match_cond): New struct.
10333         (gimple_match_op): Add a cond member function.
10334         (gimple_match_op::gimple_match_op): Update all forms to take a
10335         gimple_match_cond.
10336         * genmatch.c (expr::gen_transform): Use the same condition as res_op
10337         for the suboperation, but don't specify a particular else_value.
10338         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10339         (visit_nary_op, visit_reference_op_load): Pass
10340         gimple_match_cond::UNCOND to the gimple_match_op constructor.
10341         * gimple-match-head.c: Include tree-eh.h
10342         (convert_conditional_op): New function.
10343         (maybe_resimplify_conditional_op): Likewise.
10344         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10345         (gimple_resimplify2): Likewise.
10346         (gimple_resimplify3): Likewise.
10347         (gimple_resimplify4): Likewise.
10348         (maybe_push_res_to_seq): Return null for conditional operations.
10349         (try_conditional_simplification): New function.
10350         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
10351         constructor.
10352         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10353         IFN_COND_* call.
10354         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10355         function.
10356         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10358 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10360         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10361         DECL_FCONTEXT
10362         (hash_tree): Do not hash DECL_FCONTEXT
10363         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10364         Do not stream DECL_FCONTEXT.
10365         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10366         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10368 2018-07-12  Richard Biener  <rguenther@suse.de>
10370         PR debug/86462
10371         * dwarf2out.c (gen_block_die): Only output blocks when they have
10372         at least one !DECL_IGNORED_P variable.
10374 2018-07-12  Richard Biener  <rguenther@suse.de>
10376         PR target/84829
10377         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10378         Remove -mieee-fp handling.
10380 2018-07-12  Richard Biener  <rguenther@suse.de>
10382         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10383         left-over from last patch.
10385 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
10387         PR tree-optimization/86492
10388         * gimple-ssa-store-merging.c
10389         (imm_store_chain_info::coalesce_immediate_stores): Call
10390         check_no_overlap even for the merge_overlapping case.  Formatting fix.
10392 2018-07-12  Richard Biener  <rguenther@suse.de>
10394         PR middle-end/86479
10395         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10396         move possibly trapping operations into the conditional.
10398 2018-07-12  Richard Biener  <rguenther@suse.de>
10400         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10401         (vn_lookup_simplify_result): Remove recursion limit applied
10402         here.
10403         (vn_nary_build_or_lookup_1): Adjust.
10404         (try_to_simplify): Likewise.
10405         * gimple-match-head.c (gimple_resimplify1): Instead apply one
10406         here.
10407         (gimple_resimplify2): Likewise.
10408         (gimple_resimplify3): Likewise.
10409         (gimple_resimplify4): Likewise.
10411 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
10413         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10414         Use __mmask64 type instead of __mmask8 for __M argument.
10415         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10416         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10417         __U argument.
10418         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10419         __mmask16 for __M argument.
10420         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10421         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10422         to __mmask16 instead of __mmask8.
10423         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10424         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10425         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10426         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10427         instead of __mmask16 for __U argument.
10428         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10429         __mmask16 instead of __mmask8 for __U argument.
10430         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10431         __U argument.
10432         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10433         __mmask16.
10434         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10435         argument.
10436         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10437         __U argument.
10438         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10439         __mmask16.
10440         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
10441         of __mmask16.
10442         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
10443         __U argument.
10444         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
10445         __U argument.
10446         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
10447         __U argument.
10448         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
10449         __U argument.
10450         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
10451         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
10452         return type as well as __M argument type and all casts from __mmask8
10453         to __mmask32.
10454         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
10455         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
10456         return type as well as __M argument type and all casts from __mmask8
10457         to __mmask16.
10458         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
10459         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
10460         return type as well as __M argument type and all casts from __mmask8
10461         to __mmask32.
10462         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
10463         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
10464         return type as well as __M argument type and all casts from __mmask8
10465         to __mmask16.
10466         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
10467         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
10468         __mmask16.
10470 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
10472         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
10473         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
10474         for __U argument.
10476 2018-07-11  Paul Koning  <ni1d@arrl.net>
10478         * doc/md.texi (define_subst): Document how multiple occurrences of
10479         the same argument in the replacement pattern are handled.
10481 2018-07-11  Paul Koning  <ni1d@arrl.net>
10483         * doc/extend.texi (Common Variable Attributes): Move "mode" into
10484         alphabetical order.
10485         (Common Type Attributes): Add "mode" attribute.
10487 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
10489         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
10490         stream DECL_ORIGINAL_TYPE.
10491         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
10492         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
10493         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
10494         Do not walk original type.
10495         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
10496         external decls.
10497         (write_ts_decl_non_common_tree_pointers): Do not stream
10498         DECL_ORIGINAL_TYPE
10499         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
10500         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
10502 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10504         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
10505         thread twice from the same starting edge.
10507 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10509         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
10510         * gimple.c (gimple_call_nonnull_result_p): ...here...
10511         (gimple_call_nonnull_arg): ...and here.
10512         * gimple.h (gimple_call_nonnull_result_p): New.
10513         (gimple_call_nonnull_arg): New.
10515 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
10517         * config/arm/arm-cpus.in: Move information from fpu field of each
10518         cpu definition to the isa field.
10519         * config/arm/parsecpu.awk (fpu): Delete match rule.
10520         (gen_comm_data): Don't add bits from the CPU's FPU entry.
10522 2018-07-11  Richard Biener  <rguenther@suse.de>
10524         PR debug/86457
10525         * dwarf2out.c (init_sections_and_labels): Use
10526         output_asm_line_debug_info consistently.
10527         (dwarf2out_early_finish): Likewise.
10528         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
10529         type units.
10531 2018-07-11  Richard Biener  <rguenther@suse.de>
10533         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
10534         Rework father_bb setting in a way to avoid propagating constants
10535         multiple times on a loop body.
10537 2018-07-10  Mark Wielaard  <mark@klomp.org>
10539         PR debug/86459
10540         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
10542 2018-07-10  Richard Biener  <rguenther@suse.de>
10544         * hash-map.h (hash_map::iterator::operator*): Return
10545         references to key and value.
10547 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
10549         PR c++/86443
10550         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10551         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
10552         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
10553         (gimplify_omp_for): For composite loops, move outer
10554         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
10555         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
10556         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
10557         TREE_LIST for both the original class iterator and the "last" helper
10558         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
10559         loop, remember has_decl_expr from outer composite loops for the
10560         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
10562 2018-07-09  Martin Sebor  <msebor@redhat.com>
10564         PR middle-end/77357
10565         PR middle-end/86428
10566         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
10567         accessing implicitly initialized array elements.
10568         * expr.c (string_constant): Handle string initializers of
10569         character arrays within aggregates.
10570         * gimple-fold.c (fold_array_ctor_reference): Add argument.
10571         Store element offset.  As a special case, handle zero size.
10572         (fold_nonarray_ctor_reference): Same.
10573         (fold_ctor_reference): Add argument.  Store subobject offset.
10574         * gimple-fold.h (fold_ctor_reference): Add argument.
10576 2018-07-09  Paul Koning  <ni1d@arrl.net>
10578         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
10579         (pdp11_insn_cost): New function.
10580         (pdp11_md_asm_adjust): New function.
10581         (TARGET_INVALID_WITHIN_DOLOOP): Define.
10582         (pdp11_rtx_costs): Update to match machine better.
10583         (output_addr_const_pdp11): Correct format mismatch warnings.
10584         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
10585         * config/pdp11/pdp11.md: General change to add base_cost and/or
10586         length attributes for use by new pdp11_insn_cost function.
10587         (MIN_BRANCH): Correct definition.
10588         (MIN_SOB): Ditto.
10589         (doloop_end): Use standard pattern name for looping pattern.
10590         (doloop_end_nocc): New.
10591         (movsf): Add another constraint alternative.
10592         (zero_extendqihi2): Add constraint alternatives for not in place
10593         extend.
10594         (zero_extendhisi2): Remove.
10595         (shift patterns): Add CC handling variants.
10596         (bswaphi2): New.
10597         (bswapsi2): New.
10598         (rothi3): New.
10599         (define_peephole2): New peephole to recognize mov that sets CC for
10600         subsequent test.
10602 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10604         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
10605         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
10606         wi::add.
10608 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
10610         PR c/86420
10611         * real.c (real_nextafter): Return true if result is denormal.
10613 2018-07-09  Martin Liska  <mliska@suse.cz>
10615         * common.opt: Add back wrongly removed attribute.
10617 2018-07-09  Richard Biener  <rguenther@suse.de>
10619         PR debug/86413
10620         * dwarf2out.c (gen_block_die): For an early generated DIE
10621         always output high/low PC attributes.
10623 2018-07-09  Tom de Vries  <tdevries@suse.de>
10625         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
10626         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
10627         onto VAR_DECL with abstract origin.
10629 2018-07-07  Jim Wilson  <jimw@sifive.com>
10631         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
10633 2018-07-07  Tom de Vries  <tdevries@suse.de>
10635         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
10637 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
10639         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
10640         overflow_type.
10641         (combine_bound): Use wide-int overflow calculation instead of
10642         rolling our own.
10643         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
10644         overflow_type.
10645         * fold-const.c (int_const_binop_2): Same.
10646         (extract_muldiv_1): Same.
10647         (fold_div_compare): Same.
10648         (fold_abs_const): Same.
10649         * match.pd: Same.
10650         * poly-int.h (add): Same.
10651         (sub): Same.
10652         (neg): Same.
10653         (mul): Same.
10654         * predict.c (predict_iv_comparison): Same.
10655         * profile-count.c (slow_safe_scale_64bit): Same.
10656         * simplify-rtx.c (simplify_const_binary_operation): Same.
10657         * tree-chrec.c (tree_fold_binomial): Same.
10658         * tree-data-ref.c (split_constant_offset_1): Same.
10659         * tree-if-conv.c (idx_within_array_bound): Same.
10660         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
10661         * tree-ssa-phiopt.c (minmax_replacement): Same.
10662         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
10663         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
10664         * vr-values.c (vr_values::adjust_range_with_scev): Same.
10665         * wide-int.cc (wi::add_large): Same.
10666         (wi::mul_internal): Same.
10667         (wi::sub_large): Same.
10668         (wi::divmod_internal): Same.
10669         * wide-int.h: Change overflow type to overflow_type for neg, add,
10670         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
10671         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
10672         mul_internal, divmod_internal.
10673         (overflow_type): New enum.
10674         (accumulate_overflow): New.
10676 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
10678         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
10679         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
10681 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
10683         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
10684         argument is checked for zero before entering loop, avoid checking again.
10686 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
10688         * gimplify.h (generic_expr_could_trap_p): Set as global function.
10689         * gimplify.h (generic_expr_could_trap_p): Likwise.
10690         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
10692 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
10694         PR tree-optimization/86401
10695         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
10696         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
10697         (fold_bit_and_mask): ... here.  New helper function for match.pd.
10698         * fold-const.h (fold_bit_and_mask): Declare.
10699         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
10701 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
10703         PR target/86324
10704         * target.def (translate_mode_attribute): New hook.
10705         * targhooks.h (default_translate_mode_attribute): Declare.
10706         * targhooks.c (default_translate_mode_attribute): New function.
10707         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
10708         * doc/tm.texi: Regenerate.
10709         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
10710         (rs6000_translate_mode_attribute): New function.
10712 2018-07-06  Paul Koning  <ni1d@arrl.net>
10714         * doc/md.texi (define_split): Document DONE and FAIL.
10715         (define_peephole2): Ditto.
10717 2018-07-05  Jeff Law  <law@redhat.com>
10719         PR tree-optimization/86010
10720         * tree-ssa-dse.c (compute_trims): More aggressively trim at
10721         both the head and tail of mem* and str* calls.
10723 2018-07-05  Jim Wilson  <jimw@sifive.com>
10725         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
10727 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
10729         * config/aarch64/aarch64-simd.md: correct flags text for
10730         MIN_EXPR replacement.
10732 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
10734         * configure: Regenerated.
10736 2018-07-05  Carl Love  <cel@us.ibm.com>
10738         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
10739         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
10740         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
10741         VSX_BUILTIN_DOUBLEL_V4SF.
10743 2018-07-05  Martin Sebor  <msebor@redhat.com>
10745         PR c++/86400
10746         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
10747         than its domain to compute its the upper bound of a char array.
10749 2018-07-05  Nathan Sidwell  <nathan@acm.org>
10751         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
10752         * doc/cpp.texi: Update comment.
10753         * doc/tm.texi: Rebuilt.
10754         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
10755         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
10756         * doc/extend.texi (Backwards Compatibility): Clarify it is system
10757         headers affected by extern "C".
10758         * system.h: Poison NO_IMPLICIT_EXTERN_C.
10759         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
10760         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
10761         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
10762         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
10763         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
10764         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
10765         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
10766         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
10767         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
10768         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
10769         config/sparc/sp64-elf.h, config/spu/spu.h,
10770         config/stormy16/stormy16.h, config/v850/v850.h,
10771         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
10772         define NO_IMPLICIT_EXTERN_C.
10773         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
10775 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10777         PR target/84711
10778         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10779         instead of GET_MODE_SIZE when comparing Units.
10781 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10783         PR target/84711
10784         * rtlanal.c (set_noop_p): Constrain on mode change,
10785         include hard-reg-set.h
10787 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10789         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
10791 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
10793         Revert
10794         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10796         PR sanitizer/84250
10797         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10798         libasan.
10799         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10801 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10803         PR sanitizer/84250
10804         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10805         libasan.
10806         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10808 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
10810         PR middle-end/86380
10811         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
10813 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
10815         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
10816         neg_*_op* variables.
10818 2018-07-04  Martin Liska  <mliska@suse.cz>
10820         * tree-switch-conversion.c: Define
10821         max_ratio_for_speed and max_ratio_for_size constants.
10823 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10824             Martin Liska  <mliska@suse.cz>
10826         PR middle-end/66240
10827         PR target/45996
10828         PR c/84100
10829         * common.opt: Rename align options with 'str_' prefix.
10830         * common/config/i386/i386-common.c (set_malign_value): New
10831         function.
10832         (ix86_handle_option): Use it to set -falign-* options/
10833         * config/aarch64/aarch64-protos.h (struct tune_params): Change
10834         type from int to string.
10835         * config/aarch64/aarch64.c: Update default values from int
10836         to string.
10837         * config/alpha/alpha.c (alpha_override_options_after_change):
10838         Likewise.
10839         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
10840         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10841         max skip conditionally.
10842         * config/i386/freebsd.h (SUBALIGN_LOG): New.
10843         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10844         max skip conditionally.
10845         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10846         max skip conditionally.
10847         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
10848         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10849         max skip conditionally.
10850         * config/i386/i386.c (struct ptt): Change type from int to
10851         string.
10852         (ix86_default_align): Set default values.
10853         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
10854         max skip conditionally.
10855         * config/i386/iamcu.h (SUBALIGN_LOG): New.
10856         (ASM_OUTPUT_MAX_SKIP_ALIGN):
10857         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
10858         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10859         max skip conditionally.
10860         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
10861         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
10862         * config/i386/x86-64.h (SUBALIGN_LOG): New.
10863         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10864         max skip conditionally.
10865         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10866         * config/ia64/ia64.c (ia64_option_override): Set default values
10867         for alignment options.
10868         * config/m68k/m68k.c: Handle new str_align_* options.
10869         * config/mips/mips.c (mips_set_compression_mode): Change
10870         type of constants.
10871         (mips_option_override): Set default values for options.
10872         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
10873         Likewise.
10874         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10875         Likewise.
10876         * config/rx/rx.c (rx_option_override): Likewise.
10877         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
10878         (LABEL_ALIGN): Use align_labels_log.
10879         (LOOP_ALIGN): Use align_loops_align.
10880         * config/s390/s390.c (s390_asm_output_function_label): Use new
10881         macros.
10882         * config/sh/sh.c (sh_override_options_after_change):
10883         Change type of constants.
10884         * config/spu/spu.c (spu_sched_init): Likewise.
10885         * config/sparc/sparc.c (sparc_option_override): Set default
10886         values for options.
10887         * config/visium/visium.c (visium_option_override): Likewise.
10888         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
10889         emit p2align format with last argument if it's not needed.
10890         * doc/invoke.texi: Document extended format of -falign-*.
10891         * final.c: Use align_labels alignment.
10892         * flags.h (struct target_flag_state): Change type to use
10893         align_flags.
10894         (struct align_flags_tuple): New.
10895         (struct align_flags): Likewise.
10896         (align_loops_log): Redefine macro to use new types.
10897         (align_loops_max_skip): Redefine macro to use new types.
10898         (align_jumps_log): Redefine macro to use new types.
10899         (align_jumps_max_skip): Redefine macro to use new types.
10900         (align_labels_log): Redefine macro to use new types.
10901         (align_labels_max_skip): Redefine macro to use new types.
10902         (align_functions_log): Redefine macro to use new types.
10903         (align_loops): Redefine macro to use new types.
10904         (align_jumps): Redefine macro to use new types.
10905         (align_labels): Redefine macro to use new types.
10906         (align_functions): Redefine macro to use new types.
10907         (align_functions_max_skip): Redefine macro to use new types.
10908         (align_loops_value): New macro.
10909         (align_jumps_value): New macro.
10910         (align_labels_value): New macro.
10911         (align_functions_value): New macro.
10912         * function.c (invoke_set_current_function_hook): Propagate
10913         alignment values from flags to global variables default in
10914         topleev.h.
10915         * ipa-icf.c (sem_function::equals_wpa): Use
10916         cl_optimization_option_eq instead of memcmp.
10917         * lto-streamer.h (cl_optimization_stream_out): Support streaming
10918         of string types.
10919         (cl_optimization_stream_in): Likewise.
10920         * optc-save-gen.awk: Support strings in cl_optimization.
10921         * opth-gen.awk: Likewise.
10922         * opts.c (finish_options): Remove error checking of invalid
10923         value ranges.
10924         (MAX_CODE_ALIGN): Remove.
10925         (MAX_CODE_ALIGN_VALUE): Likewise.
10926         (parse_and_check_align_values): New function.
10927         (check_alignment_argument): Likewise.
10928         (common_handle_option): Use check_alignment_argument.
10929         * opts.h (parse_and_check_align_values): Declare.
10930         * toplev.c (init_alignments): Remove.
10931         (read_log_maxskip): New.
10932         (parse_N_M): Likewise.
10933         (parse_alignment_opts): Likewise.
10934         (backend_init_target): Remove usage of init_alignments.
10935         * toplev.h (parse_alignment_opts): Declare.
10936         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
10937         argument.
10938         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
10939         * tree.c (cl_option_hasher::equal): New.
10940         * varasm.c: Use new global macros.
10942 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10944         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
10945         Use a simpler align directive also if MAXSKIP = ALIGN-1.
10946         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10947         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10948         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10949         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10950         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
10951         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
10952         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
10953         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10954         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10955         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10956         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10958 2018-07-04  Martin Liska  <mliska@suse.cz>
10959             Jonathan Wakely  <jwakely@redhat.com>
10961         * coverage.c: Use correct type.
10962         * doc/invoke.texi: Language correction.
10964 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
10966         PR target/85620
10967         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
10968         ENDBRANCH for non-tail call which may return via indirect branch.
10969         * doc/extend.texi: Document indirect_return attribute.
10971 2018-07-03  Martin Sebor  <msebor@redhat.com>
10973         PR tree-optimization/86274
10974         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
10975         precondition.
10976         (format_floating): Correct handling of infinities and NaNs.
10978 2018-07-03  Martin Sebor  <msebor@redhat.com>
10980         * print-tree.c (print_real_cst): New function.
10981         (print_node_brief): Call it.
10982         (print_node): Ditto.
10984 2018-07-03  Jeff Law  <law@redhat.com>
10986         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
10987         into a single pattern.
10989         * config/h8300/h8300.md (ors code_iterator): New.
10990         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
10991         a single pattern and single splitter.
10992         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
10993         (iorqi3_1, xorqi3_1): Likewise.
10994         (iorqi3, xorqi3 expanders): Similarly.
10996         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
10997         (movmd_internal) into a single pattern using the P mode iterator.
10998         (movmd splitters): Similarly.
10999         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
11000         (movsd splitters): Similarly.
11002         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
11003         ADDB, ADDW and ADDL into a single ADD attribute which selects the
11004         right table based on the size of the operand.
11005         * config/h8300/h8300.md (length_table): Corresponding changes. All
11006         references to "addb", "addw" and "addl" changed to "add".
11007         (btst patterns): Merge two variants into a single pattern.
11008         (tstqi, tsthi): Likewise.
11009         (addhi3_incdec, addsi3_incdec): Likewise.
11010         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
11011         (mulhi3, mulsi3): Likewise.
11012         (udivhi3, udivsi3): Likewise.
11013         (divhi3, divsi3): Likewise.
11014         (andorqi3, andorhi3, andorsi3): Likewise.
11016 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
11018         PR target/85694
11019         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
11020         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
11022 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11024         PR tree-optimization/85694
11025         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
11026         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
11027         UNSPEC_URHADD.
11028         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
11029         (<u>avg<mode>3_ceil): New patterns.
11031 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
11033         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
11034         scan-tree-dump directive.
11035         * gcc.dg/vect/slp-perm-2.c: Likewise.
11036         * gcc.dg/vect/slp-perm-3.c: Likewise.
11037         * gcc.dg/vect/slp-perm-5.c: Likewise.
11038         * gcc.dg/vect/slp-perm-6.c: Likewise.
11039         * gcc.dg/vect/slp-perm-7.c: Likewise.
11040         * gcc.dg/vect/slp-perm-8.c: Likewise.
11042 2018-07-03  Marek Polacek  <polacek@redhat.com>
11044         PR middle-end/86202
11045         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
11047 2018-07-03  Richard Biener  <rguenther@suse.de>
11049         PR ipa/86389
11050         * tree-ssa-structalias.c (find_func_clobbers): Properly
11051         handle indirect calls.
11053 2018-07-03  Jeff Law  <law@redhat.com>
11055         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
11056         (shifts): New code iterator.
11057         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
11058         expander.  Fix HImode handling on H8/SX.
11059         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
11060         (subqi3, subhi3, subsi3 expanders): Likewise.
11061         (andqi3, andhi3, andsi3 expanders): Likewise.
11062         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
11063         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
11064         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
11065         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
11066         (zero_extendqihi2, zero_extendqisi2): Likewise.
11067         (extendqihi2, extendqisi2): Likewise.
11068         (rotlqi3, rotlhi3, rotlsi3): Likewise.
11069         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
11070         (rotlqi3_1, rotlhi3_1): Likewise.
11071         (logicalhi3_sn, logicalsi3_sn): Likewise.
11072         (logicalhi3, logicalsi3): Likewise.
11074 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11076         * tree-vect-patterns.c (vect_recog_rotate_pattern)
11077         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11078         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
11079         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
11080         type to append_pattern_def_seq instead of creating a stmt_vec_info
11081         directly.
11082         (build_mask_conversion): Likewise.  Remove vinfo argument.
11083         (vect_add_conversion_to_patterm): Likewise, renaming to...
11084         (vect_add_conversion_to_pattern): ...this.
11085         (vect_recog_mask_conversion_pattern): Update call to
11086         build_mask_conversion.  Pass the vector type to
11087         append_pattern_def_seq here too.
11088         (vect_recog_gather_scatter_pattern): Update call to
11089         vect_add_conversion_to_pattern.
11091 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11093         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
11094         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11095         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
11096         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
11097         STMT_VINFO_PATTERN_DEF_SEQ to null here.
11098         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
11099         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
11100         append_pattern_def_seq instead of new_pattern_def_seq.
11101         (vect_recog_divmod_pattern): Do both of the above.
11102         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
11103         is null.
11105 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11107         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11108         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
11109         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
11110         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
11111         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
11112         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11113         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
11114         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
11115         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
11116         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
11117         parameter with a single stmt_vec_info.
11118         (vect_recog_func_ptr): Likewise.
11119         (vect_recog_gather_scatter_pattern): Likewise, folding in...
11120         (vect_try_gather_scatter_pattern): ...this.
11121         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
11122         the stmt_vec_info of the statement to be matched.  Don't clear
11123         STMT_VINFO_RELATED_STMT.
11124         (vect_pattern_recog): Update call accordingly.
11126 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11128         PR tree-optimization/85694
11129         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
11130         (uavgM3_ceil): Document new optabs.
11131         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
11132         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
11133         functions.
11134         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
11135         (savg_ceil_optab): New optabs.
11136         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
11137         (vect_vect_recog_func_ptrs): Add it.
11138         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
11139         constant directly from the associated lhs.
11141 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11143         * tree-vect-patterns.c (vect_split_statement): New function.
11144         (vect_convert_input): Use it to try to split an existing cast.
11146 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11148         * poly-int.h (print_hex): New function.
11149         * dumpfile.h (dump_dec, dump_hex): Declare.
11150         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
11151         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
11152         min_input_precision, operation_precision and operation_sign.
11153         * tree-vect-patterns.c (vect_get_range_info): New function.
11154         (vect_same_loop_or_bb_p, vect_single_imm_use)
11155         (vect_operation_fits_smaller_type): Delete.
11156         (vect_look_through_possible_promotion): Add an optional
11157         single_use_p parameter.
11158         (vect_recog_over_widening_pattern): Rewrite to use new
11159         stmt_vec_info infomration.  Handle one operation at a time.
11160         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
11161         (vect_truncatable_operation_p, vect_set_operation_type)
11162         (vect_set_min_input_precision): New functions.
11163         (vect_determine_min_output_precision_1): Likewise.
11164         (vect_determine_min_output_precision): Likewise.
11165         (vect_determine_precisions_from_range): Likewise.
11166         (vect_determine_precisions_from_users): Likewise.
11167         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
11168         (vect_vect_recog_func_ptrs): Put over_widening first.
11169         Add cast_forwprop.
11170         (vect_pattern_recog): Call vect_determine_precisions.
11172 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
11174         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
11175         statements that have been replaced by further pattern statements.
11176         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
11178 2018-07-03  Richard Biener  <rguenther@suse.de>
11180         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
11181         always set *dt.  Dump vectype in vectype overload.
11182         * dumpfile.h (dump_gimple_expr): New function.
11183         (dump_gimple_expr_loc): Likewise.
11184         * dumpfile.c (dump_gimple_expr): New function.
11185         (dump_gimple_expr_loc): Likewise.
11187 2018-07-02  Jeff Law  <law@redhat.com>
11189         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
11190         the H8/300, H8/300H and H8/S variants into a single pattern.
11191         (movhi_h8300, movqi_h8300hs): Similarly.
11192         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
11193         (QHI mode iterator): New.
11195         * config/h8300/h8300.md: Remove trailing whitespace.
11197 2018-07-02  Jim Wilson  <jimw@sifive.com>
11199         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
11200         instead of emit_insn for interrupt returns.
11201         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
11202         (riscv_sret, riscv_uret): Likewise.
11204 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11206         * pretty-print.c (selftest::test_pp_format): Move save and restore
11207         of quotes to class auto_fix_quotes, and add an instance.
11208         * selftest.c: Include "intl.h".
11209         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
11210         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
11211         * selftest.h (selftest::auto_fix_quotes): New class.
11213 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
11215         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
11216         (aarch64_sve_prepare_conditional_op): Remove.
11217         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
11218         Allow aarch64_simd_reg_or_zero as select operand; remove
11219         the aarch64_sve_prepare_conditional_op call.
11220         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
11221         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11222         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
11223         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
11224         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
11225         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
11226         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
11227         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
11228         and a splitters to match all of the *_any patterns.
11229         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
11231         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
11232         (SVE_COND_FP_BINARY_REV): Remove.
11233         (sve_int_op_rev, sve_fp_op_rev): New.
11234         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
11235         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
11236         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
11237         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
11238         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
11239         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
11240         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
11241         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
11242         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
11244         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
11245         Remove match_dup 1 from the inner unspec.
11246         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
11248         * config/aarch64/aarch64.md (movprfx): New attr.
11249         (length): Default movprfx to 8.
11250         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
11251         (*madd<SVE_I>, *msub<SVE_I): Likewise.
11252         (*<su>mul<SVE_I>3_highpart): Likewise.
11253         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
11254         (*v<ASHIFT><SVE_I>3): Likewise.
11255         (*<su><MAXMIN><SVE_I>3): Likewise.
11256         (*<su><MAXMIN><SVE_F>3): Likewise.
11257         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
11258         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11259         (*div<SVE_F>4): Likewise.
11261 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
11263         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11264         in dump string.
11266 2018-07-02  Richard Biener  <rguenther@suse.de>
11268         PR tree-optimization/86363
11269         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11270         memset argument refers to a non-variable address.
11272 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11274         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11275         {PLUS,MINUS}_EXPR code to...
11276         (adjust_symbolic_bound): ...here,
11277         (combine_bound): ...here,
11278         (set_value_range_with_overflow): ...and here.
11280 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11282         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11283         code...
11284         (extract_range_from_abs_expr): ...here.
11286 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
11288         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11289         -fno-omit-frame-pointer when not optimizing.
11291 2018-07-02  Martin Liska  <mliska@suse.cz>
11293         PR ipa/86279
11294         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11295         (propagate_nothrow): Likewise.
11297 2018-07-02  Martin Liska  <mliska@suse.cz>
11299         PR ipa/86323
11300         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11302 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11304         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11305         function in r262149, changing "loc" param from source_location to
11306         const dump_location_t &.
11307         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11308         declaration, as above.
11310 2018-07-01  Paul Koning  <ni1d@arrl.net>
11312         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11313         -munit-asm, -mgnu-asm, -mdec-asm.
11314         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11315         (pdp11_output_labelref): New.
11316         (pdp11_output_def): New.
11317         (pdp11_output_addr_vec_elt): New.
11318         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
11319         %# and %@ format codes.
11320         (pdp11_option_override): New.
11321         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11322         (pdp11_output_ident): New.
11323         (pdp11_asm_named_section): New.
11324         (pdp11_asm_init_sections): New.
11325         (pdp11_file_start): New.
11326         (pdp11_file_end): New.
11327         (output_ascii): Use .ascii/.asciz for -mdec-asm.
11328         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
11329         %o, like %c but octal.
11330         (pdp11_option_override): New.
11331         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11332         -mdec-asm.
11333         (DATA_SECTION_ASM_OP): Ditto.
11334         (READONLY_DATA_SECTION_ASM_OP): New.
11335         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11336         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11337         (ASM_OUTPUT_LABELREF): Ditto.
11338         (ASM_OUTPUT_DEF): Ditto.
11339         (ASM_OUTPUT_EXTERNAL): New.
11340         (ASM_OUTPUT_SOURCE_FILENAME): New.
11341         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11342         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11343         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
11344         %# and %@ format codes.
11345         * config/pdp11/pdp11.opt (mgnu-asm): New.
11346         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11347         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11348         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11350 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
11352         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11353         dereferencing path[] beyond its length.
11354         (debug_path): New.
11355         (debug_all_paths): New.
11356         (rewire_first_differing_edge): New.
11357         (adjust_paths_after_duplication): New.
11358         (duplicate_thread_path): Call adjust_paths_after_duplication.
11359         Add new argument.
11360         (thread_through_all_blocks): Add new argument to
11361         duplicate_thread_path.
11363 2018-06-30  Jim Wilson  <jimw@sifive.com>
11365         * config/riscv/predicates.md (p2m1_shift_operand): New.
11366         (high_mask_shift_operand): New.
11367         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11368         pattern using p2m1_shift_operand.
11369         (lshsi3_zero_extend_3+2): New combiner pattern using
11370         high_mask_shift_operand.
11372 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11374         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11375         split out from...
11376         (vect_recog_rotate_pattern): ...here.
11377         (vect_convert_input): Try to insert casts of invariants in the
11378         preheader.
11379         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11380         preheader to be empty.
11382 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11384         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11385         vector type.  If given, install it in the new statement's
11386         STMT_VINFO_VECTYPE.
11387         (vect_element_precision): New function.
11388         (vect_unpromoted_value): New struct.
11389         (vect_unpromoted_value::vect_unpromoted_value): New function.
11390         (vect_unpromoted_value::set_op): Likewise.
11391         (vect_look_through_possible_promotion): Likewise.
11392         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11393         (vect_widened_op_tree, vect_convert_input): Likewise.
11394         (vect_convert_inputs, vect_convert_output): Likewise.
11395         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11396         to handle the optional cast of the multiplication result and
11397         vect_widened_op_tree to detect the widened multiplication itself.
11398         Do not require the input and output of promotion casts to have
11399         the same sign, but base the signedness of the operation on the
11400         input rather than the result.  If the pattern includes two
11401         promotions, check that those promotions have the same sign.
11402         Do not restrict the MULT_EXPR handling to a double-width result;
11403         handle quadruple-width results and wider.  Use vect_convert_inputs
11404         to convert the inputs to the common type.
11405         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
11406         to handle the optional cast of the ABS result.  Also allow a sign
11407         change or a sign extension between the ABS and MINUS.
11408         Use vect_widened_op_tree to detect the widened subtraction and use
11409         vect_convert_inputs to convert the inputs to the common type.
11410         (vect_handle_widen_op_by_const): Delete.
11411         (vect_recog_widen_op_pattern): New function.
11412         (vect_recog_widen_mult_pattern): Use it.
11413         (vect_recog_widen_shift_pattern): Likewise.
11414         (vect_recog_widen_sum_pattern): Use
11415         vect_look_through_possible_promotion to handle the promoted
11416         PLUS_EXPR operand.
11418 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11420         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11421         the containing gimple_seq *.
11422         * gimple-iterator.h (gsi_for_stmt): Declare it.
11423         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11424         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11425         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11426         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11427         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11428         checks.
11429         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11430         split out from...
11431         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
11432         statement being replaced is part of an existing pattern
11433         definition sequence, inserting the new pattern statements before
11434         the original one.
11435         (vect_pattern_recog_1): Don't return a bool.  If the statement
11436         is already part of a pattern, instead apply pattern matching
11437         to the pattern definition statements.  Don't clear the
11438         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11439         (vect_pattern_recog): Don't break after the first match;
11440         continue processing the pattern definition statements instead.
11441         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
11443 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11445         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11446         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11447         (vect_recog_widen_sum_pattern): Use it.
11449 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11451         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11452         phi is not a pattern statement and has not been replaced by
11453         a pattern statement.
11454         * tree-vect-patterns.c (type_conversion_p): Don't check
11455         STMT_VINFO_IN_PATTERN_P.
11456         (vect_recog_vector_vector_shift_pattern): Likewise.
11457         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
11458         the pattern statement rather than the original statement; check
11459         directly for a WIDEN_MULT_EXPR here.
11460         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
11461         vect_is_simple_use to return the pattern statement rather
11462         than the original statement; use is_pattern_stmt_p to check
11463         for such a pattern statement.
11464         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
11465         to return the pattern statement rather than the original statement;
11466         don't do the same transformation here.
11467         (vect_is_simple_use): If the defining statement has been replaced
11468         by a pattern statement, return the pattern statement instead.
11469         Remove the corresponding (local) transformation from the vectype
11470         overload.
11472 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11474         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
11475         end and default to null.
11476         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11477         (vectorizable_reduction): Update calls accordingly, dropping the
11478         gimple ** argument if the passed-back statement isn't needed.
11479         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
11480         (vect_recog_rotate_pattern): Likewise.
11481         (vect_recog_mask_conversion_pattern): Likewise.
11482         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11483         (vect_mask_constant_operand_p): Likewise.
11484         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
11485         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
11486         (get_group_load_store_type, get_load_store_type): Likewise.
11487         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
11488         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
11489         (vectorizable_conversion, vectorizable_assignment): Likewise.
11490         (vectorizable_shift, vectorizable_operation): Likewise.
11491         (vectorizable_store, vect_is_simple_cond): Likewise.
11492         (vectorizable_condition, vectorizable_comparison): Likewise.
11493         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
11494         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
11495         and move it to the end.  Cope with null def_stmt_outs.
11497 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11499         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
11501 2018-06-29  Jeff Law  <law@redhat.com>
11503         * config/v850/v850.c (v850_legitimate_address_p): Handle large
11504         displacements for TARGET_V850E2V3 and newer.
11505         (TARGET_LRA_P): Remove.  Defaults to LRA now.
11506         * config/v850/v850.md (sign23byte_load): Remove.
11507         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
11508         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
11510 2018-06-29  Martin Liska  <mliska@suse.cz>
11512         PR lto/85759
11513         * coverage.c (coverage_init): Mangle full path name.
11514         * doc/invoke.texi: Document the change.
11515         * gcov-io.c (mangle_path): New.
11516         * gcov-io.h (mangle_path): Likewise.
11517         * gcov.c (mangle_name): Use mangle_path for path mangling.
11519 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11521         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
11522         if starting source register is not even.
11524 2018-06-29  Martin Liska  <mliska@suse.cz>
11526         PR tree-optimization/86263
11527         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
11528         Make edge redirection.
11530 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
11532         * dumpfile.c (dump_loc): Add indentation based on scope depth.
11533         (dump_scope_depth): New variable.
11534         (get_dump_scope_depth): New function.
11535         (dump_begin_scope): New function.
11536         (dump_end_scope): New function.
11537         * dumpfile.h (get_dump_scope_depth): New declaration.
11538         (dump_begin_scope): New declaration.
11539         (dump_end_scope): New declaration.
11540         (class auto_dump_scope): New class.
11541         (AUTO_DUMP_SCOPE): New macro.
11542         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
11543         AUTO_DUMP_SCOPE.
11545 2018-06-29  Richard Biener  <rguenther@suse.de>
11547         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
11548         compute_all_dependences succeeds.
11549         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
11550         exceed --param loop-max-datarefs-for-datadeps.
11552 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
11554         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
11556 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
11558         PR target/86348
11559         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
11560         alternative 0 in preferred_for_speed attribute.
11562 2018-06-28  Paul Koning  <ni1d@arrl.net>
11564         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
11565         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
11566         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
11567         * config/pdp11/pdp11.md: Correct "length" attribute calculation
11568         for shift insn patterns.
11570 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
11572         * cgraph.c (cgraph_node::get_body): Replace assignments to
11573         "dump_file" with calls to set_dump_file.
11574         * dumpfile.c (alt_dump_file): Make static, and group with...
11575         (alt_flags): ...this definition.
11576         (dumps_are_enabled): New variable.
11577         (refresh_dumps_are_enabled): New function.
11578         (set_dump_file): New function.
11579         (set_alt_dump_file): New function.
11580         (gcc::dump_manager::dump_start): Replace assignments to
11581         "dump_file" and "alt_dump_file" with calls to set_dump_file and
11582         set_alt_dump_file.
11583         (gcc::dump_manager::dump_finish): Likewise.
11584         * dumpfile.h (alt_dump_file): Delete decl.
11585         (dumps_are_enabled): New variable decl.
11586         (set_dump_file): New function decl.
11587         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
11588         global.
11589         * tree-nested.c (lower_nested_functions): Replace assignments to
11590         "dump_file" with calls to set_dump_file.
11592 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
11594         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
11595         goto_locus of each outgoing edge of each basic block.
11597 2018-06-28  Richard Biener  <rguenther@suse.de>
11599         * dwarf2out.c (decl_scope_table): Remove.
11600         (push_decl_scope): Likewise.
11601         (pop_decl_scope): Likewise.
11602         (gen_type_die_for_member): Do not call push/pop_decl_scope.
11603         (gen_struct_or_union_type_die): Likewise.
11604         (gen_tagged_type_die): Likewise.
11605         (dwarf2out_init): Do not initialize decl_scope_table.
11606         (dwarf2out_c_finalize): Do not free it.
11608 2018-06-28  Richard Biener  <rguenther@suse.de>
11610         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
11611         deciding whether to not re-use a DIE.
11613 2018-06-28  Richard Biener  <rguenther@suse.de>
11615         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
11616         DW_AT_abstract_origin attribute.
11618 2018-06-28  Martin Liska  <mliska@suse.cz>
11620         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11621         Use newly introduced constants.
11622         * tree-switch-conversion.h (struct jump_table_cluster):
11623         Define max_ratio_for_size and max_ratio_for_speed.
11625 2018-06-28  Martin Liska  <mliska@suse.cz>
11627         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11628         Add new checking assert to catch invalid state.
11629         (jump_table_cluster::can_be_handled): Handle single case
11630         clusters.
11631         (jump_table_cluster::is_beneficial): Bail out for such case.
11632         (bit_test_cluster::find_bit_tests):
11633         Add new checking assert to catch invalid state.
11634         (bit_test_cluster::can_be_handled): Handle single case
11635         clusters.
11636         (bit_test_cluster::is_beneficial): Bail out for such case.
11637         (switch_decision_tree::analyze_switch_statement):
11638         Fix comment.
11640 2018-06-28  Martin Liska  <mliska@suse.cz>
11642         * common.opt: Introduce -completion option.
11643         * gcc.c (driver_handle_option): Handle it.
11644         (driver::main): Print completions if completion
11645         is set.
11646         * opt-suggestions.c (option_proposer::get_completions):
11647         New function.
11648         (option_proposer::suggest_completion): Likewise.
11649         (option_proposer::find_param_completions): Likewise.
11650         (verify_autocompletions): Likewise.
11651         (test_completion_valid_options): Likewise.
11652         (test_completion_valid_params): Likewise.
11653         (in_completion_p): Likewise.
11654         (empty_completion_p): Likewise.
11655         (test_completion_partial_match): Likewise.
11656         (test_completion_garbage): Likewise.
11657         (opt_proposer_c_tests): Likewise.
11658         * opt-suggestions.h: Declare new functions.
11659         * opts.c (common_handle_option): Handle OPT__completion_.
11660         * selftest-run-tests.c (selftest::run_tests): Add
11661         opt_proposer_c_tests.
11662         * selftest.c (assert_str_startswith): New.
11663         * selftest.h (assert_str_startswith): Likewise.
11664         (opt_proposer_c_tests): New.
11665         (ASSERT_STR_STARTSWITH): Likewise.
11667 2018-06-28  Martin Liska  <mliska@suse.cz>
11669         * Makefile.in: Add opt-suggestions.o.
11670         * gcc-main.c: Include opt-suggestions.h.
11671         * gcc.c (driver::driver): Likewise.
11672         (driver::~driver): Remove m_option_suggestions.
11673         (driver::build_option_suggestions): Moved to option_proposer.
11674         (driver::suggest_option): Likewise.
11675         (driver::handle_unrecognized_options): Use option_proposer.
11676         * gcc.h (class driver): Add new memver m_option_proposer.
11677         * opt-suggestions.c: New file.
11678         * opt-suggestions.h: New file.
11680 2018-06-28  Martin Liska  <mliska@suse.cz>
11682         * vec.h (class auto_string_vec): New (moved from auto_argvec).
11683         (auto_string_vec::~auto_string_vec): Likewise.
11685 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
11687         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
11688         prevent_decl_creation_for_types fields up and add reset_location field.
11689         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
11690         statement if id->reset_location is true.
11691         (copy_edges_for_bb): Do not set goto_locus on the new edges if
11692         id->reset_location is true.
11693         (copy_phis_for_bb): Force input_location on the arguments if
11694         id->reset_location is true.
11695         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
11696         is set on the function to be inlined.
11698 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
11700         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
11702 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
11704         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
11705         registers for Pmode.
11706         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
11707         hard registers for the clobbered pseudo.
11709 2018-06-27  Paul Koning  <ni1d@arrl.net>
11711         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11712         mutually exclusive options.
11713         * config/pdp11/constraints.md (h): New constraint.
11714         (O): Update definition to match shift code generation.
11715         (D): New constraint.
11716         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
11717         (CCFP): Remove.
11718         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
11719         function.
11720         (output_jump): Change arguments.
11721         (pdp11_fixed_cc_regs): New function.
11722         (pdp11_cc_mode): Ditto.
11723         (pdp11_expand_shift): Ditto.
11724         (pdp11_assemble_shift): Ditto.
11725         (pdp11_small_shift): Ditto.
11726         (pdp11_branch_cost): Remove.
11727         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
11728         from output.
11729         (pdp11_register_move_cost): Update for CC registers.
11730         (pdp11_rtx_costs): Add case for LSHIFTRT.
11731         (pdp11_output_jump): Add CCNZ mode conditional branches.
11732         (notice_update_cc_on_set): Remove.
11733         (pdp11_cc_mode): New function.
11734         (simple_memory_operand): Correct pre/post decrement case.
11735         (no_side_effect_operand): New function.
11736         (pdp11_regno_reg_class): Add CC_REGS class.
11737         (pdp11_fixed_cc_regs): New function.
11738         (pdp11_small_shift): New function.
11739         (pdp11_expand_shift): New function to expand shift insns.
11740         (pdp11_assemble_shift): New function to output shifts.
11741         (pdp11_branch_cost): Remove.
11742         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
11743         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
11744         (WCHAR_TYPE): Ditto.
11745         (PTRDIFF_TYPE): Ditto.
11746         (ADJUST_INSN_LENGTH): New macro.
11747         (FIXED_REGISTERS): Add CC registers.
11748         (CALL_USED_REGISTERS): Ditto.
11749         (reg_class): Ditto.
11750         (REG_CLASS_NAMES): Ditto.
11751         (REG_CLASS_CONTENTS): Ditto.
11752         (SELECT_CC_MODE): Use new function.
11753         (TARGET_FLAGS_REGNUM): New macro.
11754         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
11755         (cc0_reg_rtx): Remove.
11756         (CC_STATUS_MDEP): Remove.
11757         (CC_STATUS_MDEFP_INIT): Remove.
11758         (CC_IN_FPU): Remove.
11759         (NOTICE_UPDATE_CC): Remove.
11760         (REGISTER_NAMES): Add CC registers.
11761         (BRANCH_COST): Change to constant 1.
11762         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
11763         handling.
11764         * config/pdp11/pdp11.opt (mbcopy): Remove.
11765         (mbcopy-builtin): Remove.
11766         (mbranch-cheap): Remove.
11767         (mbranch-expensive): Remove.
11768         * config/pdp11/predicates.md (expand_shift_operand): Update to
11769         match shift code generation.
11770         (ccnz_operator): New predicate.
11771         * doc/invoke.texi (PDP-11 Options): Remove deleted options
11772         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
11773         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
11774         exclusive options.
11775         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
11776         description of O constraint.
11778 2018-06-27  Jeff Law  <law@redhat.com>
11779             Austin Law  <austinklaw@gmail.com>
11781         * config/v850/v850.md (addsi3_set_flags): New pattern.
11782         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
11783         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
11784         (zero_extendhisi2_v850_set_flags): Likewise.
11785         (zero_extendqisi2_v850_set_flags): Likewise.
11786         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
11787         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
11788         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
11790         * config/v850/v850-protos.h (notice_update_cc): Remove.
11791         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
11792         (v850_print_operand): Handle 'D' and "d".
11793         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
11794         Add handling of arithmetic/logical operations compared against zero.
11795         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
11796         Do not look at v850_compare_op, instead get mode from last argument.
11797         (v850_gen_compare): Remove
11798         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
11799         after reload for prologue insns.
11800         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
11801         patterns.
11802         (construct_save_jarl): Likewise.
11803         (TARGET_FLAGS_REGNUM): Define.
11804         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
11805         (NOTICE_UPDATE_CC): Remove.
11806         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
11807         than cc0.  Conditionalize on reload_completed.
11808         (cmpsi_insn, setfcc_insn): Likewise.
11809         (tst1 splitter): Turn into define_and_split which sets the flags
11810         after reload.
11811         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
11812         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
11813         (cstoresf4, cstoredf4): Clobber the flags.
11814         (cmpsi, cmpsf, cmpdf): Remove expanders.
11815         (setf_insn): Remove pattern.
11816         (addsi3): Turn into define_and_split which clobbers the flags after
11817         reload and a suitable pattern (addsi3_clobber_flags) for use after
11818         reload.
11819         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
11820         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
11821         (ashrsi3, ashrsi3_v850e2): Likewise.
11822         (bins): Clobber the flags.
11823         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
11824         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
11825         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
11826         (call_value_internal_short, call_value_internal_long): Likewise.
11827         (callt_save_interrupt, callt_return_interrupt): Likewise.
11828         (save_interrupt, return_interrupt): Likewise.
11829         (callt_save_all_interrupt, save_all_interrupt): Likewise.
11830         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
11831         (restore_all_interrupt, _restore_all_interrupt): Likewise.
11832         (All FP comparisons): Only allow after reload has completed.
11833         (trfsr): Likewise.
11834         (divh, divhu): Tweak output template.
11835         (branch_z_normal, branch_z_invert): Remove
11836         (branch_nz_normal, branch_nz_invert): Likewise.
11837         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
11839         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
11840         * config/v850/v850.c (notice_update_cc): Remove.
11841         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
11842         (CC_NO_CARRY): Likewise.
11843         (NOTICE_UPDATE_CC): Define to nothing.
11844         * config/v850/v850.md: Remove block comment on cc0 handling
11845         Remove "cc" attribute from all patterns.  Remove cc_status handling
11846         from all patterns.  Minor formatting fixes.
11848 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11850         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
11851         (cortex-a76.cortex-a55): Likewise.
11852         * config/aarch64/aarch64-tune.md: Regenerate.
11853         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
11854         cortex-a76.cortex-a55.
11856 2018-06-27  Jeff Law  <law@redhat.com>
11858         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
11859         (MULTILIB_DIRNAMES): Similarly.
11861 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
11863         * gimple.h (gimple_return_retbnd): Delete.
11864         (gimple_return_set_retbnd): Likewise.
11865         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
11866         gimple_return_set_retbnd.
11867         * gimple-pretty-print.c (dump_gimple_return): Remove call to
11868         gimple_return_retbnd and adjust.
11869         * tree-inline.h (struct copy_body_data): Remove retbnd field.
11870         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
11871         Explicitly return NULL in a couple more cases.  Move assertion
11872         on debug statements and remove unreachable code.
11873         (reset_debug_binding): Do not test id->retbnd.
11874         (expand_call_inline): Do not set it.
11876 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11878         * configure.ac: Add --disable-gcov option.
11879         * configure: Regenerate.
11880         * Makefile.in: Honour @enable_gcov@.
11881         * doc/install.texi: Document --disable-gcov.
11883 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11885         * config/arm/arm-cpus.in (cortex-a76): New entry.
11886         (cortex-a76.cortex-a55): Likewise.
11887         * config/arm/arm-tables.opt: Regenerate.
11888         * config/arm/arm-tune.md: Likewise.
11889         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
11890         * doc/invoke.texi (ARM Options): Document cortex-a76 and
11891         cortex-a76.cortex-a55.
11893 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
11895         PR target/85769
11896         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
11898 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
11900         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
11901         comment.
11902         (EPILOGUE_USES): Likewise.
11904 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
11906         * tree-inline.c (remap_location): New function extracted from...
11907         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
11908         (copy_phis_for_bb): ...here.  Call remap_location.
11909         (copy_cfg_body): Adjust call to copy_edges_for_bb.
11911 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
11913         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
11914         unaligned vsx for 16B memset.
11916 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
11918         PR target/86285
11919         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
11920         ieee128_float_type_node to long_double_type_node unless
11921         TARGET_LONG_DOUBLE_128 is set.
11923 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
11925         * cfgloop.c (get_loop_location): Convert return type from
11926         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
11927         by implicit construction from rtx_insn *, and using
11928         dump_user_location_t::from_function_decl for the fallback case.
11929         * cfgloop.h (get_loop_location): Convert return type from
11930         location_t to dump_user_location_t.
11931         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
11932         dump_printf_loc to pass in a dump_location_t rather than a
11933         location_t, via the gimple stmt.
11934         * coverage.c (get_coverage_counts): Update calls to
11935         dump_printf_loc to pass in dump_location_t rather than a
11936         location_t.
11937         * doc/optinfo.texi (Dump types): Convert example of
11938         dump_printf_loc from taking "locus" to taking "insn".  Update
11939         description of the "_loc" calls to cover dump_location_t.
11940         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
11941         "selftest.h".
11942         (dump_user_location_t::dump_user_location_t): New constructors,
11943         from gimple *stmt and rtx_insn *.
11944         (dump_user_location_t::from_function_decl): New function.
11945         (dump_loc): Make static.
11946         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
11947         const dump_location_t &.
11948         (dump_generic_expr_loc): Delete.
11949         (dump_printf_loc): Convert param "loc" from location_t to
11950         const dump_location_t &.
11951         (selftest::test_impl_location): New function.
11952         (selftest::dumpfile_c_tests): New function.
11953         * dumpfile.h: Include "profile-count.h".
11954         (class dump_user_location_t): New class.
11955         (struct dump_impl_location_t): New struct.
11956         (class dump_location_t): New class.
11957         (dump_printf_loc): Convert 2nd param from source_location to
11958         const dump_location_t &.
11959         (dump_generic_expr_loc): Delete.
11960         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
11961         const dump_location_t &.
11962         * gimple-fold.c (fold_gimple_assign): Update call to
11963         dump_printf_loc to pass in a dump_location_t rather than a
11964         location_t, via the gimple stmt.
11965         (gimple_fold_call): Likewise.
11966         * gimple-loop-interchange.cc
11967         (loop_cand::analyze_iloop_reduction_var): Update for change to
11968         check_reduction_path.
11969         (tree_loop_interchange::interchange): Update for change to
11970         find_loop_location.
11971         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
11972         change in return-type of find_loop_location.
11973         (graphite_regenerate_ast_isl): Likewise.
11974         * graphite-optimize-isl.c (optimize_isl): Likewise.
11975         * graphite.c (graphite_transform_loops): Likewise.
11976         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
11977         pass in a dump_location_t rather than a location_t, via the
11978         gimple stmt.
11979         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11980         * ipa.c (walk_polymorphic_call_targets): Likewise.
11981         * loop-unroll.c (report_unroll): Convert "locus" param from
11982         location_t to dump_location_t.
11983         (decide_unrolling): Update for change to get_loop_location's
11984         return type.
11985         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
11986         location_t to dump_user_location_t.
11987         (grid_find_single_omp_among_assignments_1): Updates calls to
11988         dump_printf_loc to pass in a dump_location_t rather than a
11989         location_t, via the gimple stmt.
11990         (grid_parallel_clauses_gridifiable): Convert "tloc" from
11991         location_t to dump_location_t.  Updates calls to dump_printf_loc
11992         to pass in a dump_location_t rather than a location_t, via the
11993         gimple stmt.
11994         (grid_inner_loop_gridifiable_p): Likewise.
11995         (grid_dist_follows_simple_pattern): Likewise.
11996         (grid_gfor_follows_tiling_pattern): Likewise.
11997         (grid_target_follows_gridifiable_pattern): Likewise.
11998         (grid_attempt_target_gridification): Convert initialization
11999         of local "grid" from memset to zero-initialization; FIXME: does
12000         this require C++11?  Update call to dump_printf_loc to pass in a
12001         optinfo_location rather than a location_t, via the gimple stmt.
12002         * profile.c (read_profile_edge_counts): Updates call to
12003         dump_printf_loc to pass in a dump_location_t rather than a
12004         location_t
12005         (compute_branch_probabilities): Likewise.
12006         * selftest-run-tests.c (selftest::run_tests): Call
12007         dumpfile_c_tests.
12008         * selftest.h (dumpfile_c_tests): New decl.
12009         * tree-loop-distribution.c (pass_loop_distribution::execute):
12010         Update for change in return type of find_loop_location.
12011         * tree-parloops.c (parallelize_loops): Likewise.
12012         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
12013         "locus" from location_t to dump_user_location_t.
12014         (canonicalize_loop_induction_variables): Likewise.
12015         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
12016         for change in return type of find_loop_location.
12017         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
12018         to dump_printf_loc to pass in a dump_location_t rather than a
12019         location_t, via the stmt.
12020         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
12021         Likewise.
12022         * tree-vect-loop-manip.c (find_loop_location): Convert return
12023         type from source_location to dump_user_location_t.
12024         (vect_do_peeling): Update for above change.
12025         (vect_loop_versioning): Update for change in type of
12026         vect_location.
12027         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
12028         from location_t to dump_user_location_t.
12029         (vect_estimate_min_profitable_iters): Update for change in type
12030         of vect_location.
12031         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
12032         location_t to dump_location_t.
12033         (vect_slp_bb): Update for change in type of vect_location.
12034         * tree-vectorizer.c (vect_location): Convert from source_location
12035         to dump_user_location_t.
12036         (try_vectorize_loop_1): Update for change in vect_location's type.
12037         (vectorize_loops): Likewise.
12038         (increase_alignment): Likewise.
12039         * tree-vectorizer.h (vect_location): Convert from source_location
12040         to dump_user_location_t.
12041         (find_loop_location): Convert return type from source_location to
12042         dump_user_location_t.
12043         (check_reduction_path): Convert 1st param from location_t to
12044         dump_user_location_t.
12045         * value-prof.c (check_counter): Update call to dump_printf_loc to
12046         pass in a dump_user_location_t rather than a location_t; update
12047         call to error_at for change in type of "locus".
12048         (check_ic_target): Update call to dump_printf_loc to
12049         pass in a dump_user_location_t rather than a location_t, via the
12050         call_stmt.
12052 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12054         * config/s390/s390.h (enum processor_flags): Do not use
12055         default tune parameter when -march was specified.
12057 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
12059         PR target/86314
12060         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
12061         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
12062         operands.
12064 2018-06-26  Richard Biener  <rguenther@suse.de>
12066         PR tree-optimization/86287
12067         PR bootstrap/86316
12068         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
12069         (vect_analyze_loop): Initialize n_stmts.
12071 2018-06-26  Richard Biener  <rguenther@suse.de>
12073         PR middle-end/86271
12074         * fold-const.c (fold_convertible_p): Pointer extension
12075         isn't valid.
12077 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
12079         PR debug/86064
12080         * dwarf2out.c (loc_list_has_views): Adjust comments.
12081         (dw_loc_list): Split single cross-partition range with
12082         nonzero locview.
12084 2018-06-25  Jeff Law  <law@redhat.com>
12086         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
12087         on -mbig-switch by default.
12089         * config/v850/predicates.md (const_float_1_operand): Fix match_code
12090         test.
12091         (const_float_0_operand): Remove unused predicate.
12092         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
12093         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
12094         (recipsf2): New expander.  Original pattern now called
12095         (recipsf2_insn).
12096         (recipdf2, recipdf2_insn): Similarly.
12097         (rsqrtsf2, rsqrtsf2_insn): Similarly
12098         (rsqrtdf2, rsqrtdf2_insn): Similarly
12100 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
12102         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
12103         Simplify logic for FreeBSD (twice).
12105 2018-06-25  Martin Sebor  <msebor@redhat.com>
12107         PR tree-optimization/86204
12108         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
12109         a strnlen result if it's less than the length of the string.
12111 2018-06-25  Martin Sebor  <msebor@redhat.com>
12113         PR tree-optimization/85700
12114         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
12115         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
12116         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
12118 2018-06-25  Martin Sebor  <msebor@redhat.com>
12120         * doc/extend.texi (Zero-length arrays): Update and clarify.
12122 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
12124         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
12125         added IEEE/IBM long double multilib support on PowerPC little
12126         endian Linux systems.
12127         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
12128         (MULTILIB_DEFAULTS): Likewise.
12129         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12130         Likewise.
12131         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
12132         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
12133         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
12135 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
12137         PR middle-end/86311
12138         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
12139         (REORDER_45): Likewise.
12141 2018-06-25  Jeff Law  <law@redhat.com>
12143         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
12144         dividend to 32 bits.  Adjust length.
12145         (udivmodhi4): Cleanup output template.  Fix length.
12147 2018-06-25  Carl Love  <cel@us.ibm.com>
12149         * config/rs6000/vsx.md: Change word selector to prefered location.
12151 2018-06-25  Richard Biener  <rguenther@suse.de>
12153         PR tree-optimization/86304
12154         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
12155         epilogue-if-converted loops as well.
12157 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
12159         * lto-section-out.c (lto_begin_section): Do not print section
12160         name for noaddr and unnumbered dumps.
12162 2018-06-25  Richard Biener  <rguenther@suse.de>
12164         * tree-vectorizer.h (struct vec_info_shared): New structure
12165         with parts split out from struct vec_info and loop_nest from
12166         struct _loop_vec_info.
12167         (struct vec_info): Adjust accordingly.
12168         (struct _loop_vec_info): Likewise.
12169         (LOOP_VINFO_LOOP_NEST): Adjust.
12170         (LOOP_VINFO_DATAREFS): Likewise.
12171         (LOOP_VINFO_DDRS): Likewise.
12172         (struct _bb_vec_info): Likewise.
12173         (BB_VINFO_DATAREFS): Likewise.
12174         (BB_VINFO_DDRS): Likewise.
12175         (struct _stmt_vec_info): Add dr_aux member.
12176         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
12177         (DR_MISALIGNMENT_UNINITIALIZED): New.
12178         (set_dr_misalignment): Adjust.
12179         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
12180         (vect_analyze_loop): Adjust prototype.
12181         (vect_analyze_loop_form): Likewise.
12182         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12183         Compute dependences lazily.
12184         (vect_record_base_alignments): Use shared datarefs/ddrs.
12185         (vect_verify_datarefs_alignment): Likewise.
12186         (vect_analyze_data_refs_alignment): Likewise.
12187         (vect_analyze_data_ref_accesses): Likewise.
12188         (vect_analyze_data_refs): Likewise.
12189         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
12190         constructor parameter for shared part.
12191         (vect_analyze_loop_form): Pass in shared part and adjust.
12192         (vect_analyze_loop_2): Pass in storage for the number of
12193         stmts.  Move loop nest finding to the caller.  Compute
12194         datarefs lazily.
12195         (vect_analyze_loop): Pass in shared part.
12196         (vect_transform_loop): Verify shared datarefs are unchanged.
12197         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
12198         constructor parameter for shared part.
12199         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
12200         (vect_slp_bb): Verify shared datarefs are unchanged before
12201         transform.
12202         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
12203         change.
12204         (new_stmt_vec_info): Initialize DR_AUX misalignment to
12205         DR_MISALIGNMENT_UNINITIALIZED.
12206         * tree-vectorizer.c (vec_info::vec_info): Add constructor
12207         parameter for shared part.
12208         (vec_info::~vec_info): Adjust.
12209         (vec_info_shared::vec_info_shared): New.
12210         (vec_info_shared::~vec_info_shared): Likewise.
12211         (vec_info_shared::save_datarefs): Likewise.
12212         (vec_info_shared::check_datarefs): Likewise.
12213         (try_vectorize_loop_1): Construct shared part live for analyses
12214         of a single loop for multiple vector sizes.
12215         * tree-parloops.c (gather_scalar_reductions): Adjust.
12217 2018-06-25  Richard Biener  <rguenther@suse.de>
12219         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
12220         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
12221         (vect_analyze_data_refs): Remove similar code from here and
12222         simplify accordingly.
12224 2018-06-25  Richard Biener  <rguenther@suse.de>
12226         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
12227         for reverse storage order accesses rather than asserting
12228         they cannot happen here.
12230 2018-06-25  Tom de Vries  <tdevries@suse.de>
12232         PR debug/86257
12233         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
12234         Use data16 instead of .byte for insn prefix.
12236 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
12238         PR C++/86082
12239         * parser.c (make_char_string_pack): Pass this literal chars
12240         through cpp_interpret_string.
12241         (cp_parser_userdef_numeric_literal): Check the result of
12242         make_char_string_pack.
12244 2018-06-24  Maya Rashish  <coypu@sdf.org>
12246         * ginclude/stddef.h: Simplify conditions around avoiding
12247         re-definition of __size_t.
12249 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12251         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
12252         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
12254 2018-06-22  Maya Rashish  <coypu@sdf.org>
12256         * doc/invoke.texi (mno-fancy-math-387): Update for changes
12257         made to OpenBSD and NetBSD through the years.
12259 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12261         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12262         behavior of vec_pack (vector double, vector double) to match
12263         behavior of vec_float2 (vector double, vector double).
12265 2018-06-22  Olivier Hainque  <hainque@adacore.com>
12267         * gimplify.c (gimplify_function_tree): Prevent creation
12268         of a trampoline for the address of the current function
12269         passed to entry/exit instrumentation hooks.
12271 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
12273         PR target/86222
12274         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12275         correctly.
12277 2018-06-22  Martin Liska  <mliska@suse.cz>
12279         PR tree-optimization/86263
12280         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12281         Bail out if is_enabled is false.
12282         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12283         New declaration.
12284         (jump_table_cluster::is_enabled): New function.
12286 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12288         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12289         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12290         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12291         (lto_input_ts_binfo_tree_pointers): Likewise.
12292         * tree-streamer-out.c (streamer_write_tree_bitfields,
12293         write_ts_binfo_tree_pointers): Likewise.
12294         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12296 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12298         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12300 2018-06-22  Martin Liska  <mliska@suse.cz>
12302         * symbol-summary.h (get): Make it pure and inline move
12303         functionality from ::get function.
12304         (get): Remove and inline into ::get and ::get_create.
12305         (get_create): Move code from ::get function.
12307 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12309         PR target/85994
12310         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12311         -x assembler-with-cpp.
12313 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12315         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12316         _FILE_OFFSET_BITS=64 for C++.
12318 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
12320         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12321         conversion insn that shows up when pr85657-3.c is compiled using
12322         IEEE 128-bit long double.
12323         (neg<mode>2_internal): Use the correct mode to check whether the
12324         mode is IBM extended.
12325         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12326         multiply and divide external functions from being created more
12327         than once.
12329 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12331         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12332         functions.
12333         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
12334         the edge can be forwarded.
12335         (cfg_layout_merge_blocks): Likewise.
12337 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12339         * except.c (finish_eh_generation): Commit edge insertions only after
12340         the EH edges have been redirected from post-landing to landing pads.
12342 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12344         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12345         create_tmp_var_for to create the FRAME decl.
12346         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12348 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12350         * tree-inline.c (copy_edges_for_bb): Minor tweak.
12351         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12352         debug statement when resetting its value.
12353         (expand_call_inline): Copy the locus of the call onto the assignment
12354         of the return value, if any.  Use local variable in more cases.
12356 2018-06-21  Martin Liska  <mliska@suse.cz>
12358         * ipa-pure-const.c (propagate_nothrow): Use
12359         funct_state_summaries->get.
12360         (dump_malloc_lattice): Likewise.
12361         (propagate_malloc): Likewise.
12363 2018-06-21  Richard Biener  <rguenther@suse.de>
12365         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12366         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12367         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12368         comment.
12369         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12370         BLOCK_ABSTRACT_ORIGIN unconditionally.
12372 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
12374         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12375         deleting it.
12376         * ipa-reference.c (ipa_reference_c_finalize): Delete
12377         ipa_ref_opt_sum_summaries and set it to NULL.
12379 2018-06-21  Tom de Vries  <tdevries@suse.de>
12381         PR tree-optimization/85859
12382         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12383         test with comment from bb_no_side_effects_p.
12385 2018-06-21  Richard Biener  <rguenther@suse.de>
12387         PR tree-optimization/86232
12388         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12389         max for constant niter.
12391 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12393         * config/aarch64/aarch64-simd.md
12394         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12396 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12398         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12399         Make opernads of the unspec commutative.
12401 2018-06-21  Richard Biener  <rguenther@suse.de>
12403         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12404         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12405         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12406         (vect_analyze_data_ref_dependence): Re-order checks to deal with
12407         NULL DR_STEP.
12408         (vect_record_base_alignments): Do not record base alignment
12409         for gathers or scatters.
12410         (vect_compute_data_ref_alignment): Drop return value that is always
12411         true.  Bail out early for gathers or scatters.
12412         (vect_enhance_data_refs_alignment): Bail out early for gathers
12413         or scatters.
12414         (vect_find_same_alignment_drs): Likewise.
12415         (vect_analyze_data_refs_alignment): Remove dead code.
12416         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12417         (vect_analyze_data_refs): For possible gathers or scatters do
12418         not create an alternate DR, just check their possible validity
12419         and mark them.  Adjust DECL_NONALIASED handling to not rely
12420         on DR_BASE_ADDRESS.
12421         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12422         update inits of gathers or scatters.
12423         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12424         Also copy gather/scatter flag to pattern vinfo.
12426 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12428         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12429         behavior of vec_packsu (vector unsigned long long, vector unsigned
12430         long long) to match behavior of vec_packs with same signature.
12432 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
12433             Thomas Schwinge <thomas@codesourcery.com>
12434             Cesar Philippidis  <cesar@codesourcery.com>
12436         * gimplify.c (gimplify_scan_omp_clauses): Add support for
12437         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12438         (gimplify_adjust_omp_clauses): Likewise.
12439         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
12440         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
12441         (gimplify_omp_target_update): Update handling of acc update and
12442         enter/exit data.
12443         * omp-low.c (install_var_field): Remove unused parameter
12444         base_pointers_restrict.
12445         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
12446         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
12447         FINALIZE}
12448         (omp_target_base_pointers_restrict_p): Delete.
12449         (scan_omp_target): Update call to scan_sharing_clauses.
12450         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
12451         FINALIZE}.
12452         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
12453         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12454         (convert_local_omp_clauses): Likewise.
12455         * tree-pretty-print.c (dump_omp_clause): Likewise.
12456         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
12457         FINALIZE}.
12458         (omp_clause_code_name): Likewise.
12460 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12462         PR debug/86194
12463         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
12464         be narrowed.
12466         PR tree-optimization/86231
12467         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
12468         anti-range don't overwrite *vr0min before using it to compute *vr0max.
12470 2018-06-20  Tom de Vries  <tdevries@suse.de>
12472         PR tree-optimization/86097
12473         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
12474         iv type if signedness of iv type is not the same as that of *nit.
12476 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12478         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
12479         EDGE_EH edges, verify they are all EDGE_EH.
12481 2018-06-20  Maya Rashish  <coypu@sdf.org>
12483         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
12485 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12487         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
12488         * config/aarch64/aarch64.c (xgene1_tunings): Add
12489         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
12490         (aarch64_mode_valid_for_sched_fusion_p):
12491         Allow 16-byte modes.
12492         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
12493         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
12494         128-bit modes.
12495         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
12496         New pattern.
12497         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
12498         * config/aarch64/iterators.md (VQ2): New mode iterator.
12500 2018-06-20  Martin Liska  <mliska@suse.cz>
12502         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12503         Change default ratio from 10 to 8.
12505 2018-06-20  Martin Liska  <mliska@suse.cz>
12507         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12508         New.
12509         (bit_test_cluster::find_bit_tests): Likewise.
12510         (switch_decision_tree::analyze_switch_statement): Find clusters.
12511         * tree-switch-conversion.h (struct jump_table_cluster): Document
12512         hierarchy.
12514 2018-06-20  Martin Liska  <mliska@suse.cz>
12516         * tree-switch-conversion.c (switch_conversion::collect):
12517         Record m_uniq property.
12518         (switch_conversion::expand): Bail out for special conditions.
12519         (group_cluster::~group_cluster): New.
12520         (group_cluster::group_cluster): Likewise.
12521         (group_cluster::dump): Likewise.
12522         (jump_table_cluster::emit): New.
12523         (switch_decision_tree::fix_phi_operands_for_edges): New.
12524         (struct case_node): Remove struct.
12525         (jump_table_cluster::can_be_handled): New.
12526         (case_values_threshold): Moved to header.
12527         (reset_out_edges_aux): Likewise.
12528         (jump_table_cluster::is_beneficial): New.
12529         (bit_test_cluster::can_be_handled): Likewise.
12530         (add_case_node): Remove.
12531         (bit_test_cluster::is_beneficial): New.
12532         (case_bit_test::cmp): New.
12533         (bit_test_cluster::emit): New.
12534         (expand_switch_as_decision_tree_p): Remove.
12535         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
12536         (fix_phi_operands_for_edge): Likewise.
12537         (switch_decision_tree::analyze_switch_statement): New.
12538         (compute_cases_per_edge): Move ...
12539         (switch_decision_tree::compute_cases_per_edge): ... here.
12540         (try_switch_expansion): Likewise.
12541         (switch_decision_tree::try_switch_expansion): Likewise.
12542         (record_phi_operand_mapping): Likewise.
12543         (switch_decision_tree::record_phi_operand_mapping): Likewise.
12544         (emit_case_decision_tree): Likewise.
12545         (switch_decision_tree::emit): Likewise.
12546         (balance_case_nodes): Likewise.
12547         (switch_decision_tree::balance_case_nodes): Likewise.
12548         (dump_case_nodes): Likewise.
12549         (switch_decision_tree::dump_case_nodes): Likewise.
12550         (emit_jump): Likewise.
12551         (switch_decision_tree::emit_jump): Likewise.
12552         (emit_cmp_and_jump_insns): Likewise.
12553         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
12554         (emit_case_nodes): Likewise.
12555         (switch_decision_tree::emit_case_nodes): Likewise.
12556         (conditional_probability): Remove.
12557         * tree-switch-conversion.h (enum cluster_type): New.
12558         (PRINT_CASE): New.
12559         (struct cluster): Likewise.
12560         (cluster::cluster): Likewise.
12561         (struct simple_cluster): Likewise.
12562         (simple_cluster::simple_cluster): Likewise.
12563         (struct group_cluster): Likewise.
12564         (struct jump_table_cluster): Likewise.
12565         (struct bit_test_cluster): Likewise.
12566         (struct min_cluster_item): Likewise.
12567         (struct case_tree_node): Likewise.
12568         (case_tree_node::case_tree_node): Likewise.
12569         (jump_table_cluster::case_values_threshold): Likewise.
12570         (struct case_bit_test): Likewise.
12571         (struct switch_decision_tree): Likewise.
12572         (struct switch_conversion): Likewise.
12573         (switch_decision_tree::reset_out_edges_aux): Likewise.
12575 2018-06-20  Martin Liska  <mliska@suse.cz>
12577         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
12578         (hoist_edge_and_branch_if_true): Likewise.
12579         (expand_switch_using_bit_tests_p): Likewise.
12580         (struct case_bit_test): Likewise.
12581         (case_bit_test_cmp): Likewise.
12582         (emit_case_bit_tests): Likewise.
12583         (switch_conversion::switch_conversion): New class.
12584         (struct switch_conv_info): Remove old struct.
12585         (collect_switch_conv_info): More to ...
12586         (switch_conversion::collect): ... this.
12587         (check_range): Likewise.
12588         (switch_conversion::check_range): Likewise.
12589         (check_all_empty_except_final): Likewise.
12590         (switch_conversion::check_all_empty_except_final): Likewise.
12591         (check_final_bb): Likewise.
12592         (switch_conversion::check_final_bb): Likewise.
12593         (create_temp_arrays): Likewise.
12594         (switch_conversion::create_temp_arrays): Likewise.
12595         (free_temp_arrays): Likewise.
12596         (gather_default_values): Likewise.
12597         (switch_conversion::gather_default_values): Likewise.
12598         (build_constructors): Likewise.
12599         (switch_conversion::build_constructors): Likewise.
12600         (constructor_contains_same_values_p): Likewise.
12601         (switch_conversion::contains_same_values_p): Likewise.
12602         (array_value_type): Likewise.
12603         (switch_conversion::array_value_type): Likewise.
12604         (build_one_array): Likewise.
12605         (switch_conversion::build_one_array): Likewise.
12606         (build_arrays): Likewise.
12607         (switch_conversion::build_arrays): Likewise.
12608         (gen_def_assigns): Likewise.
12609         (switch_conversion::gen_def_assigns): Likewise.
12610         (prune_bbs): Likewise.
12611         (switch_conversion::prune_bbs): Likewise.
12612         (fix_phi_nodes): Likewise.
12613         (switch_conversion::fix_phi_nodes): Likewise.
12614         (gen_inbound_check): Likewise.
12615         (switch_conversion::gen_inbound_check): Likewise.
12616         (process_switch): Use the newly created class.
12617         (switch_conversion::expand): New.
12618         (switch_conversion::~switch_conversion): New.
12619         * tree-switch-conversion.h: New file.
12621 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12623         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
12624         tree-vect-patterns.c.
12625         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
12626         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
12627         (vect_recog_sad_pattern): Likewise.
12628         (vect_recog_widen_sum_pattern): Likewise.
12629         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
12630         (vect_recog_widen_shift_pattern): Remove the type_in argument.
12631         (vect_recog_rotate_pattern): Likewise.
12632         (vect_recog_mult_pattern): Likewise.
12633         (vect_recog_vector_vector_shift_pattern): Likewise.
12634         (vect_recog_divmod_pattern): Likewise.
12635         (vect_recog_mixed_size_cond_pattern): Likewise.
12636         (vect_recog_bool_pattern): Likewise.
12637         (vect_recog_mask_conversion_pattern): Likewise.
12638         (vect_try_gather_scatter_pattern): Likewise.
12639         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
12640         (vect_recog_over_widening_pattern): Likewise.
12641         (vect_recog_gather_scatter_pattern): Likewise.
12642         (vect_recog_func_ptr): Move from tree-vectorizer.h
12643         (vect_vect_recog_func_ptrs): Move further down the file.
12644         (vect_recog_func): Likewise.  Remove the third argument.
12645         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
12646         (vect_pattern_recog_1): Expect the pattern function to do any
12647         necessary target tests.  Also expect it to provide a vector type.
12648         Remove the type_in handling.
12650 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12652         * tree-vect-patterns.c (vect_pattern_detected): New function.
12653         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
12654         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
12655         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
12656         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
12657         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12658         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12659         (vect_recog_mask_conversion_pattern)
12660         (vect_try_gather_scatter_pattern): Likewise.
12662 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12664         * tree-vect-patterns.c (vect_get_internal_def): New function.
12665         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12666         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
12667         (search_type_for_mask_1): Use it.
12669 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12671         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12672         redundant WIDEN_SUM_EXPR handling.
12673         (vect_recog_sad_pattern): Likewise.
12675 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12677         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12678         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
12679         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
12680         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
12681         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
12683 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12685         * tree-vect-stmts.c (vectorizable_call): Make sure that we
12686         use the stmt_vec_info of the original bb statement for the
12687         new zero assignment, even if the call is part of a pattern.
12689 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12691         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
12692         that the sequence is attached to the original statement rather
12693         than the pattern statement.
12694         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
12695         PATTERN_DEF_SEQ from the original statement rather than
12696         the main pattern statement.
12697         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
12698         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
12699         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
12701 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12703         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
12704         definition statements before the early exit for statements that aren't
12705         live or relevant.
12706         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
12707         split out from...
12708         (vect_transform_loop): ...here.  Process pattern definition
12709         statements without first checking whether the main pattern
12710         statement is live or relevant.
12712 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
12714         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
12715         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
12717 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
12719         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
12720         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
12721         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
12722         (expand_block_compare): Change select_block_compare_mode call.
12723         (expand_strncmp_align_check): Use new functions, fix comment.
12724         (emit_final_str_compare_gpr): New function.
12725         (expand_strn_compare): Refactor and clean up code.
12726         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
12728 2018-06-19  Tony Reix  <tony.reix@atos.com>
12729             Damien Bergamini  <damien.bergamini@atos.com>
12730             David Edelsohn  <dje.gcc@gmail.com>
12732         * collect2.c (static_obj): New variable.
12733         (static_libs): New variable.
12734         (is_in_list): Uncomment declaration.
12735         (main): Track AIX libraries linked statically.
12736         (is_in_list): Uncomment definition.
12737         (scan_prog_file): Don't add AIX shared libraries initializer
12738         to constructor list if linking statically.
12740 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
12742         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
12743         constant.
12744         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
12746 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
12748         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
12749         blocks.
12751 2018-06-19  Martin Liska  <mliska@suse.cz>
12753         * config/i386/i386.c (ix86_can_inline_p): Do not use
12754         ipa_fn_summaries::get_create.
12755         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
12756         get.
12757         (devirtualization_time_bonus): Likewise.
12758         (ipcp_propagate_stage): Likewise.
12759         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12760         (edge_set_predicate): Likewise.
12761         (evaluate_conditions_for_known_args): Likewise.
12762         (evaluate_properties_for_edge): Likewise.
12763         (ipa_call_summary::reset): Tranform to ...
12764         (ipa_call_summary::~ipa_call_summary): ... this.
12765         (ipa_fn_summary::reset): Transform to ...
12766         (ipa_fn_summary::~ipa_fn_summary): ... this.
12767         (ipa_fn_summary_t::remove): Rename to ...
12768         (ipa_fn_summary_t::remove_callees): ... this.
12769         (ipa_fn_summary_t::duplicate): Use placement new
12770         instead of memory copy.
12771         (ipa_call_summary_t::duplicate): Likewise.
12772         (ipa_call_summary_t::remove): Remove.
12773         (dump_ipa_call_summary): Change get_create to get.
12774         (ipa_dump_fn_summary): Dump only when summary exists.
12775         (analyze_function_body): Use symbol_summary::get instead
12776         of get_create.
12777         (compute_fn_summary): Likewise.
12778         (estimate_edge_devirt_benefit): Likewise.
12779         (estimate_edge_size_and_time): Likewise.
12780         (inline_update_callee_summaries): Likewise.
12781         (remap_edge_change_prob): Likewise.
12782         (remap_edge_summaries): Likewise.
12783         (ipa_merge_fn_summary_after_inlining): Likewise.
12784         (write_ipa_call_summary): Likewise.
12785         (ipa_fn_summary_write): Likewise.
12786         (ipa_free_fn_summary): Likewise.
12787         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
12788         (struct ipa_call_summary): Likewise.
12789         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
12790         of get_create.
12791         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
12792         (estimate_size_after_inlining): Likewise.
12793         (estimate_growth): Likewise.
12794         (growth_likely_positive): Likewise.
12795         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12796         (inline_call): Likewise.
12797         * ipa-inline.c (caller_growth_limits): Likewise.
12798         (can_inline_edge_p): Likewise.
12799         (can_inline_edge_by_limits_p): Likewise.
12800         (compute_uninlined_call_time): Likewise.
12801         (compute_inlined_call_time): Likewise.
12802         (want_inline_small_function_p): Likewise.
12803         (edge_badness): Likewise.
12804         (update_caller_keys): Likewise.
12805         (update_callee_keys): Likewise.
12806         (inline_small_functions): Likewise.
12807         (inline_to_all_callers_1): Likewise.
12808         (dump_overall_stats): Likewise.
12809         (early_inline_small_functions): Likewise.
12810         (early_inliner): Likewise.
12811         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12812         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12813         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12814         * ipa-split.c (execute_split_functions): Likewise.
12815         * symbol-summary.h: Likewise.
12816         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12818 2018-06-19  Richard Biener  <rguenther@suse.de>
12820         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
12821         (vectorize_loops): ... here.  Fix dbgcnt handling.
12822         (try_vectorize_loop): Wrap try_vectorize_loop_1.
12824 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
12826         PR target/86197
12827         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
12828         ieee128 argument takes up only one (vector) register, not two (floating
12829         point) registers.
12831 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
12833         * gimplify.c (gimplify_init_constructor): Really never clear for an
12834         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
12836 2018-06-19  Richard Biener  <rguenther@suse.de>
12838         PR tree-optimization/86179
12839         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
12840         after failed recognition.
12842 2018-06-18  Martin Sebor  <msebor@redhat.com>
12844         PR middle-end/85602
12845         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
12846         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
12847         Handle integer subtraction.
12848         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
12849         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
12851 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12853         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
12854         param from rtx to rtx_insn *.
12855         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
12856         param.
12857         (frv_ifcvt_modify_insn): Likwise.
12858         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
12859         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
12860         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
12861         as_a <rtx_insn *> cast to local "unprotected_region" once
12862         it's been established that it's not NULL or pc_rtx.
12863         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
12864         param "sethi" from rtx to rtx_insn *.
12865         (nds32_group_float_insns): Likewise for param "insn".
12866         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
12867         param.
12868         (vax_output_int_subtract): Likewise.
12869         * config/vax/vax.c (vax_output_int_add): Likewise for param
12870         "insn".
12871         (vax_output_int_subtract): Likewise.
12872         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
12873         (emit_pattern_after): Likewise for param "after".
12874         (emit_insn_after): Likewise.
12875         (emit_jump_insn_after): Likewise.
12876         (emit_call_insn_after): Likewise.
12877         (emit_debug_insn_after): Likewise.
12878         (emit_pattern_before): Likewise for param "before".
12879         (emit_insn_before): Likewise.
12880         (emit_jump_insn_before): Likewise.
12881         * final.c (get_insn_template): Likewise for param "insn", removing
12882         a cast.
12883         * output.h (get_insn_template): Likewise for 2nd param.
12884         * rtl.h (emit_insn_before): Likewise.
12885         (emit_jump_insn_before): Likewise.
12886         (emit_debug_insn_before_noloc): Likewise.
12887         (emit_insn_after): Likewise.
12888         (emit_jump_insn_after): Likewise.
12889         (emit_call_insn_after): Likewise.
12890         (emit_debug_insn_after): Likewise.
12891         (set_insn_deleted): Likewise for param.
12893 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
12895         PR target/85358
12896         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
12897         floating point modes, so that IFmode is numerically greater than
12898         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
12899         to declare the ordering.  This prevents IFmode from being
12900         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
12901         machine.  Include rs6000-modes.h to share the fractional values
12902         between genmodes* and the rest of the compiler.
12903         (IFmode): Likewise.
12904         (KFmode): Likewise.
12905         (TFmode): Likewise.
12906         * config/rs6000/rs6000-modes.h: New file.
12907         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
12908         meaning of rs6000_long_double_size so that 126..128 selects an
12909         appropriate 128-bit floating point type.
12910         (rs6000_option_override_internal): Likewise.
12911         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
12912         (TARGET_LONG_DOUBLE_128): Change the meaning of
12913         rs6000_long_double_size so that 126..128 selects an appropriate
12914         128-bit floating point type.
12915         (LONG_DOUBLE_TYPE_SIZE): Update comment.
12916         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
12917         source and destination to match the standard usage.
12918         (truncifkf2): Likewise.
12919         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
12920         ISA 2.07 to use an explicit clobber, instead of passing in a
12921         temporary.
12922         (copysign<mode>3_soft): Likewise.
12924 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12926         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12927         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
12928         (vect_slp_analyze_instance_dependence): Likewise.
12929         (vect_enhance_data_refs_alignment): Likewise.
12930         (vect_analyze_data_refs_alignment): Likewise.
12931         (vect_slp_analyze_and_verify_instance_alignment
12932         (vect_analyze_data_ref_accesses): Likewise.
12933         (vect_prune_runtime_alias_test_list): Likewise.
12934         (vect_analyze_data_refs): Likewise.
12935         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
12936         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12937         (vect_analyze_scalar_cycles_1): Likewise.
12938         (vect_get_loop_niters): Likewise.
12939         (vect_analyze_loop_form_1): Likewise.
12940         (vect_update_vf_for_slp): Likewise.
12941         (vect_analyze_loop_operations): Likewise.
12942         (vect_analyze_loop): Likewise.
12943         (vectorizable_induction): Likewise.
12944         (vect_transform_loop): Likewise.
12945         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
12946         * tree-vect-slp.c (vect_analyze_slp): Likewise.
12947         (vect_make_slp_decision): Likewise.
12948         (vect_detect_hybrid_slp): Likewise.
12949         (vect_slp_analyze_operations): Likewise.
12950         (vect_slp_bb): Likewise.
12951         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
12952         (vectorizable_bswap): Likewise.
12953         (vectorizable_call): Likewise.
12954         (vectorizable_simd_clone_call): Likewise.
12955         (vectorizable_conversion): Likewise.
12956         (vectorizable_assignment): Likewise.
12957         (vectorizable_shift): Likewise.
12958         (vectorizable_operation): Likewise.
12959         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
12961 2018-06-18  Martin Sebor  <msebor@redhat.com>
12963         PR tree-optimization/81384
12964         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
12965         * builtins.c (expand_builtin_strnlen): New function.
12966         (expand_builtin): Call it.
12967         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
12968         * builtins.def (BUILT_IN_STRNLEN): New.
12969         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
12970         Warn for bounds in excess of maximum object size.
12971         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
12972         single-value ranges.  Handle strnlen.
12973         (handle_builtin_strlen): Handle strnlen.
12974         (strlen_check_and_optimize_stmt): Same.
12975         * doc/extend.texi (Other Builtins): Document strnlen.
12977 2018-06-18  Maya Rashish  <coypu@sdf.org>
12979         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
12980         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
12981         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
12983         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
12984         here to ...
12985         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
12987 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12989         * tree.c (escaped_string::escape): Replace cast to char * by
12990         const_cast<char *> (unescaped).
12992 2018-06-18  Nick Clifton  <nickc@redhat.com>
12994         PR 84195
12995         * tree.c (escaped_string): New class.  Converts an unescaped
12996         string into its escaped equivalent.
12997         (warn_deprecated_use): Use the new class to convert the
12998         deprecation message, if present.
12999         (test_escaped_strings): New self test.
13000         (test_c_tests): Add test_escaped_strings.
13001         * doc/extend.texi (deprecated): Add a note that the
13002         deprecation message is affected by the -fmessage-length
13003         option, and that control characters will be escaped.
13004         (#pragma GCC error): Document this pragma.
13005         (#pragma GCC warning): Likewise.
13006         * doc/invoke.texi (-fmessage-length): Document this option's
13007         effect on the #warning and #error preprocessor directives and
13008         the deprecated attribute.
13010 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
13012         * tree.c (decl_value_expr_lookup): Revert latest change.
13013         (decl_value_expr_insert): Likewise.
13015 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
13017         * gimplify.c (nonlocal_vlas): Delete.
13018         (nonlocal_vla_vars): Likewise.
13019         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
13020         referenced VLAs.
13021         (gimplify_body): Do not create and destroy nonlocal_vlas.
13022         * tree-nested.c: Include diagnostic.h.
13023         (use_pointer_in_frame): Tweak.
13024         (lookup_field_for_decl): Add assertion and declare the transformation.
13025         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
13026         internal error when the reference is in a wrong context.  Do not
13027         create a debug decl by default.
13028         (note_nonlocal_block_vlas): Delete.
13029         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
13030         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
13031         create a debug decl by default.
13032         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
13033         call is in a wrong context.
13034         (fixup_vla_decls): New function.
13035         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
13036         debug variables were created.
13037         * tree.c (decl_value_expr_lookup): Add checking assertion.
13038         (decl_value_expr_insert): Likewise.
13040 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13042         PR middle-end/82479
13043         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
13044         * tree-scalar-evolution.c (interpret_expr): Likewise.
13045         (expression_expensive_p): Likewise.
13046         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
13047         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
13048         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
13049         (ssa_defined_by_minus_one_stmt_p): New.
13051 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
13053         PR middle-end/64946
13054         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
13055         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
13056         * dojump.c (do_jump): Likewise.
13057         * expr.c (expand_expr_real_2): Check operand type's sign.
13058         * fold-const.c (const_unop): Handle ABSU_EXPR.
13059         (fold_abs_const): Likewise.
13060         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
13061         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
13062         (strip_sign_op_1): Likesise.
13063         * match.pd: Add new pattern to generate ABSU_EXPR.
13064         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
13065         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13066         * tree-eh.c (operation_could_trap_helper_p): Likewise.
13067         * tree-inline.c (estimate_operator_cost): Likewise.
13068         * tree-pretty-print.c (dump_generic_node): Likewise.
13069         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
13070         * tree.def (ABSU_EXPR): New.
13072 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
13074         PR middle-end/86095
13075         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
13076         documented as preserved for backward compatibility only.
13077         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
13079         PR rtl-optimization/86108
13080         * bb-reorder.c (create_forwarder_block): Renamed to ...
13081         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
13082         jump from new landing pad to the second part.
13083         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
13084         Adjust callers.
13086 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
13088         PR middle-end/85878
13089         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
13090         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
13091         Only call store_expr for halves if the mode is the same.
13093         PR middle-end/86123
13094         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
13095         Fix up comment formatting.
13097 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13099         * typed-splay-tree.h (typed_splay_tree::remove): New function.
13100         (typed_splay_tree::closure,
13101         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
13102         (typed_splay_tree::typed_splay_tree,
13103         typed_splay_tree::operator =): Declared private.
13104         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
13105         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
13106         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
13107         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
13108         typed_splay_tree::splay_tree_splay,
13109         typed_splay_tree::splay_tree_foreach_helper,
13110         typed_splay_tree::splay_tree_insert,
13111         typed_splay_tree::splay_tree_remove,
13112         typed_splay_tree::splay_tree_lookup,
13113         typed_splay_tree::splay_tree_predecessor,
13114         typed_splay_tree::splay_tree_successor,
13115         typed_splay_tree::splay_tree_min,
13116         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
13117         (typed_splay_tree::root, typed_splay_tree::comp,
13118         typed_splay_tree::delete_key,
13119         typed_splay_tree::delete_value): New data members.
13120         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
13121         typed_splay_tree::remove.
13123 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
13125         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
13126         -mginv and -mno-ginv to the assembler.
13127         * config/mips/mips.opt (-mcrc): New option.
13128         (-mginv): Likewise.
13129         * doc/invoke.text (-mcrc): Document.
13130         (-mginv): Likewise.
13132 2018-06-15  Nick Clifton  <nickc@redhat.com>
13134         PR 84195
13135         * tree.c (escaped_string): New class.  Converts an unescaped
13136         string into its escaped equivalent.
13137         (warn_deprecated_use): Use the new class to convert the
13138         deprecation message, if present.
13139         (test_escaped_strings): New self test.
13140         (test_c_tests): Add test_escaped_strings.
13141         * doc/extend.texi (deprecated): Add a note that the
13142         deprecation message is affected by the -fmessage-length
13143         option, and that control characters will be escaped.
13144         (#pragma GCC error): Document this pragma.
13145         (#pragma GCC warning): Likewise.
13146         * doc/invoke.texi (-fmessage-length): Document this option's
13147         effect on the #warning and #error preprocessor directives and
13148         the deprecated attribute.
13150 2018-06-15  Richard Biener  <rguenther@suse.de>
13152         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
13153         here, also noting vector size used.
13154         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
13155         size used in MSG_OPTIMIZED_LOCATIONS dump.
13156         (pass_slp_vectorize::execute): Adjust.
13158 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
13160         PR target/85968
13161         * config/arc/arc.c (arc_return_address_register): Fix
13162         if-condition.
13164 2018-06-15  Richard Biener  <rguenther@suse.de>
13166         PR middle-end/86159
13167         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
13168         leave useless conversion stripping to force_gimple_operand_gsi.
13169         (gimplify_build2): Likewise.
13170         (gimplify_build1): Likewise.
13172 2018-06-15  Richard Biener  <rguenther@suse.de>
13174         PR middle-end/86076
13175         * tree-cfg.c (move_stmt_op): unshare invariant addresses
13176         before adjusting their block.
13178 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13180         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
13181         multilibs for *-*-rtems*.
13182         * config/riscv/t-rtems: New file.
13184 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
13186         PR middle-end/86122
13187         * match.pd ((A +- CST1) +- CST2): Punt if last resort
13188         unsigned_type_for returns NULL.
13190         PR target/85945
13191         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
13192         subregs of multi-word pseudos unless the float mode has word size.
13194 2018-06-14  Richard Biener  <rguenther@suse.de>
13196         PR middle-end/86139
13197         * tree-vect-generic.c (build_word_mode_vector_type): Remove
13198         duplicate and harmful type_hash_canon.
13199         * tree.c (type_hash_canon): Assert we didn't find ourselves.
13201 2018-06-14  Richard Biener  <rguenther@suse.de>
13203         PR ipa/86124
13204         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
13205         NULL cgraph_node.
13207 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
13209         * config/rtems.h (STDINT_LONG32): Define.
13211 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
13212             Prachi Godbole  <prachi.godbole@imgtec.com>
13214         * config/mips/mips-cpus.def: Define P6600.
13215         * config/mips/mips-tables.opt: Regenerate.
13216         * config/mips/mips.c (mips_ucbranch_type): New enum.
13217         (mips_rtx_cost_data): Add support for P6600.
13218         (mips_issue_rate): Likewise.
13219         (mips_multipass_dfa_lookahead): Likewise.
13220         (mips_avoid_hazard): Likewise.
13221         (mips_reorg_process_insns): Likewise.
13222         (mips_classify_branch_p6600): New function.
13223         * config/mips/mips.h (TUNE_P6600): New define.
13224         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
13225         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
13226         * config/mips/mips.md: Include p6600.md.
13227         (processor): Add p6600.
13228         * config/mips/p6600.md: New file.
13229         * doc/invoke.texi: Add p6600 to supported architectures.
13231 2018-06-13  Martin Sebor  <msebor@redhat.com>
13233         PR tree-optimization/86114
13234         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
13235         of integer types.
13236         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
13238 2018-06-13  Richard Biener  <rguenther@suse.de>
13240         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
13241         Properly set vector type of the intermediate stmt.
13242         * tree-vect-stmts.c (vectorizable_operation): The destination
13243         var always has vectype_out type.
13245 2018-06-13  Jeff Law  <law@redhat.com>
13247         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
13248         integer 0 for argument to print_rtl_with_bb.
13249         (rl78_reorg): Likewise.
13251 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
13253         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
13254         from rtx to rtx_insn *.
13255         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13256         "label".
13257         (add_sched_insns_for_speculation): Likewise for local "target",
13258         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13259         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13260         from rtx_insn ** to rtx_code_label **.
13261         (reorg_emit_nops): Likewise.
13262         (c6x_reorg): Likewise for local "call_labels".
13263         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13264         rtx to rtx_insn *.
13265         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13266         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13267         the loops over LABEL_REFS.
13268         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13269         braf_label.
13270         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13271         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13272         (split_branches): Strengthen local "olabel" from rtx to
13273         rtx_insn *, adding a safe_as_a cast.
13274         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13275         to "rtx_insn *".
13276         (add_insn_after): Likewise for first two params.
13277         (add_insn_before): Likewise.
13278         (remove_insn): Likewise for param.
13279         (emit_pattern_before_noloc): Likewise for second and third params.
13280         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13281         (emit_call_insn_before_noloc): Likewise.
13282         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13283         to "rtx_insn *".
13284         (emit_barrier_before): Likewise.
13285         (emit_label_before): Strengthen "label" param from "rtx" to
13286         "rtx_code_label *".  Strengthen "before" param from "rtx" to
13287         "rtx_insn *".
13288         (emit_insn_after_1): Strengthen "after" param from "rtx" to
13289         "rtx_insn *".
13290         (emit_pattern_after_noloc): Likewise.
13291         (emit_insn_after_noloc): Likewise.
13292         (emit_jump_insn_after_noloc): Likewise.
13293         (emit_call_insn_after_noloc): Likewise.
13294         (emit_debug_insn_after_noloc): Likewise.
13295         (emit_barrier_after): Likewise.
13296         (emit_label_after): Likewise for both params.
13297         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
13298         "loc" param from "int" to "location_t".
13299         (emit_insn_after_setloc): Likewise.
13300         (emit_jump_insn_after_setloc): Likewise.
13301         (emit_call_insn_after_setloc): Likewise.
13302         (emit_debug_insn_after_setloc): Likewise.
13303         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
13304         "loc" param from "int" to "location_t".
13305         (emit_pattern_before): Convert NULL_RTX to NULL.
13306         (emit_insn_before_setloc): Convert "loc" param from "int" to
13307         "location_t".
13308         (emit_jump_insn_before_setloc): Likewise.
13309         (emit_call_insn_before_setloc): Likewise.
13310         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13311         rtx_insn *.  Convert "loc" param from "int" to "location_t".
13312         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13313         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13314         Convert 3rd param from "int" to "location_t".
13315         (emit_barrier_before, emit_barrier_after, next_real_insn):
13316         Strengthen param from rtx to rtx_insn *.
13317         (emit_label_before): Strengthen 1st param from "rtx" to
13318         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
13319         "rtx_insn *".
13320         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13321         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13322         Strengthen 2nd param from "rtx" to "rtx_insn *".
13323         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13324         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13325         Likewise. Convert 3rd param from "int" to "location_t".
13326         (emit_label_after): Strengthen 1st param from "rtx" to
13327         "rtx_code_label *".
13328         (next_real_insn, remove_insn): Strengthen param from "rtx" to
13329         "rtx_insn *".
13330         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13331         from "rtx" to "rtx_insn *".
13333 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
13335         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13336         bodies streamed in with -Q.
13337         * dumpfile.c (dump_files): Add lto-stream-out dump file.
13338         * dumpfile.h (tree_dump_index): Add lto_stream_out.
13339         * gimple-streamer-out.c: Include gimple-pretty-print.h
13340         (output_bb): Dump stmts streamed.
13341         * lto-section-out.c: Include print-tree.h
13342         (lto_begin_section): Dump sections created.
13343         (lto_output_decl_index): Dump decl encoded.
13344         * lto-streamer-out.c: Include print-tree.h
13345         (create_output_block): Dump output block created.
13346         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13347         (output_function): Dump function output.
13348         (output_constructor): Dump constructor streamed.
13349         (write_global_stream): Output indexes encoded.
13350         (produce_asm_for_decls): Dump streams encoded.
13351         * lto-streamer.c (streamer_dump_file): New global var.
13352         * lto-streamer.h (streamer_dump_file): Declare.
13353         * passes.c (ipa_write_summaries): Initialize streamer dump.
13354         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13355         in.
13357 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
13359         PR target/86048
13360         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13361         offsets for register save directives.  Emit a second batch of save
13362         directives, if need be, when the function accesses prior frames.
13364 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13366         * config/arc/fpu.md (fmasf4): Force operand to register.
13367         (fnmasf4): Likewise.
13369 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13371         * config/arc/arc-protos.h (arc_pad_return): Remove.
13372         * config/arc/arc.c (machine_function): Remove force_short_suffix
13373         and size_reason.
13374         (arc_print_operand): Adjust printing of '&'.
13375         (arc_verify_short): Remove conditional printing of short suffix.
13376         (arc_final_prescan_insn): Remove reference to size_reason.
13377         (pad_return): New function.
13378         (arc_reorg): Call pad_return.
13379         (arc_pad_return): Remove.
13380         (arc_init_machine_status): Remove reference to force_short_suffix.
13381         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13382         (attr length): When attribute iscompact is true force to 2
13383         regardless; in the case of maybe check if we want to force the
13384         instruction to have 4 bytes length.
13385         (nopv): Change it to generate 4 byte long nop as well.
13386         (blockage): New pattern.
13387         (simple_return): Remove call to arc_pad_return.
13388         (p_return_i): Likewise.
13390 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13392         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13394 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13396         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13397         ARC cores.
13399 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13401         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13402         for ARC700 and ARCv2.
13404 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
13406         PR target/86076
13407         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13408         operands[2] instead of operands[1].
13411 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13413         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13414         case, check whether the outer register overlaps an unallocatable
13415         register, not just whether it fits the required class.
13417 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13419         * poly-int.h (can_div_trunc_p): Add new overload in which all values
13420         are poly_ints.
13421         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13422         (memrefs_conflict_p): Likewise.
13423         (init_alias_analysis): Likewise.
13424         * cfgexpand.c (expand_debug_expr): Likewise.
13425         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13426         * cse.c (fold_rtx): Likewise.
13427         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13428         * expr.c (emit_block_move_hints): Likewise.
13429         (clear_storage_hints, push_block, emit_push_insn): Likewise.
13430         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13431         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13432         (emit_group_store): Likewise.
13433         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
13434         to read the PRE/POST_MODIFY increment.
13435         * calls.c (store_one_arg): Use strip_offset.
13436         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13437         poly_int_rtx_p.
13438         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13439         by a VEC_SELECT.
13440         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
13441         (simplify_binary_operation_1): Extend CONST_INT handling to
13442         poly_int_rtx_p.
13443         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
13444         than a HOST_WIDE_INT.
13445         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
13446         poly_int64.
13447         (adjust_mems, add_stores): Update accodingly.
13448         (vt_canonicalize_addr): Track polynomial offsets.
13449         (emit_note_insn_var_location): Likewise.
13450         (vt_add_function_parameter): Likewise.
13451         (vt_initialize): Likewise.
13453 2018-06-12  Jeff Law  <law@redhat.com>
13455         * config.gcc (alpha*-*-freebsd*): Remove.
13456         * config/alpha/freebsd.h: Remove.
13458 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
13460         PR other/69968
13461         * spellcheck-tree.c (levenshtein_distance): Rename to...
13462         (get_edit_distance): ...this, and update for underlying renaming.
13463         * spellcheck-tree.h (levenshtein_distance): Rename to...
13464         (get_edit_distance): ...this.
13465         * spellcheck.c (levenshtein_distance): Rename to...
13466         (get_edit_distance): ...this.  Convert from Levenshtein distance
13467         to Damerau-Levenshtein distance by supporting transpositions of
13468         adjacent characters.  Rename "v1" to "v_next" and "v0" to
13469         "v_one_ago".
13470         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
13471         (selftest::test_edit_distance_unit_test_oneway): ...this, and
13472         update for underlying renaming.
13473         (selftest::levenshtein_distance_unit_test): Rename to...
13474         (selftest::test_get_edit_distance_unit): ...this, and update for
13475         underlying renaming.
13476         (selftest::test_find_closest_string): Add example from PR 69968
13477         where transposition helps
13478         (selftest::test_metric_conditions): Update for renaming.
13479         (selftest::test_metric_conditions): Likewise.
13480         (selftest::spellcheck_c_tests): Likewise.
13481         * spellcheck.h (levenshtein_distance): Rename both overloads to...
13482         (get_edit_distance): ...this.
13483         (best_match::consider): Update for renaming.
13485 2018-06-12  Martin Sebor  <msebor@redhat.com>
13487         PR tree-optimization/85259
13488         * builtins.c (compute_objsize): Handle constant offsets.
13489         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
13490         true iff a warning has been issued.
13491         * gimple.h (gimple_nonartificial_location): New function.
13492         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
13493         gimple_nonartificial_location and handle -Wno-system-headers.
13494         (handle_builtin_stxncpy): Same.
13496 2018-06-12  Martin Sebor  <msebor@redhat.com>
13498         PR c/85931
13499         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
13501 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
13503         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
13504         BUILTIN_VEC_XST entries for pointer to double and long long.
13506 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
13508         PR target/85990
13509         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13510         Update comments.
13511         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13512         Likewise.
13514 2018-06-12  Martin Liska  <mliska@suse.cz>
13516         * doc/options.texi: Document IntegerRange.
13518 2018-06-12  Martin Liska  <mliska@suse.cz>
13520         * config/i386/i386.opt: Make MPX-related options as Deprecated.
13521         * opt-functions.awk: Handle Deprecated flag.
13522         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
13523         and report error.
13524         (read_cmdline_option): Report warning for a deprecated option.
13525         * opts.h (struct cl_option): Add new field cl_deprecated.
13526         (CL_ERR_DEPRECATED): New.
13528 2018-06-12  Martin Liska  <mliska@suse.cz>
13530         * doc/options.texi: Document Deprecated option flag.
13532 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13534         * config/arc/arc-arch.h (arc_extras): New enum.
13535         (arc_cpu_t):Add field extra.
13536         (arc_cpu_types): Consider the extras.
13537         * config/arc/arc-cpus.def: Add extras info.
13538         * config/arc/arc-opts.h (processor_type): Consider extra field.
13539         * config/arc/arc.c (arc_override_options): Handle extra field.
13541 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13543         * config/arc/arc-arch.h: Update ARC_OPTX macro.
13544         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
13545         field.
13546         * config/arc/arc.c (arc_init): Update pic warning.
13547         (irq_range): Update irq range parsing warnings.
13548         (arc_override_options): Update various warning messages.
13549         (arc_handle_aux_attribute): Likewise.
13551 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
13553         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
13555 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13557         * doc/sourcebuild.texi: Document usage of line number 0 in verify
13558         compiler messages directives.
13560 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
13562         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
13563         * config/mips/mips-tables.opt: Regenerate.
13564         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
13565         mips64r6.
13566         * doc/invoke.texi: Document -march=i6500.
13568 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
13570         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
13571         (i6400_gpmul): Add cpu_unit.
13572         (i6400_gpdiv): Likewise.
13573         (i6400_msa_add_d): Update reservations.
13574         (i6400_msa_int_add) Likewise.
13575         (i6400_msa_short_logic3) Likewise.
13576         (i6400_msa_short_logic2) Likewise.
13577         (i6400_msa_short_logic) Likewise.
13578         (i6400_msa_move) Likewise.
13579         (i6400_msa_cmp) Likewise.
13580         (i6400_msa_short_float2) Likewise.
13581         (i6400_msa_div_d) Likewise.
13582         (i6400_msa_long_logic1) Likewise.
13583         (i6400_msa_long_logic2) Likewise.
13584         (i6400_msa_mult) Likewise.
13585         (i6400_msa_long_float2) Likewise.
13586         (i6400_msa_long_float4) Likewise.
13587         (i6400_msa_long_float5) Likewise.
13588         (i6400_msa_long_float8) Likewise.
13589         (i6400_fpu_fadd): Include frint type.
13590         (i6400_fpu_store): New define_insn_reservation.
13591         (i6400_fpu_load): Likewise.
13592         (i6400_fpu_move): Likewise.
13593         (i6400_fpu_fcmp): Likewise.
13594         (i6400_fpu_fmadd): Likewise.
13595         (i6400_int_mult): Include imul3nc type and update reservation.
13596         (i6400_int_div): Include idiv3 type and update reservation.
13597         (i6400_int_load): Update to check type not move_type.
13598         (i6400_int_store): Likewise.
13599         (i6400_int_prefetch): Set zero latency.
13601 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
13603         * gcc.c: Document new %@{...} sequence.
13604         (LINK_COMMAND_SPEC): Use it for the -L switches.
13605         (cpp_unique_options): Use it for the -I switches.
13606         (at_file_argbuf): New global variable.
13607         (in_at_file): Likewise.
13608         (alloc_args): Create at_file_argbuf.
13609         (clear_args): Truncate at_file_argbuf.
13610         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
13611         (open_at_file): New function.
13612         (close_at_file): Likewise.
13613         (create_at_file): Delete.
13614         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
13615         <'o'>: Likewise.
13616         <'@'>: New case.
13617         (validate_switches_from_spec): Deal with %@{...} sequence.
13618         (validate_switches): Likewise.
13619         (driver::finalize): Call clear_args.
13621 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
13623         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
13625 2018-06-11  Martin Sebor  <msebor@redhat.com>
13627         * doc/invoke.texi (-Wall): List -Wc++17-compat.
13628         (Wno-class-memaccess): Add @opindex.
13629         (Wno-templates, Wno-multiple-inheritance): Same.
13630         (Wno-virtual-inheritance, Wno-namespaces): Same.
13631         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
13632         (Wno-format-overflow, Wno-format-truncation): Same.
13633         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
13634         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
13635         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
13636         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
13637         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
13638         (Wno-misspelled-isr): Same.
13640 2018-06-11  Martin Sebor  <msebor@redhat.com>
13642         * PR tree-optimization/86083
13643         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
13645 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
13647         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
13649 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
13651         PR target/85755
13652         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
13653         on the correct operand.
13654         (*movdi_internal64): Ditto.
13656 2018-06-11  Martin Liska  <mliska@suse.cz>
13658         PR tree-optimization/86089
13659         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
13661 2018-06-11  Julia Koval  <julia.koval@intel.com>
13663         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
13664         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
13665         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
13667 2018-06-11  Olivier Hainque  <hainque@adacore.com>
13669         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
13670         for Ada with strict dwarf2.
13672 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
13674         PR target/85755
13675         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
13676         addresses.
13678 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
13680         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
13682 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
13684         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
13685         TARGET_ELF.
13687 2018-06-08  Martin Liska  <mliska@suse.cz>
13689         * tree-cfg.h (debug_function): Fix argument type to match
13690         implementation.
13692 2018-06-08  Martin Liska  <mliska@suse.cz>
13694         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
13695         Remove usage of MPX-related (and removed) fields.
13696         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
13698 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
13700         * cfg.c (debug): Use TDF_NONE rather than 0.
13701         * cfghooks.c (debug): Likewise.
13702         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
13703         (struct dump_option_value_info): Convert to...
13704         (struct kv_pair): ...this template type.
13705         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
13706         rather than 0.
13707         (optinfo_verbosity_options): Likewise.
13708         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
13709         OPTGROUP_NONE.
13710         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
13711         than int for "optgroup_flags" param.
13712         (dump_generic_expr_loc): Use dump_flags_t rather than int for
13713         "dump_kind" param.
13714         (dump_dec): Likewise.
13715         (dump_finish): Use TDF_NONE rather than 0.
13716         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
13717         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
13718         than 0.  Update for change to option_ptr.
13719         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
13720         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
13721         0.  Update for changes to optinfo_verbosity_options and
13722         optgroup_options.
13723         (opt_info_switch_p): Convert optgroup_flags from int to
13724         optgroup_flags_t.
13725         (dump_basic_block): Use dump_flags_t rather than int
13726         for "dump_kind" param.
13727         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
13728         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
13729         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
13730         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
13731         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
13732         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
13733         TDF_NONE): Convert from macros to...
13734         (enum dump_flag): ...this new enum.
13735         (dump_flags_t): Update to use enum.
13736         (operator|, operator&, operator~, operator|=, operator&=):
13737         Implement for dump_flags_t.
13738         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
13739         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
13740         Convert from macros to...
13741         (enum optgroup_flag): ...this new enum.
13742         (optgroup_flags_t): New typedef.
13743         (operator|, operator|=): Implement for optgroup_flags_t.
13744         (struct dump_file_info): Convert field "alt_flags" to
13745         dump_flags_t.  Convert field "optgroup_flags" to
13746         optgroup_flags_t.
13747         (dump_basic_block): Use dump_flags_t rather than int for param.
13748         (dump_generic_expr_loc): Likewise.
13749         (dump_dec): Likewise.
13750         (dump_register): Convert param "optgroup_flags" to
13751         optgroup_flags_t.
13752         (opt_info_enable_passes): Likewise.
13753         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
13754         than 0.
13755         * gimple-pretty-print.c (debug): Likewise.
13756         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
13757         (merged_store_group::apply_stores): Likewise.
13758         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
13759         * gimple.c (verify_gimple_pp): Likewise.
13760         * graphite-poly.c (print_pbb_body): Likewise.
13761         * passes.c (pass_manager::register_one_dump_file): Convert
13762         local "optgroup_flags" to optgroup_flags_t.
13763         * print-tree.c (print_node): Use TDF_NONE rather than 0.
13764         (debug): Likewise.
13765         (debug_body): Likewise.
13766         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
13767         to optgroup_flags_t.
13768         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
13769         than 0.
13770         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
13771         (convert_mult_to_fma): Likewise.
13772         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
13773         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
13774         * tree-vect-data-refs.c (dump_lower_bound): Convert param
13775         "dump_kind" to dump_flags_t.
13777 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13779         * config/rs6000/rs6000.c (min, max): Delete.
13781 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13783         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
13784         -mabi=spe and -mabi=no-spe.
13786 2018-06-08  Martin Liska  <mliska@suse.cz>
13788         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
13789         where we expect an existing summary.
13791 2018-06-08  Martin Liska  <mliska@suse.cz>
13793         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
13794         * ipa-inline.h (estimate_edge_growth): Likewise.
13796 2018-06-08  Martin Liska  <mliska@suse.cz>
13798         * cgraph.c (function_version_hasher::hash): Use
13799         cgraph_node::get_uid ().
13800         (function_version_hasher::equal):
13801         * cgraph.h (cgraph_node::get_uid): New method.
13802         * ipa-inline.c (update_caller_keys): Use
13803         cgraph_node::get_uid ().
13804         (update_callee_keys): Likewise.
13805         * ipa-utils.c (searchc): Likewise.
13806         (ipa_reduced_postorder): Likewise.
13807         * lto-cgraph.c (input_node): Likewise.
13808         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
13809         * symbol-summary.h (symtab_insertion): Likewise.
13810         (symtab_removal): Likewise.
13811         (symtab_duplication): Likewise.
13812         * tree-pretty-print.c (dump_function_header): Likewise.
13813         * tree-sra.c (convert_callers_for_node): Likewise.
13815 2018-06-08  Martin Liska  <mliska@suse.cz>
13817         * cgraph.c (symbol_table::create_edge): Always assign a new
13818         unique number.
13819         (symbol_table::free_edge): Do not recycle numbers.
13820         * cgraph.h (cgraph_edge::get): New method.
13821         * symbol-summary.h (symtab_removal): Use it.
13822         (symtab_duplication): Likewise.
13823         (call_summary::hashable_uid): Remove.
13825 2018-06-08  Martin Liska  <mliska@suse.cz>
13827         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
13828         (initialize_growth_caches): Remove.
13829         (free_growth_caches): Likewise.
13830         (do_estimate_edge_time): Use edge_growth_cache.
13831         (do_estimate_edge_size): Likewise.
13832         (do_estimate_edge_hints): Likewise.
13833         * ipa-inline.c (reset_edge_caches): Likewise.
13834         (recursive_inlining): Likewise.
13835         (inline_small_functions): Likewise.
13836         * ipa-inline.h (initialize_growth_caches): Remove.
13837         (estimate_edge_size): Likewise.
13838         (estimate_edge_time): Likewise.
13839         (estimate_edge_hints): Likewise.
13840         (reset_edge_growth_cache): Likewise.
13841         * symbol-summary.h (call_summary::remove): New method.
13843 2018-06-08  Martin Liska  <mliska@suse.cz>
13845         * ipa-cp.c (class edge_clone_summary): New summary.
13846         (grow_edge_clone_vectors): Remove.
13847         (ipcp_edge_duplication_hook): Remove.
13848         (class edge_clone_summary_t): New call_summary class.
13849         (ipcp_edge_removal_hook): Remove.
13850         (edge_clone_summary_t::duplicate): New function.
13851         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
13852         (create_specialized_node): Likewise.
13853         (ipcp_driver): Initialize edge_clone_summaries and do not
13854         register hooks.
13856 2018-06-08  Martin Liska  <mliska@suse.cz>
13858         * symbol-summary.h (get): New function.
13859         (call_summary::m_initialize_when_cloning): New class member.
13861 2018-06-08  Martin Liska  <mliska@suse.cz>
13863         * cgraph.c (cgraph_node::remove): Do not recycle uid.
13864         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
13865         (symbol_table::allocate_cgraph_symbol): Do not set uid.
13866         * passes.c (uid_hash_t): Record removed_nodes by their uids.
13867         (remove_cgraph_node_from_order): Use the removed_nodes set.
13868         (do_per_function_toporder): Likwise.
13869         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
13870         instead of summary_uid.
13871         (symtab_removal): Likewise.
13872         (symtab_duplication): Likewise.
13874 2018-06-08  Martin Liska  <mliska@suse.cz>
13876         * ipa-cp.c (ipcp_store_bits_results): Use
13877         ipcp_transformation_sum.
13878         (ipcp_store_vr_results): Likewise.
13879         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
13880         to ...
13881         (ipcp_transformation_initialize): ... this.
13882         (ipa_set_node_agg_value_chain):
13883         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
13884         (write_ipcp_transformation_info): Likewise.
13885         (read_ipcp_transformation_info): Likewise.
13886         (ipcp_update_bits): Likewise.
13887         (ipcp_update_vr): Likewise.
13888         (ipcp_transform_function): Likewise.
13889         * ipa-prop.h: Rename ipcp_transformation_summary to
13890         ipcp_transformation.
13891         (class ipcp_transformation_t): New function summary.
13892         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
13893         (ipa_get_agg_replacements_for_node): Likewise.
13895 2018-06-08  Martin Liska  <mliska@suse.cz>
13897         * ipa-pure-const.c (struct funct_state_d): Do it class instead
13898         of struct.
13899         (class funct_state_summary_t): New function_summary class.
13900         (has_function_state): Remove.
13901         (get_function_state): Likewise.
13902         (set_function_state): Likewise.
13903         (add_new_function): Likewise.
13904         (funct_state_summary_t::insert): New function.
13905         (duplicate_node_data): Remove.
13906         (remove_node_data): Remove.
13907         (funct_state_summary_t::duplicate): New function.
13908         (register_hooks): Create new funct_state_summaries.
13909         (pure_const_generate_summary): Use it.
13910         (pure_const_write_summary): Likewise.
13911         (pure_const_read_summary): Likewise.
13912         (propagate_pure_const): Likewise.
13913         (propagate_nothrow): Likewise.
13914         (dump_malloc_lattice): Likewise.
13915         (propagate_malloc): Likewise.
13916         (execute): Do not register hooks, just remove summary
13917         instead.
13918         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
13919         constructor.
13921 2018-06-08  Martin Liska  <mliska@suse.cz>
13923         * ipa-reference.c (remove_node_data): Remove.
13924         (duplicate_node_data): Likewise.
13925         (class ipa_ref_var_info_summary_t): New class.
13926         (class ipa_ref_opt_summary_t): Likewise.
13927         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
13928         (get_reference_optimization_summary): Use
13929         ipa_ref_opt_sum_summaries.
13930         (set_reference_vars_info): Remove.
13931         (set_reference_optimization_summary): Likewise.
13932         (ipa_init): Create summaries.
13933         (init_function_info): Use function summary.
13934         (ipa_ref_opt_summary_t::duplicate): New function.
13935         (ipa_ref_opt_summary_t::remove): New function.
13936         (get_read_write_all_from_node): Fix GNU coding style.
13937         (propagate): Use function summary.
13938         (write_node_summary_p): Fix GNU coding style.
13939         (stream_out_bitmap): Likewise.
13940         (ipa_reference_read_optimization_summary): Use function summary.
13941         (ipa_reference_c_finalize): Do not release hooks.
13943 2018-06-08  Martin Liska  <mliska@suse.cz>
13945         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
13946         (analyze_function_body): Extract multiple calls of get_create.
13947         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13948         * ipa-inline.c (recursive_inlining): Use ::get method.
13949         * ipa-inline.h (estimate_edge_growth): Likewise.
13951 2018-06-08  Martin Liska  <mliska@suse.cz>
13953         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
13954         HSA_INVALID.
13955         (hsa_function_summary::hsa_function_summary): Use the new enum
13956         value.
13957         (hsa_gpu_implementation_p): Use hsa_summaries::get.
13958         * hsa-gen.c (hsa_get_host_function): Likewise.
13959         (get_brig_function_name): Likewise.
13960         * ipa-hsa.c (process_hsa_functions): Likewise.
13961         (ipa_hsa_write_summary): Likewise.
13962         * symbol-summary.h (symtab_duplication): Use ::get function/
13963         (get): New function.
13965 2018-06-08  Martin Liska  <mliska@suse.cz>
13967         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
13968         of get.
13969         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
13970         (hsa_register_kernel): Likewise.
13971         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
13972         * hsa-gen.c (hsa_get_host_function): Likewise.
13973         (get_brig_function_name): Likewise.
13974         (generate_hsa): Likewise.
13975         (pass_gen_hsail::execute): Likewise.
13976         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
13977         (devirtualization_time_bonus): Likewise.
13978         (ipcp_propagate_stage): Likewise.
13979         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13980         (edge_set_predicate): Likewise.
13981         (evaluate_conditions_for_known_args): Likewise.
13982         (evaluate_properties_for_edge): Likewise.
13983         (ipa_fn_summary::reset): Likewise.
13984         (ipa_fn_summary_t::duplicate): Likewise.
13985         (dump_ipa_call_summary): Likewise.
13986         (ipa_dump_fn_summary): Likewise.
13987         (analyze_function_body): Likewise.
13988         (compute_fn_summary): Likewise.
13989         (estimate_edge_devirt_benefit): Likewise.
13990         (estimate_edge_size_and_time): Likewise.
13991         (estimate_calls_size_and_time): Likewise.
13992         (estimate_node_size_and_time): Likewise.
13993         (inline_update_callee_summaries): Likewise.
13994         (remap_edge_change_prob): Likewise.
13995         (remap_edge_summaries): Likewise.
13996         (ipa_merge_fn_summary_after_inlining): Likewise.
13997         (ipa_update_overall_fn_summary): Likewise.
13998         (read_ipa_call_summary): Likewise.
13999         (inline_read_section): Likewise.
14000         (write_ipa_call_summary): Likewise.
14001         (ipa_fn_summary_write): Likewise.
14002         (ipa_free_fn_summary): Likewise.
14003         * ipa-hsa.c (process_hsa_functions): Likewise.
14004         (ipa_hsa_write_summary): Likewise.
14005         (ipa_hsa_read_section): Likewise.
14006         * ipa-icf.c (sem_function::merge): Likewise.
14007         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
14008         (do_estimate_edge_time): Likewise.
14009         (estimate_size_after_inlining): Likewise.
14010         (estimate_growth): Likewise.
14011         (growth_likely_positive): Likewise.
14012         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14013         (inline_call): Likewise.
14014         * ipa-inline.c (caller_growth_limits): Likewise.
14015         (can_inline_edge_p): Likewise.
14016         (can_inline_edge_by_limits_p): Likewise.
14017         (compute_uninlined_call_time): Likewise.
14018         (compute_inlined_call_time): Likewise.
14019         (want_inline_small_function_p): Likewise.
14020         (edge_badness): Likewise.
14021         (update_caller_keys): Likewise.
14022         (update_callee_keys): Likewise.
14023         (recursive_inlining): Likewise.
14024         (inline_small_functions): Likewise.
14025         (inline_to_all_callers_1): Likewise.
14026         (dump_overall_stats): Likewise.
14027         (early_inline_small_functions): Likewise.
14028         (early_inliner): Likewise.
14029         * ipa-inline.h (estimate_edge_growth): Likewise.
14030         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14031         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14032         * ipa-prop.h (IPA_NODE_REF): Likewise.
14033         (IPA_EDGE_REF): Likewise.
14034         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14035         (propagate_malloc): Likewise.
14036         * ipa-split.c (execute_split_functions): Likewise.
14037         * symbol-summary.h: Rename get to get_create.
14038         (get): Likewise.
14039         (get_create): Likewise.
14040         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14042 2018-06-08  Martin Liska  <mliska@suse.cz>
14044         * symbol-summary.h (release): Move definition out of class
14045         declaration.
14046         (symtab_removal): Likewise.
14047         (symtab_duplication): Likewise.
14049 2018-06-08  Martin Liska  <mliska@suse.cz>
14051         * symbol-summary.h (function_summary): Move constructor
14052         implementation out of class declaration.
14053         (release): Likewise.
14054         (symtab_insertion): Likewise.
14055         (symtab_removal): Likewise.
14056         (symtab_duplication): Likewise.
14057         (get): Likewise.
14059 2018-06-08  Martin Liska  <mliska@suse.cz>
14061         * Makefile.in: Remove support for MPX (macros, related functions,
14062         fields in cgraph_node, ...).
14063         * builtin-types.def (BT_BND): Likewise.
14064         (BT_FN_BND_CONST_PTR): Likewise.
14065         (BT_FN_CONST_PTR_BND): Likewise.
14066         (BT_FN_VOID_PTR_BND): Likewise.
14067         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
14068         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
14069         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
14070         (expand_builtin_mempcpy_with_bounds): Likewise.
14071         (expand_builtin_memset_with_bounds): Likewise.
14072         (expand_builtin_memset_args): Likewise.
14073         (std_expand_builtin_va_start): Likewise.
14074         (expand_builtin): Likewise.
14075         (expand_builtin_with_bounds): Likewise.
14076         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
14077         (DEF_LIB_BUILTIN_CHKP): Likewise.
14078         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
14079         (DEF_CHKP_BUILTIN): Likewise.
14080         (BUILT_IN_MEMCPY): Likewise.
14081         (BUILT_IN_MEMMOVE): Likewise.
14082         (BUILT_IN_MEMPCPY): Likewise.
14083         (BUILT_IN_MEMSET): Likewise.
14084         (BUILT_IN_STPCPY): Likewise.
14085         (BUILT_IN_STRCAT): Likewise.
14086         (BUILT_IN_STRCHR): Likewise.
14087         (BUILT_IN_STRCPY): Likewise.
14088         (BUILT_IN_STRLEN): Likewise.
14089         (BUILT_IN_MEMCPY_CHK): Likewise.
14090         (BUILT_IN_MEMMOVE_CHK): Likewise.
14091         (BUILT_IN_MEMPCPY_CHK): Likewise.
14092         (BUILT_IN_MEMSET_CHK): Likewise.
14093         (BUILT_IN_STPCPY_CHK): Likewise.
14094         (BUILT_IN_STRCAT_CHK): Likewise.
14095         (BUILT_IN_STRCPY_CHK): Likewise.
14096         * calls.c (store_bounds): Likewise.
14097         (emit_call_1): Likewise.
14098         (special_function_p): Likewise.
14099         (maybe_warn_nonstring_arg): Likewise.
14100         (initialize_argument_information): Likewise.
14101         (finalize_must_preallocate): Likewise.
14102         (compute_argument_addresses): Likewise.
14103         (expand_call): Likewise.
14104         * cfgexpand.c (expand_call_stmt): Likewise.
14105         (expand_return): Likewise.
14106         (expand_gimple_stmt_1): Likewise.
14107         (pass_expand::execute): Likewise.
14108         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
14109         (cgraph_node::remove): Likewise.
14110         (cgraph_node::dump): Likewise.
14111         (cgraph_node::verify_node): Likewise.
14112         * cgraph.h (chkp_function_instrumented_p): Likewise.
14113         (symtab_node::get_alias_target): Likewise.
14114         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
14115         (cgraph_local_p): Likewise.
14116         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
14117         (cgraph_edge::rebuild_references): Likewise.
14118         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
14119         (walk_polymorphic_call_targets): Likewise.
14120         (cgraph_node::expand_thunk): Likewise.
14121         (symbol_table::output_weakrefs): Likewise.
14122         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
14123         (ix86_handle_option): Likewise.
14124         * config/i386/constraints.md: Likewise.
14125         * config/i386/i386-builtin-types.def (BND): Likewise.
14126         (VOID): Likewise.
14127         (PVOID): Likewise.
14128         (ULONG): Likewise.
14129         * config/i386/i386-builtin.def (BDESC_END): Likewise.
14130         (BDESC_FIRST): Likewise.
14131         (BDESC): Likewise.
14132         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
14133         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
14134         * config/i386/i386.c (enum reg_class): Likewise.
14135         (ix86_target_string): Likewise.
14136         (ix86_option_override_internal): Likewise.
14137         (ix86_conditional_register_usage): Likewise.
14138         (ix86_valid_target_attribute_inner_p): Likewise.
14139         (ix86_set_indirect_branch_type): Likewise.
14140         (ix86_set_current_function): Likewise.
14141         (ix86_function_arg_regno_p): Likewise.
14142         (init_cumulative_args): Likewise.
14143         (ix86_function_arg_advance): Likewise.
14144         (ix86_function_arg): Likewise.
14145         (ix86_pass_by_reference): Likewise.
14146         (ix86_function_value_regno_p): Likewise.
14147         (ix86_function_value_1): Likewise.
14148         (ix86_function_value_bounds): Likewise.
14149         (ix86_return_in_memory): Likewise.
14150         (ix86_setup_incoming_vararg_bounds): Likewise.
14151         (ix86_va_start): Likewise.
14152         (indirect_thunk_need_prefix): Likewise.
14153         (print_reg): Likewise.
14154         (ix86_print_operand): Likewise.
14155         (ix86_expand_call): Likewise.
14156         (ix86_output_function_return): Likewise.
14157         (reg_encoded_number): Likewise.
14158         (BDESC_VERIFYS): Likewise.
14159         (ix86_init_mpx_builtins): Likewise.
14160         (ix86_init_builtins): Likewise.
14161         (ix86_emit_cmove): Likewise.
14162         (ix86_emit_move_max): Likewise.
14163         (ix86_expand_builtin): Likewise.
14164         (ix86_builtin_mpx_function): Likewise.
14165         (ix86_get_arg_address_for_bt): Likewise.
14166         (ix86_load_bounds): Likewise.
14167         (ix86_store_bounds): Likewise.
14168         (ix86_load_returned_bounds): Likewise.
14169         (ix86_store_returned_bounds): Likewise.
14170         (ix86_class_likely_spilled_p): Likewise.
14171         (ix86_hard_regno_mode_ok): Likewise.
14172         (x86_order_regs_for_local_alloc): Likewise.
14173         (ix86_mitigate_rop): Likewise.
14174         (ix86_bnd_prefixed_insn_p): Likewise.
14175         (ix86_mpx_bound_mode): Likewise.
14176         (ix86_make_bounds_constant): Likewise.
14177         (ix86_initialize_bounds): Likewise.
14178         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
14179         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
14180         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
14181         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
14182         (TARGET_CHKP_BOUND_MODE): Likewise.
14183         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
14184         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
14185         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
14186         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
14187         * config/i386/i386.h (TARGET_MPX): Likewise.
14188         (TARGET_MPX_P): Likewise.
14189         (VALID_BND_REG_MODE): Likewise.
14190         (FIRST_BND_REG): Likewise.
14191         (LAST_BND_REG): Likewise.
14192         (enum reg_class): Likewise.
14193         (BND_REG_P): Likewise.
14194         (BND_REGNO_P): Likewise.
14195         (BNDmode): Likewise.
14196         (ADJUST_INSN_LENGTH): Likewise.
14197         * config/i386/i386.md: Likewise.
14198         * config/i386/i386.opt: Likewise.
14199         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
14200         (defined): Likewise.
14201         (LINK_MPX): Likewise.
14202         (MPX_SPEC): Likewise.
14203         (LIBMPX_SPEC): Likewise.
14204         (LIBMPXWRAPPERS_SPEC): Likewise.
14205         (CHKP_SPEC): Likewise.
14206         * config/i386/predicates.md: Likewise.
14207         * dbxout.c (dbxout_type): Likewise.
14208         * doc/extend.texi: Likewise.
14209         * doc/invoke.texi: Likewise.
14210         * doc/md.texi: Likewise.
14211         * doc/tm.texi: Likewise.
14212         * doc/tm.texi.in: Likewise.
14213         * dwarf2out.c (is_base_type): Likewise.
14214         (gen_formal_types_die): Likewise.
14215         (gen_subprogram_die): Likewise.
14216         (gen_type_die_with_usage): Likewise.
14217         (gen_decl_die): Likewise.
14218         (dwarf2out_late_global_decl): Likewise.
14219         * expr.c (expand_assignment): Likewise.
14220         (emit_storent_insn): Likewise.
14221         (store_expr_with_bounds): Likewise.
14222         (store_expr): Likewise.
14223         (expand_expr_real_1): Likewise.
14224         * expr.h (store_expr_with_bounds): Likewise.
14225         * function.c (use_register_for_decl): Likewise.
14226         (struct bounds_parm_data): Likewise.
14227         (assign_parms_augmented_arg_list): Likewise.
14228         (assign_parm_find_entry_rtl): Likewise.
14229         (assign_parm_is_stack_parm): Likewise.
14230         (assign_parm_load_bounds): Likewise.
14231         (assign_bounds): Likewise.
14232         (assign_parms): Likewise.
14233         (expand_function_start): Likewise.
14234         * gcc.c (CHKP_SPEC): Likewise.
14235         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
14236         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
14237         (wrestrict_dom_walker::check_call): Likewise.
14238         * gimple.c (gimple_build_call_from_tree): Likewise.
14239         * gimple.h (enum gf_mask): Likewise.
14240         (gimple_call_with_bounds_p): Likewise.
14241         (gimple_call_set_with_bounds): Likewise.
14242         * gimplify.c (gimplify_init_constructor): Likewise.
14243         * ipa-cp.c (initialize_node_lattices): Likewise.
14244         (propagate_constants_across_call): Likewise.
14245         (find_more_scalar_values_for_callers_subset): Likewise.
14246         * ipa-hsa.c (process_hsa_functions): Likewise.
14247         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
14248         * ipa-icf.c (sem_function::merge): Likewise.
14249         * ipa-inline.c (early_inliner): Likewise.
14250         * ipa-pure-const.c (warn_function_noreturn): Likewise.
14251         (warn_function_cold): Likewise.
14252         (propagate_pure_const): Likewise.
14253         * ipa-ref.h (enum GTY): Likewise.
14254         * ipa-split.c (find_retbnd): Likewise.
14255         (consider_split): Likewise.
14256         (split_function): Likewise.
14257         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
14258         * ipa.c (walk_polymorphic_call_targets): Likewise.
14259         (symbol_table::remove_unreachable_nodes): Likewise.
14260         (process_references): Likewise.
14261         (cgraph_build_static_cdtor_1): Likewise.
14262         * lto-cgraph.c (lto_output_node): Likewise.
14263         (output_refs): Likewise.
14264         (compute_ltrans_boundary): Likewise.
14265         (input_overwrite_node): Likewise.
14266         (input_node): Likewise.
14267         (input_cgraph_1): Likewise.
14268         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14269         * passes.c (pass_manager::execute_early_local_passes): Likewise.
14270         (class pass_chkp_instrumentation_passes): Likewise.
14271         (make_pass_chkp_instrumentation_passes): Likewise.
14272         * passes.def: Likewise.
14273         * rtl.h (struct GTY): Likewise.
14274         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14275         * stor-layout.c (layout_type): Likewise.
14276         * symtab.c: Likewise.
14277         * target.def: Likewise.
14278         * targhooks.c (default_chkp_bound_type): Likewise.
14279         (default_chkp_bound_mode): Likewise.
14280         (default_builtin_chkp_function): Likewise.
14281         (default_chkp_function_value_bounds): Likewise.
14282         (default_chkp_make_bounds_constant): Likewise.
14283         (default_chkp_initialize_bounds): Likewise.
14284         * targhooks.h (default_chkp_bound_type): Likewise.
14285         (default_chkp_bound_mode): Likewise.
14286         (default_builtin_chkp_function): Likewise.
14287         (default_chkp_function_value_bounds): Likewise.
14288         (default_chkp_make_bounds_constant): Likewise.
14289         (default_chkp_initialize_bounds): Likewise.
14290         * toplev.c (compile_file): Likewise.
14291         (process_options): Likewise.
14292         * tree-core.h (DEF_BUILTIN): Likewise.
14293         (DEF_BUILTIN_CHKP): Likewise.
14294         * tree-inline.c (declare_return_variable): Likewise.
14295         (remap_gimple_stmt): Likewise.
14296         (copy_bb): Likewise.
14297         (initialize_inlined_parameters): Likewise.
14298         (expand_call_inline): Likewise.
14299         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14300         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14301         (make_pass_ipa_chkp_produce_thunks): Likewise.
14302         (make_pass_chkp): Likewise.
14303         (make_pass_chkp_opt): Likewise.
14304         (make_pass_chkp_instrumentation_passes): Likewise.
14305         * tree-pretty-print.c (dump_generic_node): Likewise.
14306         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14307         * tree-ssa-dce.c (propagate_necessity): Likewise.
14308         (eliminate_unnecessary_stmts): Likewise.
14309         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14310         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14311         * tree-ssa-sccvn.h: Likewise.
14312         * tree-ssa-strlen.c (get_string_length): Likewise.
14313         (valid_builtin_call): Likewise.
14314         (adjust_last_stmt): Likewise.
14315         (handle_builtin_strchr): Likewise.
14316         (handle_builtin_strcpy): Likewise.
14317         (handle_builtin_stxncpy): Likewise.
14318         (handle_builtin_memcpy): Likewise.
14319         (handle_builtin_strcat): Likewise.
14320         (strlen_check_and_optimize_stmt): Likewise.
14321         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14322         * tree-streamer-in.c: Likewise.
14323         * tree-streamer.c (record_common_node): Likewise.
14324         * tree.c (tree_code_size): Likewise.
14325         (wide_int_to_tree_1): Likewise.
14326         (type_contains_placeholder_1): Likewise.
14327         (build_common_tree_nodes): Likewise.
14328         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14329         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14330         (POINTER_BOUNDS_P): Likewise.
14331         (BOUNDED_TYPE_P): Likewise.
14332         (BOUNDED_P): Likewise.
14333         (CALL_WITH_BOUNDS_P): Likewise.
14334         (pointer_bounds_type_node): Likewise.
14335         * value-prof.c (gimple_ic): Likewise.
14336         * var-tracking.c (vt_add_function_parameters): Likewise.
14337         * varasm.c (make_decl_rtl): Likewise.
14338         (assemble_start_function): Likewise.
14339         (output_constant): Likewise.
14340         (maybe_assemble_visibility): Likewise.
14341         * varpool.c (ctor_for_folding): Likewise.
14342         * chkp-builtins.def: Remove.
14343         * ipa-chkp.c: Remove.
14344         * ipa-chkp.h: Remove.
14345         * rtl-chkp.c: Remove.
14346         * rtl-chkp.h: Remove.
14347         * tree-chkp-opt.c: Remove.
14348         * tree-chkp.c: Remove.
14349         * tree-chkp.h: Remove.
14351 2018-06-07  Carl Love  <cel@us.ibm.com>
14353         * config/rs6000/vsx.md (vextract_fp_from_shorth,
14354         vextract_fp_from_shortl): Add BE support.
14356 2018-06-07  Paul Koning  <ni1d@arrl.net>
14358         * compare-elim.c (try_merge_compare): Don't merge compare if
14359         address contains a side effect.
14360         (try_eliminate_compare): Likewise.
14362 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
14364         * config.gcc: Support "tremont".
14365         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14366         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14367         PROCESSOR_TREMONT.
14368         * config/i386/i386.c (m_TREMONT): Define.
14369         (processor_target_table): Add "tremont".
14370         (PTA_TREMONT): Define.
14371         (ix86_lea_outperforms): Add TARGET_TREMONT.
14372         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14373         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14374         and M_INTEL_GOLDMONT_PLUS.
14375         (fold_builtin_cpu): Add "tremont".
14376         (ix86_add_stmt_cost): Add TARGET_TREMONT.
14377         (ix86_option_override_internal): Add "tremont".
14378         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14379         (processor_type): Add PROCESSOR_TREMONT.
14380         * config/i386/x86-tune.def: Add m_TREMONT.
14381         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14383 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14385         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14386         symbol defined for msp430i* devices to be lower case.
14388 2018-06-07  Richard Biener  <rguenther@suse.de>
14390         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14391         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
14392         Properly wrap signed arithmetic if overflow wraps.
14394 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
14396         PR tree-optimization/69615
14397         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14398         of a cast from a same precision integral SSA_NAME in a bb dominated
14399         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
14400         cast to utype if rhs2 has already a compatible type.
14402 2018-06-07  Richard Biener  <rguenther@suse.de>
14404         PR tree-optimization/85935
14405         * graphite-scop-detection.c (find_params_in_bb): Analyze
14406         condition operands with respect to the correct loop.  Assert
14407         the analysis doesn't fail.
14409 2018-06-04  Carl Love  <cel@us.ibm.com>
14411         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14412         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
14413         as it is slightly cheaper.
14414         (first_match_or_eos_index_<mode>):
14415         Calculate index using natural element order.
14416         (first_match_index_<mode>):
14417         Calculate index using natural element order.
14418         (first_match_or_eos_index_<mode>):
14419         Calculate index using natural order.
14420         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14421         for BE and LE modes.
14422         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14423         P9V_BUILTIN_VCLZLSBB_V16QI.
14424         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14425         specific.
14427 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14429         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14430         indentation and line wrap for many prototypes.  Add missing
14431         @smallexample directives around block of prototypes for vec_xl and
14432         vec_xst.
14434 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
14436         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14437         track if we pass or return IEEE 128-bit floating point.
14438         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14439         C++ mangling that is compatible with GCC 8.1.
14440         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
14441         (init_cumulative_args): Note if we pass or return IEEE 128-bit
14442         floating point types.
14443         (rs6000_function_arg_advance_1): Likewise.
14444         (rs6000_mangle_type): Optionally generate mangled names that match
14445         what GCC 8.1 generated for IEEE 128-bit floating point types.
14446         (rs6000_globalize_decl_name): If we have an external function that
14447         passes or returns IEEE 128-bit types, generate a weak reference
14448         from the mangled name used in GCC 8.1 to the current mangled
14449         name.
14450         (rs6000_init_builtins): Make __ibm128 use the long double type if
14451         long double is IBM extended double.  Make __float128 use the long
14452         double type if long double is IEEE 128-bit.
14454         PR target/85657
14455         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
14456         macro for __ibm128 built-in functions.
14457         (PACK_IF): Add __ibm128 pack/unpack functions.
14458         (UNPACK_IF): Likewise.
14459         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
14460         enable long double built-in functions if long double is IEEE
14461         128-bit floating point.
14462         (rs6000_invalid_builtin): Update long double built-in function
14463         error message.
14464         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
14465         functions, adjust the built-in function to use the long double
14466         built-in function if __ibm128 and long double are the same type.
14467         * doc/extend.texi (PowerPC builtins): Update documention for
14468         __builtin_{,un}pack_longdouble.  Add documentation for
14469         __builtin_{,un}pack_ibm128.
14471 2018-06-06  Jim Wilson  <jimw@sifive.com>
14473         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
14474         (struct machine_function): New field interrupt_mode.
14475         (riscv_handle_type_attribute): New function.  Add forward declaration.
14476         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
14477         (riscv_expand_epilogue): Check interrupt_mode field.
14478         (riscv_set_current_function): Check interrupt attribute args and
14479         set interrupt_mode field.
14480         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
14481         (riscv_sret, riscv_uret): New.
14482         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
14483         new arguments to interrupt attribute.
14485 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
14487         PR target/63177
14488         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
14489         Don't handle -mcpu=power8 if -mpower9-vector is also used.
14491 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14493         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
14494         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
14495         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14496         several redundant entries.
14498 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
14500         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
14501         type from "rtx" to "rtx_insn *".
14502         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
14503         for local "call_insn", removing cast.
14504         (ix86_expand_call): Likewise, introducing a "call_insn" local.
14506 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
14508         PR tree-optimization/86066
14509         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
14510         for BIT_INSERT_EXPR stores.
14512 2018-06-06  Richard Biener  <rguenther@suse.de>
14514         PR tree-optimization/86062
14515         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14516         component refs ontop
14517         of to be offsetted base.
14519 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14521         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
14522         to be static and remove check on interrupt attribute name.
14524 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14526         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14527         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
14529 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
14531         PR target/79924
14532         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
14533         second argument.
14534         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
14535         Remove second argument, change how error is called.
14536         (aarch64_layout_arg): Remove second argument from
14537         aarch64_err_no_fpadvsimd call.
14538         (aarch64_init_cumulative_args): Ditto.
14539         (aarch64_gimplify_va_arg_expr): Ditto.
14540         * config/aarch64/aarch64.md (mov<mode>): Ditto.
14542 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
14544         * config/i386/i386.md (simple_return_indirect_internal): New expander.
14545         (*simple_return_indirect_internal<mode>): Rename from
14546         simple_return_indirect_internal.  Use W mode iterator.
14547         (rstorssp): New expander.
14548         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
14549         (clrssbsy): New expander.
14550         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
14552 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14554         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
14555         __typeof__.
14556         (cmse_check_pointed_object): Likewise.
14558 2018-06-05  Martin Liska  <mliska@suse.cz>
14560         PR gcov-profile/47618
14561         * doc/invoke.texi: Document how -fprofile-dir format
14562         is extended.
14564 2018-06-05  Richard Biener  <rguenther@suse.de>
14566         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
14567         removal pretend DOM info isn't available so we do not update
14568         it and only remove edges, not dominated blocks.  Actually free
14569         DOM info in case we removed something.  Remove unreachable blocks.
14570         (mfb_keep_latches): Work with either DOM info or marked backedges.
14571         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
14572         first.  Mark backedges if DOM info isn't available.
14573         (Re-)compute DOM info after cleanup_control_flow_pre.
14575 2018-06-05  Richard Biener  <rguenther@suse.de>
14577         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
14578         (locus_discrim_hasher::hash): Adjust.
14579         (locus_discrim_hasher::equal): Likewise.
14580         (next_discriminator_for_locus): Work on line directly.
14581         (same_line_p): Pass in expanded locus1 as well.
14582         (assign_discriminators): Avoid redundant location expansions.
14584 2018-06-05  Richard Biener  <rguenther@suse.de>
14586         PR tree-optimization/86046
14587         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
14588         if required after clearing TREE_ADDRESSABLE.
14590 2018-06-05  Richard Biener  <rguenther@suse.de>
14592         PR tree-optimization/86047
14593         * tree-ssa-loop.c (for_each_index): Glob handling of all
14594         decls and constants and really handle all of them.
14596 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14598         PR target/81497
14599         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
14600         qualifier_void_pointer and qualifier_const_void_pointer.
14601         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
14602         (arm_init_builtins): Handle the above.
14603         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
14604         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
14605         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
14606         void intrinsics.
14608 2018-06-05  Martin Liska  <mliska@suse.cz>
14610         * auto-profile.c (read_autofdo_file): Do not use
14611         gcov_ctr_summary struct.
14612         (afdo_callsite_hot_enough_for_early_inline): Likewise.
14613         * coverage.c (struct counts_entry): Likewise.
14614         (read_counts_file): Read just single summary entry.
14615         (get_coverage_counts): Use gcov_summary struct.
14616         * coverage.h (get_coverage_counts): Likewise.
14617         * gcov-dump.c (dump_working_sets): Likewise.
14618         (tag_summary): Dump just single summary.
14619         * gcov-io.c (gcov_write_summary): Write just histogram
14620         summary.
14621         (gcov_read_summary): Read just single summary.
14622         (compute_working_sets): Use gcov_summary struct.
14623         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
14624         of GCOV_COUNTERS_SUMMABLE.
14625         (GCOV_COUNTERS_SUMMABLE): Remove.
14626         (GCOV_FIRST_VALUE_COUNTER): Replace with
14627         GCOV_COUNTER_V_INTERVAL.
14628         (struct gcov_ctr_summary): Remove.
14629         (struct gcov_summary): Directly use fields of former
14630         gcov_ctr_summary.
14631         (compute_working_sets): Use gcov_summary struct.
14632         * gcov.c (read_count_file): Do not use ctrs fields.
14633         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
14634         struct.
14635         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
14636         struct.
14637         * profile.c: Likewise.
14638         * profile.h: Likewise.
14640 2018-06-05  Martin Liska  <mliska@suse.cz>
14642         PR gcov-profile/84846
14643         * gcov.c (output_lines): Print working directory only
14644         in intermediate format.
14646 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
14648         * config/s390/s390-builtin-types.def: Add void function type.
14649         * config/s390/s390-builtins.def: Use the function type for the
14650         tbeginc builtin.
14652 2018-06-04  Jim Wilson  <jimw@sifive.com>
14654         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
14655         to int.
14656         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
14657         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
14658         handle EH_RETURN_DATA_REGNO registers properly.
14659         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
14660         (riscv_expand_epilogue): Update comment.  Change argument name and
14661         type.  Update code to use new name and type.  Pass new args to
14662         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
14663         EXCEPTION_RETURN.
14664         * config/riscv/riscv.md (NORMAL_RETURN): New.
14665         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
14666         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
14667         (eh_return): Call gen_eh_return_internal and emit barrier.
14668         (eh_return_internal): Call riscv_expand_epilogue.
14670 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
14672         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
14673         bit_insertion field and declare can_be_merged_into method.
14674         (merged_store_group::can_be_merged_into): New method.
14675         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
14676         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
14677         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
14679 2018-06-04  Richard Biener  <rguenther@suse.de>
14681         PR tree-optimization/85955
14682         * builtins.c (fold_builtin_sincos): Convert pointers to
14683         destination to appropriate type before dereferencing.
14685 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
14687         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
14689 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
14691         * expr.c (expand_expr_real_1): Force the operand into memory if
14692         its TYPE_MODE is BLKmode and if there is no integer mode for
14693         the number of bits being extracted.
14695 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
14697         PR target/85832
14698         PR target/86036
14699         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14700         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
14702 2018-06-04  Richard Biener  <rguenther@suse.de>
14704         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
14705         (cleanup_tree_cfg_noloop): ... single caller.  Do
14706         start_recording_case_labels later.
14708 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
14710         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
14711         to _IMMINTRIN_H_INCLUDED.
14712         * config/i386/pconfigintrin.h: Ditto.
14713         * config/i386/waitpkgintrin.h: Ditto.
14714         * config/i386/immintrin.h: Add includes for sgxintrin.h,
14715         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
14716         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
14717         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
14718         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
14719         waitpkgintrin.h and cldemoteintrin.h.
14721 2018-06-04  Richard Biener  <rguenther@suse.de>
14723         PR tree-optimization/86038
14724         * tracer.c (find_best_successor): Check probability for
14725         being initialized, bail out if not.
14727 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
14729         PR target/86003
14730         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
14731         of bits to ignore when comparing architectures.
14733 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
14735         PR tree-optimization/69615
14736         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
14737         maximum or minimum of the type, try to merge it also as if
14738         range1 is + [-, x - 1] or + [x + 1, -].
14740         PR c++/86025
14741         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
14743 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
14745         PR tree-optimization/86034
14746         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
14747         the unsigned bitfield type in a bit insertion sequence if it does not
14748         have a larger precision than the bitfield size.
14749         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
14751 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
14753         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
14755 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14757         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
14758         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
14759         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
14760         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
14762 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14764         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
14765         Disable -fdelete-null-pointer-checks for ELF toolchain.
14767 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14768             Kito Cheng  <kito.cheng@gmail.com>
14770         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
14771         (nds32le-*-*, nds32be-*-*): Integrate checking process.
14772         (nds32*-*-*): Add glibc and uclibc conditions.
14773         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
14774         (TARGET_EXCEPT_UNWIND_INFO): Define.
14775         * config/nds32/elf.h: New file.
14776         * config/nds32/linux.h: New file.
14777         * config/nds32/nds32-elf.opt: New file.
14778         * config/nds32/nds32-linux.opt: New file.
14779         * config/nds32/nds32-fp-as-gp.c
14780         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
14781         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
14782         TARGET_LINUX_ABI.
14783         (nds32_asm_file_end): Ditto.
14784         (nds32_print_operand): Ditto.
14785         (nds32_insert_attributes): Ditto.
14786         (nds32_init_libfuncs): New function.
14787         (TARGET_HAVE_TLS): Define.
14788         (TARGET_INIT_LIBFUNCS): Define.
14789         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
14790         spec content.
14791         (TARGET_ELF): Apply different mcmodel setting.
14792         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
14793         been migrated into elf.h and linux.h files.
14794         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
14795         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
14796         (mcmodel): The content has been migrated into nds32-elf.opt and
14797         nds32-linux.opt files.
14798         * config/nds32/t-elf: New file.
14799         * config/nds32/t-linux: New file.
14801 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14802             Shiva Chen  <shiva0217@gmail.com>
14804         * config/nds32/constants.md (unspec_volatile_element): Add
14805         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
14806         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
14807         optimization.
14808         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
14809         (make_pass_nds32_fp_as_gp): Declare.
14810         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
14811         optmization pass.
14812         (nds32_asm_function_end_prologue): Remove unused asm output.
14813         (nds32_asm_function_begin_epilogue): Remove unused asm output.
14814         (nds32_asm_file_start): Output necessary fp_as_gp information.
14815         (nds32_option_override): Adjust register usage.
14816         (nds32_expand_prologue): Consider fp_as_gp situation.
14817         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
14818         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
14819         (epilogue): Ditto.
14820         (return): Ditto.
14821         (simple_return): Ditto.
14822         (omit_fp_begin): Output special directive for fp_as_gp.
14823         (omit_fp_end): Output special directive for fp_as_gp.
14824         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
14825         mforbid-fp-as-gp): New options.
14827 2018-06-01  Mark Wielaard  <mark@klomp.org>
14829         * dwarf2out.c (dwarf2out_finish): Remove generation of
14830         DW_AT_loclists_base.
14832 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
14834         * gimple-ssa-store-merging.c: Include gimple-fold.h.
14835         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
14836         (struct merged_store_group): Add bit_insertion field.
14837         (dump_char_array): Use standard hexadecimal format.
14838         (merged_store_group::merged_store_group): Set bit_insertion to false.
14839         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
14840         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
14841         also print the mask in the dump file.
14842         (pass_store_merging::gate): Minor tweak.
14843         (imm_store_chain_info::coalesce_immediate): Fix wrong association
14844         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
14845         stores with INTEGER_CST stores.
14846         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
14847         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
14848         and use it throughout.  Generate bit insertion sequences if need be.
14849         (pass_store_merging::process_store): Remove redundant condition.
14850         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
14852 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
14854         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
14855         the 128-bit floating point types.  Fix function comment.
14857 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14859         * config/aarch64/aarch64-simd.md
14860         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
14861         mnemonics.
14862         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
14863         mnemonics.
14865 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14867         PR tree-optimization/85989
14868         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
14869         variable.
14870         (backprop::intersect_uses): Check it when deciding whether this
14871         is a backedge reference.
14872         (backprop::process_block): Add each phi to m_visited_phis
14873         after visiting it, then clear it at the end.
14875 2018-06-01  Richard Biener  <rguenther@suse.de>
14877         * tree-vectorizer.h (vect_dr_stmt): New function.
14878         (vect_get_load_cost): Adjust.
14879         (vect_get_store_cost): Likewise.
14880         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
14881         Use vect_dr_stmt instead of DR_SMTT.
14882         (vect_record_base_alignments): Likewise.
14883         (vect_calculate_target_alignment): Likewise.
14884         (vect_compute_data_ref_alignment): Likewise and make static.
14885         (vect_update_misalignment_for_peel): Likewise.
14886         (vect_verify_datarefs_alignment): Likewise.
14887         (vector_alignment_reachable_p): Likewise.
14888         (vect_get_data_access_cost): Likewise.  Pass down
14889         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
14890         (vect_get_peeling_costs_all_drs): Likewise.
14891         (vect_peeling_hash_get_lowest_cost): Likewise.
14892         (vect_enhance_data_refs_alignment): Likewise.
14893         (vect_find_same_alignment_drs): Likewise.
14894         (vect_analyze_data_refs_alignment): Likewise.
14895         (vect_analyze_group_access_1): Likewise.
14896         (vect_analyze_group_access): Likewise.
14897         (vect_analyze_data_ref_access): Likewise.
14898         (vect_analyze_data_ref_accesses): Likewise.
14899         (vect_vfa_segment_size): Likewise.
14900         (vect_small_gap_p): Likewise.
14901         (vectorizable_with_step_bound_p): Likewise.
14902         (vect_prune_runtime_alias_test_list): Likewise.
14903         (vect_analyze_data_refs): Likewise.
14904         (vect_supportable_dr_alignment): Likewise.
14905         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14906         (vect_gen_prolog_loop_niters): Likewise.
14907         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14908         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
14909         modify DR_STMT.
14910         (vect_recog_mask_conversion_pattern): Likewise.
14911         (vect_try_gather_scatter_pattern): Likewise.
14912         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
14913         to vect_get_store_cost.
14914         (vect_get_store_cost): Get stmt_info instead of DR.
14915         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
14916         (vect_get_load_cost): Get stmt_info instead of DR.
14918 2018-06-01  Richard Biener  <rguenther@suse.de>
14920         PR middle-end/86017
14921         * gimple-fold.c (var_decl_component_p): Also allow offsetted
14922         vars wrapped in MEM_REFs.
14924 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14926         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14927         Fix subreg tests so that we only return a choice between
14928         GENERAL_REGS and FP_REGS if the original classes included both.
14930 2018-06-01  Richard Biener  <rguenther@suse.de>
14932         PR ipa/85960
14933         * tree-ssa-structalias.c (get_function_part_constraint):
14934         Handle NULL fi->decl.
14935         (find_func_aliases_for_call): Properly handle indirect
14936         fi from direct call.
14937         (find_func_clobbers): Likewise.
14938         (ipa_pta_execute): Likewise.
14939         (create_variable_info_for): For functions that are ifunc_resolver
14940         resolve to a varinfo that contains the result of the resolver call.
14941         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
14942         aliases.
14944 2018-05-31  Michael Collison  <michael.collison@arm.com>
14946         * config/aarch64/aarch64.md:
14947         (*fix_to_zero_extenddfdi2): New pattern.
14948         * gcc.target/aarch64/fix_extend1.c: New testcase.
14950 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
14952         PR middle-end/78809
14953         PR middle-end/83026
14954         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
14955         and BUILT_IN_STRNCMP_EQ.
14956         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
14957         BUILT_IN_STRNCMP_EQ.
14958         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
14959         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14960         (gimple_fold_builtin): Likewise.
14961         * tree-ssa-strlen.c (compute_string_length): New function.
14962         (determine_min_obsize): New function.
14963         (handle_builtin_string_cmp): New function to handle calls to
14964         string compare functions.
14965         (strlen_optimize_stmt): Add handling to builtin string compare
14966         calls.
14967         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
14968         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14969         * tree.c (build_common_builtin_nodes): Add new defines of
14970         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
14972 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
14974         PR target/85984
14975         * bb-reorder.c (pass_partition_blocks::gate): Return false for
14976         functions with naked attribute.
14978 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
14980         * config/i386/sse.md (avx_vec_concat<mode>):
14981         Substitute concat_tg_mode mode attribute with xtg_mode.
14982         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
14983         (concat_tg_mode): Remove mode attribute.
14985 2018-05-31  Martin Sebor  <msebor@redhat.com>
14987         PR c/82063
14988         * calls.c (alloc_max_size): Correct a logic error/typo.
14989         Treat excessive arguments as infinite.  Warn for invalid arguments.
14990         * doc/invoke.texi (-Walloc-size-larger-than): Update.
14992 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
14994         PR target/85829
14995         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
14996         and movx for Haswell.
14998 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
14999             Cesar Philippidis  <cesar@codesourcery.com>
15001         PR middle-end/85879
15002         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
15003         when emitting error on private/firstprivate reductions.
15004         * omp-low.c (lower_omp_target): Avoid reference-type processing
15005         on pointers for firstprivate clause.
15007 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15009         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
15010         (st1x2): Likewise.
15011         (st1x3): Likewise.
15012         * config/aarch64/aarch64-simd.md
15013         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
15014         (aarch64_ld1_x3_<mode>): Likewise
15015         (aarch64_st1x2<VALLDIF:mode>): Likewise
15016         (aarch64_st1_x2_<mode>): Likewise
15017         (aarch64_st1x3<VALLDIF:mode>): Likewise
15018         (aarch64_st1_x3_<mode>): Likewise
15019         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
15020         (vld1_s8_x3): Likewise.
15021         (vld1_u16_x3): Likewise.
15022         (vld1_s16_x3): Likewise.
15023         (vld1_u32_x3): Likewise.
15024         (vld1_s32_x3): Likewise.
15025         (vld1_u64_x3): Likewise.
15026         (vld1_s64_x3): Likewise.
15027         (vld1_f16_x3): Likewise.
15028         (vld1_f32_x3): Likewise.
15029         (vld1_f64_x3): Likewise.
15030         (vld1_p8_x3): Likewise.
15031         (vld1_p16_x3): Likewise.
15032         (vld1_p64_x3): Likewise.
15033         (vld1q_u8_x3): Likewise.
15034         (vld1q_s8_x3): Likewise.
15035         (vld1q_u16_x3): Likewise.
15036         (vld1q_s16_x3): Likewise.
15037         (vld1q_u32_x3): Likewise.
15038         (vld1q_s32_x3): Likewise.
15039         (vld1q_u64_x3): Likewise.
15040         (vld1q_s64_x3): Likewise.
15041         (vld1q_f16_x3): Likewise.
15042         (vld1q_f32_x3): Likewise.
15043         (vld1q_f64_x3): Likewise.
15044         (vld1q_p8_x3): Likewise.
15045         (vld1q_p16_x3): Likewise.
15046         (vld1q_p64_x3): Likewise.
15047         (vst1_s64_x2): Likewise.
15048         (vst1_u64_x2): Likewise.
15049         (vst1_f64_x2): Likewise.
15050         (vst1_s8_x2): Likewise.
15051         (vst1_p8_x2): Likewise.
15052         (vst1_s16_x2): Likewise.
15053         (vst1_p16_x2): Likewise.
15054         (vst1_s32_x2): Likewise.
15055         (vst1_u8_x2): Likewise.
15056         (vst1_u16_x2): Likewise.
15057         (vst1_u32_x2): Likewise.
15058         (vst1_f16_x2): Likewise.
15059         (vst1_f32_x2): Likewise.
15060         (vst1_p64_x2): Likewise.
15061         (vst1q_s8_x2): Likewise.
15062         (vst1q_p8_x2): Likewise.
15063         (vst1q_s16_x2): Likewise.
15064         (vst1q_p16_x2): Likewise.
15065         (vst1q_s32_x2): Likewise.
15066         (vst1q_s64_x2): Likewise.
15067         (vst1q_u8_x2): Likewise.
15068         (vst1q_u16_x2): Likewise.
15069         (vst1q_u32_x2): Likewise.
15070         (vst1q_u64_x2): Likewise.
15071         (vst1q_f16_x2): Likewise.
15072         (vst1q_f32_x2): Likewise.
15073         (vst1q_f64_x2): Likewise.
15074         (vst1q_p64_x2): Likewise.
15075         (vst1_s64_x3): Likewise.
15076         (vst1_u64_x3): Likewise.
15077         (vst1_f64_x3): Likewise.
15078         (vst1_s8_x3): Likewise.
15079         (vst1_p8_x3): Likewise.
15080         (vst1_s16_x3): Likewise.
15081         (vst1_p16_x3): Likewise.
15082         (vst1_s32_x3): Likewise.
15083         (vst1_u8_x3): Likewise.
15084         (vst1_u16_x3): Likewise.
15085         (vst1_u32_x3): Likewise.
15086         (vst1_f16_x3): Likewise.
15087         (vst1_f32_x3): Likewise.
15088         (vst1_p64_x3): Likewise.
15089         (vst1q_s8_x3): Likewise.
15090         (vst1q_p8_x3): Likewise.
15091         (vst1q_s16_x3): Likewise.
15092         (vst1q_p16_x3): Likewise.
15093         (vst1q_s32_x3): Likewise.
15094         (vst1q_s64_x3): Likewise.
15095         (vst1q_u8_x3): Likewise.
15096         (vst1q_u16_x3): Likewise.
15097         (vst1q_u32_x3): Likewise.
15098         (vst1q_u64_x3): Likewise.
15099         (vst1q_f16_x3): Likewise.
15100         (vst1q_f32_x3): Likewise.
15101         (vst1q_f64_x3): Likewise.
15102         (vst1q_p64_x3): Likewise.
15104 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15106         * config/msp430/msp430.c (msp430_output_labelref): Prepend
15107         user_label_prefix to name.
15109         * tree-core.h: Update comment about the format of NAME string
15110         passed to handler in attribute_spec.
15112         * config/msp430/msp430.md: Remove erroneous subreg expression from
15113         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
15114         zero_extend{q,h}isi2.
15116 2018-05-30  Borislav Petkov  <bp@suse.de>
15118         * doc/extend.texi: Document some architecture specific
15119         constraints and sort entries.
15121 2018-05-30  Martin Sebor  <msebor@redhat.com>
15123         PR middle-end/85369
15124         * builtins.c (expand_builtin_stpcpy_1): New function.
15125         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
15126         only if the former succeeds.
15128 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
15130         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
15131         in saphira.
15133 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15135         * doc/invoke.texi (-flinker-output): Document
15137 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15139         * passes.c (ipa_write_summaries): Only modify statements if body
15140         is in memory.
15141         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
15142         incrementally linking.
15143         (ipa_passes): Likewise.
15144         * lto-cgraph.c (lto_output_node): When incrementally linking do not
15145         pass down resolution info.
15146         * common.opt (flag_incremental_link): Update info.
15147         * gcc.c (plugin specs): Turn flinker-output=* to
15148         -plugin-opt=-linker-output-known
15149         * toplev.c (compile_file): Also cut compilation when doing incremental
15150         link.
15151         * flag-types. (enum lto_partition_model): Add
15152         LTO_LINKER_OUTPUT_NOLTOREL.
15153         (invoke.texi): Add -flinker-output docs.
15154         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
15155         link same way as WPA; do not stream in dead initializers.
15157         * dwarf2out.c (dwarf2out_die_ref_for_decl,
15158         darf2out_register_external_decl): Support incremental link.
15160 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15162         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
15164 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
15166         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
15167         it down to simple_object_copy_lto_debug_sections.
15168         (run_gcc): Determine incremental LTO link time and configure
15169         lto1 into non-wpa mode, disable renaming of debug sections.
15171 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15173         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15174         descriptions of various incorrectly documented functions.
15176 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15178         Revert:
15179         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15180         address check not strict.
15182 2018-05-30  Richard Biener  <rguenther@suse.de>
15184         PR tree-optimization/85964
15185         * tracer.c (better_p): Drop initialized count check, we only
15186         call the function with initialized counts now.
15187         (find_best_successor): Do find a best edge if one
15188         has uninitialized count.
15189         (find_best_predecessor): Likewise.  Do BB frequency check only
15190         if count is initialized.
15192 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
15194         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
15195         (aarch64_ldrstr_offset_compare): New.
15196         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
15197         load/store orderings.
15198         (aarch64_gen_adjusted_ldpstp): Likewise.
15200 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
15202         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
15203         Check for subset of GENERAL_REGS and FP_REGS.
15204         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
15205         r=w alternative.
15207 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
15209         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
15210         and wi::to_poly_offset.  Add the current offset and then check
15211         whether the sum fits, rather than using an unchecked addition of
15212         a checked term.  Check for a shwi rather than a uhwi.
15213         * expr.c (get_bit_range): Use tree_to_poly_uint64.
15214         (store_constructor): Use poly_int_tree_p.
15215         (expand_expr_real_1): Likewise.
15216         * function.c (assign_temp): Likewise.
15217         * fold-const.c (const_binop): Use poly_int_tree_p and
15218         wi::to_poly_offset.
15219         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
15220         division.
15221         * ipa-icf-gimple.c (func_checker::compare_operand): Use
15222         to_poly_offset for MEM offsets.
15223         * ipa-icf.c (sem_variable::equals): Likewise.
15224         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
15225         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
15226         wi::to_poly_offset for BIT_FIELD_REF offsets.
15227         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
15228         wi::to_poly_offset.
15229         * var-tracking.c (emit_note_insn_var_location): Use
15230         tree_to_poly_uint64.
15232 2018-05-29  Jim Wilson  <jimw@sifive.com>
15234         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
15236 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
15238         PR target/85950
15239         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
15240         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
15241         sequence.
15242         (sse4_1_round<mode>2): Use nonimmediate_operand
15243         for operand 1 predicate.
15245 2018-05-29  Martin Sebor  <msebor@redhat.com>
15246             Richard Biener  <rguenther@suse.de>
15248         PR testsuite/85888
15249         * calls.c (get_size_range): Call determine_value_range instead
15250         of get_value_range..
15251         * tree-vrp.h (determine_value_range): Declared new function.
15252         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
15254 2018-05-29  Richard Biener  <rguenther@suse.de>
15256         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
15257         sure to use non-pattern stmts for get_earlier_stmt arguments.
15258         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15259         called on pattern stmts.
15260         (get_later_stmt): Likewise.
15262 2018-05-29  Martin Liska  <mliska@suse.cz>
15264         PR gcov-profile/85759
15265         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15266         env variables.
15268 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
15270         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15271         VEC_UNPACK_*_EXPR.
15272         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15273         VEC_PACK_*_EXPR.
15275         PR target/85918
15276         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15277         VEC_PACK_FLOAT_EXPR): New tree codes.
15278         * tree-pretty-print.c (op_code_prio): Handle
15279         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15280         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15281         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15282         * tree-inline.c (estimate_operator_cost): Likewise.
15283         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15284         * fold-const.c (const_binop): Likewise.
15285         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15286         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15287         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15288         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15289         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15290         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15291         * expr.c (expand_expr_real_2): Likewise.
15292         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15293         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15294         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15295         optabs.
15296         * optabs.c (expand_widen_pattern_expr): For
15297         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15298         sign from result type rather than operand's type.
15299         (expand_binop_directly): For vec_packu_float_optab and
15300         vec_packs_float_optab allow result type to be different from operand's
15301         type.
15302         * optabs-tree.c (optab_for_tree_code): Handle
15303         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15304         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
15305         * tree-vect-generic.c (expand_vector_operations_1):  Handle
15306         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15307         VEC_PACK_FLOAT_EXPR.
15308         * tree-vect-stmts.c (supportable_widening_operation): Handle
15309         FIX_TRUNC_EXPR.
15310         (supportable_narrowing_operation): Handle FLOAT_EXPR.
15311         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15312         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15313         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
15314         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15315         mode attributes.
15316         (vec_pack<floatprefix>_float_<mode>): New expander.
15317         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15318         attributes.
15319         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15320         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15321         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15322         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15323         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15324         Document.
15325         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15326         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15327         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15328         VEC_PACK_FLOAT_EXPR): Document.
15330 2018-05-29  Richard Biener  <rguenther@suse.de>
15332         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15333         member.
15334         (stmt_vec_info_vec): Make pointer.
15335         (init_stmt_vec_info_vec): Remove.
15336         (free_stmt_vec_info_vec): Likewise.
15337         (set_stmt_vec_info_vec): New function.
15338         (free_stmt_vec_infos): Likewise.
15339         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15340         (set_vinfo_for_stmt): Likewise.
15341         (get_earlier_stmt): Likewise.
15342         (get_later_stmt): Likewise.
15343         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15344         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15345         (vec_info::~vec_info): Free stmt_vec_infos.
15346         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15347         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15348         (pass_slp_vectorize::execute): Likewise.
15349         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15350         (free_stmt_vec_info_vec): Likewise.
15351         (set_stmt_vec_info_vec): New function.
15352         (free_stmt_vec_infos): Likewise.
15353         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15354         the global stmt_vec_info_vec.
15355         * tree-parloops.c (gather_scalar_reductions): Use
15356         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15357         vector.
15359 2018-05-29  Richard Biener  <rguenther@suse.de>
15361         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15363 2018-05-29  Martin Liska  <mliska@suse.cz>
15364             David Malcolm  <dmalcolm@redhat.com>
15366         * vec.c (test_reverse): New.
15367         (vec_c_tests): Add new test.
15368         * vec.h (vl_ptr>::reverse): New function.
15370 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
15372         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15374         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15375         and later.
15377 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15379         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15381 2018-05-28  Richard Biener  <rguenther@suse.de>
15383         PR tree-optimization/85933
15384         * tree-vect-data-refs.c (vect_record_base_alignments): Only
15385         look at stmts marked as vectorizable.
15387 2018-05-28  Richard Biener  <rguenther@suse.de>
15389         PR tree-optimization/85934
15390         * tree-vect-generic.c (expand_vector_operations_1): Hoist
15391         vector boolean check before scalar optimization.
15393 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
15395         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15396         for armv5te.
15398 2018-05-28  Mark Wielaard  <mark@klomp.org>
15400         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15401         if it is an expression containing a minus sign.
15403 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
15405         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15407 2018-05-27  Paul Koning  <ni1d@arrl.net>
15409         * config/pdp11/pdp11.md (truncsihi2): Remove.
15411 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
15412             Chung-Ju Wu  <jasonwucj@gmail.com>
15414         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15415         implementation.
15416         (unaligned_store_dw): Ditto.
15417         * config/nds32/nds32-memory-manipulation.c
15418         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15419         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15420         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15421         (emit_setmem_word_loop): Rename to ...
15422         (emit_setmem_doubleword_loop): ... this.
15423         (nds32_gen_dup_4_byte_to_word_value): New function.
15424         (nds32_gen_dup_8_byte_to_double_word_value): New function.
15425         (nds32_expand_setmem_loop): Refine implementation.
15426         (nds32_expand_setmem_loop_v3m): Ditto.
15427         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15428         pattern.
15430 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15432         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15434 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15436         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15437         (nds32_init_machine_status): Initialize machine->attr_naked_p and
15438         machine->attr_no_prologue_p.
15439         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
15440         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
15441         (nds32_expand_epilogue): Consider attr_naked_p.
15442         (nds32_expand_epilogue_v3pop): Likewise.
15443         (nds32_can_use_return_insn): Likewise.
15444         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
15445         attr_no_prologue_p fields.
15446         * config/nds32/nds32.opt (mret-in-naked-func): New option.
15448 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
15450         PR target/85918
15451         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
15452         attributes.
15453         * config/i386/sse.md
15454         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
15455         Rename to ...
15456         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
15457         ... this.
15458         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
15459         Rename to ...
15460         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
15461         ... this.
15462         (*<floatsuffix>floatv2div2sf2): Rename to ...
15463         (*float<floatunssuffix>v2div2sf2): ... this.
15464         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
15465         (float<floatunssuffix>v2div2sf2_mask): ... this.
15466         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
15467         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
15468         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
15469         to ...
15470         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
15471         ... this.
15472         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15473         Rename to ...
15474         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15475         ... this.
15476         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15477         Rename to ...
15478         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15479         ... this.
15480         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
15481         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
15482         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
15483         gen_ufix_truncv8dfv8si2.
15484         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
15485         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
15486         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
15487         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
15488         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
15489         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
15490         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
15491         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
15493 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15495         PR target/85900
15496         PR target/85345
15497         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
15499 2018-05-25  Jim Wilson  <jimw@sifive.com>
15501         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
15502         * config/riscv/riscv.c (struct machine_function): Add
15503         interrupt_handler_p and attribute_checked_p fields.
15504         (riscv_attribute_table): Add interrupt.
15505         (riscv_interrupt_type_p): New.
15506         (riscv_save_reg_p): Save extra regs for interrupt handler.
15507         (riscv_use_save_libcall): Return false  for interrupt handler.
15508         (riscv_first_stack_step): Add forward declaration.
15509         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
15510         for interrupt handler with large frame.  Use it for saved reg list.
15511         (riscv_expand_prologue): Move flag_stack_usage_info support to
15512         eliminate duplication.
15513         (riscv_expand_epilogue): Generate mret for interrupt handler.
15514         (riscv_epilogue_uses): New.
15515         (riscv_can_use_return_insn): Return false for interrupt handler.
15516         (riscv_function_ok_for_sibcall): Likewise.
15517         (riscv_set_current_function): Add interrupt handler support.
15518         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
15519         * config/riscv/riscv.md (UNSPECV_MRET): New.
15520         (GP_REGNUM): New.
15521         (riscv_frflags, riscv_fsflags): Use tab after opcode.
15522         (riscv_mret): New.
15523         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
15525 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
15527         PR tree-optimization/85712
15528         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
15529         this candidate has already been replaced in-situ by a copy.
15531 2018-05-25  Jason Merrill  <jason@redhat.com>
15533         PR c++/80485 - inline function non-zero address.
15534         * symtab.c (nonzero_address): Check DECL_COMDAT.
15536 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
15538         PR target/83628
15539         * config/alpha/alpha.md (ashlsi3): New insn pattern.
15540         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
15541         extension of SImode operation.  Use const123_operand predicate.
15542         (*saddsi_1): Remove.
15543         (*saddl_se_1): Ditto.
15544         (*ssubsi_1): Ditto.
15545         (*ssubl_se_1): Ditto.
15546         * config/alpha/predicates.md (const123_operand): New predicate.
15547         * config/alpha/constraints.md (P): Use IN_RANGE.
15549 2018-05-25  Richard Biener  <rguenther@suse.de>
15551         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
15552         defaulted to true.
15553         (ref_maybe_used_by_stmt_p): Likewise.
15554         (stmt_may_clobber_ref_p): Likewise.
15555         (stmt_may_clobber_ref_p_1): Likewise.
15556         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
15557         and pass it along.
15558         (ref_maybe_used_by_stmt_p): Likewise.
15559         (stmt_may_clobber_ref_p): Likewise.
15560         (stmt_may_clobber_ref_p_1): Likewise.
15561         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
15562         the alias oracle to disambiguate DRs with stmts DR analysis
15563         couldn't handle.
15564         (vect_analyze_data_refs): Do not give up on not analyzable
15565         DRs for BB vectorization.  Remove code truncating the dataref
15566         vector.
15568 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
15570         PR target/85832
15571         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15572         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
15573         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
15575 2018-05-25  Richard Biener  <rguenther@suse.de>
15577         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
15578         function, combining stmt data ref gathering and fatal analysis
15579         parts.
15580         (vect_analyze_data_refs): Remove now redudnant code and simplify.
15581         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
15582         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
15583         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
15584         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
15586 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15588         PR tree-optimization/85720
15589         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
15590         SCC if all partitions are builtins.
15591         (version_loop_by_alias_check): New parameter.  Generate cancelable
15592         runtime alias check if all partitions are builtins.
15593         (distribute_loop): Update call to above function.
15595 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15597         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
15598         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
15599         (parm_default_def_partition_arg): Ditto.
15600         (set_parm_default_def_partition): Ditto.
15601         (get_parm_default_def_partitions): Ditto and make it static.
15602         (get_undefined_value_partitions): Ditto and make it static.
15603         (remove_ssa_form): Refactor call to init_var_map here.
15604         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
15605         computation for loop region.
15606         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
15607         (register_default_def): Delete.
15608         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
15609         (parm_default_def_partition_arg): Ditto.
15610         (set_parm_default_def_partition): Ditto.
15611         (get_parm_default_def_partitions): Ditto and make it static.
15612         (get_undefined_value_partitions): Ditto and make it static.
15613         (coalesce_with_default, coalesce_with_default): Update comment.
15614         (create_coalesce_list_for_region): New func factored out from
15615         create_outofssa_var_map.
15616         (populate_coalesce_list_for_outofssa): New func factored out from
15617         create_outofssa_var_map and coalesce_ssa_name.
15618         (create_outofssa_var_map): Delete.
15619         (coalesce_ssa_name): Refactor to support live range computation.
15620         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
15621         (get_parm_default_def_partitions): Delete.
15622         (get_undefined_value_partitions): Ditto.
15623         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
15624         computation for loop region.
15625         (new_tree_live_info, loe_visit_block): Ditto.
15626         (live_worklist, set_var_live_on_entry): Ditto.
15627         (calculate_live_on_exit, verify_live_on_entry): Ditto.
15628         * tree-ssa-live.h (struct _var_map): New fields.
15629         (init_var_map): Change decl.
15630         (region_contains_p): New.
15632 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15634         * tree-ssa-live.h (live_merge_and_clear): Delete.
15636 2018-05-25  Richard Biener  <rguenther@suse.de>
15638         PR c++/85912
15639         * tree-dump.c (dequeue_and_dump): Remove access to removed
15640         operand 2 of a SWITCH_EXPR.
15642 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15644         * doc/sourcebuild.texi (vect_double_cond_arith): Include
15645         multiplication and division.
15646         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
15647         (cond_udiv@var{m}, cond_umod@var{m}): Document.
15648         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
15649         (cond_udiv_optab, cond_umod_optab): New optabs.
15650         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
15651         (IFN_COND_RDIV): New internal functions.
15652         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
15653         TRUNC_MOD_EXPR and RDIV_EXPR.
15654         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
15655         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
15656         New unspecs.
15657         (SVE_INT_BINARY): Include mult.
15658         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
15659         (optab, sve_int_op): Handle mult.
15660         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
15661         UNSPEC_COND_DIV.
15662         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
15663         for SVE_INT_BINARY_SD.
15665 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15667         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
15668         (optab, sve_int_op): Handle div and udiv.
15669         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
15670         for SVE_INT_BINARY_SD.
15671         (*<optab><mode>3): New insn for the same.
15673 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15675         * tree-vect-patterns.c: Include predict.h.
15676         (vect_recog_divmod_pattern): Restrict check for division support
15677         to when optimizing for size.
15679 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15681         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
15682         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
15683         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
15684         (gimple_match_op::set_op): Likewise.
15685         (gimple_resimplify4): Declare.
15686         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
15687         (expr::gen_transform): Likewise.
15688         (decision_tree::gen): Generate a simplification routine for 4 operands.
15689         * gimple-match-head.c (gimple_simplify): Add an overload for
15690         4 operands.  In the top-level function, handle up to 4 call
15691         arguments and call gimple_resimplify4.
15692         (gimple_resimplify4): New function.
15693         (build_call_internal): Pass a fourth operand.
15694         (maybe_push_to_seq): Likewise.
15695         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
15696         Fold VEC_COND_EXPRs of an operation and a default value into
15697         an IFN_COND_* function if possible.
15698         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
15699         New unspecs.
15700         (SVE_COND_FP_BINARY): Include them.
15701         (optab, sve_fp_op): Handle them.
15702         (SVE_INT_BINARY_REV): New code iterator.
15703         (SVE_COND_FP_BINARY_REV): New int iterator.
15704         (commutative): New int attribute.
15705         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
15706         Declare.
15707         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
15708         function.
15709         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
15710         (*cond_<optab><mode>): New patterns for reversed operands.
15712 2018-05-25  Richard Biener  <rguenther@suse.de>
15714         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
15715         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
15716         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
15717         (STMT_VINFO_GROUPED_ACCESS): Adjust.
15718         * tree-vect-data-refs.c (everywhere): Adjust users.
15719         * tree-vect-loop.c (everywhere): Likewise.
15720         * tree-vect-slp.c (everywhere): Likewise.
15721         * tree-vect-stmts.c (everywhere): Likewise.
15722         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
15724 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15726         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
15727         Rename to...
15728         (gcc_cv_as_section_exclude): ... this.
15729         Try Solaris as #exclude syntax.
15730         * configure: Regenerate.
15731         * config.in: Regenerate.
15732         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
15733         SECTION_EXCLUDE.
15734         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
15735         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
15737         * varasm.c (default_elf_asm_named_section): Don't check if
15738         HAVE_GAS_SECTION_EXCLUDE is defined.
15740 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15742         * doc/md.texi: Update the documentation of the cond_* optabs
15743         to mention the new final operand.  Fix GET_MODE_NUNITS call.
15744         Describe the scalar case too.
15745         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
15746         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
15747         instead of 2.
15748         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
15749         (get_conditional_internal_fn): Update comment.
15750         * tree-vect-loop.c (vectorizable_reduction): Pass the original
15751         accumulator value as a final argument to conditional functions.
15752         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
15753         a define_expand and add an "else" operand.  Assert for now that
15754         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
15755         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
15756         (*cond_<optab><mode>): New patterns.
15757         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
15758         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15759         (UNSPEC_COND_EOR): Delete.
15760         (optab): Remove associated mappings.
15761         (SVE_INT_BINARY): New code iterator.
15762         (sve_int_op): Remove int attribute and add "minus" to the code
15763         attribute.
15764         (SVE_COND_INT_OP): Delete.
15765         (SVE_COND_FP_OP): Rename to...
15766         (SVE_COND_FP_BINARY): ...this.
15768 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15770         * optabs.c (can_reuse_operands_p): New function.
15771         (maybe_legitimize_operands): Try to reuse the results for
15772         earlier operands.
15774 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15776         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
15777         Add {q} suffix to insn mnemonic.
15779 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15781         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
15782         (msp430_warn_func_return): New.
15784 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
15786         * fold-const.c (tree_nonzero_bits): New function.
15787         * fold-const.h (tree_nonzero_bits): Likewise.
15788         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
15789         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
15791 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15793         PR target/85900
15794         PR target/85345
15795         * varasm.c (assemble_alias): Check ifunc_resolver only on
15796         FUNCTION_DECL.
15798 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15800         PR target/85903
15801         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
15802         when memory input operand is handled.
15804 2018-05-24  Luis Machado  <luis.machado@linaro.org>
15806         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
15807         global.
15808         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
15810 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15812         * match.pd: Delay FMA folds until after vectorization.
15814 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15816         PR target/83009
15817         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15818         address check not strict.
15820 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15822         * gimple-match.h (gimple_match_op): New class.
15823         (mprts_hook): Replace parameters with a gimple_match_op *.
15824         (maybe_build_generic_op): Likewise.
15825         (gimple_simplified_result_is_gimple_val): Replace parameters with
15826         a const gimple_match_op *.
15827         (gimple_simplify): Replace code_helper * and tree * parameters with
15828         a gimple_match_op * parameter.
15829         (gimple_resimplify1): Replace code_helper *, tree and tree *
15830         parameters with a gimple_match_op * parameter.
15831         (gimple_resimplify2): Likewise.
15832         (gimple_resimplify3): Likewise.
15833         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
15834         parameters with a gimple_match_op * parameter.
15835         * gimple-match-head.c (gimple_simplify): Change prototypes of
15836         auto-generated functions to take a gimple_match_op * instead of
15837         separate code_helper * and tree * parameters.  Make the same
15838         change in the top-level overload and update calls to the
15839         gimple_resimplify routines.  Update calls to the auto-generated
15840         functions and to maybe_push_res_to_seq in the publicly-facing
15841         operation-specific gimple_simplify overloads.
15842         (gimple_match_op::MAX_NUM_OPS): Define.
15843         (gimple_resimplify1): Replace rcode and ops with a single res_op
15844         parameter.  Update call to gimple_simplify.
15845         (gimple_resimplify2): Likewise.
15846         (gimple_resimplify3): Likewise.
15847         (mprts_hook): Replace parameters with a gimple_match_op *.
15848         (maybe_build_generic_op): Likewise.
15849         (build_call_internal): Replace type, nargs and ops with
15850         a gimple_match_op *.
15851         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
15852         with a single gimple_match_op *.  Update calls to mprts_hook,
15853         build_call_internal and gimple_simplified_result_is_gimple_val.
15854         Factor out code that is common to the tree_code and combined_fn cases.
15855         * genmatch.c (expr::gen_transform): Replace tem_code and
15856         tem_ops with a gimple_match_op called tem_op.  Update calls
15857         to the gimple_resimplify functions and maybe_push_res_to_seq.
15858         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
15859         res_ops.  Update call to the gimple_resimplify functions.
15860         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
15861         (decision_tree::gen): Make the functions take a gimple_match_op *
15862         called res_op instead of separate res_code and res_ops parameters.
15863         Update call accordingly.
15864         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
15865         and ops with a single res_op parameter.  Update calls to
15866         maybe_build_generic_op and maybe_push_res_to_seq.
15867         (fold_stmt_1): Update calls to gimple_simplify and
15868         replace_stmt_with_simplification.
15869         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
15870         and gimple_simplified_result_is_gimple_val.
15871         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
15872         gimple_simplify.
15873         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
15874         with a gimple_match_op *.
15875         (vn_nary_build_or_lookup): Likewise.  Update call to
15876         vn_nary_build_or_lookup_1.
15877         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
15878         gimple_match_op *.  Update calls to the gimple_resimplify routines
15879         and to gimple_simplified_result_is_gimple_val.
15880         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
15881         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
15882         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
15883         (visit_nary_op): Likewise.
15884         (visit_reference_op_load): Likewise.
15886 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15888         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
15889         modifier for printing the step amount.
15891 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
15893         PR target/78849
15894         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
15895         types.
15897 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
15899         * doc/sourcebuild.texi (Endianness): New subsubsection.
15901 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15903         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15904         <prefetch_dynamic_strides>: New const bool field.
15905         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15906         prefetch_dynamic_strides.
15907         (exynosm1_prefetch_tune): Likewise.
15908         (thunderxt88_prefetch_tune): Likewise.
15909         (thunderx_prefetch_tune): Likewise.
15910         (thunderx2t99_prefetch_tune): Likewise.
15911         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
15912         false.
15913         (aarch64_override_options_internal): Update to set
15914         PARAM_PREFETCH_DYNAMIC_STRIDES.
15915         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15916         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15917         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15918         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15919         prefetch-dynamic-strides setting.
15921 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15923         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15924         <minimum_stride>: New const int field.
15925         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15926         minimum_stride field defaulting to -1.
15927         (exynosm1_prefetch_tune): Likewise.
15928         (thunderxt88_prefetch_tune): Likewise.
15929         (thunderx_prefetch_tune): Likewise.
15930         (thunderx2t99_prefetch_tune): Likewise.
15931         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
15932         <default_opt_level>: Set to 3.
15933         (aarch64_override_options_internal): Update to set
15934         PARAM_PREFETCH_MINIMUM_STRIDE.
15935         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15936         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15937         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15938         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15939         stride is constant and is below the minimum stride threshold.
15941 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15943         * config/arm/arm-cpus.in (mode26): Delete.
15944         (armv4): Delete mode26 reference.
15945         * config/arm/arm.c (arm_configure_build_target): Delete use of
15946         isa_bit_mode26.
15948 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15950         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
15951         New insn pattern.
15952         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
15953         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
15954         for non-SSE modes.
15955         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
15956         (floatunsdidf2): Ditto.
15958 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15960         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
15961         (fixuns_trunc<mode>si2_avx512f): Ditto.
15962         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
15963         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
15964         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
15966 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15968         PR rtl-optimization/79985
15969         * df-scan.c (df_insn_refs_collect): Remove special case for
15970         global registers and asm statements.
15972 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15974         * extend.texi (Global Register Variables): Rewrite the bullet list.
15975         Note that the register is available for allocation. Note that access
15976         via inline asm must use constraints. Add note about async-signal
15977         handlers. Remove paragraph about automagic register selection.
15979 2018-05-23  Richard Biener  <rguenther@suse.de>
15981         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
15982         of fixed offset from memset VN.
15984 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
15986         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
15987         first_interp field.
15988         (alloc_cand_and_find_basis): Initialize first_interp field.
15989         (slsr_process_mul): Modify first_interp field.
15990         (slsr_process_add): Likewise.
15991         (slsr_process_cast): Modify first_interp field for each new
15992         interpretation.
15993         (slsr_process_copy): Likewise.
15994         (dump_candidate): Dump first_interp field.
15995         (replace_mult_candidate): Process all interpretations, not just
15996         subsequent ones.
15997         (replace_rhs_if_not_dup): Likewise.
15998         (replace_one_candidate): Likewise.
16000 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
16002         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
16003         Add new boolean.
16004         (aarch64_needs_frame_chain): New function.
16005         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
16007 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
16009         PR target/84882
16010         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
16011         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
16012         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
16013         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
16014         as true for strict-align.
16015         (aarch64_can_inline_p): Perform checks even when callee has no
16016         attributes to check for strict alignment.
16017         * doc/extend.texi (AArch64 Function Attributes): Document
16018         no-strict-align.
16019         * doc/invoke.texi: (AArch64 Options): Likewise.
16021 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
16023         PR tree-optimization/85853
16024         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
16025         the handling of the root of the node to...
16026         (vect_slp_analyze_node_operations_1): ...this new function,
16027         and run the whole thing with the child nodes' def types
16028         set according to their SLP node's def type.
16030 2018-05-23  Richard Biener  <rguenther@suse.de>
16032         PR middle-end/85874
16033         * tree-data-ref.c (create_runtime_alias_checks): Defer
16034         and ignore overflow warnings.
16036 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
16038         PR tree-optimization/85822
16039         * tree-vrp.c (is_masked_range_test): Fix handling of negative
16040         constants.
16042 2018-05-23  Richard Biener  <rguenther@suse.de>
16044         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16045         memset constants via native_interpret_expr.
16047 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
16049         PR target/85345
16050         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
16051         attribute.
16052         (cgraph_node::create_alias): Likewise.
16053         (cgraph_node::get_availability): Check ifunc_resolver instead
16054         of looking up ifunc attribute.
16055         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
16056         * varasm.c (do_assemble_alias): Likewise.
16057         (assemble_alias): Likewise.
16058         (default_binds_local_p_3): Likewise.
16059         * cgraph.h (cgraph_node): Add ifunc_resolver.
16060         (cgraph_node::only_called_directly_or_aliased_p): Return false
16061         for IFUNC resolver.
16062         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
16063         attribute.
16064         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
16065         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
16066         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
16067         instead of looking up ifunc attribute.
16069 2018-05-22  Luis Machado  <luis.machado@linaro.org>
16071         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
16073 2018-05-22  Martin Sebor  <msebor@redhat.com>
16075         PR middle-end/85359
16076         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
16077         only when expasion succeeds.
16078         (expand_builtin_strcmp): Same.
16079         (expand_builtin_strncmp): Same.
16081 2018-05-22  Martin Sebor  <msebor@redhat.com>
16083         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
16085 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16086             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16088         * config/aarch64/aarch64-ldpstp.md: Replace uses of
16089         aarch64_mem_pair_operand with memory_operand and delete operand swapping
16090         code.
16091         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16092         Add check for legitimate_address.
16093         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
16094         (aarch64_swap_ldrstr_operands): New.
16095         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
16096         Define prototype.
16098 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
16099             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16101         * config/aarch64/aarch64.md: New patterns to generate stp
16102         and ldp.
16103         (store_pair_sw, store_pair_dw): New patterns to generate stp for
16104         single words and double words.
16105         (load_pair_sw, load_pair_dw): Likewise.
16106         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
16107         Delete.
16108         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
16109         Delete.
16110         * config/aarch64/aarch64-ldpstp.md: Modify peephole
16111         for different mode ldpstp and add peephole for merged zero stores.
16112         Likewise for loads.
16113         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
16114         Add size check.
16115         (aarch64_gen_store_pair): Rename calls to match new patterns.
16116         (aarch64_gen_load_pair): Rename calls to match new patterns.
16117         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
16118         (load_pair<DREG:mode><DREG2:mode>): ... This.
16119         (store_pair<mode>): Rename to...
16120         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
16121         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
16122         New mode iterators.
16123         (V_INT_EQUIV): Handle SImode.
16124         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
16125         New predicate.
16127 2018-05-22  Martin Sebor  <msebor@redhat.com>
16129         PR c/85623
16130         * calls.c (maybe_warn_nonstring_arg): Use string length to set
16131         or ajust the presumed bound on an operation to avoid unnecessary
16132         warnings.
16134 2018-05-22  Martin Sebor  <msebor@redhat.com>
16136         PR tree-optimization/85826
16137         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
16138         assuming that a DECL necesarily has a constant size.
16140 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
16142         PR middle-end/85862
16143         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
16145 2018-05-22  Richard Biener  <rguenther@suse.de>
16147         PR tree-optimization/85834
16148         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
16149         non-constant and non-zero memset arguments.
16151 2018-05-22  Martin Liska  <mliska@suse.cz>
16153         PR ipa/85607
16154         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
16156 2018-05-22  Richard Biener  <rguenther@suse.de>
16158         PR tree-optimization/85863
16159         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16160         comparisons when vectype is specified.
16161         (vectorizable_condition): Do not specify vectype for
16162         vect_is_simple_cond when SLP vectorizing.
16164 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
16166         PR target/85657
16167         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
16168         define __ibm128 as long double.
16169         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
16170         as a distinct type when IEEE 128-bit support is enabled.
16171         (init_float128_ieee): Fix up conversions between IFmode and IEEE
16172         128-bit types to use the correct functions.
16173         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
16174         convert between 128-bit floating point types that have different
16175         modes but the same representation, instead of using gen_lowpart to
16176         makean alias.
16177         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16178         KFmode.
16179         (IFKF_reg): New attributes to give the register constraints for
16180         IFmode and KFmode.
16181         (extend<mode>tf2_internal): New insns to mark an explicit
16182         conversion between 128-bit floating point types that have a
16183         different mode but share the same representation.
16185 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
16187         PR tree-optimization/85814
16188         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
16189         a null return from get_strinfo when unsharing the next
16190         strinfo in the chain.
16192 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
16194         PR gcc/84923
16195         * varasm.c (weak_finish): Clean up weak_decls.
16197 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16199         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
16200         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
16201         UNSPEC_UADALP values.
16202         * config/aarch64/iterators.md (ABAL): New int iterator.
16203         (ABDL2): Likewise.
16204         (ADALP): Likewise.
16205         (sur): Add mappings for the above.
16206         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
16207         New define_insn.
16208         (aarch64_<sur>abal<mode>_4): Likewise.
16209         (aarch64_<sur>adalp<mode>_3): Likewise.
16210         (<sur>sadv16qi): New define_expand.
16212 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
16214         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
16215         (*movdf_internal): Ditto.
16216         (*rcpsf2_sse): Ditto.
16217         (*rsqrtsf2_sse): Ditto.
16218         (*sqrt<mode>2_sse): Ditto.
16220 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
16222         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
16223         eor3q<mode>4.
16224         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
16225         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
16226         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
16227         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
16228         vbcaxq_s64): New.
16229         * config/aarch64/arm_neon.h: Likewise.
16230         * config/aarch64/iterators.md (VQ_I): New.
16232 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
16234         * config.gcc: Add arc/t-multilib-linux to tmake_file for
16235         arc*-*-linux*.
16236         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
16237         MULTILIB_DIRNAMES
16239 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16241         * config/nds32/constraints.md (S): New constraint.
16242         * config/nds32/nds32.md (call_internal): Use constraint S.
16243         (call_value_internal): Likewise.
16244         (sibcall_internal): Likewise.
16245         (sibcall_value_internal): Likewise.
16247 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16248             Chung-Ju Wu  <jasonwucj@gmail.com>
16250         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
16251         into consideration.
16253 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
16254             Chung-Ju Wu  <jasonwucj@gmail.com>
16256         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
16257         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
16258         (nds32_rtx_costs_impl): Simplify.
16259         (nds32_address_cost_impl): Simplify.
16260         (nds32_init_rtx_costs): New function.
16261         (nds32_rtx_costs_speed_prefer): Likewise.
16262         (nds32_rtx_costs_size_prefer): Likewise.
16263         (nds32_address_cost_speed_prefer): Likewise.
16264         (nds32_address_cost_speed_fwprop): Likewise.
16265         (nds32_address_cost_size_prefer): Likewise.
16266         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16267         * config/nds32/nds32.c (nds32_option_override): Use
16268         nds32_init_rtx_costs function.
16270 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16272         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16273         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16274         (TARGET_PIPELINE_N8): Likewise.
16275         (TARGET_PIPELINE_N10): Likewise.
16276         (TARGET_PIPELINE_N13): Likewise.
16277         (TARGET_PIPELINE_GRAYWOLF): Likewise.
16279 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
16281         * config/nds32/nds32-fpu.md: Update copyright year.
16283 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16285         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16287 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16289         * config/nds32/nds32.c
16290         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16291         * config/nds32/nds32.opt (minline-asm-r15): New option.
16293 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16295         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16296         MASK_HW_ABS.
16297         * config/nds32/nds32.md (abssi2): New pattern.
16299 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
16301         * config/i386/i386.md (rex64namesuffix): New mode attribute.
16302         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16303         Merge insn pattern from sse_cvtsi2ss<round_name> and
16304         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16305         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16306         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16307         using SWI48 mode iterator.
16308         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16309         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16310         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16311         pattern from sse_cvttss2si<round_saeonly_name>
16312         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16313         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16314         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16315         using SWI48 mode iterator.
16316         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16317         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16318         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16319         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16320         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16321         using SWI48 mode iterator.
16322         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16323         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16324         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16325         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16326         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16327         SWI48 mode iterator.
16328         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16329         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16330         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16331         pattern from sse_cvttsd2si<round_saeonly_name>
16332         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16334 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16336         * config/nds32/nds32-md-auxiliary.c
16337         (nds32_valid_smw_lwm_base_p): Refine.
16338         (nds32_output_smw_single_word): Refine.
16339         (nds32_output_smw_double_word): New.
16340         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16342 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16344         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16345         (nds32_output_stack_pop): Refine.
16346         (nds32_expand_unaligned_load): Refine.
16347         (nds32_expand_unaligned_store): Refine.
16349 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16350             Chung-Ju Wu  <jasonwucj@gmail.com>
16352         * config/nds32/constants.md: Add TP_REGNUM constant.
16353         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16354         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16355         UNSPEC_ADD32.
16356         * config/nds32/nds32-doubleword.md: Consider flag_pic.
16357         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16358         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16359         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16360         and PIC code generation.
16361         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16362         code generation.
16363         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16364         optimization.
16365         * config/nds32/nds32.md: Support TLS and PIC.
16366         * config/nds32/nds32.c: Support TLS and PIC.
16367         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16368         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16369         predicate.
16371 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16373         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16374         mode with E_ prefix.
16376 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16377             Chung-Ju Wu  <jasonwucj@gmail.com>
16379         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16380         * config/nds32/nds32-md-auxiliary.c
16381         (symbolic_reference_mentioned_p): New.
16382         (nds32_legitimize_ict_address): New.
16383         (nds32_expand_ict_move): New.
16384         (nds32_indirect_call_referenced_p): New.
16385         (nds32_symbol_binds_local_p): Delete.
16386         (nds32_long_call_p): Modify.
16387         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16388         * config/nds32/nds32-protos.h
16389         (symbolic_reference_mentioned_p): Declare.
16390         (nds32_legitimize_ict_address): Declare.
16391         (nds32_expand_ict_move): Declare.
16392         (nds32_indirect_call_referenced_p): Declare.
16393         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16394         (nds32_relax_group): Use nds32_ict_const_p as condition.
16395         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16396         (nds32_asm_file_start): Output ict_model directive in asm code.
16397         (nds32_legitimate_address_p): Consider indirect call.
16398         (nds32_print_operand): Consider indirect call.
16399         (nds32_print_operand_address): Consider indirect call.
16400         (nds32_insert_attributes): Handle "indirect_call" attribute.
16401         (TARGET_LEGITIMATE_ADDRESS_P): Define.
16402         (TARGET_LEGITIMATE_CONSTANT_P): Define.
16403         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16404         (TARGET_DELEGITIMIZE_ADDRESS): Define.
16405         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16406         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16407         (TARGET_ICT_MODEL_SMALL): Define.
16408         (TARGET_ICT_MODEL_LARGE): Define.
16409         * config/nds32/nds32.md (movsi): Consider ict model.
16410         (call, call_value): Consider ict model.
16411         (sibcall, sibcall_value): Consider ict model.
16412         * config/nds32/nds32.opt (mict-model): New option.
16413         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16414         model.
16416 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
16417             Monk Chiang  <sh.chiang04@gmail.com>
16418             Jim Wilson <jimw@sifive.com>
16420         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16421         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
16422         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16423         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16424         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
16425         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16426         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16427         compute save_libcall_adjustment properly.
16428         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16429         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16430         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16431         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16432         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16433         (ABI_SPEC): Handle mabi=ilp32e.
16434         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16435         (RVE): Add RVE mask.
16436         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16437         <-march>: Add rv32e as an example.
16439 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
16441         PR c++/82899
16442         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
16443         (intra_create_variable_infos): Handle C++ constructors.
16445 2018-05-18  Martin Liska  <mliska@suse.cz>
16447         * passes.def: Remove a redundant pass.
16449 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
16451         PR bootstrap/85838
16452         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
16454 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16456         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
16457         (ARMv4): Update.
16458         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16459         (ARMv6m): Update.
16460         (armv2, armv2a, armv3, armv3m): Delete architectures.
16461         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
16462         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
16463         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
16464         Delete cpus.
16465         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
16466         (*mulsidi3adddi): Likewise.
16467         (mulsidi3): Likewise.
16468         (*mulsidi3_nov6): Likewise.
16469         (umulsidi3): Likewise.
16470         (umulsidi3_nov6): Likewise.
16471         (umaddsidi4): Likewise.
16472         (*umulsidi3adddi): Likewise.
16473         (smulsi3_highpart): Likewise.
16474         (*smulsi3_highpart_nov6): Likewise.
16475         (umulsi3_highpart): Likewise.
16476         (*umulsi3_highpart_nov6): Likewise.
16477         * config/arm/arm.h (arm_arch3m): Delete.
16478         * config/arm/arm.c (arm_arch3m): Delete.
16479         (arm_option_override_internal): Update armv3-related comment.
16480         (arm_configure_build_target): Delete use of isa_bit_mode32.
16481         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
16482         (arm_rtx_costs_internal): Delete check of arm_arch3m.
16483         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
16484         (mulsa3): Likewise.
16485         (mulusa3): Likewise.
16486         * config/arm/arm-protos.h (arm_arch3m): Delete.
16487         * config/arm/arm-tables.opt: Regenerate.
16488         * config/arm/arm-tune.md: Likewise.
16489         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
16490         deleted architectures.
16492 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16494         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
16495         (armv5t, armv5te): New features.
16496         (ARMv5, ARMv5e): Delete fgroups.
16497         (ARMv5t, ARMv5te): Adjust for above changes.
16498         (ARMv6m): Likewise.
16499         (armv5, armv5e): Delete arches.
16500         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
16501         arm_arch5.
16502         (*call_reg_arm): Likewise.
16503         (*call_value_reg_armv5): Likewise.
16504         (*call_value_reg_arm): Likewise.
16505         (*call_symbol): Likewise.
16506         (*call_value_symbol): Likewise.
16507         (*sibcall_insn): Likewise.
16508         (*sibcall_value_insn): Likewise.
16509         (clzsi2): Likewise.
16510         (prefetch): Likewise.
16511         (define_split and define_peephole2 dependent on arm_arch5):
16512         Likewise.
16513         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
16514         arm_arch5e.
16515         (TARGET_ARM_QBIT): Likewise.
16516         (TARGET_DSP_MULTIPLY): Likewise.
16517         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
16518         (arm_arch5, arm_arch5e): Delete.
16519         (arm_arch5t, arm_arch5te): Declare.
16520         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
16521         (arm_arch5t): Declare.
16522         (arm_option_reconfigure_globals): Update for the above.
16523         (arm_options_perform_arch_sanity_checks): Update comment, replace
16524         use of arm_arch5 with arm_arch5t.
16525         (use_return_insn): Likewise.
16526         (arm_emit_call_insn): Likewise.
16527         (output_return_instruction): Likewise.
16528         (arm_final_prescan_insn): Likewise.
16529         (arm_coproc_builtin_available): Likewise.
16530         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
16531         arm_arch5e with arm_arch5t and arm_arch5te.
16532         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
16533         (arm_arch5t, arm_arch5te): Declare.
16534         * config/arm/arm-tables.opt: Regenerate.
16535         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
16536         * config/arm/t-multilib: Likewise.
16537         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
16538         instead of arm_arch5.
16539         (*call_reg_thumb1): Likewise.
16540         (*call_value_reg_thumb1_v5): Likewise.
16541         (*call_value_reg_thumb1): Likewise.
16542         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
16543         unreachable path.
16544         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
16546 2018-05-18  Martin Liska  <mliska@suse.cz>
16548         PR gcov-profile/84846
16549         * doc/gcov.texi: Document -t option of gcov tool.
16551 2018-05-18  Martin Liska  <mliska@suse.cz>
16553         PR gcov-profile/84846
16554         * gcov.c (print_usage): Add new -t option.
16555         (process_args): Handle the option.
16556         (generate_results): Use stdout as output when requested by
16557         the option.
16559 2018-05-18  Martin Liska  <mliska@suse.cz>
16561         PR gcov-profile/84846
16562         * coverage.c (coverage_init): Write PWD to .gcno file.
16563         * doc/gcov.texi: Document how working directory is printed.
16564         * gcov-dump.c (dump_gcov_file): Print PWD.
16565         * gcov.c (output_intermediate_file): Likewise.
16566         (read_graph_file): Read PWD string.
16567         (output_lines): Print PWD.
16569 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16571         PR middle-end/85817
16572         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
16573         for retval and return false if all args to phi are zero.
16575 2018-05-18  Richard Biener  <rguenther@suse.de>
16577         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
16578         method.
16579         (evrp_dom_walker::before_dom_children): Call it.
16581 2018-05-18  Richard Biener  <rguenther@suse.de>
16583         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
16584         results when processing array refs with variable index.
16586 2018-05-18  Toon Moene  <toon@moene.org>
16588         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
16589         directly after that of -floop-interchange. Indicate that both
16590         options are enabled by default when specifying -O3.
16592 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16594         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
16595         iterator.  Delete separate integer-mode vec_set<mode> expander.
16596         (aarch64_simd_vec_setv2di): Delete.
16597         (vec_setv2di): Delete.
16598         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
16599         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
16600         the "w, r" alternative.
16602 2018-05-18  Martin Liska  <mliska@suse.cz>
16604         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
16605         * tree-pass.h (make_pass_lower_switch_O0): New function.
16606         * tree-switch-conversion.c (node_has_low_bound): Remove.
16607         (node_has_high_bound): Likewise.
16608         (node_is_bounded): Likewise.
16609         (class pass_lower_switch): Make it a template type and create
16610         two instances.
16611         (pass_lower_switch::execute): Add template argument.
16612         (make_pass_lower_switch): New function.
16613         (make_pass_lower_switch_O0): New function.
16614         (do_jump_if_equal): Remove.
16615         (emit_case_nodes): Simplify to just handle all 3 cases and leave
16616         all the hard work to tree optimization passes.
16618 2018-05-18  Martin Liska  <mliska@suse.cz>
16620         * dbgcnt.c (limit_low): Renamed from limit.
16621         (limit_high): New variable.
16622         (dbg_cnt_is_enabled): Check for upper limit.
16623         (dbg_cnt): Adjust dumping.
16624         (dbg_cnt_set_limit_by_index): Add new argument for high
16625         value.
16626         (dbg_cnt_set_limit_by_name): Likewise.
16627         (dbg_cnt_process_single_pair): Parse new format.
16628         (dbg_cnt_process_opt): Use strtok.
16629         (dbg_cnt_list_all_counters): Remove 'value' and add
16630         'limit_high'.
16631         * doc/invoke.texi: Document changes.
16633 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
16635         * doc/sourcebuild.texi (scalar_all_fma): Document.
16636         * tree.def (FMA_EXPR): Delete.
16637         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
16638         * internal-fn.c (ternary_direct): New macro.
16639         (expand_ternary_optab_fn): Likewise.
16640         (direct_ternary_optab_supported_p): Likewise.
16641         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
16642         * builtins.c (fold_builtin_fma): Delete.
16643         (fold_builtin_3): Don't call it.
16644         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
16645         * expr.c (expand_expr_real_2): Likewise.
16646         * fold-const.c (operand_equal_p): Likewise.
16647         (fold_ternary_loc): Likewise.
16648         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
16649         * gimple.c (DEFTREECODE): Likewise.
16650         * gimplify.c (gimplify_expr): Likewise.
16651         * optabs-tree.c (optab_for_tree_code): Likewise.
16652         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
16653         * tree-eh.c (operation_could_trap_p): Likewise.
16654         (stmt_could_throw_1_p): Likewise.
16655         * tree-inline.c (estimate_operator_cost): Likewise.
16656         * tree-pretty-print.c (dump_generic_node): Likewise.
16657         (op_code_prio): Likewise.
16658         * tree-ssa-loop-im.c (stmt_cost): Likewise.
16659         * tree-ssa-operands.c (get_expr_operands): Likewise.
16660         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
16661         * fold-const-call.h (fold_fma): Delete.
16662         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
16663         CFN_FNMA and CFN_FNMS.
16664         (fold_fma): Delete.
16665         * genmatch.c (combined_fn): New enum.
16666         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
16667         (commutative_op): New function.
16668         (commutate): Use it.  Handle more than 2 operands.
16669         (dt_operand::gen_gimple_expr): Use commutative_op.
16670         (parser::parse_expr): Allow :c to be used with non-binary
16671         operators if the commutative operand is known.
16672         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
16673         CFN_FMS, CFN_FNMA and CFN_FNMS.
16674         (backprop::process_assign_use): Remove FMA_EXPR handling.
16675         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
16676         (gen_hsa_fma): New function.
16677         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
16678         IFN_FNMA and IFN_FNMS.
16679         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
16680         * gimple-fold.h (follow_all_ssa_edges): Declare.
16681         * gimple-fold.c (follow_all_ssa_edges): New function.
16682         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
16683         gimple_build interface and use follow_all_ssa_edges to fold the result.
16684         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
16685         instead of checking for optabs directly.
16686         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
16687         rather than FMA_EXPRs.
16688         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
16689         call to IFN_FMA instead of an FMA_EXPR.
16691 2018-05-17  Jim Wilson  <jimw@sifive.com>
16693         * expr.c (do_tablejump): When converting index to Pmode, if we have a
16694         sign extended promoted subreg, and the range does not have the sign bit
16695         set, then do a sign extend.
16697         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
16698         test, check for sign extended subreg and/or constant operands, and
16699         do a sign extend in that case.
16701 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
16703         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
16704         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
16705         Add untyped.
16706         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
16707         Change logics_shift_reg to logics_shift_imm.
16708         (thunderx2t99_fp_loadpair_basic): Delete.
16709         (thunderx2t99_fp_storepair_basic): Delete.
16710         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
16711         (thunderx2t99_asimd_polynomial): Delete.
16712         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
16713         and neon_fp_mul_d_scalar_q.
16714         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
16715         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
16716         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
16717         (thunderx2t99_asimd_lut): Add missing tbl types.
16718         (thunderx2t99_asimd_ext): Delete.
16719         (thunderx2t99_asimd_load1_1_mult): Delete.
16720         (thunderx2t99_asimd_load1_2_mult): Delete.
16721         (thunderx2t99_asimd_load1_ldp): New.
16722         (thunderx2t99_asimd_load1): New.
16723         (thunderx2t99_asimd_load2): Add missing *load2* types.
16724         (thunderx2t99_asimd_load3): New.
16725         (thunderx2t99_asimd_load4): New.
16726         (thunderx2t99_asimd_store1_1_mult): Delete.
16727         (thunderx2t99_asimd_store1_2_mult): Delete.
16728         (thunderx2t99_asimd_store2_mult): Delete.
16729         (thunderx2t99_asimd_store2_onelane): Delete.
16730         (thunderx2t99_asimd_store_stp): New.
16731         (thunderx2t99_asimd_store1): New.
16732         (thunderx2t99_asimd_store2): New.
16733         (thunderx2t99_asimd_store3): New.
16734         (thunderx2t99_asimd_store4): New.
16736 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
16738         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
16739         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
16741 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
16742             Segher Boessenkool  <segher@kernel.crashing.org>
16744         PR target/85698
16745         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
16746         operand.
16748 2018-05-17  Richard Biener  <rguenther@suse.de>
16750         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
16751         for pruning loop and prune defs feeding only already visited PHIs.
16753 2018-05-17  Richard Biener  <rguenther@suse.de>
16755         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
16757 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
16758             Richard Biener  <rguenther@suse.de>
16760         PR tree-optimization/85793
16761         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
16762         for VMAT_ELEMENTWISE.
16764 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16766         * internal-fn.h (lookup_internal_fn): Declare
16767         * internal-fn.c (lookup_internal_fn): New function.
16768         * gimple.c (gimple_build_call_from_tree): Handle calls to
16769         internal functions.
16770         * gimple-pretty-print.c (dump_gimple_call): Print "." before
16771         internal function names.
16772         * tree-pretty-print.c (dump_generic_node): Likewise.
16773         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
16775 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16777         * gimple-fold.h (gimple_build): Make the function forms take
16778         combined_fn rather than built_in_function.
16779         (gimple_simplify): Likewise.
16780         * gimple-match-head.c (gimple_simplify): Likewise.
16781         * gimple-fold.c (gimple_build): Likewise.
16782         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
16783         rather than gimple_build_call_internal.
16784         (get_initial_defs_for_reduction): Likewise.
16785         (vect_create_epilog_for_reduction): Likewise.
16786         (vectorizable_live_operation): Likewise.
16788 2018-05-17  Martin Liska  <mliska@suse.cz>
16790         * gimple-ssa-sprintf.c (format_directive): Do not use
16791         space in between 'G_' and '('.
16793 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
16795         PR target/85323
16796         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
16797         even if the mask is not all ones.
16799         PR target/85323
16800         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
16801         vector.
16802         (ix86_gimple_fold_builtin): Likewise.
16804         PR target/85323
16805         * config/i386/i386.c: Include tree-vector-builder.h.
16806         (ix86_vector_shift_count): New function.
16807         (ix86_fold_builtin): Fold shift builtins by scalar count.
16808         (ix86_gimple_fold_builtin): Likewise.
16810         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
16811         _mm512_setzero): New intrinsics.
16813 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
16814             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16816         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
16817         code generation for cases where splatting a value is not useful.
16818         * simplify-rtx.c (simplify_ternary_operation): Simplify
16819         vec_merge across a vec_duplicate and a paradoxical subreg forming
16820         a vector mode to a vec_concat.
16822 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
16824         * config.gcc: Support "goldmont-plus".
16825         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16826         "goldmont-plus".
16827         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16828         PROCESSOR_GOLDMONT_PLUS.
16829         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
16830         (processor_target_table): Add "goldmont-plus".
16831         (PTA_GOLDMONT_PLUS): Define.
16832         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
16833         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
16834         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
16835         (fold_builtin_cpu): Add "goldmont-plus".
16836         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
16837         (ix86_option_override_internal): Add "goldmont-plus".
16838         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
16839         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
16840         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
16841         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
16843 2018-05-17  Richard Biener  <rguenther@suse.de>
16845         PR tree-optimization/85757
16846         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
16847         remove defs that only feed that PHI from further processing.
16849 2018-05-16  Jim Wilson  <jimw@sifive.com>
16851         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
16852         asterisk to name.
16853         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
16855 2018-05-16  Mark Wielaard  <mark@klomp.org>
16857         * dwarf2out.c (count_index_strings): New function.
16858         (output_indirect_strings): Call count_index_strings and generate
16859         header for dwarf_version >= 5.
16861 2018-05-16  Mark Wielaard  <mark@klomp.org>
16863         * dwarf2out.c (dwarf_FORM): New function.
16864         (set_indirect_string): Use dwarf_FORM.
16865         (reset_indirect_string): Likewise.
16866         (size_of_die): Likewise.
16867         (value_format): Likewise.
16868         (output_die): Likewise.
16869         (add_skeleton_AT_string): Likewise.
16870         (output_macinfo_op): Likewise.
16871         (index_string): Likewise.
16872         (output_index_string_offset): Likewise.
16873         (output_index_string): Likewise.
16874         (count_index_strings): Likewise.
16876 2018-05-16  Carl Love  <cel@us.ibm.com>
16878         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16879         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
16881 2018-05-16  Martin Jambor  <mjambor@suse.cz>
16883         * ipa-prop.c (ipa_free_all_edge_args): Remove.
16884         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
16886 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
16888         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
16889         (fnma<mode>4): Likewise.
16890         (fms<mode>4): Likewise.
16891         (fnms<mode>4): Likewise.
16892         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
16893         (aarch64_fnma<mode>4): Likewise.
16894         (aarch64_fms<mode>4): Likewise.
16895         (aarch64_fnms<mode>4): Likewise.
16896         (aarch64_fnmadd<mode>4): Likewise.
16898 2018-05-16  Jason Merrill  <jason@redhat.com>
16900         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
16902 2018-05-16  Richard Biener  <rguenther@suse.de>
16904         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
16905         (dump_stmt_cost): Declare.
16906         (add_stmt_cost): Dump cost we add.
16907         (add_stmt_costs): New function.
16908         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
16909         No longer exported.
16910         (vect_analyze_stmt): Adjust prototype.
16911         (vectorizable_condition): Likewise.
16912         (vectorizable_live_operation): Likewise.
16913         (vectorizable_reduction): Likewise.
16914         (vectorizable_induction): Likewise.
16915         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
16916         cost vector to pass to vectorizable_ and record afterwards.
16917         (vect_model_reduction_cost): Take cost vector argument and adjust.
16918         (vect_model_induction_cost): Likewise.
16919         (vectorizable_reduction): Likewise.
16920         (vectorizable_induction): Likewise.
16921         (vectorizable_live_operation): Likewise.
16922         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
16923         SLP_TREE_NUMBER_OF_VEC_STMTS.
16924         (vect_analyze_slp_cost_1): Remove.
16925         (vect_analyze_slp_cost): Likewise.
16926         (vect_slp_analyze_node_operations): Take visited args and
16927         a target cost vector.  Avoid processing already visited stmt sets.
16928         (vect_slp_analyze_operations): Use a local cost vector to gather
16929         costs and register those of non-discarded instances.
16930         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
16931         (vect_schedule_slp_instance): Remove copying of
16932         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
16933         zero.
16934         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
16935         adding cost.  Record cost entry location.
16936         (vect_prologue_cost_for_slp_op): Function to compute cost of
16937         a constant or invariant generated for SLP vect in the prologue,
16938         split out from vect_analyze_slp_cost_1.
16939         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
16940         (vect_model_promotion_demotion_cost): Likewise.
16941         (vect_model_store_cost): Likewise, make static.
16942         (vect_model_load_cost): Likewise.
16943         (vectorizable_bswap): Add cost vector arg and adjust.
16944         (vectorizable_call): Likewise.
16945         (vectorizable_simd_clone_call): Likewise.
16946         (vectorizable_conversion): Likewise.
16947         (vectorizable_assignment): Likewise.
16948         (vectorizable_shift): Likewise.
16949         (vectorizable_operation): Likewise.
16950         (vectorizable_store): Likewise.
16951         (vectorizable_load): Likewise.
16952         (vectorizable_condition): Likewise.
16953         (vectorizable_comparison): Likewise.
16954         (can_vectorize_live_stmts): Likewise.
16955         (vect_analyze_stmt): Likewise.
16956         (vect_transform_stmt): Adjust calls to vectorizable_*.
16957         * tree-vectorizer.c: Include gimple-pretty-print.h.
16958         (dump_stmt_cost): New function.
16960 2018-05-16  Richard Biener  <rguenther@suse.de>
16962         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
16963         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
16964         * tree-ssa-dse.c: Include tree-ssa-loop.h.
16965         (check_name): New callback.
16966         (dse_classify_store): Track cycles via a visited bitmap of PHI
16967         defs and simplify handling of in-loop and across loop dead stores
16968         and properly fail for loop-variant refs.  Handle byte-tracking with
16969         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
16970         limiting the walk.
16972 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
16974         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16975         (vect_get_mask_type_for_stmt): Likewise.
16976         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
16977         split out from...
16978         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
16979         to determine the statement's vector type and the vector type that
16980         should be used for calculating nunits.  Deal with cases in which
16981         the type has to be deferred.
16982         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
16983         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
16984         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
16985         (vect_determine_vf_for_stmt): New functions, split out from...
16986         (vect_determine_vectorization_factor): ...here.
16987         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
16988         (vect_get_mask_type_for_stmt): New functions, split out from
16989         vect_determine_vectorization_factor.
16991 2018-05-16  Richard Biener  <rguenther@suse.de>
16993         * tree-cfg.c (verify_gimple_assign_ternary): Properly
16994         verify the [VEC_]COND_EXPR embedded comparison.
16996 2018-05-15  Martin Sebor  <msebor@redhat.com>
16998         PR tree-optimization/85753
16999         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
17000         RECORD_TYPE in addition to ARRAY_TYPE.
17002 2018-05-15  Martin Sebor  <msebor@redhat.com>
17004         PR middle-end/85643
17005         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
17007 2018-05-15  Richard Biener  <rguenther@suse.de>
17009         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
17010         add by_clobber_p one.  Change algorithm to collect all defs
17011         representing uses we need to walk and try reducing them to
17012         a single one before failing.
17013         (dse_dom_walker::dse_optimize_stmt): Adjust.
17015 2018-05-13  Mark Wielaard  <mark@klomp.org>
17017         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
17018         (size_of_loc_descr): Likewise.
17019         (output_loc_operands): Likewise.
17020         (output_loc_operands_raw): Likewise.
17021         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
17022         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
17023         (hash_loc_operands): Likewise.
17024         (compare_loc_operands): Likewise.
17026 2018-05-14  Mark Wielaard  <mark@klomp.org>
17028         * dwarf2out.c (count_index_addrs): New function.
17029         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
17031 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17033         PR tree-optimization/83648
17034         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
17035         return value as malloc candidate.
17037 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17039         PR ipa/85734
17040         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
17041         param as true in call to suggest_attribute.
17043 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
17045         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
17046         -mreadonly-in-sdata.
17048 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17050         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
17051         New pattern.
17052         (aarch64_crypto_aesd_fused): Likewise.
17054 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
17056         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
17057         (movsi_aarch64): Likewise.
17058         (load_pairsi): Likewise.
17059         (load_pairdi): Likewise.
17060         (store_pairsi): Likewise.
17061         (store_pairdi): Likewise.
17062         (load_pairsf): Likewise.
17063         (load_pairdf): Likewise.
17064         (store_pairsf): Likewise.
17065         (store_pairdf): Likewise.
17066         (zero_extend): Likewise.
17067         (trunc): Swap alternatives.
17068         (fcvt_target): Add '?' to prefer w over r.
17070 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
17072         PR target/85756
17073         * config/i386/i386.md: Disallow non-commutative arithmetics in
17074         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
17075         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
17076         in the peephole2 before it.
17078 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
17080         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
17081         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
17082         (ix86_handle_option): Handle -mcldemote.
17083         * config.gcc: New header.
17084         * config/i386/cldemoteintrin.h: New file.
17085         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
17086         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
17087         -mcldemote.
17088         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17089         OPTION_MASK_ISA_CLDEMOTE.
17090         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
17091         (ix86_valid_target_attribute_inner_p): Ditto.
17092         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
17093         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
17094         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
17095         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
17096         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
17097         (cldemote): New.
17098         * config/i386/i386.opt: Add -mcldemote.
17099         * config/i386/x86intrin.h: New header.
17100         * doc/invoke.texi: Add -mcldemote.
17102 2018-05-14  Richard Biener  <rguenther@suse.de>
17104         * doc/match-and-simplify.texi: Adjust :s documentation.
17106 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
17108         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
17109         intended memcpy size.
17110         (REORDER_45): Likewise.
17112 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
17114         * sort.cc: New file.
17115         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
17116         * vec.c (qsort_chk): Use gcc_qsort.
17117         * Makefile.in (OBJS-libcommon): Add sort.o.
17118         (build/sort.o): New target.  Use it...
17119         (BUILD_RTL): ... here, and...
17120         (build/gencfn-macros): ... here, and...
17121         (build/genmatch): ... here.
17123 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17124             Chung-Ju Wu  <jasonwucj@gmail.com>
17126         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
17127         * config/nds32/nds32-graywolf.md: New file.
17128         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
17129         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
17130         pipeline.
17131         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
17132         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
17133         * config/nds32/nds32.md (pipeline_model): Add graywolf.
17134         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
17135         * config/nds32/pipelines.md: Include n15 settings.
17137 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17138             Chung-Ju Wu  <jasonwucj@gmail.com>
17140         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
17141         * config/nds32/nds32-n13.md: New file.
17142         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
17143         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
17144         pipeline.
17145         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
17146         * config/nds32/nds32.md (pipeline_model): Add n13.
17147         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
17148         * config/nds32/pipelines.md: Include n13 settings.
17150 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
17151             Chung-Ju Wu  <jasonwucj@gmail.com>
17153         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
17154         * config/nds32/nds32-n10.md: New file.
17155         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
17156         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
17157         pipeline.
17158         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
17159         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
17160         * config/nds32/nds32.md (pipeline_model): Add n10.
17161         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
17162         * config/nds32/pipelines.md: Include n10 settings.
17164 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
17165             Kito Cheng  <kito.cheng@gmail.com>
17166             Chung-Ju Wu  <jasonwucj@gmail.com>
17168         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
17169         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17170         Add enum values for DSP extension instructions.
17171         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
17172         New constraints.
17173         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
17174         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
17175         New code iterators.
17176         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
17177         * config/nds32/nds32-dspext.md: New file for DSP implementation.
17178         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
17179         * config/nds32/nds32-intrinsic.md: Likewise.
17180         * config/nds32/nds32_intrinsic.h: Likewise.
17181         * config/nds32/nds32-md-auxiliary.c: Likewise.
17182         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
17183         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
17184         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
17185         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
17186         * config/nds32/nds32-protos.h: New declarations for DSP extension.
17187         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
17188         TYPE_DMAC in switch statement.
17189         * config/nds32/nds32.c: New checking and implementation for DSP
17190         extension instructions.
17191         * config/nds32/nds32.h: Likewise.
17192         * config/nds32/nds32.md: Likewise.
17193         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
17194         * config/nds32/predicates.md: Implement new predicates for DSP
17195         extension.
17197 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
17199         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
17200         Reformat alternatives and attributes so it is easier to identify
17201         which constraints/attributes go with which instruction.
17202         (mov<mode>_hardfloat32, FMOVE64): Likewise.
17203         (mov<mode>_softfloat32, FMOVE64): Likewise.
17204         (mov<mode>_hardfloat64, FMOVE64): Likewise.
17205         (mov<mode>_softfloat64, FMOVE64): Likewise.
17207 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17209         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17210         subsection.
17211         (Basic PowerPC Built-in Functions): The new name of the
17212         subsection previously known as "PowerPC Built-in Functions".
17213         (Basic PowerPC Built-in Functions Available on all Configurations):
17214         New subsubsection.
17215         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
17216         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
17217         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
17218         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
17220 2018-05-11  Martin Jambor  <mjambor@suse.cz>
17222         PR ipa/85655
17223         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
17224         single const.
17226 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17228         PR target/85733
17229         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
17231 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
17233         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
17234         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
17235         (ix86_handle_option): Handle -mwaitpkg.
17236         * config.gcc: New header.
17237         * config/i386/cpuid.h (bit_WAITPKG): New bit.
17238         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
17239         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
17240         function type.
17241         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17242         OPTION_MASK_ISA_WAITPKG.
17243         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
17244         (ix86_option_override_internal): Add PTA_WAITPKG.
17245         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
17246         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
17247         IX86_BUILTIN_TPAUSE.
17248         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
17249         __builtin_ia32_umwait and __builtin_ia32_tpause.
17250         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
17251         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
17252         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
17253         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
17254         UNSPECV_TPAUSE): New.
17255         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
17256         * config/i386/i386.opt: Add -mwaitpkg.
17257         * config/i386/waitpkgintrin.h: New file.
17258         * config/i386/x86intrin.h: New header.
17259         * doc/invoke.texi: Add -mwaitpkg.
17261 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17263         PR target/85606
17264         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17265         equivalent.
17266         (cortex-m0): Use armv6s-m isa.
17267         (cortex-m0plus): Likewise.
17268         (cortex-m1): Likewise.
17269         (cortex-m0.small-multiply): Likewise.
17270         (cortex-m0plus.small-multiply): Likewise.
17271         (cortex-m1.small-multiply): Likewise.
17273 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
17274             Jakub Jelinek  <jakub@redhat.com>
17276         PR tree-optimization/85692
17277         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17278         source permute as well.
17280 2018-05-11  Martin Liska  <mliska@suse.cz>
17282         PR sanitizer/85556
17283         * doc/extend.texi: Document LLVM style format for no_sanitize
17284         attribute.
17286 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17288         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17289         mode_supports_vsx_dform_quad to mode_supports_dq_form.
17290         (mode_supports_vsx_dform_quad): Likewise.
17291         (mode_supports_vmx_dform): Move these functions to be next to the
17292         other mode_supports functions.
17293         (mode_supports_dq_form): Likewise.
17294         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17295         mode_supports_dq_form.
17296         (reg_offset_addressing_ok_p): Likewise.
17297         (offsettable_ok_by_alignment): Likewise.
17298         (rs6000_legitimate_offset_address_p): Likewise.
17299         (legitimate_lo_sum_address_p): Likewise.
17300         (rs6000_legitimize_address): Likewise.
17301         (rs6000_legitimize_reload_address): Likewise.
17302         (rs6000_secondary_reload_inner): Likewise.
17303         (rs6000_preferred_reload_class): Likewise.
17304         (rs6000_output_move_128bit): Likewise.
17306 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17308         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17309         Generate SImode target register for null target.
17310         <case IX86_BUILTIN_XGETBV>: Ditto.
17311         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17312         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17314 2018-05-10  Carl Love  <cel@us.ibm.com>
17316         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17317         dcbtt and dcbtstt if operands[2] is 0.
17319 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17321         PR target/85693
17322         * config/i386/sse.md (usadv64qi): New expander.
17324 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
17326         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17327         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17328         -maltivec=be support.
17329         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17330         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17331         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17332         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17333         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17334         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17335         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17336         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17337         altivec_vsumsws): Adjust.
17338         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17339         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17340         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17341         support.
17342         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17343         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17344         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17345         (altivec_lve<VI_char>x): Delete expand.
17346         (*altivec_lve<VI_char>x_internal): Rename to...
17347         (altivec_lve<VI_char>x): ... this.
17348         (altivec_lvxl_<mode>): Delete expand.
17349         (*altivec_lvxl_<mode>_internal): Rename to ...
17350         (altivec_lvxl_<mode>): ... this.
17351         (altivec_stvxl_<mode>): Delete expand.
17352         (*altivec_stvxl_<mode>_internal): Rename to ...
17353         (altivec_stvxl_<mode>): ... this.
17354         (altivec_stve<VI_char>x): Delete expand.
17355         (*altivec_stve<VI_char>x_internal): Rename to ...
17356         (altivec_stve<VI_char>x): ... this.
17357         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17358         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17359         reduc_plus_scal_<mode>): Adjust.
17360         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17361         comment.
17362         (rs6000_cpu_cpp_builtins): Adjust.
17363         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17364         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17365         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17366         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17367         -maltivec=be support.
17368         (rs6000_split_vec_extract_var): Adjust.
17369         (rs6000_split_v4si_init): Adjust.
17370         (swap_selector_for_mode): Delete.
17371         (altivec_expand_lvx_be, altivec_expand_stvx_be,
17372         altivec_expand_stvex_be): Delete.
17373         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17374         -maltivec=be support.
17375         (rs6000_gimple_fold_builtin): Ditto.
17376         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17377         Adjust.
17378         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17379         (TARGET_DIRECT_MOVE_64BIT): Adjust.
17380         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17381         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17382         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17383         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17384         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17385         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17386         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17387         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17388         anonymous split): Adjust.
17389         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17390         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17392 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
17394         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17395         when --with-gxx-include-dir is also specified.
17396         * configure: Regenerate.
17398 2018-05-09  Jim Wilson  <jimw@sifive.com>
17400         PR target/84797
17401         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17402         * config/riscv/t-withmultilib: New.
17403         * config/riscv/withmultilib.h: New.
17404         * doc/install.texi: Document RISC-V --with-multilib-list support.
17406 2018-05-09  Richard Biener  <rguenther@suse.de>
17408         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17409         vector.
17410         (vect_bb_vectorization_profitable_p): Adjust.  Compute
17411         actual scalar cost using the cost vector and the add_stmt_cost
17412         machinery.
17414 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17416         PR rtl-optimization/85645
17417         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17418         in the REG_CFA_REGISTER note for LR, don't leave it empty.
17420 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17422         PR rtl-optimization/85645
17423         * shrink-wrap.c (spread_components): Return a boolean saying if
17424         anything was changed.
17425         (try_shrink_wrapping_separate): Iterate spread_components until
17426         nothing changes anymore.
17428 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17430         PR rtl-optimization/85645
17431         * regrename.c (build_def_use): Also kill the chains that include the
17432         destination of a REG_CFA_REGISTER note.
17434 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17436         PR rtl-optimization/85645
17437         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17438         insn that has a REG_CFA_REGISTER note.
17440 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
17442         * cfgexpand.c (expand_clobber): New function.
17443         (expand_gimple_stmt_1): Use it.
17444         * tree-vect-stmts.c (vect_clobber_variable): New function,
17445         split out from...
17446         (vectorizable_simd_clone_call): ...here.
17447         (vectorizable_store): Emit a clobber either side of an
17448         IFN_STORE_LANES sequence.
17449         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
17451 2018-05-09  Tom de Vries  <tom@codesourcery.com>
17453         PR target/85626
17454         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
17455         (define_insn "trap_if_false"): Add exit after trap.
17457 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
17459         PR rtl-optimization/85638
17460         * bb-reorder.c: Include common/common-target.h.
17461         (create_forwarder_block): New function extracted from...
17462         (fix_up_crossing_landing_pad): ...here.  Rename into...
17463         (dw2_fix_up_crossing_landing_pad): ...this.
17464         (sjlj_fix_up_crossing_landing_pad): New function.
17465         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
17466         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
17467         from both partitions and exit the loop after one iteration.
17469 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17471         Revert:
17472         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17473         subsection.
17474         (Basic PowerPC Built-in Functions): The new name of the
17475         subsection previously known as "PowerPC Built-in Functions".
17476         (Basic PowerPC Built-in Functions Available on all Configurations):
17477         New subsubsection.
17478         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17479         subsubsection.
17480         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17481         subsubsection.
17482         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17483         subsubsection.
17484         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17485         subsubsection.
17487 2018-05-08  Jim Wilson  <jimw@sifive.com>
17489         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
17490         (LD_EMUL_SUFFIX): New.
17491         (LINK_SPEC): Use it.
17493 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17495         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17496         subsection.
17497         (Basic PowerPC Built-in Functions): The new name of the
17498         subsection previously known as "PowerPC Built-in Functions".
17499         (Basic PowerPC Built-in Functions Available on all Configurations):
17500         New subsubsection.
17501         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17502         subsubsection.
17503         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17504         subsubsection.
17505         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17506         subsubsection.
17507         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17508         subsubsection.
17510 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17512         PR target/85683
17513         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
17514         after cmpelim optimization.
17516 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
17518         * config.gcc: Support "goldmont".
17519         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
17520         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17521         PROCESSOR_GOLDMONT.
17522         * config/i386/i386.c (m_GOLDMONT): Define.
17523         (processor_target_table): Add "goldmont".
17524         (PTA_GOLDMONT): Define.
17525         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
17526         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
17527         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
17528         (fold_builtin_cpu): Add "goldmont".
17529         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
17530         (ix86_option_override_internal): Add "goldmont".
17531         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
17532         (processor_type): Add PROCESSOR_GOLDMONT.
17533         * config/i386/i386.md: Add CPU "glm".
17534         * config/i386/glm.md: New file.
17535         * config/i386/x86-tune.def: Add m_GOLDMONT.
17536         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
17538 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17540         PR target/85572
17541         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
17542         E_V4DImode.
17543         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
17544         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
17545         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
17547         PR target/85317
17548         * config/i386/i386.c (ix86_fold_builtin): Handle
17549         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
17551         PR target/85480
17552         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
17553         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
17555 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
17557         PR target/85658
17558         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
17559         (check_arch): Likewise.
17560         (check_fpu): Return the result rather than printing it.
17561         (end arch): Fix operator precedence.
17562         (end cpu): Likewise.
17563         (END): Print the result from check_fpu.
17565 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17566             Alan Hayward  <alan.hayward@arm.com>
17567             David Sherwood  <david.sherwood@arm.com>
17569         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
17570         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
17571         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
17572         (*fcmuo<mode>_and): New patterns.
17574 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17576         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
17577         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
17578         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
17579         (cmp_op, sve_imm_con): New code attributes.
17580         (SVE_COND_INT_CMP, imm_con): Delete.
17581         (cmp_op): Remove above unspecs from int attribute.
17582         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
17583         to...
17584         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
17585         comparison-specific unspecs.
17586         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
17587         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
17588         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
17589         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
17590         (*vec_fcm<cmp_op><mode>): Rename to...
17591         (*fcm<cmp_op><mode>): ...this and adjust likewise.
17592         (*vec_fcmuo<mode>): Rename to...
17593         (*fcmuo<mode>): ...this and adjust likewise.
17594         (*pred_fcm<cmp_op><mode>): New pattern.
17595         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
17596         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
17597         functions.
17598         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
17599         and UNORDERED.
17600         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
17601         (aarch64_emit_sve_predicated_cond): New function.
17602         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
17603         (aarch64_emit_unspec_cond_or): Replace with...
17604         (aarch64_emit_sve_or_conds): ...this new function.  Use
17605         aarch64_emit_sve_ptrue_op for the individual comparisons and
17606         aarch64_emit_binop to OR them together.
17607         (aarch64_emit_inverted_unspec_cond): Replace with...
17608         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
17609         aarch64_emit_sve_ptrue_op for the comparison and
17610         aarch64_emit_unop to invert the result.
17611         (aarch64_expand_sve_vec_cmp_float): Update after the above
17612         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
17614 2018-05-07  Nathan Sidwell  <nathan@acm.org>
17616         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
17617         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
17618         (Backwards Compatibility): Likewise.
17620 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17622         PR bootstrap/85681
17623         Revert:
17624         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17626         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17627         <prefetch_dynamic_strides>: New const bool field.
17628         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17629         prefetch_dynamic_strides.
17630         (exynosm1_prefetch_tune): Likewise.
17631         (thunderxt88_prefetch_tune): Likewise.
17632         (thunderx_prefetch_tune): Likewise.
17633         (thunderx2t99_prefetch_tune): Likewise.
17634         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17635         to false.
17636         (aarch64_override_options_internal): Update to set
17637         PARAM_PREFETCH_DYNAMIC_STRIDES.
17638         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17639         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17640         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17641         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17642         prefetch-dynamic-strides setting.
17644         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17646         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17647         <minimum_stride>: New const int field.
17648         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17649         minimum_stride field.
17650         (exynosm1_prefetch_tune): Likewise.
17651         (thunderxt88_prefetch_tune): Likewise.
17652         (thunderx_prefetch_tune): Likewise.
17653         (thunderx2t99_prefetch_tune): Likewise.
17654         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17655         (aarch64_override_options_internal): Update to set
17656         PARAM_PREFETCH_MINIMUM_STRIDE.
17657         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17658         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17659         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17660         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17661         stride is constant and is below the minimum stride threshold.
17663 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17665         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
17666         to 512.
17668 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17670         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17671         <prefetch_dynamic_strides>: New const bool field.
17672         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17673         prefetch_dynamic_strides.
17674         (exynosm1_prefetch_tune): Likewise.
17675         (thunderxt88_prefetch_tune): Likewise.
17676         (thunderx_prefetch_tune): Likewise.
17677         (thunderx2t99_prefetch_tune): Likewise.
17678         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17679         to false.
17680         (aarch64_override_options_internal): Update to set
17681         PARAM_PREFETCH_DYNAMIC_STRIDES.
17682         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17683         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17684         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17685         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17686         prefetch-dynamic-strides setting.
17688 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17690         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17691         <minimum_stride>: New const int field.
17692         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17693         minimum_stride field.
17694         (exynosm1_prefetch_tune): Likewise.
17695         (thunderxt88_prefetch_tune): Likewise.
17696         (thunderx_prefetch_tune): Likewise.
17697         (thunderx2t99_prefetch_tune): Likewise.
17698         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17699         (aarch64_override_options_internal): Update to set
17700         PARAM_PREFETCH_MINIMUM_STRIDE.
17701         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17702         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17703         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17704         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17705         stride is constant and is below the minimum stride threshold.
17707 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
17709         PR c++/85659
17710         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
17711         the type is addressable.  Don't force op into register if it has
17712         BLKmode.
17714 2018-05-05  Roland McGrath  <mcgrathr@google.com>
17716         PR other/77609
17717         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
17718         any section for which we don't know a specific type it should have,
17719         regardless of name.  Previously this was done only for the exact
17720         names ".init_array", ".fini_array", and ".preinit_array".
17721         (default_elf_asm_named_section): Add comment about
17722         relationship with default_section_type_flags and SECTION_NOTYPE.
17723         (get_section): Don't consider it a type conflict if one side has
17724         SECTION_NOTYPE and the other doesn't, as long as neither has the
17725         SECTION_BSS et al used in the default_section_type_flags logic.
17727 2018-05-05  Tom de Vries  <tom@codesourcery.com>
17729         PR target/85653
17730         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
17731         (workaround_barsyncs): New function.
17732         (nvptx_reorg): Use workaround_barsyncs.
17733         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17734         (define_expand "nvptx_membar_cta"): New define_expand.
17735         (define_insn "*nvptx_membar_cta"): New insn.
17737 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
17739         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
17740         To improve optimization opportunities.
17741         * builtin-types.def: The new needed builtin types for the above.
17743 2018-05-04  Richard Biener  <rguenther@suse.de>
17745         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
17746         * gimple-ssa-store-merging.c
17747         (imm_store_chain_info::output_merged_store): Remove redundant create,
17748         release split_store vector contents on failure.
17749         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
17750         scalar stmt vector on cache hit.
17752 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
17754         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
17755         Xilinx FP support.
17756         * config.gcc (powerpc-xilinx-eabi*): Remove.
17757         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
17758         support.
17759         (fusion_addis_mem_combo_load): Ditto.
17760         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
17761         FP support.
17762         (rs6000_cpu_cpp_builtins): Ditto.
17763         * config/rs6000/rs6000-linux.c
17764         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
17765         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
17766         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
17767         support.
17768         (rs6000_setup_reg_addr_masks): Ditto.
17769         (rs6000_init_hard_regno_mode_ok): Ditto.
17770         (rs6000_option_override_internal): Ditto.
17771         (legitimate_lo_sum_address_p): Ditto.
17772         (rs6000_legitimize_address): Ditto.
17773         (rs6000_legitimize_reload_address): Ditto.
17774         (rs6000_legitimate_address_p): Ditto.
17775         (abi_v4_pass_in_fpr): Ditto.
17776         (setup_incoming_varargs): Ditto.
17777         (rs6000_gimplify_va_arg): Ditto.
17778         (rs6000_split_multireg_move): Ditto.
17779         (rs6000_savres_strategy): Ditto.
17780         (rs6000_emit_prologue_components): Ditto.
17781         (rs6000_emit_epilogue_components): Ditto.
17782         (rs6000_emit_prologue): Ditto.
17783         (rs6000_emit_epilogue): Ditto.
17784         (rs6000_elf_file_end): Ditto.
17785         (rs6000_function_value): Ditto.
17786         (rs6000_libcall_value): Ditto.
17787         * config/rs6000/rs6000.h: Ditto.
17788         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
17789         (TARGET_MINMAX): ... this.  New.
17790         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
17791         * config/rs6000/rs6000.md: Remove Xilinx FP support.
17792         (*movsi_internal1_single): Delete.
17793         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
17794         mfpu=, mxilinx-fpu): Delete.
17795         * config/rs6000/singlefp.h: Delete.
17796         * config/rs6000/sysv4.h: Remove Xilinx FP support.
17797         * config/rs6000/t-rs6000: Ditto.
17798         * config/rs6000/t-xilinx: Delete.
17799         * config/rs6000/titan.md: Adjust for fp_type removal.
17800         * config/rs6000/vsx.md: Remove Xilinx FP support.
17801         (VStype_simple): Delete.
17802         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
17803         * config/rs6000/xfpu.h: Delete.
17804         * config/rs6000/xfpu.md: Delete.
17805         * config/rs6000/xilinx.h: Delete.
17806         * config/rs6000/xilinx.opt: Delete.
17807         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
17808         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
17810 2018-05-04  Tom de Vries  <tom@codesourcery.com>
17812         PR libgomp/85639
17813         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
17814         if ignore == 0.
17816 2018-05-04  Richard Biener  <rguenther@suse.de>
17818         PR middle-end/85627
17819         * tree-complex.c (update_complex_assignment): We are always in SSA form.
17820         (expand_complex_div_wide): Likewise.
17821         (expand_complex_operations_1): Likewise.
17822         (expand_complex_libcall): Preserve EH info of the original stmt.
17823         (tree_lower_complex): Handle removed blocks.
17824         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
17825         on complex multiplication and division libcall builtins.
17827 2018-05-04  Richard Biener  <rguenther@suse.de>
17829         PR middle-end/85574
17830         * fold-const.c (negate_expr_p): Restrict negation of operand
17831         zero of a division to when we know that can happen without
17832         overflow.
17833         (fold_negate_expr_1): Likewise.
17835 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
17837         PR libstdc++/85466
17838         * real.h (real_nextafter): Declare.
17839         * real.c (real_nextafter): New function.
17840         * fold-const-call.c (fold_const_nextafter): New function.
17841         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
17842         CASE_CFN_NEXTTOWARD.
17843         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
17844         even when arg1_mode is different from arg0_mode.
17846 2018-05-03  Nathan Sidwell  <nathan@acm.org>
17848         * doc/extend.texi (Deprecated Features): Remove
17849         -ffriend-injection.
17850         (Backwards Compatibility): Likewise.
17851         * doc/invoke.texi (C++ Language Options): Likewise.
17852         (C++ Dialect Options): Likewise.
17854 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
17856         PR target/85530
17857         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
17858         _mm512_mask_mullox_epi64): New intrinsics.
17860 2018-05-03  Tom de Vries  <tom@codesourcery.com>
17862         PR testsuite/85106
17863         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17864         dump files): Add offload-tree.
17866 2018-05-03  Richard Biener  <rguenther@suse.de>
17868         PR tree-optimization/85615
17869         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
17870         to loops not nested in BBs loop father to avoid creating multi-entry
17871         loops.
17873 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17875         PR tree-optimization/70291
17876         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
17877         arguments.  Change return type to tree.  Emit libcall as a new
17878         statement rather than replacing existing one when inplace_p is true.
17879         (expand_complex_multiplication_components): New function.
17880         (expand_complex_multiplication): Expand floating-point complex
17881         multiplication using the above.
17882         (expand_complex_division): Rename inner_type parameter to type.
17883         Update expand_complex_libcall call-site.
17884         (expand_complex_operations_1): Update expand_complex_multiplication
17885         and expand_complex_division call-sites.
17887 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17889         PR target/85582
17890         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17891         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17892         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
17893         the highest significant bit of the shift count mask is clear.  In
17894         check whether and[sq]i3 is needed verify that all significant bits
17895         of the shift count other than the highest are set.
17897 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17899         PR libgomp/82428
17900         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
17901         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
17902         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
17903         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
17904         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
17905         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
17906         __builtin_goacc_parlevel_size.
17908 2018-05-02  Richard Biener  <rguenther@suse.de>
17910         PR tree-optimization/85597
17911         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
17912         do not use split vect_get_vec_defs call but call vect_get_slp_defs
17913         directly.
17915 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17917         PR testsuite/85106
17918         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17919         dump files): Add ltrans-tree.
17921 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17923         PR testsuite/85106
17924         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17925         dump files): Add wpa-ipa.
17927 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
17929         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
17930         powerpc*-*-linux*paired* target.
17931         * config/rs6000/750cl.h: Delete.
17932         * config/rs6000/paired.h: Delete.
17933         * config/rs6000/paired.md: Delete.
17934         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
17935         float support.
17936         * config/rs6000/rs6000-builtin.def: Remove paired float support.
17937         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
17938         comment.  Remove paired float support.
17939         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
17940         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
17941         VECTOR_PAIRED.
17942         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
17943         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
17944         declarations.
17945         * config/rs6000/rs6000.c: Remove paired float support.
17946         (paired_expand_vector_init, paired_expand_vector_move,
17947         paired_emit_vector_compare, paired_emit_vector_cond_expr,
17948         (paired_expand_lv_builtin, paired_expand_stv_builtin,
17949         paired_expand_builtin, paired_expand_predicate_builtin,
17950         paired_init_builtins): Delete.
17951         * config/rs6000/rs6000.h: Remove paired float support.
17952         * config/rs6000/rs6000.md: Remove paired float support.
17953         (move_from_CR_ov_bit): Delete.
17954         * config/rs6000/rs6000.opt (mpaired): Delete.
17955         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
17956         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
17958 2018-05-02  Richard Biener  <rguenther@suse.de>
17960         PR middle-end/85567
17961         * gimplify.c (gimplify_save_expr): When in SSA form allow
17962         SAVE_EXPRs to compute to SSA vars.
17964 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17966         PR target/85582
17967         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17968         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17969         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
17970         clobber operands[2], instead use a new pseudo.  Formatting fixes.
17972 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
17974         PR tree-optimization/85586
17975         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
17976         exit early for statements in the same group if the accesses are
17977         not strided.
17979 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17981         PR lto/85451
17982         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
17983         error message.
17985 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
17987         PR tree-optimization/85143
17988         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
17990 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17992         PR lto/85451
17993         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
17994         not found" error message.
17996 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17998         PR other/83786
17999         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
18000         * vec.c (test_ordered_remove_if): New function.
18001         (vec_c_tests): Call test_ordered_remove_if.
18002         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
18003         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
18004         * tree-vect-patterns.c (vect_pattern_recog_1): Use
18005         VEC_ORDERED_REMOVE_IF.
18007 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18009         PR tree-optimization/82665
18010         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
18011         pointer subtraction where arguments come from a memchr call.
18013 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
18015         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
18016         --push-state --as-needed and --pop-state instead of --as-needed and
18017         --no-as-needed if ld supports it.
18018         * configure: Regenerated.
18020         PR web/85578
18021         * doc/install.texi2html: Replace _002d with - and _002a with * in
18022         generated html files using sed.
18024 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18026         PR c++/85523
18027         * gcc-rich-location.c (blank_line_before_p): New function.
18028         (use_new_line): New function.
18029         (gcc_rich_location::add_fixit_insert_formatted): New function.
18030         * gcc-rich-location.h
18031         (gcc_rich_location::add_fixit_insert_formatted): New function.
18033 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18035         * selftest.c (assert_streq): Rename "expected" and "actual" to
18036         "val1" and "val2".  Extend NULL-handling to cover both inputs
18037         symmetrically, while still requiring both to be non-NULL for a pass.
18038         * selftest.h (assert_streq): Rename "expected" and "actual" to
18039         "val1" and "val2".
18040         (ASSERT_EQ): Likewise.
18041         (ASSERT_EQ_AT): Likewise.
18042         (ASSERT_KNOWN_EQ): Likewise.
18043         (ASSERT_KNOWN_EQ_AT): Likewise.
18044         (ASSERT_NE): Likewise.
18045         (ASSERT_MAYBE_NE): Likewise.
18046         (ASSERT_MAYBE_NE_AT): Likewise.
18047         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
18048         the assertion to pass.
18049         (ASSERT_STREQ_AT): Likewise.
18051 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18053         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
18054         interaction with -pie.
18056 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18058         * selftest.h: Fix alphabetization of per-source-file selftest
18059         declarations.
18061 2018-04-30  Jason Merrill  <jason@redhat.com>
18063         PR c++/61982 - dead stores to destroyed objects.
18064         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
18065         of clobber.
18067 2018-04-30  Jason Merrill  <jason@redhat.com>
18069         * tree.c (build_clobber): New.
18070         * tree.h: Declare it.
18071         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
18073 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18075         * diagnostic-show-locus.c (layout::layout): Update for
18076         location_get_source_line returning a char_span.
18077         (struct char_span): Move to input.h.
18078         (struct correction): Update for fields in char_span becoming
18079         private.
18080         (struct source_line): Update for location_get_source_line
18081         returning a char_span.
18082         (layout::print_line): Likewise.
18083         * edit-context.c (edited_file::print_content): Likewise.
18084         (edited_file::print_diff_hunk): Likewise.
18085         (edited_file::print_run_of_changed_lines): Likewise.
18086         (edited_file::get_num_lines): Likewise.
18087         (edited_line::edited_line): Likewise.
18088         * final.c (asm_show_source): Likewise.
18089         * input.c (location_get_source_line): Convert return type
18090         from const char * to char_span, losing the final "line_len"
18091         param.
18092         (dump_location_info): Update for the above.
18093         (get_substring_ranges_for_loc): Likewise.  Use a char_span
18094         when handling the literal within the line.
18095         (test_reading_source_line): Update for location_get_source_line
18096         returning a char_span.
18097         * input.h (class char_span): Move here from
18098         diagnostic-show-locus.c, converting from a struct to a class.
18099         Make data members private.
18100         (char_span::operator bool): New.
18101         (char_span::length): New.
18102         (char_span::get_buffer): New.
18103         (char_span::operator[]): New.
18104         (char_span::subspan): Make const.
18105         (char_span::xstrdup): New.
18106         (location_get_source_line): Convert return type from const char *
18107         to char_span, losing the final "line_size" param.
18109 2018-04-30  Jan Hubicka  <jh@suse.cz>
18111         * lto-wrapper.c (ltrans_priorities): New static var.
18112         (cmp_priority): New.
18113         (run_gcc): Read priorities and if doing parallel build order
18114         the Makefile by them.
18116 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
18118         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
18120 2018-04-30  Richard Biener  <rguenther@suse.de>
18122         * tree-cfg.c (verify_address): Remove base argument, add
18123         flag whether to check TREE_ADDRESSABLE and do that.
18124         (verify_expr): Remove.
18125         (verify_types_in_gimple_reference): Add pieces from verify_expr.
18126         (verify_gimple_assign_single): Likewise.
18127         (verify_gimple_switch): Likewise.
18128         (verify_expr_location_1): Dereference tp once.  Add (disabled)
18129         piece from verify_expr.
18130         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
18132 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18134         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
18136 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18138         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
18139         (small_data_pattern): Likewise.
18140         (arc_rewrite_small_data): Likewise.
18141         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
18142         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
18143         (get_symbol_alignment): New function.
18144         (legitimate_small_data_address_p): Likewise.
18145         (legitimate_scaled_address): Update, call
18146         legitimate_small_data_address_p.
18147         (output_sdata): New static variable.
18148         (arc_print_operand): Update how we handle small data operands.
18149         (arc_print_operand_address): Likewise.
18150         (arc_legitimate_address_p): Update, use
18151         legitimate_small_data_address_p.
18152         (arc_rewrite_small_data_p): Remove.
18153         (arc_rewrite_small_data_1): Likewise.
18154         (arc_rewrite_small_data): Likewise.
18155         (small_data_pattern): Likewise.
18156         (compact_sda_memory_operand): Update to use
18157         legitimate_small_data_address_p and get_symbol_alignment.
18158         (prepare_move_operands): Don't rewite sdata pattern.
18159         (prepare_extend_operands): Remove.
18160         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
18161         pattern.
18162         (zero_extendqisi2): Likewise.
18163         (zero_extendhisi2): Likewise.
18164         (extendqihi2): Likewise.
18165         (extendqisi2): Likewise.
18166         (extendhisi2): Likewise.
18167         (addsi3): Likewise.
18168         (subsi3): Likewise.
18169         (andsi3): Likewise.
18170         * config/arc/constraints.md (Usd): Change it to memory constraint.
18172 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
18174         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
18175         as source of std instructions.
18176         * config/arc/arc.md (movsi_insn): Update pattern predicate to
18177         allow 6-bit constants as source for store instructions.
18178         (movdi_insn): Update instruction pattern to allow 6-bit constants
18179         as source for store instructions.
18181 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
18183         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
18185 2018-04-30  Nathan Sidwell  <nathan@acm.org>
18186             Sandra Loosemore <sandra@codesourcery.com>
18188         * dumpfile.c (dump_open): Allow '-' for stdout.
18189         * doc/invoke.texi (Developer Options): Document dump filename
18190         determination early.  Document stdin/stdout selection.
18192 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
18194         Microblaze Target: PIC data text relative
18196         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
18197         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
18198         Add declaration.
18199         * config/microblaze/microblaze.h (microblaze_constant_address_p):
18200         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
18201         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
18202         New addressing mode for data-text relative position indepenedent code.
18203         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
18204         'ADDRESS_SYMBOLIC_TXT_REL'.
18205         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
18206         (microblaze_legitimate_pic_operand): Exclude function calls from
18207         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
18208         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
18209         addresses cases.
18210         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18211         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
18212         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
18213         for 'address + offset'.
18214         (microblaze_expand_prologue): Add new function prologue call for
18215         'r20' assignation.
18216         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
18217         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
18218         table in case of TARGET_PIC_DATA_TEXT_REL.
18219         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
18220         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
18221         Add new macros 'UNSPEC_TEXT',
18222         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
18223         + exclude function calls from 'UNSPEC_PLT' in case of data text
18224         relative mode.
18225         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18226         new target hook for generating address diff vector tables in case of
18227         flag_pic.
18228         * doc/tm.texi : Regenerate.
18229         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
18230         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
18231         of addr diff vector generation.
18232         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
18233         target hook definition.
18234         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
18235         Add default function for generate_pic_addr_diff_vec -> flag_pic.
18236         * doc/invoke.texi (Add new pic option): Add new microblaze pic
18237         option for data text relative.
18239 2018-04-30  Richard Biener  <rguenther@suse.de>
18241         * tree-chrec.h (evolution_function_is_constant_p): Remove
18242         redundant check.
18243         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
18245 2018-04-30  Richard Biener  <rguenther@suse.de>
18247         PR bootstrap/85571
18248         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
18250 2018-04-30  Richard Biener  <rguenther@suse.de>
18252         PR tree-optimization/28364
18253         PR tree-optimization/85275
18254         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
18255         copying first exit test.
18257 2018-04-28  Mark Wielaard  <mark@klomp.org>
18259         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18260         dwarf_version >= 5.
18261         (dwarf_AT): Handle DW_AT_addr_base.
18262         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18264 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
18266         PR target/84431
18267         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18268         (*ashl<dwi>3_doubleword_mask_1): Ditto.
18269         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18270         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18272 2018-04-28  Richard Biener  <rguenther@suse.de>
18274         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18275         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18276         to reflect use.  Only add interesting stmts.
18278 2018-04-27  Martin Jambor  <mjambor@suse.cz>
18280         PR ipa/85549
18281         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18282         the jump function allows for passing through aggregate values.
18284 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
18286         * input.h (in_system_header_at): Convert from macro to inline
18287         function.
18288         (from_macro_expansion_at): Likewise.
18289         (from_macro_definition_at): Likewise.
18291 2018-04-27  Jeff Law  <law@redhat.com>
18293         * config.gcc: Mark tile* targets as deprecated/obsolete.
18295 2018-04-27  Richard Biener  <rguenther@suse.de>
18297         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18298         fix for ILP32.
18300 2018-04-27  Richard Biener  <rguenther@suse.de>
18302         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18304 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
18306         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18307         with Yd constraint. Set "preferred_for_speed" attribute from
18308         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18309         with Yd constraint.
18310         (*movdi_internal): Ditto.
18311         (movti_interunit splitters): Remove
18312         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18313         (movdi_interunit splitters): Ditto.
18314         * config/i386/constraints.md (Ye): Remove.
18315         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18317 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18319         PR target/85512
18320         * config/aarch64/constraints.md (Usg): Limit to 31.
18321         (Usj): Limit to 63.
18323 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
18325         PR tree-optimization/85529
18326         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18327         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18328         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18329         zero extension or masking of the MSB bit.
18330         (optimize_range_tests): Add FIRST_BB argument, pass it through
18331         to optimize_range_tests_var_bound.
18332         (maybe_optimize_range_tests, reassociate_bb): Adjust
18333         optimize_range_tests callers.
18335 2018-04-26  Richard Biener  <rguenther@suse.de>
18336             Jakub Jelinek  <jakub@redhat.com>
18338         * cgraph.h (symbol_table): Just declare debug method here.
18339         * symtab.c (symbol_table::debug): Define.
18341 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
18343         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18345 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
18347         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18348         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18349         (*movdi_internal): Substitute Yi and Yj constraint with x
18350         and Ym and Yn constraint with y constraint.  Update "isa"
18351         attribute and set "preferred_for_speed" attribute from
18352         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18353         (*movsi_internal): Ditto.
18354         (*movdf_internal): Ditto.
18355         (*movsf_internal): Ditto.
18356         (*zero_extendsidi2): Ditto.
18357         * config/i386/sse.md (vec_set<mode>_0): Ditto.
18358         (sse2_loadld): Ditto.
18359         (*vec_extract<ssevecmodelower>_0): Ditto.
18360         (*vec_extractv4si_0_zext_sse4): Ditto.
18361         (vec_concatv2di): Ditto.
18362         (*vec_dup<mode>): Ditto.
18363         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18364         * config/i386/constraints.md (Yi): Remove.
18365         (Yj): Remove.
18366         (Ym): Remove.
18367         (Yn): Remove.
18369 2018-04-26  Nathan Sidwell  <nathan@acm.org>
18371         * dumpfile.c (dump_open): New.
18372         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18373         (dump_finish): Detect stdio/stderr by value not name.
18375 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
18377         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18379 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18381         PR target/84952
18382         * config/nvptx/nvptx.c (verify_neutering_jumps)
18383         (verify_neutering_labels): New function
18384         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18386 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18388         PR target/84025
18389         * config/nvptx/nvptx.c (needs_neutering_p): New function.
18390         (nvptx_single): Use needs_neutering_p to skip over insns that do not
18391         need neutering.
18393 2018-04-26  Richard Biener <rguenther@suse.de>
18394             Tom de Vries  <tom@codesourcery.com>
18396         PR lto/85422
18397         * lto-streamer-out.c (output_function): Fixup loops if required to match
18398         discovery done in the reader.
18400 2018-04-26  Richard Biener  <rguenther@suse.de>
18402         PR tree-optimization/85116
18403         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18404         have a loop exit from the single latch predecessor.  Remove
18405         case of header with just condition.
18406         (ch_base::copy_headers): Exclude infinite loops from any
18407         processing.
18408         (pass_ch::execute): Record exits.
18410 2018-04-26  Richard Biener  <rguenther@suse.de>
18412         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18413         prologue cost vector and pass it to vect_get_load_cost.
18414         (vect_get_peeling_costs_all_drs): Likewise.
18415         (vect_peeling_hash_get_lowest_cost): Likewise.
18416         (vect_enhance_data_refs_alignment): Likewise.
18418 2018-04-26  Richard Biener  <rguenther@suse.de>
18420         PR middle-end/85450
18421         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18422         checking of integer<->pointer conversions.
18423         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18424         sign-/zero-extending pointer types.
18425         (expand_omp_for_static_chunk): Likewise.
18427 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
18428             Jean Lee  <xiaoyur347@gmail.com>
18430         * config/mips/mips.c (mips_asan_shadow_offset): New function.
18431         (TARGET_ASAN_SHADOW_OFFSET): Define.
18432         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18433         true for -fsanitize=address.
18435 2018-04-25  Mark Wielaard  <mark@klomp.org>
18437         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18438         shorter ones.
18440 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18442         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
18443         than "alu", remove explicit "memory" and "imm_disp" attributes.
18444         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
18446         PR middle-end/85414
18447         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
18448         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
18449         gen_lowpart_no_emit.
18451 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
18453         PR target/85473
18454         * config/i386/i386.c (ix86_expand_builtin): Change memory
18455         operand to XI, extend p0 to Pmode.
18456         * config/i386/i386.md: Change unspec volatile and operand
18457         1 mode to XI, change operand 0 mode to P.
18459 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18461         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
18462         GET_MODE_MASK before any checking.
18463         (nds32_can_use_bset_p): Likewise.
18464         (nds32_can_use_btgl_p): Likewise.
18466 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18468         * config/nds32/nds32-doubleword.md: New define_split pattern for
18469         illegal register number.
18471 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18473         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
18475 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18477         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
18479 2018-04-25  Richard Biener  <rguenther@suse.de>
18481         * lto-streamer.h (LTO_major_version): Bump to 8.
18483 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18485         * BASE-VER: Set to 9.0.0.
18487 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
18489         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
18490         in __abskf2 and __powikf2.
18492 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18494         PR target/85512
18495         * config/aarch64/constraints.md (Usg, Usj): New constraints.
18496         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
18497         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
18498         Use the above on operand 2.  Reindent.
18499         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
18501 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18503         PR target/85485
18504         * common/config/i386/i386-common.c (ix86_handle_option): Don't
18505         handle OPT_mcet.
18506         * config/i386/i386.opt (mcet): Removed.
18507         * doc/install.texi: Remove -mcet documentation.
18508         * doc/invoke.texi: Likewise.
18510 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18512         PR target/85485
18513         * doc/install.texi: Remove -mcet from bootstrap-cet.
18515 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18517         PR target/85511
18518         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
18519         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
18520         if TARGET_64BIT.
18522         PR target/85503
18523         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
18524         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
18525         containing a CONST_VECTOR.
18527 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
18529         * doc/install.texi: Update newlib dependency for nvptx.
18531 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18533         PR target/85508
18534         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
18535         instead of INTVAL when shifting x left.
18537 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
18539         PR tree-optimization/85478
18540         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
18541         vect_grouped_store_supported for single element vectors.
18543 2018-04-24  Richard Biener  <rguenther@suse.de>
18545         PR target/85491
18546         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
18547         load cost increase to the case of non-constant step.
18549 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18551         PR target/84828
18552         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
18553         destination if any_malformed_asm.
18555 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
18557         PR middle-end/85496
18558         * expr.c (store_field): In the bitfield case, if the value comes from
18559         a function call and is returned in registers by means of a PARALLEL,
18560         do not change the mode of the temporary unless BLKmode and VOIDmode.
18562 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
18564         PR rtl-optimization/85423
18565         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
18566         dependencies to debug insns when the previous insn is non-debug.
18568 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18570         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
18571         enums into a single definition.
18572         (fls): Fix predicates and printing.
18573         (seti): Likewise.
18575 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18577         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
18578         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
18579         and short u6 immediate.
18580         (check_if_valid_sleep_operand): Remove.
18581         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
18583 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18585         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
18586         flag_always_save_lp condition.
18587         * config/nds32/nds32.opt (malways-save-lp): New option.
18589 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18591         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
18592         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
18593         * config/nds32/nds32.h
18594         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
18595         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
18597 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18599         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
18600         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
18602 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18603             Chung-Ju Wu  <jasonwucj@gmail.com>
18605         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
18606         Declare.
18607         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
18608         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
18610 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18612         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
18614 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18616         * config/nds32/nds32-protos.h (nds32_data_alignment,
18617         nds32_local_alignment): Declare.
18618         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
18619         nds32_local_alignment): New functions.
18620         (TARGET_CONSTANT_ALIGNMENT): Define.
18621         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
18623 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18625         * config/nds32/nds32.c
18626         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
18627         (TARGET_MODES_TIEABLE_P): Likewise.
18629 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18631         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
18632         level Ofast and Og.
18634 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
18635             Chung-Ju Wu  <jasonwucj@gmail.com>
18637         * config/nds32/constants.md (unspec_volatile_element): Add enum values
18638         for unaligned access.
18639         * config/nds32/nds32-intrinsic.c: Implementation of expanding
18640         unaligned access.
18641         * config/nds32/nds32-intrinsic.md: Likewise.
18642         * config/nds32/nds32_intrinsic.h: Likewise.
18643         * config/nds32/nds32.h (nds32_builtins): Likewise.
18644         * config/nds32/nds32.opt (munaligned-access): New option.
18645         * config/nds32/nds32.c (nds32_asm_file_start): Display
18646         flag_unaligned_access status.
18648 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
18650         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
18651         -mno-relax is present.
18652         * config/riscv/linux.h (LINK_SPEC): Ditto.
18654 2018-04-20  Martin Sebor  <msebor@redhat.com>
18656         PR c/85365
18657         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
18658         for null pointers.
18659         (gimple_fold_builtin_stxcpy_chk): Same.
18660         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
18662 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
18664         PR target/85456
18665         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
18666         __powikf2 when long double is IEEE 128-bit.
18668 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
18670         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
18671         step to make sure stack always aligned.
18673 2018-04-20  Carl Love  <cel@us.ibm.com>
18675         PR target/83402
18676         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
18677         size check for arg0.
18679 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
18680             Tom de Vries  <tom@codesourcery.com>
18682         PR target/85445
18683         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
18684         Emit insns for calls too.
18685         (nvptx_find_par): Always look for worker-level predecessor insn.
18686         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
18687         calls.
18688         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
18689         (nvptx_process_pars): Propagate frames for calls.
18691 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
18693         PR target/85469
18694         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
18695         Removed.
18696         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
18697         (ix86_handle_option): Don't handle OPT_mibt.
18698         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
18699         __SHSTK__.
18700         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
18701         has_ibt and ibt.
18702         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
18703         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
18704         (ix86_target_macros): Define __CET__ with flag_cf_protection
18705         for -fcf-protection.
18706         * config/i386/i386.c (isa2_opts): Remove -mibt.
18707         * config/i386/i386.h (TARGET_IBT): Removed.
18708         (TARGET_IBT_P): Likewise.
18709         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
18710         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
18711         * config/i386/i386.opt (mcet): Update help message.
18712         (mshstk): Likewise.
18713         (mibt): Removed.
18714         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
18715         -mcet as an alias for -mshstk.
18717 2018-04-20  Richard Biener <rguenther@suse.de>
18719         PR middle-end/85475
18720         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
18721         complexity by forcing a single use of the multiply operand.
18723 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18725         ipa/85449
18726         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
18727         recursion dependency to only apply to non-clones.
18729 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18731         ipa/85447
18732         * ipa-cp.c (create_specialized_node): Check that clones of
18733         self-recursive edges exist during IPA-CP.
18735 2018-04-19  Toon Moene  <toon@moene.org>
18737         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
18738         by -O3.
18740 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18742         PR tree-optimization/85467
18743         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
18744         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
18745         VECTOR_CST element to type.
18747 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18749         PR target/85397
18750         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
18751         * config/i386/i386.md (builtin_setjmp_setup): Removed.
18752         (builtin_longjmp): Likewise.
18753         (save_stack_nonlocal): New pattern.
18754         (restore_stack_nonlocal): Likewise.
18756 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18758         PR target/85404
18759         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18760         Replace ASM_OUTPUT_LABEL with fprintf.
18762 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18764         PR target/85417
18765         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18766         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
18767         * config/i386/i386-c.c (ix86_target_macros_internal): Also
18768         define __IBT__ and __SHSTK__ for -fcf-protection.
18769         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
18770         TARGET_IBT.
18771         (ix86_trampoline_init): Likewise.
18772         (x86_output_mi_thunk): Likewise.
18773         (ix86_notrack_prefixed_insn_p): Likewise.
18774         (ix86_option_override_internal): Don't disallow -fcf-protection.
18775         * config/i386/i386.md (rdssp<mode>): Also enable for
18776         -fcf-protection.
18777         (incssp<mode>): Likewise.
18778         (nop_endbr): Likewise.
18779         * config/i386/i386.opt (mcet): Change help message to built-in
18780         functions only.
18781         (mibt): Likewise.
18782         (mshstk): Likewise.
18783         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
18784         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
18785         enable CET built-in functions.
18787 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
18789         * common/config/i386/i386-common.c
18790         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
18791         OPTION_MASK_ISA_MOVDIRI_UNSET,
18792         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
18793         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
18794         * config.gcc (movdirintrin.h): New header.
18795         * config/i386/cpuid.h (bit_MOVDIRI,
18796         bit_MOVDIR64B): New bits.
18797         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
18798         and -mmvodir64b.
18799         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
18800         (VOID, PVOID, PCVOID)): New function types.
18801         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
18802         __builtin_ia32_directstoreu_u64,
18803         __builtin_ia32_movdir64b): New builtins.
18804         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
18805         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
18806         and -mmovdiri.
18807         (ix86_valid_target_attribute_inner_p): Ditto.
18808         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
18809         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
18810         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
18811         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
18812         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
18813         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
18814         (movdiri<mode>, movdir64b_<mode>): New.
18815         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
18816         * config/i386/immintrin.h: Include movdirintrin.h.
18817         * config/i386/movdirintrin.h: New file.
18818         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
18820 2018-04-19  Richard Biener  <rguenther@suse.de>
18822         PR middle-end/85455
18823         * cfg.c (clear_bb_flags): When loop state says we have
18824         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
18826 2018-04-19  Richard Biener  <rguenther@suse.de>
18828         PR tree-optimization/84737
18829         * tree-vect-data-refs.c (vect_copy_ref_info): New function
18830         copying restrict info.
18831         (vect_setup_realignment): Use it.
18832         * tree-vectorizer.h (vect_copy_ref_info): Declare.
18833         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
18834         the first DR to all generated stores.
18835         (vectorizable_load): Likewise for loads.
18837 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18839         PR tree-optimization/85446
18840         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
18841         the integral and pointer types to have the same precision.
18843         * doc/install.texi: Document --disable-cet being the default and
18844         --enable-cet=auto.
18846 2018-04-18  Martin Liska  <mliska@suse.cz>
18848         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
18849         style.
18851 2018-04-18  Martin Liska  <mliska@suse.cz>
18853         Revert
18854         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
18856         PR ipa/83983
18857         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18858         arguments if they are comparable.
18860 2018-04-18  Martin Liska  <mliska@suse.cz>
18862         Revert
18863         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
18865         PR lto/84805
18866         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18867         incomplete types.
18869 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
18871         PR target/85388
18872         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
18873         ENDBR after calling __morestack.
18875 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
18877         PR jit/85384
18878         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
18879         by using gcc_base_ver to generate a gcc_driver_version, and use
18880         it when generating GCC_DRIVER_NAME.
18881         * configure: Regenerate.
18883 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
18885         PR target/81084
18886         * config.gcc: Obsolete powerpc*-*-*spe*.
18888 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18890         PR debug/84637
18891         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
18892         (stabstr_D): Change type of unum from unsigned int to
18893         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
18894         type.
18896 2018-04-17  Jim Wilson  <jimw@sifive.com>
18898         PR 84856
18899         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
18900         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
18901         Set arg_pointer_offset after using pretend_args_size.
18903 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18905         PR rtl-optimization/85431
18906         * dse.c (record_store): Ignore zero width stores.
18908         PR sanitizer/85230
18909         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
18910         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
18911         __builtin_stack_restore rather than after it.
18912         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
18913         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
18914         argument instead of virtual_dynamic_stack_rtx.
18916 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18918         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
18919         New prototype.
18920         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
18921         Add note to error message to explain internal mapping of overloaded
18922         built-in function name to non-overloaded built-in function name.
18923         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
18924         function.
18926 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
18928         PR target/85424
18929         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
18930         where the inputs overlap with the output.
18932 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18934         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
18935         (=v, v) alternative and explicit "memory" attribute.
18936         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
18937         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18938         attributes.
18939         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
18940         "sselog1" type instead of "sselog".
18941         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
18942         "sselog".  Remove explicit "memory" attribute.
18943         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
18944         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18945         attributes.
18946         (vec_extract_hi_v32hi): Merge all alternatives into one, use
18947         "sselog1" type instead of "sselog".  Remove explicit "memory"
18948         attribute.
18949         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
18950         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18951         attribute.
18952         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
18953         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18954         attributes.
18955         (vec_extract_hi_v64qi): Merge all alternatives into one, use
18956         "sselog1" type instead of "sselog".  Remove explicit "memory"
18957         attribute.
18958         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
18959         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18960         attribute.
18962         PR target/85430
18963         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
18965         PR middle-end/85414
18966         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
18967         on a SUBREG.
18969 2018-04-17  Martin Jambor  <mjambor@suse.cz>
18971         PR ipa/85421
18972         * ipa-cp.c (create_specialized_node): Call
18973         expand_all_artificial_thunks if necessary.
18975 2018-04-17  Martin Liska  <mliska@suse.cz>
18977         PR lto/85405
18978         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
18979         in message, remote space in between '_G' and '('.
18981 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18983         PR target/85281
18984         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
18985         avx512f_vmcmp<mode>3<round_saeonly_name>,
18986         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
18987         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
18988         avx512f_rndscale<mode><round_saeonly_name>,
18989         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
18990         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
18991         Use %<iptr>2 instead of %2 for -masm=intel.
18992         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
18993         avx512f_vcvttss2usi<round_saeonly_name>,
18994         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
18995         -masm=intel.
18996         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
18997         avx512f_vcvttsd2usi<round_saeonly_name>,
18998         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
18999         Use %q1 instead of %1 for -masm=intel.
19000         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
19001         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
19002         of %3 for -masm=intel.
19003         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
19004         -masm=intel.
19005         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
19006         -masm=intel.
19007         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
19008         -masm=intel.
19009         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
19010         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
19011         %g1.
19012         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
19013         -masm=intel.
19014         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
19015         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
19016         %g1 and one with %0 and %1.
19017         (avx512er_vmrcp28<mode><round_saeonly_name>,
19018         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
19019         %1 for -masm=intel.
19020         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
19021         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
19022         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
19023         of %0 and %{%4%} for -masm=intel.
19024         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
19025         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
19026         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
19027         order of %0 and %{%5%}%{z%} for -masm=intel.
19029 2018-04-17  Jan Hubicka  <jh@suse.cz>
19031         PR lto/85405
19032         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
19034 2018-04-17  Martin Liska  <mliska@suse.cz>
19036         PR ipa/85329
19037         * multiple_target.c (create_dispatcher_calls): Set apostrophes
19038         for target_clone error message.  Make default implementation
19039         clone to be a local declaration.
19040         (separate_attrs): Add new argument and check for an empty
19041         string.
19042         (expand_target_clones): Handle it.
19043         (ipa_target_clone): Make redirection just for target_clones
19044         functions.
19046 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
19047             Tom de Vries  <tom@codesourcery.com>
19049         PR middle-end/84955
19050         * omp-expand.c (expand_oacc_for): Add dummy false branch for
19051         tiled basic blocks without omp continue statements.
19053 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
19055         PR target/83660
19056         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
19057         vec_extract expression as having side effects to make sure it gets
19058         a cleanup point.
19060 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
19062         PR target/85403
19063         * config/i386/i386.c (get_builtin_code_for_version): Check
19064         error_mark_node.
19066 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
19068         PR target/84331
19069         * config.gcc: Support "skylake".
19070         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19071         PROCESSOR_SKYLAKE.
19072         * config/i386/i386.c (m_SKYLAKE): Define.
19073         (processor_target_table): Add "skylake".
19074         (ix86_option_override_internal): Add "skylake".
19075         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
19076         PROCESSOR_CANNONLAKE.
19077         (get_builtin_code_for_version): Fix priority for
19078         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
19079         PROCESSOR_SKYLAKE-AVX512.
19080         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
19081         (processor_type): Add PROCESSOR_SKYLAKE.
19083 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
19084             Jason Merrill  <jason@redhat.com>
19086         PR c++/85112
19087         * convert.c (convert_to_integer_1): Use direct recursion for
19088         enumeral types and types with a precision less than the number
19089         of bits in their mode.
19091 2018-04-16  Julia Koval  <julia.koval@intel.com>
19093         PR target/84413
19094         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
19095         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
19097 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
19099         PR target/85293
19100         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
19101         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
19102         and -mno-direct-move.
19104 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
19106         PR target/83402
19107         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
19108         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
19109         Ensure negative shifts result in {0}.
19111 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
19113         PR rtl-optimization/79916
19114         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19115         regs (if any) to define how to gnerate SD moves when LRA is in
19116         progress.
19118 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
19120         PR rtl-optimization/85393
19121         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
19122         * except.c (expand_dw2_landing_pad_for_region): Make static.
19123         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
19124         a label and unconditional jump to old_bb, rather than
19125         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
19126         basic block.
19128         PR rtl-optimization/85376
19129         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
19130         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
19131         instead of a specific value.
19133 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19134             Bin Cheng  <bin.cheng@arm.com>
19136         PR tree-optimization/82965
19137         PR tree-optimization/83991
19138         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
19139         by_profile_only parameter.
19140         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
19141         information if the loop was predicted to iterate too many times.
19142         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
19144 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
19146         PR lto/71991
19147         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
19148         always inline.
19150 2018-04-13  Martin Liska  <mliska@suse.cz>
19151             Jakub Jelinek  <jakub@redhat.com>
19153         PR middle-end/81657
19154         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
19155         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
19156         * builtins.c (expand_builtin_memory_copy_args): Use
19157         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
19158         handle dest_addr == pc_rtx.
19160 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
19162         PR target/85291
19163         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
19164         asked to not generate direct moves.
19165         (fix_trunc<mode>si2_stfiwx): Similar.
19166         (fix_trunc<mode>si2_internal): Similar.
19168 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19170         PR debug/83157
19171         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
19172         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
19173         lookup if dest in some wider mode is known to be const0_rtx and
19174         if so, record permanent equivalence for it to be ZERO_EXTEND of
19175         the narrower mode destination.
19177 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19179         * lto-streamer-out.c (output_function): Revert 259346.
19180         * omp-expand.c (expand_oacc_for): Likewise.
19182 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
19184         PR rtl-optimization/85354
19185         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
19186         * sel-sched.c (sel_global_init): ... here.
19188 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
19190         PR target/85238
19191         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
19192         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
19193         mode for PE-COFF targets.
19194         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
19195         (i386_pe_asm_lto_end): Likewise.
19196         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
19197         (TARGET_ASM_LTO_END): Likewise.
19198         * config/i386/winnt.c (saved_debug_info_level): New static variable.
19199         (i386_pe_asm_lto_start): New function.
19200         (i386_pe_asm_lto_end): Likewise.
19202 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
19203             Richard Biener  <rguenther@suse.de>
19205         PR middle-end/84955
19206         * lto-streamer-out.c (output_function): Fix CFG loop state before
19207         streaming out.
19208         * omp-expand.c (expand_oacc_for): Handle calls to internal
19209         functions like regular functions.
19211 2018-04-12  Richard Biener  <rguenther@suse.de>
19213         PR lto/85371
19214         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
19215         for the early LTO debug to properly generate references to it
19216         during DIE emission.  Do not re-use that for the skeleton for
19217         split-dwarf.
19218         (dwarf2out_early_finish): Likewise.
19220 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19222         PR target/85328
19223         * config/i386/sse.md
19224         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
19225         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
19226         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
19227         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
19228         and output is a reg, avoid creating invalid lowpart subreg, but
19229         instead split into a 512-bit move.  Don't split if not AVX512VL,
19230         input is xmm16+ reg and output is a mem.
19231         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
19232         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
19233         xmm16+ reg and output is a mem.
19235 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19237         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19238         also for flag_dwarf2_cfi_asm.
19240 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
19242         PR rtl-optimization/85342
19243         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
19244         a bool scalar var inside of the loop instead.  Don't try to update
19245         recog_data.operand after failed apply_change_group.
19247 2018-04-12  Tom de Vries  <tom@codesourcery.com>
19249         PR target/85296
19250         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
19251         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
19252         array with flexible array member as array without given dimension.
19253         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
19254         argument for undefined param to true.
19256 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
19258         PR target/85321
19259         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19260         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19261         from PowerPC section.
19262         * config/rs6000/sysv4.opt (mcall-): Improve help text.
19263         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19264         help text that is too long.
19265         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19266         help text that is too long.
19267         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19268         help text that is too long.
19270 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
19272         * config/alpha/alpha.md (stack_probe_internal): Rename
19273         from "probe_stack".  Update all callers.
19275 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19277         PR rtl-optimization/84566
19278         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19279         sched_macro_fuse_insns.
19281 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19283         PR target/84301
19284         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19285         (compute_block_dependences): ... from here.
19287 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19289         PR tree-optimization/85331
19290         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19291         from int to HOST_WIDE_INT.
19293 2018-04-11  Martin Jambor  <mjambor@suse.cz>
19295         PR ipa/84149
19296         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19297         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19298         not the same as the source val.
19299         (cgraph_edge_brings_value_p): New parameter.
19300         (gather_edges_for_value): Pass destination value to
19301         cgraph_edge_brings_value_p.
19302         (perhaps_add_new_callers): Likewise.
19303         (get_info_about_necessary_edges): Likewise and exclude values brought
19304         only by self-recursive edges.
19305         (create_specialized_node): Redirect only clones of self-calling edges.
19306         (+self_recursive_pass_through_p): New function.
19307         (find_more_scalar_values_for_callers_subset): Use it.
19308         (find_aggregate_values_for_callers_subset): Likewise.
19309         (known_aggs_to_agg_replacement_list): Removed.
19310         (decide_whether_version_node): Re-calculate known constants for all
19311         remaining context clones.
19313 2018-04-11  Richard Biener  <rguenther@suse.de>
19315         PR lto/85339
19316         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19317         from early DWARF output.
19318         (dwarf2out_early_finish): Output line info unconditionally into
19319         early DWARF and add reference to it.
19321 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19323         PR target/85281
19324         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19325         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19326         other than V2DFmode using iptr mode attribute.
19327         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19329 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19331         PR rtl-optimization/84659
19332         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19334 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19336         PR debug/85302
19337         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19338         SIZEP is NULL.
19339         (output_loc_list): Pass address of a dummy size variable even in the
19340         locview handling loop.
19341         (index_location_lists): Add comment on why skip_loc_list_entry can't
19342         call size_of_locs.
19344 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19346         PR target/85261
19347         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19348         into register.
19350 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
19352         PR target/85321
19353         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19354         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19355         and -mstring-compare-inline-limit.
19357 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19359         PR target/85287
19360         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19361         for stack clash protection in a register whenever we need it to be in
19362         a register.
19364 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19366         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19367         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19369 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19371         PR target/85321
19372         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19373         the help text.
19374         (mlong-double-): Ditto.
19375         * config/rs6000/sysv4.opt (msdata=): Ditto.
19376         (mtls-size=): Ditto.
19378 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19380         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19381         erroneous entries for
19382         "vector int vec_ldl (int, long int *)", and
19383         "vector unsigned int vec_ldl (int, unsigned long int *)".
19384         Add comments and entries for
19385         "vector bool char vec_ldl (int, bool char *)",
19386         "vector bool short vec_ldl (int, bool short *)",
19387         "vector bool int vec_ldl (int, bool int *)",
19388         "vector bool long long vec_ldl (int, bool long long *)",
19389         "vector pixel vec_ldl (int, pixel *)",
19390         "vector long long vec_ldl (int, long long *)",
19391         "vector unsigned long long vec_ldl (int, unsigned long long *)".
19392         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19393         type tree bool_long_long_type_node and correct definition of
19394         bool_V2DI_type_node to make reference to this new type tree.
19395         (rs6000_mangle_type): Replace erroneous reference to
19396         bool_long_type_node with bool_long_long_type_node.
19397         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19398         comments to emphasize sign distinctions for char and int types and
19399         replace RS6000_BTI_bool_long constant with
19400         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
19401         use of RS6000_BTI_pixel.
19402         (bool_long_type_node): Remove this macro definition.
19403         (bool_long_long_type_node): New macro definition
19405 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19407         PR rtl-optimization/85300
19408         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19409         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19410         simplify_unary_operation fails.
19412 2018-04-10  Martin Liska  <mliska@suse.cz>
19414         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19415         cgraph_edge and ipa_ref.
19417 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19419         PR target/85177
19420         PR target/85255
19421         * config/i386/sse.md
19422         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19423         computation of the VEC_MERGE selector from mask.
19424         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19425         Fix decoding of the VEC_MERGE selector into mask.
19427 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19429         PR tree-optimization/85286
19430         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19432 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19434         * final.c (final_1): Set insn_last_address as well as
19435         insn_current_address.
19437 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19439         PR target/85173
19440         * explow.c (emit_stack_probe): Call validize_mem on memory location
19441         before passing it to gen_probe_stack.  Create address operand and
19442         legitimize it for the probe_stack_address case.
19444 2018-04-09  Jan Hubicka  <jh@suse.cz>
19446         PR lto/85078
19447         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
19448         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
19449         * tree.c (free_lang_data_in_type): Fix handling of binfos;
19450         walk basetypes.
19451         (free_lang_data): Rebuild type inheritance graph.
19453 2018-04-09  Martin Sebor  <msebor@redhat.com>
19455         * invoke.texi (-finline-small-functions): Mention other optimization
19456         options.
19457         (-findirect-inlining, -fpartial-inlining): Same.
19458         (-finline-functions-called-once): Same.
19459         (-freorder-blocks-and-partition): Same.
19461 2018-04-09  Jan Hubicka  <jh@suse.cz>
19463         PR rtl/84058
19464         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
19465         jumps; choose last target that matches the criteria (i.e.
19466         no partition changes for non-crossing jumps).
19467         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
19468         support for redirecting crossing jumps to non-crossing.
19470 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
19472         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
19473         also for naked functions.
19475 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
19477         * config/arc/arc.md (add_shift): New pattern.
19478         (add_shift2): Likewise.
19479         (sub_shift): Likewise.
19480         (sub_shift_cmp0_noout): Likewise.
19481         (compare_si_ashiftsi): Likewise.
19482         (xbfu_cmp0_noout): New combine pattern.
19483         (xbfu_cmp0"): Likewise.
19484         (movsi_set_cc_insn): Place the predicable variant first.
19485         (commutative_binary_cmp0_noout): Remove clobber.
19486         (commutative_binary_cmp0): New pattern.
19487         (noncommutative_binary_cmp0): Likewise.
19488         (noncommutative_binary_cmp0_noout): Likewise.
19489         (noncommutative_binary_comparison_result_used): Removed.
19490         (rsub_cmp0): New pattern.
19491         (rsub_cmp0_noout): Likewise.
19492         (extzvsi): Changed, keep only meaningful variants.
19493         (SQH, SEZ): New iterators.
19494         (SQH_postfix): New mode attribute.
19495         (SEZ_prefix): New code attribute.
19496         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
19497         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
19498         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
19499         of numerical value.
19500         (noncommutative_operator): Check the availability of barrel
19501         shifter option.
19503 2018-04-09  Richard Biener  <rguenther@suse.de>
19505         PR tree-optimization/85284
19506         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
19507         Only use the niter constraining form of simple_iv when the exit
19508         is always executed.
19510 2018-04-09  Tom de Vries  <tom@codesourcery.com>
19512         PR target/84041
19513         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19514         (define_expand "*memory_barrier"): New define_expand.
19515         (define_insn "memory_barrier"): New insn.
19517 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19519         PR rtl-optimization/80463
19520         PR rtl-optimization/83972
19521         PR rtl-optimization/83480
19523         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
19524         correct producer for the insn.
19525         (tidy_control_flow): Fixup seqnos in case of debug insns.
19527 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19529         PR rtl-optimization/83913
19531         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
19532         different sched-times when merging exprs.
19534 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19536         PR rtl-optimization/83962
19538         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
19539         tidy_fallthru_edge and tidy_control_flow.
19541 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19543         PR rtl-optimization/83530
19545         * sel-sched.c (force_next_insn): New global variable.
19546         (remove_insn_for_debug): When force_next_insn is true, also leave only
19547         next insn in the ready list.
19548         (sel_sched_region): When the region wasn't scheduled, make another pass
19549         over it with force_next_insn set to 1.
19551 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19553         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
19554         into tm_file.
19555         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19556         for interrupt control.
19557         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19558         functions for interrupt control.
19559         * config/nds32/nds32-intrinsic.md: Likewise.
19560         * config/nds32/nds32_intrinsic.h: Likewise.
19561         * config/nds32/nds32.h (nds32_builtins): Likewise.
19563 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19565         * config/nds32/nds32.c (nds32_init_machine_status,
19566         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
19567         strict_aligned_p field.
19568         (nds32_expand_to_rtl_hook): New function.
19569         (TARGET_EXPAND_TO_RTL_HOOK): Define.
19570         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
19572 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19573             Chung-Ju Wu  <jasonwucj@gmail.com>
19575         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
19576         * config/nds32/nds32-n7.md: New file.
19577         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
19578         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
19579         pipeline.
19580         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
19581         * config/nds32/nds32.md (pipeline_model): Add n7.
19582         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
19583         * config/nds32/pipelines.md: Include n7 settings.
19585 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19586             Chung-Ju Wu  <jasonwucj@gmail.com>
19588         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
19589         * config/nds32/nds32-e8.md: New file.
19590         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
19591         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
19592         pipeline.
19593         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
19594         * config/nds32/nds32.md (pipeline_model): Add e8.
19595         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
19596         * config/nds32/pipelines.md: Include e8 settings.
19598 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19599             Chung-Ju Wu  <jasonwucj@gmail.com>
19601         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
19602         * config/nds32/nds32-n8.md: New file.
19603         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
19604         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
19605         pipeline.
19606         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
19607         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
19608         * config/nds32/nds32.md (pipeline_model): Add n8.
19609         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
19610         * config/nds32/pipelines.md: Include n8 settings.
19612 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19613             Chung-Ju Wu  <jasonwucj@gmail.com>
19615         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
19616         * config/nds32/nds32-n9-2r1w.md: New file.
19617         * config/nds32/nds32-n9-3r2w.md: New file.
19618         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
19619         nds32_register_ports): New or modify for cpu n9.
19620         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
19621         pipeline.
19622         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
19623         * config/nds32/nds32-utils.c: New file.
19624         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
19625         TARGET_MUL_SLOW): Define.
19626         * config/nds32/nds32.md (pipeline_model): New attribute.
19627         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
19628         New options that support cpu n9.
19629         * config/nds32/pipelines.md: Include n9 settings.
19630         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
19632 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19634         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
19635         information if necessary.
19636         (output_cond_branch_compare_zero): Likewise.
19637         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
19638         (nds32_target_alignment): Refine for alignment.
19639         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
19640         (FUNCTION_BOUNDARY): Modify.
19641         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
19642         align case.
19643         * config/nds32/nds32.opt (malways-align, malign-functions): New.
19645 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19647         * config/nds32/constants.md (unspec_volatile_element): Add values for
19648         TLB operation and data prefetch.
19649         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19650         functions for TLB operation and data prefetch.
19651         * config/nds32/nds32-intrinsic.md: Likewise.
19652         * config/nds32/nds32_intrinsic.h: Likewise.
19653         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
19654         (nds32_print_operand): Likewise.
19655         * config/nds32/nds32.h (nds32_builtins): Likewise.
19657 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
19658         Andrew Pinski <pinsika@gcc.gnu.org>
19660         PR middle-end/82976
19661         * match.pd: Use constant_boolean_node of correct type instead of
19662         boolean_true_node or boolean_false_node for simplifying
19663         pointer comparisons to zero.
19665 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
19667         PR tree-optimization/80021
19668         * tree.c (verify_type_variant): Make error call in verify_variant_match
19669         translatable and remove final full stop.
19671 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19673         * config/nds32/constants.md (unspec_volatile_element): Add
19674         UNSPEC_VOLATILE_EH_RETURN.
19675         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
19676         nds32_output_stack_pop): Support dwarf exception handling process.
19677         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
19678         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
19679         exception handling process.
19680         (nds32_compute_stack_frame): Likewise.
19681         (nds32_return_addr_rtx): Likewise.
19682         (nds32_initial_elimination_offset): Likewise.
19683         (nds32_expand_prologue): Likewise.
19684         (nds32_expand_epilogue): Likewise.
19685         (nds32_dynamic_chain_address): New function.
19686         * config/nds32/nds32.h (machine_function): Add fields for dwarf
19687         exception handling.
19688         (DYNAMIC_CHAIN_ADDRESS): Define.
19689         (EH_RETURN_DATA_REGNO): Define.
19690         (EH_RETURN_STACKADJ_RTX): Define.
19691         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
19692         patterns for dwarf exception handling.
19694 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19696         * config/nds32/nds32.h: Clean up obsolete macros.
19698 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19700         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19701         Add enum values for particular instructions.
19702         * config/nds32/nds32-intrinsic.c: Implementation of expanding
19703         particular intrinsic functions.
19704         * config/nds32/nds32-intrinsic.md: Likewise.
19705         * config/nds32/nds32_intrinsic.h: Likewise.
19706         * config/nds32/nds32.h (nds32_builtins): Likewise.
19707         * config/nds32/nds32.md (type): Add pbsad and pbsada.
19708         (btst, ave): New patterns for particular instructions.
19710 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19712         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19713         Add enum values for atomic load/store and memory sync.
19714         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
19715         and memory sync.
19716         * config/nds32/nds32-intrinsic.md: Likewise.
19717         * config/nds32/nds32_intrinsic.h: Likewise.
19718         * config/nds32/nds32.h (nds32_builtins): Likewise.
19720 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
19722         PR tree-optimization/85257
19723         * fold-const.c (native_encode_vector): If not all elts could fit
19724         and off is -1, return 0 rather than offset.
19725         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
19726         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
19727         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
19728         adjust buffer in native_interpret_expr call.
19730 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19732         * config/nds32/constants.md (unspec_volatile_element): Add cache
19733         control enum values.
19734         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
19735         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
19736         * config/nds32/nds32.c (nds32_cctl_names): New.
19737         (nds32_print_operand): Handle cache control register names.
19738         * config/nds32/nds32.h (nds32_builtins): New enum values.
19739         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
19740         macros.
19741         * config/nds32/nds32.md (type): Add mmu.
19742         * config/nds32/pipelines.md (simple_insn): Add mmu.
19744 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19746         * config/nds32/nds32.md (type): Remove call.
19747         * config/nds32/pipelines.md (simple_insn): Likewise.
19749 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19751         * config/nds32/constants.md (unspec_volatile_element): Add
19752         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
19753         UNSPEC_VOLATILE_FMFCFG.
19754         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
19755         description for fmfcfg and fmfcsr.
19756         (bdesc_1arg): Add fmtcsr.
19757         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
19758         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
19759         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
19760         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
19761         unspec_fmfcfg): New patterns.
19762         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
19763         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
19764         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
19765         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
19766         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
19767         __nds32__fmfcfg): Define.
19769 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19771         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
19772         intrinsic register names.
19773         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
19774         intrinsic register enum values and macros.
19776 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19778         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
19779         for load/store addressing form.
19780         (nds32_print_operand_address): Likewise.
19782 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
19784         PR target/85196
19785         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
19786         based on LABEL_REF.  Remove useless assertion.
19787         (pic_address_needs_scratch): Fix formatting.
19788         (sparc_legitimize_pic_address): Minor tweaks.
19789         (sparc_delegitimize_address): Adjust assertion accordingly.
19790         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
19791         into symbolic_operand.
19792         (movsi_high_pic_label_ref): Likewise.
19793         (movsi_lo_sum_pic_label_ref): Likewise.
19794         (movdi_pic_label_ref): Likewise.
19795         (movdi_high_pic_label_ref): Likewise.
19796         (movdi_lo_sum_pic_label_ref): Likewise.
19798 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
19800         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
19801         custom LIB_SPEC setup.
19803 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
19804             Kito Cheng  <kito.cheng@gmail.com>
19806         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
19807         * config/riscv/freebsd.h: New.
19809 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19811         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
19812         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
19813         file.
19815 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19816             Kito Cheng  <kito.cheng@gmail.com>
19818         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
19819         nds32_output_call, nds32_symbol_binds_local_p): New functions.
19820         * config/nds32/nds32-protos.h (nds32_output_call,
19821         nds32_output_return): Declare.
19822         * config/nds32/nds32.md: Refine all the call and return patterns.
19824 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19826         PR debug/85252
19827         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
19828         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
19830         PR rtl-optimization/84872
19831         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
19832         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
19833         EDGE_CROSSING edge.
19835 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
19837         * expr.c (copy_blkmode_to_reg): Revert 254862.
19838         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
19840 2018-04-06  Richard Biener  <rguenther@suse.de>
19842         PR middle-end/85244
19843         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
19844         after seeing a component reference with an adjacent field.  Treat
19845         refs to arrays at struct end of external decls similar to
19846         refs to unconstrained commons.
19848 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19850         PR sanitizer/85213
19851         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
19852         look through SAVE_EXPRs with non-side-effects argument.  Adjust
19853         recursive calls.
19854         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
19855         save_p here.
19857 2018-04-06  Richard Biener  <rguenther@suse.de>
19859         PR middle-end/85180
19860         * alias.c (find_base_term): New wrapper around find_base_term
19861         unwinding CSELIB_VAL_PTR changes.
19862         (find_base_term): Do not restore CSELIB_VAL_PTR during the
19863         recursion.
19865 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19867         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
19868         instructions.
19869         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
19870         constant definitions.
19871         ("nop"): lr 0,0 -> nopr r0
19872         ("nop_lr0", "nop_lr1"): New insn definitions.
19874 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19876         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
19877         NDS32_V3PUSH_AVAILABLE_P macro.
19879 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
19880             Chung-Ju Wu  <jasonwucj@gmail.com>
19882         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
19883         (nds32*-*-*): Add float and fpu_config into supported_defaults.
19884         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
19885         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
19886         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
19887         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
19888         * config/nds32/constraints.md: New constraints and checking for hard
19889         float configuration.
19890         * config/nds32/iterators.md: New mode iterator and attribute for hard
19891         float configuration.
19892         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
19893         patterns.
19894         * config/nds32/nds32-fpu.md: New file.
19895         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
19896         deal with hard float code generation.
19897         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
19898         ARCH_V3S.
19899         (abi_type, float_reg_number): New enum type.
19900         * config/nds32/nds32-predicates.c: New predicates for hard float.
19901         * config/nds32/nds32-protos.h: Declare functions for hard float.
19902         * config/nds32/nds32.c: Implementation for hard float configuration.
19903         * config/nds32/nds32.h: Definitions for hard float configuration.
19904         * config/nds32/nds32.md: Include hard float machine description and
19905         modify patterns for hard float configuration.
19906         * config/nds32/nds32.opt: New options for hard float configuration.
19907         * config/nds32/predicates.md: New predicates for hard float
19908         configuration.
19910 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19912         * common/config/nds32/nds32-common.c
19913         (nds32_option_optimization_table): Enable -mreleax-hint by default.
19915 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
19917         PR middle-end/85195
19918         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
19919         CONSTRUCTOR_ELT (ctor, ...)->value.
19921 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
19923         PR target/85193
19924         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
19926 2018-04-05  Tom de Vries  <tom@codesourcery.com>
19928         PR target/85204
19929         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
19930         cond jump.
19932 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19933             Kito Cheng  <kito.cheng@gmail.com>
19935         * config/nds32/constraints.md (U33): Fine-tune checking condition.
19936         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
19937         * config/nds32/nds32.h (nds32_16bit_address_type): Add
19938         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
19940 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19941             Kito Cheng  <kito.cheng@gmail.com>
19943         * config/nds32/constraints.md (Ufe): New memory constraint.
19944         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
19945         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
19946         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
19947         operands.
19948         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
19949         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
19951 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19953         * config/nds32/nds32.md: Use optimize_size in the condition for
19954         alu-shift instructions.
19956 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19958         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
19960 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19962         * config/nds32/nds32.md (negsi2): Refine pattern.
19964 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19965             Chung-Ju Wu  <jasonwucj@gmail.com>
19967         * config/nds32/iterators.md (shift_rotate): New code iterator.
19968         (shift): New code attribute.
19969         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
19970         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
19971         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
19972         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
19973         bit-wise operations.
19974         (andsi3, *andsi3): Ditto.
19975         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
19976         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
19977         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
19978         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
19979         nds32_ior_operand, nds32_xor_operand): New predicates.
19981 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19983         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
19984         (addsi3, subsi3): ... this.
19986 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19988         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
19990 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19992         * config/nds32/nds32.md: Adjust indention.
19994 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19996         * config/nds32/nds32.md (feature): New attribute.
19998 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
20000         * config/nds32/nds32.md (subtype): New attribute.
20002 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20004         PR target/85203
20005         * config/arm/arm-builtins.c (arm_expand_builtin): Change
20006         expansion to perform a bitwise AND of the argument followed by a
20007         boolean negation of the result.
20009 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
20011         PR rtl-optimization/84878
20012         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
20013         the basic block.  Assert the use reference is not artificial and that
20014         it has an associated insn.
20016 2018-04-04  Michael Matz  <matz@suse.de>
20018         * builtins.c (compute_objsize): Pass correct operand
20019         to array_at_struct_end_p.
20021 2018-04-04  Richard Biener  <rguenther@suse.de>
20023         PR lto/85176
20024         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
20025         from contexts for DINFO_LEVEL_TERSE and below.
20027 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20029         * config/nds32/nds32-doubleword.md (move_<mode>): Require
20030         resiter_operand condition.
20031         * config/nds32/nds32.md (*move<mode>): Ditto.
20033 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20034             Monk Chiang  <sh.chiang04@gmail.com>
20036         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
20038 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20040         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
20042 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20043             Kito Cheng  <kito.cheng@gmail.com>
20045         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
20046         nds32_cond_code_str, output_cond_branch,
20047         output_cond_branch_compare_zero, nds32_expand_cbranch,
20048         nds32_expand_cstore, nds32_expand_movcc,
20049         nds32_output_cbranchsi4_equality_zero,
20050         nds32_output_cbranchsi4_equality_reg,
20051         nds32_output_cbranchsi4_equality_reg_or_const_int,
20052         nds32_output_cbranchsi4_greater_less_zero: New functions.
20053         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
20054         nds32_expand_cstore, nds32_expand_movcc,
20055         nds32_output_cbranchsi4_equality_zero,
20056         nds32_output_cbranchsi4_equality_reg,
20057         nds32_output_cbranchsi4_equality_reg_or_const_int,
20058         nds32_output_cbranchsi4_greater_less_zero): Declare.
20059         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
20060         nds32_rimm11s_operand): New predicates.
20061         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
20062         * config/nds32/nds32.md: Rewrite all the branch and conditional move
20063         patterns.
20065 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
20067         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
20068         * config/nds32/nds32.md: Ditto.
20069         * config/nds32/pipelines.md: Ditto.
20071 2018-04-04  Richard Biener  <rguenther@suse.de>
20073         PR tree-optimization/85168
20074         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
20075         propagating abnormals.
20077 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20079         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
20081 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
20082             Kito Cheng  <kito.cheng@gmail.com>
20084         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
20085         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
20086         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
20087         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
20088         * config/nds32/nds32.md (sibcall_internal): New.
20089         (sibcall_register): Remove.
20090         (sibcall_immediate): Remove.
20091         (sibcall_value_internal): New.
20092         (sibcall_value_register): Remove.
20093         (sibcall_value_immediate): Remove.
20094         * config/nds32/predicates.md (nds32_general_register_operand): New.
20095         (nds32_call_address_operand): New.
20097 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20099         PR rtl-optimization/85167
20100         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
20101         bb_defs if *split_p, instead preinitialize it to NULL.
20103         PR tree-optimization/85156
20104         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
20105         evaluating the argument multiple times.
20107 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
20109         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
20110         than vector.
20111         (_mm_cvtpd_ps): Likewise.
20112         (_mm_cvttpd_epi32): Likewise.
20113         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
20114         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
20115         vector, pixel, and bool following altivec.h include.
20117 2018-04-03  Martin Sebor  <msebor@redhat.com>
20119         * doc/extend.texi (Common Function Attributes): Clarify.
20120         (const attribute): Likewise.
20121         (pure attribute): Likewise.
20123 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
20125         PR target/85169
20126         * config/i386/i386.c (ix86_expand_vector_set): Use
20127         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
20129 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
20131         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
20132         instructions when changing rounding bits to preserve precision bits
20133         in the x87 control word.
20135 2018-04-03  Martin Liska  <mliska@suse.cz>
20137         PR tree-optimization/82491
20138         * rtl.h (strip_offset_and_add): Replace += suboffset with
20139         poly_uint64 () + suboffset.
20141 2018-03-29  Martin Liska  <mliska@suse.cz>
20142             Martin Jambor  <mjambor@suse.cz>
20144         PR ipa/84947
20145         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
20146         param_type is not an integral or pointer type.
20148 2018-04-03  Richard Biener  <rguenther@suse.de>
20150         * sese.h (recompute_all_dominators): Remove.
20152 2018-04-02  Martin Sebor  <msebor@redhat.com>
20154         * doc/invoke.texi (-Wrestrict): Fix typos.
20156 2018-04-02  Jim Wilson  <jimw@sifive.com>
20158         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
20159         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
20160         (<optab>di3, <optab>si3_extend): Likewise.
20161         (<optab>si3_mask, <optab>si3_mask_1): New.
20162         (<optab>di3_mask, <optab>di3_mask_1): New.
20163         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
20164         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
20165         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
20167 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
20169         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
20170         example.
20172 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
20174         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
20175         (nds32_canonicalize_comparison): New function.
20177 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20178             Kito Cheng  <kito.cheng@gmail.com>
20179             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
20181         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
20182         * config/nds32/constants.md (unspec_volatile_element): Add
20183         UNSPEC_VOLATILE_RELAX_GROUP.
20184         * config/nds32/nds32-relax-opt.c: New file.
20185         * config/nds32/nds32-predicates.c
20186         (nds32_symbol_load_store_p): New function.
20187         * config/nds32/nds32-protos.h
20188         (nds32_symbol_load_store_p): Declare function.
20189         (make_pass_nds32_relax_opt): Declare new rtl pass function.
20190         * config/nds32/nds32.c
20191         (nds32_register_pass): New function to register pass.
20192         (nds32_register_passes): New function to register passes.
20193         * config/nds32/nds32.md (relax_group): New pattern.
20194         * config/nds32/nds32.opt (mrelax-hint): New option.
20195         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
20197 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
20199         * config/nds32/t-nds32: Modify files dependency.
20201 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20203         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
20204         (PROFILE_HOOK): Define its implementation.
20206 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
20208         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
20209         type and 32-bit size.
20211 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
20213         PR middle-end/85090
20214         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
20215         (V_128_256): New mode iterator.
20216         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
20217         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
20218         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
20219         of V.
20220         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
20221         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
20223 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
20225         PR target/83315
20226         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
20227         NaN inputs correctly.
20229 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
20231         PR target/80546
20232         * config/rs6000/vsx.md (??r): New mode attribute.
20233         (*vsx_mov<mode>_64bit): Use it.
20234         (*vsx_mov<mode>_32bit): Likewise.
20236 2018-03-30  Martin Sebor  <msebor@redhat.com>
20238         PR tree-optimization/84818
20239         * builtins.c (check_access): Use warning_n.
20241 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20243         PR target/83822
20244         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
20245         condition.
20246         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
20247         condition.
20249 2018-03-30  Julia Koval  <julia.koval@intel.com>
20251         PR target/84413
20252         * x86-tune.def (movx, partial_reg_dependency): Enable for
20253         m_SKYLAKE_AVX512.
20255 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
20257         PR inline-asm/84985
20258         * lra-constraints.c (process_alt_operands): Move setting
20259         this_alternative_matches below.
20261 2018-03-29  Martin Liska  <mliska@suse.cz>
20263         PR lto/84995.
20264         * doc/invoke.texi: Document how LTO works with debug info.
20265         Describe auto-load support of binutils.  Mention 'x86-64'
20266         as valid option value of -march option.
20268 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
20270         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20272         PR c/85094
20273         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20274         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20275         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20276         checking.
20278 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20280         PR target/84912
20281         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20282         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20283         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20284         for RS6000_BTM_POWERPC64.
20285         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20286         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20287         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20288         definition.
20289         (DIVDE): Use it.
20290         (DIVDEU): Likewise.
20292 2018-03-28  Carl Love  <cel@us.ibm.com>
20294         Revert
20295         2017-09-27  Carl Love  <cel@us.ibm.com>
20297         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20298         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20299         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20300         fctiw instruction.
20302 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20304         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20305         instead of __vector bool.
20306         (_mm_max_pu8): Likewise.
20307         (_mm_min_pi16): Likewise.
20309 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20311         PR target/84912
20312         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20313         (DIVWEUO): Likewise.
20314         (DIVDEO): Likewise.
20315         (DIVDEUO): Likewise.
20316         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20317         DIVWEUO and DIVDEUO.
20318         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20319         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20320         (div_extend): Likewise.
20321         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20322         builtin function.
20323         (__builtin_divweuo): Likewise.
20324         (__builtin_divdeo): Likewise.
20325         (__builtin_divdeuo): Likewise.
20327 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20329         PR target/85095
20330         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20331         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20333         PR tree-optimization/82004
20334         * gimple-match-head.c (optimize_pow_to_exp): New function.
20335         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20336         Don't fold to exp if optimize_pow_to_exp is false.
20338 2018-03-28  Martin Liska  <mliska@suse.cz>
20340         PR other/84819
20341         * calls.c (initialize_argument_information): Fix trailing space.
20342         * common.opt: Fix typo and provide better explanation for
20343         -fsanitize-coverage option.
20344         * config/i386/i386.opt: Fix typo.
20346 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20347             Martin Liska  <mliska@suse.cz>
20349         PR sanitizer/85081
20350         * gimplify.c (asan_poison_variable): Don't do the check for
20351         gimplify_omp_ctxp here.
20352         (gimplify_decl_expr): Do it here.
20353         (gimplify_target_expr): Likewise.
20355 2018-03-28  Martin Liska  <mliska@suse.cz>
20357         PR target/84988
20358         * config/i386/i386.c (ix86_function_arg_advance): Do not call
20359         chkp_type_bounds_count if MPX is not enabled.
20361 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20363         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20365 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
20367         PR target/84914
20368         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20369         function to create the function decl for complex long double
20370         multiply and divide for -mabi=ieeelongdouble.
20371         (init_float128_ieee): Call it.
20373 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
20375         PR target/85044
20376         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20377         -fcf-protection=branch -mibt.
20378         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20380 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20382         PR target/81863
20383         * config/arm/arm.c (arm_valid_symbolic_address): Handle
20384         arm_word_relocations.
20386 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
20388         PR target/85056
20389         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20390         extern array declarations.
20392 2018-03-27  Richard Biener  <rguenther@suse.de>
20394         PR middle-end/84067
20395         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20396         explicit single_use checks.
20398 2018-03-27  Richard Biener  <rguenther@suse.de>
20400         PR tree-optimization/85082
20401         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20402         Valueize the VUSE.
20404 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20406         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20407         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20408         Turn on fasynchronous-unwind-tables and funwind-tables.
20410 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
20412         PR target/85073
20413         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20414         (*bmi_blsr_<mode>_ccz): Ditto.
20416 2018-03-26  Tom de Vries  <tom@codesourcery.com>
20418         PR tree-optimization/85063
20419         * omp-general.c (offloading_function_p): New function.  Factor out
20420         of ...
20421         * omp-offload.c (pass_omp_target_link::gate): ... here.
20422         * omp-general.h (offloading_function_p): Declare.
20423         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20424         with attribute omp declare target for offloading functions.
20426 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
20428         PR tree-optimization/84005
20429         * tree-data-ref.h (get_base_for_alignment): Declare.
20430         * tree-data-ref.c (get_base_for_alignment_1): New function.
20431         (get_base_for_alignment): Likewise.
20432         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20433         get_base_for_alignment to find a suitable base object, instead
20434         of always using drb->base_address.
20436 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20438         PR inline-asm/85022
20439         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
20440         known size by default.
20442 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
20444         PR inline-asm/85030
20445         * lra-constraints.c (process_alt_operands): Don't match BLKmode
20446         and non BLKmode operands.
20448 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20450         PR target/85026
20451         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
20452         Clean up attributes.
20454 2018-03-23  Richard Biener  <rguenther@suse.de>
20456         PR debug/85020
20457         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
20458         we are going to emit early debug for LTO.
20460 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20462         PR inline-asm/85034
20463         * function.c (match_asm_constraints_1): Don't optimize if input
20464         doesn't satisfy general_operand predicate for output's mode.
20466         PR inline-asm/85022
20467         * alias.c (write_dependence_p): Don't require for x_canonicalized
20468         non-VOIDmode if x has VOIDmode.
20470         PR sanitizer/85029
20471         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
20472         just don't try to optimize it rather than assert it never happens.
20474 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20476         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
20477         macro expansions for definition of ST_INTERNAL_<mode> and
20478         LD_INTERNAL_<mode> builtins.
20479         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
20480         Remove prototype.
20481         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
20482         function.
20483         (altivec_expand_st_builtin): Likewise.
20484         (altivec_expand_builtin): Remove calls to deleted functions.
20485         (rs6000_address_for_altivec): Delete this function.
20486         * config/rs6000/vector.md: Remove expands for
20487         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
20489 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
20491         PR target/84826
20492         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
20493         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
20494         re-computing once computed.
20495         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
20496         (arm_init_machine_status): Initialize
20497         machine->static_chain_stack_bytes.
20499 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20501         PR target/84760
20502         * doc/extend.texi: Add four new prototypes for vec_ld.
20503         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
20504         definitions for more logical presentation.
20505         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
20506         entries for V1TI variants of __builtin_altivec_ld builtin.
20507         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
20508         handling of V1TI variant of LVX icode pattern.
20509         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
20510         (rs6000_gimple_fold_builtin): Likewise.
20511         (altivec_init_builtins): Add code to define
20512         __builtin_altivec_lvx_v1ti function.
20514 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20516         PR inline-asm/84941
20517         * function.c (match_asm_constraints_1): Don't do the optimization
20518         if input isn't a REG, SUBREG, MEM or constant.
20520 2018-03-22  Tom de Vries  <tom@codesourcery.com>
20522         PR tree-optimization/84956
20523         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
20524         bb_has_abnormal_pred.
20526 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20528         PR sanitizer/85018
20529         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
20530         DECL_INITIAL (decl) to decl at the end.
20531         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
20532         adjust the comment.
20534 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
20536         * doc/extend.texi (__builtin_tgmath): Document when complex
20537         integer types are treated as _Complex _Float64.
20539 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20541         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
20543 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20545         PR tree-optimization/84960
20546         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
20547         if it is ENTRY block, move them into single succ of ENTRY in that case.
20549 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
20551         PR tree-optimization/84811
20552         * poly-int.h (poly_span_traits): Remove the T3 parameter and
20553         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
20554         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
20555         (known_subrange_p): Update accordingly.  Cast each value involved
20556         in the size comparison, rather than casting the result of the
20557         subtraction.
20559 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20561         PR tree-optimization/84982
20562         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
20563         by flipping the least significant bit rather than all bits from
20564         bitpos to bitpos + bitsize - 1.
20566 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20568         * doc/extend.texi (Deprecated Features): Remove mention of
20569         long-deleted deprecations.
20571 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20573         PR jit/84288
20574         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
20575         * configure: Regenerate.
20577 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20579         PR tree-optimization/83126
20580         * tree-parloops.c (num_phis): New function.
20581         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
20583 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20585         * doc/extend.texi (Deprecated Features): Update deprecated flags,
20586         mention anon-struct/union members and trailing attributes.
20588 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
20590         PR tree-optimization/84969
20591         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
20592         builtin memset partitions if they set different rhs values.
20594 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20596         PR rtl-optimization/84989
20597         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
20598         VEC_DUPLICATE with scalar result mode.
20600 2018-03-21  Martin Liska  <mliska@suse.cz>
20602         PR ipa/84963
20603         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
20604         not intended return statement.
20606 2018-03-21  Martin Liska  <mliska@suse.cz>
20608         PR target/84988
20609         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
20610         (chkp_find_bound_slots_1): Limit number of iterations.
20612 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
20614         PR target/84838
20615         * Minor grammar fixes for x86 options.
20617 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20619         PR debug/84875
20620         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
20621         holding REG_CFA_RESTORE notes, instead turn them into a USE.
20623 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
20625         PR target/83789
20626         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
20627         (altivec_lvx_<mode>_1op): Likewise.
20628         (altivec_stvx_<mode>_2op): Likewise.
20629         (altivec_stvx_<mode>_1op): Likewise.
20630         (altivec_lvx_<VM2:mode>): New define_expand.
20631         (altivec_stvx_<VM2:mode>): Likewise.
20632         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
20633         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20634         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
20635         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20636         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
20637         (rs6000_gen_lvx): Likewise.
20638         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
20639         (altivec_expand_stv_builtin): Likewise.
20640         (altivec_expand_builtin): Likewise.
20641         * config/rs6000/vector.md: Likewise.
20643 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20645         PR target/82518
20646         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
20647         BYTES_BIG_ENDIAN.
20649 2018-03-20  Richard Biener  <rguenther@suse.de>
20651         PR target/84986
20652         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
20653         sign-conversions as zero, fall back to standard scalar_stmt
20654         cost for the rest.
20656 2018-03-20  Martin Liska  <mliska@suse.cz>
20658         PR ipa/84825
20659         * predict.c (rebuild_frequencies): Handle case when we have
20660         PROFILE_ABSENT, but flag_guess_branch_prob is false.
20662 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20664         PR target/84990
20665         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
20666         flag_section_anchors.
20667         * varasm.c (use_blocks_for_decl_p): Remove hack for
20668         dw2_force_const_mem.
20670         PR target/84845
20671         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
20672         to ...
20673         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
20674         be created, use lowpart_subreg of operands[0] rather than operands[0]
20675         itself.
20676         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
20677         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
20678         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
20679         and n constraint instead of aarch64_shift_imm_di and Usd.
20680         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
20681         (*aarch64_<optab>_reg_minus<mode>3): ... this.
20683 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
20685         PR target/82989
20686         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
20687         to favor GPR over NEON registers.
20688         (<shift>di3_neon): Likewise.
20690 2018-03-20  Tom de Vries  <tom@codesourcery.com>
20692         PR target/84952
20693         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
20694         (nvptx_process_pars): Emit bar.sync asap and alap.
20696 2018-03-20  Tom de Vries  <tom@codesourcery.com>
20698         PR target/84954
20699         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
20700         seen_label if seen_label is already set.
20702 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20704         PR target/84945
20705         * config/i386/i386.c (fold_builtin_cpu): For features above 31
20706         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
20707         Use 1U instead of 1.  Formatting fixes.
20709         PR c/84953
20710         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
20711         instead of TREE_TYPE (s1) for the return value.
20713 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
20715         PR tree-optimization/84946
20716         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
20717         bitsize + bitsize in poly_uint64 rather than poly_int64.
20719         PR sanitizer/78651
20720         * dwarf2asm.c: Include fold-const.c.
20721         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
20722         of decl rather than decl itself.
20724         PR rtl-optimization/84643
20725         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
20727 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
20729         PR sanitizer/78651
20730         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
20731         calling assemble_variable.
20733 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
20735         PR target/81647
20736         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
20737         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
20739 2018-03-19  Jim Wilson  <jimw@sifive.com>
20741         PR bootstrap/84856
20742         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
20743         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
20744         (riscv_first_stack_step): Likewise.
20745         (riscv_option_override): Use STACK_BOUNDARY instead of
20746         MIN_STACK_BOUNDARY.
20747         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
20748         MIN_STACK_BOUNDARY.
20749         (BIGGEST_ALIGNMENT): Set to 128.
20750         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
20751         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
20752         STACK_BOUNDARY.
20754 2018-03-19  Richard Biener  <rguenther@suse.de>
20756         PR tree-optimization/84933
20757         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
20758         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
20760 2018-03-19  Richard Biener  <rguenther@suse.de>
20762         PR tree-optimization/84859
20763         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
20764         (cond_if_else_store_replacement): Perform sinking operation on
20765         single-store BBs regardless of MAX_STORES_TO_SINK setting.
20766         Generalize what a BB with a single eligible store is.
20768 2018-03-19  Richard Biener  <rguenther@suse.de>
20770         PR tree-optimization/84929
20771         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
20772         chrec_is_positive against non-chrec arg.
20774 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
20776         PR target/84711
20777         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
20779 2018-03-18  Martin Liska  <mliska@suse.cz>
20781         PR rtl-optimization/84635
20782         * regrename.c (build_def_use): Use matches_mode only when
20783         matches >= 0.
20785 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
20787         PR tree-optimization/84913
20788         * tree-vect-loop.c (vectorizable_reduction): Don't try to
20789         vectorize chains of COND_EXPRs.
20791 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20793         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
20795 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20797         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
20799 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20801         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
20803 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20804             Kito Cheng  <kito.cheng@gmail.com>
20806         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
20807         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
20808         (nds32_adjust_reg_alloc_order): New function.
20809         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
20811 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
20813         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
20814         nds32_print_operand, nds32_print_operand_address): Use
20815         HOST_WIDE_INT_PRINT_DEC instead.
20817 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20819         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
20821 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
20823         PR target/84902
20824         * config/i386/i386.c (initial_ix86_tune_features,
20825         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
20826         unsigned long long.
20827         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
20828         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
20829         rather than 1u << ix86_tune.  Formatting fix.
20830         (ix86_option_override_internal): Change ix86_arch_mask from
20831         unsigned int to unsigned HOST_WIDE_INT, initialize to
20832         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
20833         (ix86_function_specific_restore): Likewise.
20835 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20837         PR target/84899
20838         * postreload.c (reload_combine_recognize_pattern): Perform
20839         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
20840         truncate_int_for_mode the result for the destination's mode.
20842         PR c/84909
20843         * hsa-gen.c (mem_type_for_type): Fix comment typo.
20844         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
20845         Likewise.
20846         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20847         Likewise.
20849 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
20851         PR target/84876
20852         * lra-assigns.c (lra_split_hard_reg_for): Don't use
20853         regno_allocno_class_array and sorted_pseudos.
20854         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
20855         insns where regno is used.
20857 2018-03-16  Martin Liska  <mliska@suse.cz>
20859         PR ipa/84833
20860         * multiple_target.c (create_dispatcher_calls): Redirect
20861         reference in the symbol table.
20863 2018-03-16  Martin Liska  <mliska@suse.cz>
20865         PR ipa/84722
20866         * multiple_target.c (create_dispatcher_calls): Redirect also
20867         an alias.
20869 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20871         PR c++/79937
20872         PR c++/82410
20873         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
20874         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
20875         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
20877 2018-03-16  Julia Koval  <julia.koval@intel.com>
20879         * doc/invoke.texi (Skylake Server): Add CLWB.
20880         Cannonlake): Remove CLWB.
20882 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20884         PR tree-optimization/84841
20885         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
20886         1 << 3.
20887         (FLOAT_ONE_CONST_TYPE): Define.
20888         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
20889         (sort_by_operand_rank): Put entries with higher constant_type last
20890         rather than first to match comments.
20892 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
20894         * config/nios2/nios2.md (movsi_internal): Fix thinko in
20895         split predicate.
20897 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20899         PR c++/79085
20900         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
20901         check and use address of target always.
20903 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
20905         PR target/84574
20906         * config/i386/i386.c (indirect_thunk_needed): Update comments.
20907         (indirect_thunk_bnd_needed): Likewise.
20908         (indirect_thunks_used): Likewise.
20909         (indirect_thunks_bnd_used): Likewise.
20910         (indirect_return_needed): New.
20911         (indirect_return_bnd_needed): Likewise.
20912         (output_indirect_thunk_function): Add a bool argument for
20913         function return.
20914         (output_indirect_thunk_function): Don't generate alias for
20915         function return thunk.
20916         (ix86_code_end): Call output_indirect_thunk_function to generate
20917         function return thunks.
20918         (ix86_output_function_return): Set indirect_return_bnd_needed
20919         and indirect_return_needed instead of indirect_thunk_bnd_needed
20920         and indirect_thunk_needed.
20922 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
20924         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
20925         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
20926         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
20928 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
20929             Paul Hua <paul.hua.gm@gmail.com>
20931         PR c/84852
20932         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
20934 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
20936         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
20937         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
20938         resp. SFmode cases.
20940 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
20942         PR target/84711
20943         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
20944         instead of GET_MODE_SIZE when comparing Units.
20946 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
20948         PR target/68256
20949         * varasm.c (hash_section): Return an unchangeble hash value
20950         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
20951         Return !aarch64_can_use_per_function_literal_pools_p ().
20953 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20955         PR target/84860
20956         * optabs.c (emit_conditional_move): Pass address of cmode's copy
20957         rather than address of cmode as last argument to prepare_cmp_insn.
20959 2018-03-15  Julia Koval  <julia.koval@intel.com>
20961         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
20962         F_AVX512VNNI, F_AVX512BITALG): New.
20964 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
20966         PR target/83451
20967         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
20968         insn for floating-point loads and stores.
20970 2018-03-14  Carl Love  <cel@us.ibm.com>
20972         * config/rs6000/rs6000-c.c: Add macro definitions for
20973         ALTIVEC_BUILTIN_VEC_PERMXOR.
20974         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
20975         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
20976         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
20977         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
20978         UNSPEC_VPERMXOR.
20979         * config/doc/extend.texi: Add prototypes for vec_permxor.
20981 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
20983         PR c/84852
20984         * diagnostic-show-locus.c (class layout_point): Convert m_line
20985         from int to linenum_type.
20986         (line_span::comparator): Use linenum "compare" function when
20987         comparing line numbers.
20988         (test_line_span): New function.
20989         (layout_range::contains_point): Convert param "row" from int to
20990         linenum_type.
20991         (layout_range::intersects_line_p): Likewise.
20992         (layout::will_show_line_p): Likewise.
20993         (layout::print_source_line): Likewise.
20994         (layout::should_print_annotation_line_p): Likewise.
20995         (layout::print_annotation_line): Likewise.
20996         (layout::print_leading_fixits): Likewise.
20997         (layout::annotation_line_showed_range_p): Likewise.
20998         (struct line_corrections): Likewise for field m_row.
20999         (line_corrections::line_corrections): Likewise for param "row".
21000         (layout::print_trailing_fixits): Likewise.
21001         (layout::get_state_at_point): Likewise.
21002         (layout::get_x_bound_for_row): Likewise.
21003         (layout::print_line): Likewise.
21004         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
21005         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
21006         * input.c (selftest::test_linenum_comparisons): New function.
21007         (selftest::input_c_tests): Call it.
21008         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
21009         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
21010         * selftest.h (ASSERT_GT): New macro.
21011         (ASSERT_GT_AT): New macro.
21012         (ASSERT_LT): New macro.
21013         (ASSERT_LT_AT): New macro.
21015 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
21017         PR rtl-optimization/84780
21018         * combine.c (distribute_links): Don't make a link based on pc_rtx.
21020 2018-03-14  Martin Liska  <mliska@suse.cz>
21022         * tree.c (record_node_allocation_statistics): Use
21023         get_stats_node_kind.
21024         (get_stats_node_kind): New function extracted from
21025         record_node_allocation_statistics.
21026         (free_node): Use get_stats_node_kind.
21028 2018-03-14  Richard Biener  <rguenther@suse.de>
21030         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
21031         that the value-set of ANTIC_IN doesn't grow.
21033         Revert
21034         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21035         member.
21036         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21037         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21039 2018-03-14  Julia Koval  <julia.koval@intel.com>
21041         * config.gcc (icelake-client, icelake-server): New.
21042         (icelake): Remove.
21043         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
21044         (initial_ix86_arch_features): Ditto.
21045         (PTA_SKYLAKE): Add SGX.
21046         (PTA_ICELAKE): Remove.
21047         (PTA_ICELAKE_CLIENT): New.
21048         (PTA_ICELAKE_SERVER): New.
21049         (ix86_option_override_internal): Split up icelake on icelake client and
21050         icelake server.
21051         (get_builtin_code_for_version): Ditto.
21052         (fold_builtin_cpu): Ditto.
21053         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
21054         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
21055         * config/i386/i386.h (processor_type): Ditto.
21056         * doc/invoke.texi: Ditto.
21058 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
21060         PR sanitizer/83392
21061         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
21062         INTEGER_CST offset, add it together with bitpos / 8 and
21063         sign extend based on POINTER_SIZE.
21065         PR target/84844
21066         Revert
21067         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
21069         PR target/78090
21070         * config/i386/constraints.md (Yc): New register constraint.
21071         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
21072         Use Yc constraint for alternative 2 of operand 0.  Remove
21073         preferred_for_speed attribute.
21075 2018-03-14  Richard Biener  <rguenther@suse.de>
21077         PR tree-optimization/84830
21078         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
21079         with the old one to avoid oscillations.
21081 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
21083         PR target/83712
21084         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
21085         pseudos.
21086         (assign_by_spills): Return a flag of reload assignment failure.
21087         Do not process the reload assignment failures.  Do not spill other
21088         reload pseudos if they has the same reg class.  Update n if
21089         necessary.
21090         (lra_assign): Add a return arg.  Set up from the result of
21091         assign_by_spills call.
21092         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21093         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21094         usage_insns if it is not NULL.
21095         (spill_hard_reg_in_range): New function.
21096         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21097         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21098         function prototypes.
21099         (lra_assign): Change prototype.
21100         * lra.c (lra): Add code to deal with fails by splitting hard reg
21101         live ranges.
21103 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
21105         * config/riscv/riscv.opt (mrelax): New option.
21106         * config/riscv/riscv.c (riscv_file_start): Emit ".option
21107         "norelax" when riscv_mrelax is disabled.
21108         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
21110 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21112         PR target/84743
21113         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
21114         reassociation for int modes.
21116 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21118         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
21119         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
21120         for big-endian.
21121         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
21122         * config/aarch64/aarch64-sve.md
21123         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
21124         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
21125         (*extend<mode><Vwide>2): Rename to...
21126         (aarch64_sve_extend<mode><Vwide>2): ...this.
21127         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
21128         renaming the old pattern to...
21129         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
21130         unsigned packs.
21131         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
21132         define_expand, renaming the old pattern to...
21133         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
21134         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
21135         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
21136         account when deciding which SVE instruction the optab should use.
21137         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
21139 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21141         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
21142         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
21143         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
21144         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
21145         (tlsdesc_small_<mode>): Turn a define_expand and use
21146         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
21147         (tlsdesc_small_advsimd_<mode>): ...this.
21148         (tlsdesc_small_sve_<mode>): New pattern.
21150 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
21152         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
21153         (UNSPEC_UMUL_HIGHPART): New constants.
21154         (MUL_HIGHPART): New int iteraor.
21155         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
21156         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
21157         define_expand.
21158         (*<su>mul<mode>3_highpart): New define_insn.
21160 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21162         PR lto/84805
21163         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21164         incomplete types.
21166 2018-03-13  Martin Liska  <mliska@suse.cz>
21168         PR ipa/84658.
21169         * (sem_item_optimizer::sem_item_optimizer): Initialize new
21170         vector.
21171         (sem_item_optimizer::~sem_item_optimizer): Release it.
21172         (sem_item_optimizer::merge_classes): Register variable aliases.
21173         (sem_item_optimizer::fixup_pt_set): New function.
21174         (sem_item_optimizer::fixup_points_to_sets): Likewise.
21175         * ipa-icf.h: Declare new variables and functions.
21177 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
21179         PR middle-end/84834
21180         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
21181         integer_pow2p@2 and test integer_pow2p in condition.
21182         (A < 0 ? C : 0): Similarly for @1.
21184         PR middle-end/84831
21185         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
21186         characters starting at p contain '\0' character, don't look beyond
21187         that.
21189         PR target/84827
21190         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
21191         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
21193         PR target/84828
21194         * reg-stack.c (change_stack): Change update_end var from int to
21195         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
21196         also call set_block_for_insn on the newly added insns and rescan.
21198         PR target/84786
21199         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
21200         on the last operand.
21202         PR c++/84704
21203         * tree.c (stabilize_reference_1): Return save_expr (e) for
21204         STATEMENT_LIST even if it doesn't have side-effects.
21206 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
21208         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
21210 2018-03-12  Renlin Li  <renlin.li@arm.com>
21212         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
21213         aarch64_output_scalar_simd_mov_immediate.
21215 2018-03-12  Martin Sebor  <msebor@redhat.com>
21217         PR tree-optimization/83456
21218         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
21219         for perfectly overlapping calls to memcpy.
21220         (gimple_fold_builtin_memory_chk): Same.
21221         (gimple_fold_builtin_strcpy): Handle no-warning.
21222         (gimple_fold_builtin_stxcpy_chk): Same.
21223         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
21225 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21227         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
21228         parameter.  Use it for SFmode.
21229         (rs6000_function_arg_advance_1): Adjust.
21230         (rs6000_function_arg): Adjust.
21231         (rs6000_gimplify_va_arg): Pass false for that new parameter.
21233 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
21235         PR rtl-optimization/84169
21236         PR rtl-optimization/84780
21237         * combine.c (can_combine_p): Check for a 2-insn combination whether
21238         the destination register is used between the two insns, too.
21240 2018-03-12  Richard Biener  <rguenther@suse.de>
21242         PR tree-optimization/84803
21243         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
21244         for refs DR analysis didn't process.
21246 2018-03-12  Richard Biener  <rguenther@suse.de>
21248         PR tree-optimization/84777
21249         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
21250         force-vectorize loops ignore whether we are optimizing for size.
21252 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
21254         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
21255         (TARGET_MD_ASM_ADJUST): Define.
21257 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
21258             Kito Cheng  <kito.cheng@gmail.com>
21259             Chung-Ju Wu  <jasonwucj@gmail.com>
21261         * config/nds32/nds32.c (nds32_compute_stack_frame,
21262         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21263         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21264         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21265         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21266         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21267         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21268         * config/nds32/nds32.md (prologue, epilogue): Use macro
21269         NDS32_V3PUSH_AVAILABLE_P to do checking.
21271 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
21273         PR debug/58150
21274         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21275         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21276         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
21277         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
21278         addition of most attributes on !orig_type_die or the attribute not
21279         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21281 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21282             Chung-Ju Wu  <jasonwucj@gmail.com>
21284         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21285         __NDS32_VH__ macro.
21286         * config/nds32/nds32.opt (mvh): New option.
21288 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21289             Chung-Ju Wu  <jasonwucj@gmail.com>
21291         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21292         function.
21293         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21294         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21295         definition.
21297 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21298             Chung-Ju Wu  <jasonwucj@gmail.com>
21300         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21301         function.
21302         * config/nds32/nds32-multiple.md (strlensi): New pattern.
21303         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21305 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
21306             Kito Cheng  <kito.cheng@gmail.com>
21307             Chung-Ju Wu  <jasonwucj@gmail.com>
21309         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21310         UNSPEC_FFMISM and UNSPEC_FLMISM.
21311         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21312         for ffb, ffmism and flmism.
21313         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21314         (unspec_ffmism): Ditto.
21315         (unspec_flmism): Ditto.
21316         (nds32_expand_builtin_impl): Check if string extension is available.
21317         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21318         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21320 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
21322         Reverting patch:
21323         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21325         PR target/83712
21326         * lra-assigns.c (assign_by_spills): Return a flag of reload
21327         assignment failure.  Do not process the reload assignment
21328         failures.  Do not spill other reload pseudos if they has the same
21329         reg class.
21330         (lra_assign): Add a return arg.  Set up from the result of
21331         assign_by_spills call.
21332         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21333         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21334         usage_insns if it is not NULL.
21335         (spill_hard_reg_in_range): New function.
21336         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21337         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21338         function prototypes.
21339         (lra_assign): Change prototype.
21340         * lra.c (lra): Add code to deal with fails by splitting hard reg
21341         live ranges.
21343 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
21345         PR target/84807
21346         * config/i386/i386.opt: Replace Enforcment with Enforcement.
21348 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
21350         PR debug/84620
21351         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21352         (dw_val_node): Add val_symbolic_view.
21353         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21354         (symview_upper_bound): New.
21355         (new_line_info_table): Initialize symviews_since_reset.
21356         (dwarf2out_source_line): Count symviews_since_reset and set
21357         symview_upper_bound.
21358         (dw_val_equal_p): Handle symview.
21359         (add_AT_symview): New.
21360         (print_dw_val): Handle symview.
21361         (attr_checksum, attr_checksum_ordered): Likewise.
21362         (same_dw_val_p, size_of_die): Likewise.
21363         (value_format, output_die): Likewise.
21364         (add_high_low_attributes): Use add_AT_symview for entry_view.
21365         (dwarf2out_finish): Reset symview_upper_bound, clear
21366         zero_view_p.
21368 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
21370         PR target/83969
21371         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21372         Add strict argument and use it.
21373         (rs6000_split_multireg_move): Update for new strict argument.
21374         (mem_operand_gpr): Disallow all non-offsettable addresses.
21375         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21377 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
21379         PR target/84772
21380         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21381         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21382         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21384         PR c++/84767
21385         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21386         decl, use remap_type if we want to use the type.
21388 2018-03-09  Martin Sebor  <msebor@redhat.com>
21390         PR tree-optimization/84526
21391         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21392         Remove dead code.
21393         (builtin_access::generic_overlap): Be prepared to handle non-array
21394         base objects.
21396 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
21398         PR rtl-optimization/84682
21399         * lra-constraints.c (process_address_1): Check is_address flag
21400         for address constraints.
21401         (process_alt_operands): Likewise.
21402         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21403         preprocess_constraints.
21404         * recog.h (preprocess_constraints): Add oploc parameter.
21405         Adjust callers.
21406         * recog.c (preprocess_constraints): Test address_operand for
21407         CT_ADDRESS constraints.
21409 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21411         PR target/83712
21412         * lra-assigns.c (assign_by_spills): Return a flag of reload
21413         assignment failure.  Do not process the reload assignment
21414         failures.  Do not spill other reload pseudos if they has the same
21415         reg class.
21416         (lra_assign): Add a return arg.  Set up from the result of
21417         assign_by_spills call.
21418         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21419         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21420         usage_insns if it is not NULL.
21421         (spill_hard_reg_in_range): New function.
21422         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21423         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21424         function prototypes.
21425         (lra_assign): Change prototype.
21426         * lra.c (lra): Add code to deal with fails by splitting hard reg
21427         live ranges.
21429 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21431         PR target/83193
21432         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21433         Accept complain bool parameter.  Only emit errors if it is true.
21434         (arm_parse_cpu_option_name): Likewise.
21435         (arm_target_thumb_only): Adjust callers of the above.
21436         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21437         prototype to take a default true bool parameter.
21438         (arm_parse_arch_option_name): Likewise.
21440 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
21441             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
21443         PR jit/64089
21444         PR jit/84288
21445         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
21446         * configure: Regenerate.
21447         * configure.ac ("linker --version-script option"): New.
21448         ("linker soname option"): New.
21450 2018-03-09  Richard Biener  <rguenther@suse.de>
21452         PR tree-optimization/84775
21453         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
21454         immediate uses of predicate stmts and mark them modified.
21456         Revert
21457         PR tree-optimization/84178
21458         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21459         to caller.
21460         (version_loop_for_if_conversion): Delay update_ssa call.
21461         (tree_if_conversion): Delay update_ssa until after predicate
21462         insertion.
21464 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
21466         PR target/84763
21467         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
21468         when the function accesses prior frames.
21470 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21472         PR debug/84456
21473         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
21474         gen_llsym, otherwise call maybe_gen_llsym.
21476         PR inline-asm/84742
21477         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
21478         has ',' character inside of it.
21480 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21482         PR target/84748
21483         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21484         as clobbering CC_REGNUM.
21486 2018-03-08  Richard Biener  <rguenther@suse.de>
21488         PR middle-end/84552
21489         * tree-scalar-evolution.c: Include tree-into-ssa.h.
21490         (follow_copies_to_constant): Do not follow SSA names registered
21491         for update.
21493 2018-03-08  Richard Biener  <rguenther@suse.de>
21495         PR tree-optimization/84178
21496         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21497         to caller.
21498         (version_loop_for_if_conversion): Delay update_ssa call.
21499         (tree_if_conversion): Delay update_ssa until after predicate
21500         insertion.
21502 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
21504         PR tree-optimization/84178
21505         * tree-if-conv.c (release_bb_predicate): Remove the
21506         the assertion that the stmts have NULL use_ops.
21507         Discard the statements, asserting that they haven't
21508         yet been added to a BB.
21510 2018-03-08  Richard Biener  <rguenther@suse.de>
21512         PR tree-optimization/84746
21513         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
21514         (phi_translate): Pass in destination ANTIC_OUT set.
21515         (phi_translate_1): Likewise.  For a simplified result lookup
21516         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
21517         (phi_translate_set): Adjust.
21518         (do_pre_regular_insertion): Likewise.
21519         (do_pre_partial_partial_insertion): Likewise.
21521 2018-03-08  Martin Liska  <mliska@suse.cz>
21523         PR gcov-profile/84735
21524         * doc/gcov.texi: Document usage of profile files.
21525         * gcov-io.h: Document changes in the format.
21527 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
21529         PR debug/84404
21530         PR debug/84408
21531         * dwarf2out.c (struct dw_line_info_table): Update comments for
21532         view == -1.
21533         (FORCE_RESET_NEXT_VIEW): New.
21534         (FORCE_RESETTING_VIEW_P): New.
21535         (RESETTING_VIEW_P): Check for -1 too.
21536         (ZERO_VIEW_P): Likewise.
21537         (new_line_info_table): Force-reset next view.
21538         (dwarf2out_begin_function): Likewise.
21539         (dwarf2out_source_line): Simplify zero_view_p initialization.
21540         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
21541         view directly.  Omit view when omitting .loc at line 0.
21543 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21545         PR tree-optimization/84740
21546         * tree-switch-conversion.c (process_switch): Call build_constructors
21547         only if info.phi_count is non-zero.
21549         PR tree-optimization/84739
21550         * tree-tailcall.c (find_tail_calls): Check call arguments against
21551         DECL_ARGUMENTS (current_function_decl) rather than
21552         DECL_ARGUMENTS (func) when checking for tail recursion.
21554 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21556         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
21557         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
21558         Volker Reichelt's entry and add entries for people that perform
21559         GCC fuzzy testing and report numerous bugs.
21561 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
21563         PR target/82411
21564         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
21565         readonly data in sdata, if that is disabled.
21566         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
21567         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
21568         -mreadonly-in-sdata option.
21570 2018-03-07  Martin Sebor  <msebor@redhat.com>
21572         PR tree-optimization/84468
21573         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
21574         basic block when looking for nul assignment.
21576 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
21578         PR target/84277
21579         * except.h (output_function_exception_table): Adjust prototype.
21580         * except.c (output_function_exception_table): Remove FNNAME parameter
21581         and add SECTION parameter.  Ouput one part of the table at a time.
21582         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
21583         the first part of the exception table and emit unwind directives.
21584         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
21585         (i386_pe_seh_cold_init): Likewise.
21586         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
21587         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
21588         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
21589         (ix86_output_call_insn): Emit a nop in one more case for SEH.
21590         * config/i386/winnt.c: Include except.h.
21591         (struct seh_frame_state): Add reg_offset, after_prologue and
21592         in_cold_section fields.
21593         (i386_pe_seh_end_prologue): Set seh->after_prologue.
21594         (i386_pe_seh_cold_init): New function.
21595         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
21596         to seh->in_cold_section.
21597         (seh_emit_push): Record the offset of the push.
21598         (seh_emit_save): Record the offet of the save.
21599         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
21600         Test seh->after_prologue to disregard the epilogue.
21601         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
21602         (i386_pe_end_cold_function): New function.
21604 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21606         PR fortran/84565
21607         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
21608         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
21610         PR c++/84704
21611         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
21612         on tmp_var.
21613         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
21614         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
21616         PR middle-end/84723
21617         * multiple_target.c: Include tree-inline.h and intl.h.
21618         (expand_target_clones): Diagnose and fail if node->definition and
21619         !tree_versionable_function_p (node->decl).
21621 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
21623         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
21624         sprint_ul.
21625         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
21626         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
21627         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
21629 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
21631         PR target/84710
21632         * combine.c (try_combine): Use reg_or_subregno instead of handling
21633         just paradoxical SUBREGs and REGs.
21635 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
21637         * config/arc/arc.c (arc_finalize_pic): Remove function.
21638         (arc_must_save_register): We use single base PIC register, remove
21639         checks to save/restore the PIC register.
21640         (arc_expand_prologue): Likewise.
21641         * config/arc/arc-protos.h (arc_set_default_type_attributes):
21642         Remove.
21643         (arc_verify_short): Likewise.
21644         (arc_attr_type): Likewise.
21645         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
21646         (walk_stores): Likewise.
21647         (arc_address_cost): Make it static.
21648         (arc_verify_short): Likewise.
21649         (branch_dest): Likewise.
21650         (arc_attr_type): Likewise.
21651         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
21652         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
21653         (arc_final_prescan_insn): Remove inserting the nops due to
21654         hardware hazards.  It is done in reorg step.
21655         (insn_length_variant_t): Remove.
21656         (insn_length_parameters_t): Likewise.
21657         (arc_insn_length_parameters): Likewise.
21658         (arc_get_insn_variants): Likewise.
21659         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
21661 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
21663         PR inline-asm/84683
21664         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
21665         assertion failure.
21667         PR tree-optimization/84687
21668         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
21669         on new_node->decl.
21670         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
21672 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21674         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
21675         Rename to ppc_speculation_barrier.
21676         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
21677         __builtin_ppc_speculation_barrier.
21679 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
21681         PR target/84700
21682         * combine.c (combine_simplify_rtx): Don't try to simplify if
21683         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
21684         are equal to x.
21686 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
21688         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
21689         to 32 bytes when compiling for POWER9.
21691 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
21693         PR target/84564
21694         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
21695         regparm >= 3 with no arg reg available also for calls with
21696         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
21698         PR target/84524
21699         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
21700         orig,vex.
21701         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
21703 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
21705         PR target/84264
21706         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
21708 2018-03-05  Richard Biener  <rguenther@suse.de>
21710         PR tree-optimization/84486
21711         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
21712         When inserting a __builtin_assume_aligned call set the LHS
21713         SSA name alignment info accordingly.
21715 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
21717         PR tree-optimization/84114
21718         * config/aarch64/aarch64.c (aarch64_reassociation_width)
21719         Avoid reassociation of FLOAT_MODE addition.
21721 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
21723         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
21724         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
21725         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
21726         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
21727         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
21728         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
21729         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
21730         and -mwbnoinvd.
21731         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
21732         __builtin_ia32_wbinvd): New builtins.
21733         (SPECIAL_ARGS2): New.
21734         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
21735         (SPECIAL_ARGS2): New.
21736         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
21737         (ix86_valid_target_attribute_inner_p): Ditto.
21738         (ix86_init_mmx_sse_builtins): Add special_args2.
21739         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
21740         TARGET_WBNOINVD_P): New.
21741         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
21742         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
21743         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
21744         * config/i386/immintrin.h (_wbinvd): New intrinsic.
21745         * config/i386/pconfigintrin.h: New file.
21746         * config/i386/wbnoinvdintrin.h: Ditto.
21747         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
21748         wbnoinvdintrin.h.
21749         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
21751 2018-03-05  Richard Biener  <rguenther@suse.de>
21753         PR tree-optimization/84670
21754         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21755         member.
21756         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21757         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21758         (compute_antic_aux): Only assert the number of values in ANTIC_IN
21759         doesn't grow if all successors (recursively) were visited at least
21760         once.
21762 2018-03-05  Richard Biener  <rguenther@suse.de>
21764         PR tree-optimization/84650
21765         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
21766         if executed in the loop pipeline.
21768 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
21770         * doc/configfiles.texi (Configuration Files): Move info about
21771         conditionalizing $target-protos.h to...
21772         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
21773         differs from $target-protos.h.
21775 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
21776             Chung-Ju Wu  <jasonwucj@gmail.com>
21778         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
21779         * config/nds32/nds32-multiple.md (setmemsi): Define.
21780         * config/nds32/nds32-memory-manipulation.c
21781         (nds32_gen_dup_4_byte_to_word_value): New.
21782         (emit_setmem_word_loop): New.
21783         (emit_setmem_byte_loop): New.
21784         (nds32_expand_setmem_loop): New.
21785         (nds32_expand_setmem_loop_v3m): New.
21786         (nds32_expand_setmem_unroll): New.
21787         (nds32_expand_setmem): New.
21789 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21790             Chung-Ju Wu  <jasonwucj@gmail.com>
21792         * config/nds32/nds32-memory-manipulation.c
21793         (nds32_emit_load_store): New.
21794         (nds32_emit_post_inc_load_store): New.
21795         (nds32_emit_mem_move): New.
21796         (nds32_emit_mem_move_block): New.
21797         (nds32_expand_movmemsi_loop_unknown_size): New.
21798         (nds32_expand_movmemsi_loop_known_size): New.
21799         (nds32_expand_movmemsi_loop): New.
21800         (nds32_expand_movmemsi_unroll): New.
21801         (nds32_expand_movmemqi): Rename ...
21802         (nds32_expand_movmemsi): ... to this.
21803         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
21804         (movmemsi): ... to this.
21805         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
21806         (nds32_expand_movmemsi): ... to this.
21808 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21809             Monk Chiang  <sh.chiang04@gmail.com>
21810             Chung-Ju Wu  <jasonwucj@gmail.com>
21812         * config/nds32/nds32-protos.h
21813         (nds32_expand_load_multiple): New arguments.
21814         (nds32_expand_store_multiple): Ditto.
21815         (nds32_valid_multiple_load_store): Rename ...
21816         (nds32_valid_multiple_load_store_p): ... to this.
21817         * config/nds32/nds32-memory-manipulation.c
21818         (nds32_expand_load_multiple): Refine implementation.
21819         (nds32_expand_store_multiple): Ditto.
21820         * config/nds32/nds32-multiple.md
21821         (load_multiple): Update nds32_expand_load_multiple interface.
21822         (store_multiple): Update nds32_expand_store_multiple interface.
21823         * config/nds32/nds32-predicates.c
21824         (nds32_valid_multiple_load_store): Rename ...
21825         (nds32_valid_multiple_load_store_p): ... to this and refine
21826         implementation.
21827         * config/nds32/predicates.md
21828         (nds32_load_multiple_and_update_address_operation): New predicate.
21829         (nds32_store_multiple_and_update_address_operation): New predicate.
21831 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21832             Chung-Ju Wu  <jasonwucj@gmail.com>
21834         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
21835         (combo): New attribute.
21836         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
21838 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
21840         * config/nds32/nds32.opt: Change -mcmodel= default value.
21842 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
21843             Monk Chiang  <sh.chiang04@gmail.com>
21844             Chung-Ju Wu  <jasonwucj@gmail.com>
21846         * config/nds32/constants.md (unspec_element): New enum.
21847         * config/nds32/constraints.md (Umw): New constraint.
21848         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
21849         * config/nds32/nds32-intrinsic.md: Likewise.
21850         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
21851         (nds32_valid_smw_lwm_base_p): New.
21852         (nds32_output_smw_single_word): New.
21853         (nds32_output_lmw_single_word): New.
21854         (nds32_expand_unaligned_load): New.
21855         (nds32_expand_unaligned_store): New.
21856         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
21857         (nds32_output_smw_single_word): Declare.
21858         (nds32_output_lmw_single_word): Declare.
21859         (nds32_expand_unaligned_load): Declare.
21860         (nds32_expand_unaligned_store): Declare.
21861         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
21862         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
21863         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
21864         NDS32_BUILTIN_UASTORE_DW.
21865         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
21866         predicate.
21868 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
21869             Kito Cheng  <kito.cheng@gmail.com>
21870             Chung-Ju Wu  <jasonwucj@gmail.com>
21872         * config/nds32/nds32-intrinsic.c
21873         (nds32_expand_builtin_null_ftype_reg): Delete.
21874         (nds32_expand_builtin_reg_ftype_imm): Ditto.
21875         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
21876         (nds32_read_argument): New.
21877         (nds32_legitimize_target): Ditto.
21878         (nds32_legitimize_argument): Ditto.
21879         (nds32_check_constant_argument): Ditto.
21880         (nds32_expand_unop_builtin): Ditto.
21881         (nds32_expand_unopimm_builtin): Ditto.
21882         (nds32_expand_binop_builtin): Ditto.
21883         (nds32_builtin_decl_impl): Ditto.
21884         (builtin_description): Ditto.
21885         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
21886         (nds32_init_builtins_impl): Ditto.
21887         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
21888         (nds32_builtin_decl): New.
21889         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
21890         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
21892 2018-03-02  Jeff Law  <law@redhat.com>
21894         * reorg.c (stop_search_p): Handle DEBUG_INSN.
21895         (redundant_insn, fill_simple_delay_slots): Likewise.
21896         (fill_slots_from_thread): Likewise.
21897         * resource.c (mark_referenced_resources): Likewise.
21898         (mark_set_resources, find_dead_or_set_registers): Likewise.
21900 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21902         * substring-locations.h (format_warning_va): Formatting fix for
21903         ATTRIBUTE_GCC_DIAG.
21904         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
21905         argument.
21906         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
21907         * substring-locations.c: Include intl.h.
21908         (format_warning_va): Turned into small wrapper around
21909         format_warning_n_va, renamed to ...
21910         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
21911         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
21912         use ngettext.
21913         (format_warning_at_substring_n): New function.
21914         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
21915         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
21916         format_warning_at_substring with just a shorter name instead of
21917         const function pointer.
21918         (fmtwarn_n): New function.
21919         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
21920         appropriate, get rid of all the fmtstr temporaries, move conditionals
21921         with G_() wrapped string literals directly into fmtwarn arguments,
21922         cast dir.len to (int), formatting fixes.
21924 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
21926         * doc/invoke.texi: Remove "Cilk Plus" references.
21928 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21929             Richard Biener  <rguenther@suse.de>
21931         PR ipa/84628
21932         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
21933         for error or warning attributes if CALL_FROM_THUNK_P is set.
21934         Formatting fixes.
21936 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21938         PR target/56540
21939         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
21940         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
21942         PR target/56540
21943         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
21944         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
21946         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
21947         instead of -1U in last predictors element's probability member.
21949 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21951         PR ipa/83983
21952         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21953         arguments if they are comparable.
21955 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21957         PR tree-optimization/84634
21958         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
21959         masks and masked_loop_p with a single loop_masks, making sure it's
21960         null for bb vectorization.
21962 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21964         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
21965         (vect_analyze_data_ref_access): Use loop->safe_len rather than
21966         loop->force_vectorize to check whether there is no alias.
21968 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21970         PR target/84614
21971         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
21972         prototypes.
21973         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21974         comments.
21975         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
21976         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
21977         instead of a loop around prev_real_insn.
21978         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
21979         prev_real_insn.
21981         PR inline-asm/84625
21982         * config/i386/i386.c (ix86_print_operand): Use conditional
21983         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
21984         zero vector.
21986 2018-03-02  Richard Biener  <rguenther@suse.de>
21988         PR tree-optimization/84427
21989         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
21990         (bitmap_set_subtract_values): Rewrite to handle multiple
21991         exprs per value.
21992         (clean): Likewise.
21993         (prune_clobbered_mems): Likewise.
21994         (phi_translate): Take edge instead of pred/phiblock.
21995         (phi_translate_1): Likewise.
21996         (phi_translate_set): Likewise.  Insert all translated
21997         exprs for a value into the set, keeping possibly multiple
21998         expressions per value.
21999         (compute_antic_aux): Adjust for phi_translate changes.
22000         When intersecting union the expressions and prune those
22001         not in the final value set, keeping possibly multiple
22002         expressions per value.  Do not use value-insertion
22003         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
22004         all expressions.  Add verification that the value-sets
22005         only shrink during iteration.
22006         (compute_partial_antic_aux): Adjust for the phi_translate changes.
22007         (do_pre_regular_insertion): Likewise.
22008         (do_pre_partial_partial_insertion): Likewise.
22010 2018-03-02  Richard Biener  <rguenther@suse.de>
22012         PR target/82005
22013         * config/darwin.c (saved_debug_info_level): New static global.
22014         (darwin_asm_lto_start): Disable debug info generation for LTO out.
22015         (darwin_asm_lto_end): Restore debug info generation settings.
22017 2018-03-01  Martin Liska  <mliska@suse.cz>
22019         PR sanitizer/82484
22020         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
22021         volatile arguments.
22023 2018-03-01  Richard Biener  <rguenther@suse.de>
22025         PR debug/84645
22026         * dwarf2out.c (gen_variable_die): Properly handle late VLA
22027         type annotation with LTO when debug was disabled at compile-time.
22029 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
22031         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
22032         XINT with INTVAL.
22033         (mips_final_postscan_insn): Likewise.
22035 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
22037         PR rtl-optimization/84528
22038         * alias.c (init_alias_target): Add commentary.
22039         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
22040         a unique base value if the frame pointer is not eliminated
22041         to the stack pointer.
22043 2018-03-01  Tom de Vries  <tom@codesourcery.com>
22045         PR rtl-optimization/83327
22046         * lra-int.h (hard_regs_spilled_into): Declare.
22047         * lra.c (hard_regs_spilled_into): Define.
22048         (init_reg_info): Init hard_regs_spilled_into.
22049         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
22050         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
22051         (process_bb_lives): Handle hard_regs_spilled_into.
22052         (lra_create_live_ranges_1): Before doing liveness propagation, clear
22053         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
22055 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
22057         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
22058         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
22059         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
22060         * config/rs6000/aix72.h: New file.
22062 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
22064         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
22065         instead of warning_at with conditional singular and plural messages
22066         where possible.
22068         PR target/52991
22069         * stor-layout.c (update_alignment_for_field): For
22070         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
22071         && !DECL_PACKED (field), do the alignment update, just use
22072         only desired_align instead of MAX (type_align, desired_align)
22073         as the alignment.
22074         (place_field): Don't do known_align < desired_align handling
22075         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
22076         is non-NULL, instead do it after rli->prev_field handling and
22077         only if not within a bitfield word.  For DECL_PACKED (field)
22078         use type_align of BITS_PER_UNIT.
22080 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
22082         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
22083         superfluous parentheses and trailing spaces.
22085 2018-02-28  Richard Biener  <rguenther@suse.de>
22087         PR tree-optimization/84584
22088         * graphite-scop-detection.c (scop_detection::add_scop): Discard
22089         SCoPs with fake exit edge.
22091 2018-02-28  Martin Liska  <mliska@suse.cz>
22093         PR testsuite/84597
22094         * timevar.c (timer::print): Fix format to properly print 100%
22095         values.
22097 2018-02-28  Richard Biener  <rguenther@suse.de>
22099         PR middle-end/84607
22100         * genmatch.c (capture_info::walk_match): Do not mark
22101         captured expressions without operands as expr_p given
22102         they act more like predicates and should be subject to
22103         "lost tail" side-effect preserving.
22105 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
22107         PR rtl-optimization/81611
22108         * auto-inc-dec.c (attempt_change): Move dead note from
22109         mem_insn if it's the next use of regno
22110         (find_address): Take address use of reg holding
22111         non-incremented value.  Add parm to limit search to the named
22112         reg only.
22113         (merge_in_block): Attempt to use a mem insn that is the next
22114         use of the original regno.
22116 2018-02-27  Martin Sebor  <msebor@redhat.com>
22118         PR c++/83871
22119         * doc/invoke.texi (-Wmissing-attributes): New option.
22120         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
22122 2018-02-27  Martin Sebor  <msebor@redhat.com>
22124         PR translation/84207
22125         * diagnostic-core.h (warning_n, error_n, inform_n): Change
22126         n argument to unsigned HOST_WIDE_INT.
22127         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
22128         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
22129         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
22130         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
22132 2018-02-27  Richard Biener  <rguenther@suse.de>
22134         PR tree-optimization/84512
22135         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22136         Do not use the estimate returned from record_stmt_cost for
22137         the scalar iteration cost but sum properly using add_stmt_cost.
22139 2018-02-27  Richard Biener  <rguenther@suse.de>
22141         PR tree-optimization/84466
22142         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22143         Adjust last change to less strictly validate use operands.
22145 2018-02-27  Martin Liska  <mliska@suse.cz>
22147         PR gcov-profile/84548
22148         * gcov.c (process_file): Allow partial overlap and consider it
22149         also as group functions.
22150         (output_lines): Properly calculate range of lines for a group.
22152 2018-02-27  Martin Liska  <mliska@suse.cz>
22154         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
22155         'ggc' suffixes.  Change first column width.
22156         (timer::print): Fix formatting of the column.
22158 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
22160         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
22161         preserve inline entry blocks for the sake of debug inline
22162         entry point markers alone.
22163         (remove_unused_locals): Suggest in comments a better place to
22164         force the preservation of inline entry blocks that are
22165         otherwise unused, but do not preserve them.
22167 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22169         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
22171 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22173         PR target/84039
22174         * config/i386/constraints.md (Bs): Replace
22175         ix86_indirect_branch_register with
22176         TARGET_INDIRECT_BRANCH_REGISTER.
22177         (Bw): Likewise.
22178         * config/i386/i386.md (indirect_jump): Likewise.
22179         (tablejump): Likewise.
22180         (*sibcall_memory): Likewise.
22181         (*sibcall_value_memory): Likewise.
22182         Peepholes of indirect call and jump via memory: Likewise.
22183         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
22184         (*sibcall_value_GOT_32): Likewise.
22185         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22186         (GOT_memory_operand): Likewise.
22187         (call_insn_operand): Likewise.
22188         (sibcall_insn_operand): Likewise.
22189         (GOT32_symbol_operand): Likewise.
22190         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
22192 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22194         PR rtl-optimization/83496
22195         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
22196         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
22197         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
22198         redundant insn, if any.
22199         (relax_delay_slots): Likewise.
22200         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
22202 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
22204         PR tree-optimization/83965
22205         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
22206         that grouped statements are part of a reduction chain.  Return
22207         true if the statement is not marked as a reduction itself but
22208         is part of a group.
22209         (vect_recog_dot_prod_pattern): Don't check whether the statement
22210         is part of a group here.
22211         (vect_recog_sad_pattern): Likewise.
22212         (vect_recog_widen_sum_pattern): Likewise.
22214 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
22216         PR debug/84545
22217         * final.c (rest_of_clean_state): Also look for calls inside sequences.
22219 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
22221         PR target/84530
22222         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
22223         the bool argument.
22224         (ix86_output_indirect_function_return): New prototype.
22225         (ix86_split_simple_return_pop_internal): Likewise.
22226         * config/i386/i386.c (indirect_return_via_cx): New.
22227         (indirect_return_via_cx_bnd): Likewise.
22228         (indirect_thunk_name): Handle return va CX_REG.
22229         (output_indirect_thunk_function): Create alias for
22230         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
22231         (ix86_output_indirect_jmp): Remove the bool argument.
22232         (ix86_output_indirect_function_return): New function.
22233         (ix86_split_simple_return_pop_internal): Likewise.
22234         * config/i386/i386.md (*indirect_jump): Don't pass false
22235         to ix86_output_indirect_jmp.
22236         (*tablejump_1): Likewise.
22237         (simple_return_pop_internal): Change it to define_insn_and_split.
22238         Call ix86_split_simple_return_pop_internal to split it for
22239         -mfunction-return=.
22240         (simple_return_indirect_internal): Call
22241         ix86_output_indirect_function_return instead of
22242         ix86_output_indirect_jmp.
22244 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
22246         PR bootstrap/84405
22247         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
22248         memset and value initialization afterwards.
22250 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
22252         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
22254 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
22256         PR target/84521
22257         * common/config/aarch64/aarch64-common.c
22258         (aarch_option_optimization_table[]): Switch
22259         off fomit-frame-pointer
22261 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22262             Chung-Ju Wu  <jasonwucj@gmail.com>
22264         * config/nds32/nds32-multiple.md (load_multiple): Disallow
22265         volatile memory.
22266         (store_multiple): Ditto.
22268 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22270         * config.gcc: Add --with-cpu support for nds32 target.
22271         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22272         * config/nds32/nds32.opt: Add -mcpu= option.
22274 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
22276         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22277         isel=yes): Warn for these deprecated options.
22279 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
22281         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22282         ISA_2_5_MASKS_EMBEDDED.
22284 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
22286         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22287         p->max as pointers rather than using iterative_hash_expr.
22289 2018-02-23  Carl Love  <cel@us.ibm.com>
22291         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22292         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22293         BU_P8V_OVERLOAD_2.
22294         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22295         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22296         P8V_BUILTIN_VEC_VUNSIGNED2.
22298 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
22300         PR target/81572
22301         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22302         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22303         LRA_UNKNOWN_ALT.
22304         * lra-constraints.c (curr_insn_transform): Set up
22305         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
22306         LRA_UNKNOWN_ALT.
22307         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22308         * lra-eliminations.c (spill_pseudos): Ditto.
22309         (process_insn_for_elimination): Ditto.
22310         * lra-lives.c (reg_early_clobber_p): Use the new macros.
22311         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22312         LRA_NON_CLOBBERED_ALT.
22314 2018-02-22  Martin Sebor  <msebor@redhat.com>
22316         PR tree-optimization/84480
22317         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22318         to maybe_diag_stxncpy_trunc.  Call it.
22319         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22320         from gimple_fold_builtin_strcpy.  Print inlining stack.
22321         (handle_builtin_stxncpy): Print inlining stack.
22322         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22324 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
22326         PR target/84176
22327         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22328         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22329         and -fcheck-pointer-bounds are used together.
22330         (indirect_thunk_prefix): New enum.
22331         (indirect_thunk_need_prefix): New function.
22332         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
22333         "_nt" instead of "_bnd" for NOTRACK prefix.
22334         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22335         (output_indirect_thunk_function): Likewise.
22336         (): Likewise.
22337         (ix86_code_end): Update output_indirect_thunk_function calls.
22338         (ix86_output_indirect_branch_via_reg): Replace
22339         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22340         (ix86_output_indirect_branch_via_push): Likewise.
22341         (ix86_output_function_return): Likewise.
22342         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22343         incompatible with -fcf-protection=branch and
22344         -fcheck-pointer-bounds.
22346 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22348         PR target/83335
22349         * config/aarch64/aarch64.c (aarch64_print_address_internal):
22350         Change gcc_assert call to output_operand_lossage.
22352 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22354         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22356 2018-02-22  DJ Delorie  <dj@redhat.com>
22357             Sebastian Perta  <sebastian.perta@renesas.com>
22358             Oleg Endo  <olegendo@gcc.gnu.org>
22360         * config/rx/rx.c (rx_rtx_costs): New function.
22361         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22363 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22365         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22367 2018-02-22  Martin Liska  <mliska@suse.cz>
22369         PR driver/83193
22370         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22371         Add "native" as a possible value.
22373 2018-02-22  Martin Liska  <mliska@suse.cz>
22375         PR driver/83193
22376         * config/i386/i386.c (ix86_option_override_internal):
22377         Add "native" as a possible value for -march and -mtune.
22379 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
22381         PR target/84502
22382         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22383         to all type variants.
22385         PR tree-optimization/84503
22386         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22387         width as info->bitpos + info->bitsize - start.
22388         (merged_store_group::merge_overlapping): Simplify width computation.
22389         (check_no_overlap): New function.
22390         (imm_store_chain_info::try_coalesce_bswap): Compute expected
22391         start + width and last_order of the group, fail if check_no_overlap
22392         fails.
22393         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22394         to group if check_no_overlap fails.
22396 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22398         * config/rs6000/altivec.md: Delete contraint arguments to
22399         define_expand, define_split, and define_peephole2, and in
22400         define_insn_and_split if always unused.
22401         * config/rs6000/darwin.md: Ditto.
22402         * config/rs6000/dfp.md: Ditto.
22403         * config/rs6000/rs6000.md: Ditto.
22404         * config/rs6000/sync.md: Ditto.
22405         * config/rs6000/vector.md: Ditto.
22406         * config/rs6000/vsx.md: Ditto.
22408 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22410         * config/rs6000/altivec.md: Write output control strings as braced
22411         blocks instead of double-quoted strings.
22412         * config/rs6000/darwin.md: Ditto.
22413         * config/rs6000/rs6000.md: Ditto.
22414         * config/rs6000/vector.md: Ditto.
22415         * config/rs6000/vsx.md: Ditto.
22417 2018-02-21  Jason Merrill  <jason@redhat.com>
22419         PR c++/84314 - ICE with templates and fastcall attribute.
22420         * attribs.c (build_type_attribute_qual_variant): Remove assert.
22422 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22424         * ipa-cp.c (determine_versionability): Fix comment typos.
22426 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22428         PR c/84229
22429         * ipa-cp.c (determine_versionability): Do not version functions caling
22430         va_arg_pack.
22432 2018-02-21  Martin Liska  <mliska@suse.cz>
22434         PR driver/83193
22435         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22436         Add "native" as a possible value.
22437         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
22438         the macro when native cpu detection is available.
22440 2018-02-21  Martin Liska  <mliska@suse.cz>
22442         PR driver/83193
22443         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
22444         Add "native" as a possible value.
22445         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
22446         when native cpu detection is available.
22448 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
22449             Martin Sebor  <msebor@redhat.com>
22451         PR tree-optimization/84478
22452         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22453         false.
22454         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
22455         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
22456         support which is conservatively correct, for 2 only stay conservative
22457         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
22458         argument to the 2 argument get_range_strlen, adjust 6 arg
22459         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
22460         false.
22461         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
22462         (gimple_fold_builtin_strlen): Pass true as last argument to
22463         get_range_strlen.
22465 2018-02-20  Martin Sebor  <msebor@redhat.com>
22467         PR middle-end/84095
22468         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
22469         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
22470         (builtin_memref::builtin_memref): Factor out parts into
22471         set_base_and_offset and call it.
22473 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22475         PR middle-end/84406
22476         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
22477         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
22478         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
22479         search at the associated MODE_INT.
22481 2018-02-20  Jeff Law  <law@redhat.com>
22483         PR middle-end/82123
22484         PR tree-optimization/81592
22485         PR middle-end/79257
22486         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
22487         for range data rather than using global data.
22488         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
22489         range data rather than using global data.
22490         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
22491         pass it to children as needed.
22492         (struct directive::fmtresult): Similarly.
22493         (struct directive::set_width): Similarly.
22494         (struct directive::set_precision): Similarly.
22495         (format_integer, format_directive, parse_directive): Similarly.
22496         (format_none): Accept unnamed vr_values parameter.
22497         (format_percent, format_floating, format_character): Similarly.
22498         (format_string, format_plain): Similarly.
22499         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
22500         the EVRP range analyzer for range data rather than using global data.
22501         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
22502         gimple-ssa-evrp-analyze.h
22503         (class sprintf_dom_walker): Add after_dom_children member function.
22504         Add evrp_range_analyzer member.
22505         (sprintf_dom_walker::before_dom_children): Call into the EVRP
22506         range analyzer as needed.
22507         (sprintf_dom_walker::after_dom_children): New member function.
22508         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
22509         if not optimizing.
22510         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
22511         (evrp_range_analyzer::pop_to_marker): Likewise.
22513 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22515         PR tree-optimization/84419
22516         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
22517         with the required type if its current type is compatible but
22518         different.
22520 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
22522         PR middle-end/82004
22523         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
22524         after vectorization.
22526 2018-02-20  Martin Liska  <mliska@suse.cz>
22528         PR driver/83193
22529         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
22530         possible values if we don't have a hint.
22532 2018-02-20  Martin Liska  <mliska@suse.cz>
22534         PR c/84310
22535         PR target/79747
22536         * final.c (shorten_branches): Build align_tab array with one
22537         more element.
22538         * opts.c (finish_options): Add alignment option limit check.
22539         (MAX_CODE_ALIGN): Likewise.
22540         (MAX_CODE_ALIGN_VALUE): Likewise.
22541         * doc/invoke.texi: Document maximum allowed option value for
22542         all -falign-* options.
22544 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
22546         PR target/84146
22547         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
22548         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
22549         * var-tracking.c (emit_note_insn_var_location): Remove all references
22550         to NOTE_INSN_CALL_ARG_LOCATION.
22551         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
22552         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
22553         Use copy_rtx_if_shared.
22554         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
22555         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
22556         (dwarf2out_var_location): Remove handling of
22557         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
22558         on call_insn.
22559         * final.c (final_scan_insn): Remove all references to
22560         NOTE_INSN_CALL_ARG_LOCATION.
22561         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
22562         before dumping final insns.
22563         * except.c (emit_note_eh_region_end): Remove all references to
22564         NOTE_INSN_CALL_ARG_LOCATION.
22565         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
22566         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
22567         * config/arc/arc.c (hwloop_optimize): Likewise.
22568         * config/arm/arm.c (create_fix_barrier): Likewise.
22569         * config/s390/s390.c (s390_chunkify_start): Likewise.
22570         * config/sh/sh.c (find_barrier): Likewise.
22571         * config/i386/i386.c (rest_of_insert_endbranch,
22572         ix86_seh_fixup_eh_fallthru): Likewise.
22573         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
22574         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
22575         * config/frv/frv.c (frv_function_prologue): Likewise.
22576         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
22577         reg note.
22578         (note_outside_basic_block_p): Remove all references to
22579         NOTE_INSN_CALL_ARG_LOCATION.
22580         * gengtype.c (adjust_field_rtx_def): Likewise.
22581         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
22582         Likewise.
22583         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
22584         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
22586         PR c++/84444
22587         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
22588         is ADDR_EXPR.
22590         PR tree-optimization/84452
22591         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
22592         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
22593         is NULL.
22595 2018-02-19  Martin Liska  <mliska@suse.cz>
22597         PR sanitizer/82183
22598         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
22600 2018-02-19  Martin Liska  <mliska@suse.cz>
22601             Richard Sandiford  <richard.sandiford@linaro.org>
22603         PR tree-optimization/82491
22604         * gimple-fold.c (get_base_constructor): Make earlier bail out
22605         to prevent ubsan.
22607 2018-02-19  Carl Love  <cel@us.ibm.com>
22609         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
22610         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
22611         BU_P8V_OVERLOAD_1.
22612         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
22613         P8V_BUILTIN_VEC_NEG.
22615 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
22617         * config/rl78/rl78.md (movdf): New define expand.
22619 2018-02-19  Martin Liska  <mliska@suse.cz>
22621         PR other/80589
22622         * doc/invoke.texi: Fix typo.
22623         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
22625 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
22627         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
22628         handle rs6000_single_float and rs6000_double_float specially for
22629         e500 family CPUs.
22631 2018-02-16  Jeff Law  <law@redhat.com>
22633         * config/rx/rx.c (add_pop_cfi_notes): New function.;
22634         (pop_regs): Use it.
22636 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22638         PR ipa/84425
22639         * ipa-inline.c (inline_small_functions): Fix a typo.
22641 2018-02-16  Nathan Sidwell  <nathan@acm.org>
22643         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
22645 2018-02-16  Carl Love  <cel@us.ibm.com>
22647         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
22648         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
22649         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
22650         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
22651         expansion to P8V_BUILTIN_VEC_FLOAT2.
22653 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
22655         PR rtl-optimization/70023
22656         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
22657         src_regno into account.
22659 2018-02-16  Carl Love  <cel@us.ibm.com>
22661         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
22662         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
22663         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
22664         * config/rs6000/rs6000.c: Remove case statements for
22665         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
22666         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
22667         and P9V_BUILTIN_VEC_VINSERT4B.
22668         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
22669         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
22670         * config/rs6000/vsx.md:
22671         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
22672         vec_insert4b.
22674 2018-02-16  Carl Love  <cel@us.ibm.com>
22676         * config/rs6000/altivec.h: Add builtin names vec_extract4b
22677         vec_insert4b.
22678         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
22679         definitions.
22680         * config/rs6000/rs6000-c.c: Add the definitions for
22681         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
22682         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
22683         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
22684         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
22685         definition for insert4b and define insn *insert3b_internal.
22686         * doc/extend.texi: Add documentation for vec_extract4b.
22688 2018-02-16  Nathan Sidwell  <nathan@acm.org>
22690         * doc/extend.texi (Backwards Compatibility): Mention friend
22691         injection.  Note for-scope is deprecated.
22692         * doc/invoke.texi (-ffriend-injection): Deprecate.
22694 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
22696         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
22697         that moved to I2, also allow destinations that are a paradoxical
22698         subreg (instead of a normal reg).
22700 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
22702         PR target/83831
22703         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
22704         to QImode.
22706 2018-02-16  Richard Biener  <rguenther@suse.de>
22708         PR tree-optimization/84037
22709         PR tree-optimization/84016
22710         PR target/82862
22711         * config/i386/i386.c (ix86_builtin_vectorization_cost):
22712         Adjust vec_construct for the fact we need additional higher latency
22713         128bit inserts for AVX256 and AVX512 vector builds.
22714         (ix86_add_stmt_cost): Scale vector construction cost for
22715         elementwise loads.
22717 2018-02-16  Richard Biener  <rguenther@suse.de>
22719         PR tree-optimization/84417
22720         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
22721         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
22722         (non_rewritable_lvalue_p): Likewise, use poly-ints.
22724 2018-02-16  Martin Liska  <mliska@suse.cz>
22726         PR sanitizer/84307
22727         * internal-fn.def (ASAN_CHECK): Set proper flags.
22728         (ASAN_MARK): Likewise.
22730 2018-02-16  Julia Koval  <julia.koval@intel.com>
22732         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
22733         from PTA_CANNONLAKE.
22735 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22737         PR target/84272
22738         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
22739         Use ++iter rather than iter++ for std::list iterators.
22740         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
22741         defer deleting them until all nodes in the forest are processed.  Do
22742         free even leaf nodes.  Change to_process into auto_vec.
22744         PR bootstrap/84405
22745         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
22746         * vec.h (vec_default_construct): Use memset instead of placement new
22747         if BROKEN_VALUE_INITIALIZATION is defined.
22748         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
22749         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
22750         is defined.
22752         PR rtl-optimization/83723
22753         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
22754         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
22755         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
22756         recursive calls.
22757         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
22758         callers.
22759         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
22761 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
22763         PR rtl-optimization/81443
22764         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
22765         from inner REGs to paradoxical SUBREGs.
22767 2018-02-16  Richard Biener  <rguenther@suse.de>
22769         PR tree-optimization/84399
22770         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22771         For operands we can analyze at their definition make sure we can
22772         analyze them at each use as well.
22774 2018-02-16  Richard Biener  <rguenther@suse.de>
22776         PR tree-optimization/84190
22777         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
22778         volatile accesses if the decl isn't volatile.
22780 2018-02-15  Jason Merrill  <jason@redhat.com>
22782         PR c++/84314 - ICE with templates and fastcall attribute.
22783         * attribs.c (build_type_attribute_qual_variant): Don't clobber
22784         TYPE_CANONICAL on an existing type.
22786 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
22788         PR tree-optimization/84383
22789         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
22790         dstoff nor call operand_equal_p if dstbase is NULL.
22792         PR tree-optimization/84334
22793         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
22794         also a CONSTANT_CLASS_P, punt.
22796 2018-02-14  Jim Wilson  <jimw@sifive.com>
22798         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
22799         first SMALL_OPERAND check.  New local min_second_step.  Move assert
22800         to where locals are set.  Add TARGET_RVC support.
22801         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
22803 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
22805         * doc/invoke.texi: Correct -Wformat-overflow code sample.
22807 2018-02-14  Martin Sebor  <msebor@redhat.com>
22809         PR tree-optimization/83698
22810         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
22811         arrays constrain the offset range to their bounds.
22812         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
22813         (builtin_access::overlap): Avoid setting the size of overlap if it's
22814         already been set.
22815         (maybe_diag_overlap): Also consider arrays when deciding what values
22816         of offsets to include in diagnostics.
22818 2018-02-14  Martin Sebor  <msebor@redhat.com>
22820         PR c/84108
22821         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
22822         that correspond to the kind of a declaration.
22824 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
22826         PR target/83984
22827         * config/pa/pa.md: Load address of PIC label using the linkage table
22828         if the label is nonlocal.
22830 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22832         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
22833         warning message if user requests -maltivec=be.
22834         * doc/invoke.texi: Document deprecation of -maltivec=be.
22836 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
22838         PR target/84220
22839         * config/rs6000/rs6000-c.c: Update definitions for
22840         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
22841         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
22843 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22845         PR target/84239
22846         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
22847         add _get_ssp intrinsics. Remove argument from
22848         __builtin_ia32_rdssp[d|q].
22849         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
22850         * config/i386/i386-builtin.def: Remove argument from
22851         __builtin_ia32_rdssp[d|q].
22852         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
22853         ix86_expand_special_args_builtin for _rdssp[d|q].
22854         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
22855         Clear register before usage.
22856         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
22857         Add documentation for new _get_ssp and _inc_ssp intrinsics.
22859 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
22861         PR tree-optimization/84357
22862         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
22863         operand 1 of an ARRAY_REF too.
22865 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
22867         PR target/83831
22868         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
22869         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22870         declarations.
22871         (set_of_reg): New struct.
22872         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
22873         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
22874         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22875         functions.
22876         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
22877         Split into bitclr, bitset, bitinvert patterns if appropriate.
22878         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
22879         use rx_fuse_in_memory_bitop.
22880         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
22881         to named insn, correct maximum insn length.
22883 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
22885         PR target/79242
22886         * machmode.def: Define a complex mode for PARTIAL_INT.
22887         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
22888         MODE_PARTIAL_INT.
22889         * doc/rtl.texi: Document CSPImode.
22890         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
22891         handling.
22892         (msp430_hard_regno_nregs_with_padding): Likewise.
22894 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
22896         PR target/84279
22897         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
22899 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
22901         PR rtl-optimization/84169
22902         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
22903         we generated a parallel as new i3 and we split that to new i2 and i3
22904         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
22905         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
22906         those to i2, not i1.  Partially rewrite this scan code.
22908 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22910         PR c/82210
22911         * stor-layout.c (place_field): For variable length fields, adjust
22912         offset_align afterwards not just based on the field's alignment,
22913         but also on the size.
22915         PR middle-end/84309
22916         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
22917         of exps and logs in the use_exp2 case.
22919 2018-02-13  Jeff Law  <law@redhat.com>
22921         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
22922         entry for "vector".
22924         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
22925         ARGS as unused.
22927 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
22929         PR debug/84342
22930         PR debug/84319
22931         * common.opt (gas-loc-support, gas-locview-support): New.
22932         (ginline-points, ginternal-reset-location-views): New.
22933         * doc/invoke.texi: Document them.  Use @itemx where intended.
22934         (gvariable-location-views): Adjust.
22935         * target.def (reset_location_view): New.
22936         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
22937         (TARGET_RESET_LOCATION_VIEW): New.
22938         * doc/tm.texi: Rebuilt.
22939         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
22940         (dwarf2out_default_as_locview_support): New.
22941         (output_asm_line_debug_info): Use option variables.
22942         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
22943         (output_loc_list): Likewise.
22944         (add_high_low_attributes): Check option variables.
22945         Don't output entry view attribute in strict mode.
22946         (gen_inlined_subroutine_die): Check option variables.
22947         (dwarf2out_inline_entry): Likewise.
22948         (init_sections_and_labels): Likewise.
22949         (dwarf2out_early_finish): Likewise.
22950         (maybe_reset_location_view): New, from...
22951         (dwarf2out_var_location): ... here.  Call it.
22952         * debug.h (dwarf2out_default_as_loc_support): Declare.
22953         (dwarf2out_default_as_locview_support): Declare.
22954         * hooks.c (hook_int_rtx_insn_0): New.
22955         * hooks.h (hook_int_rtx_insn_0): Declare.
22956         * toplev.c (process_options): Take -gas-loc-support and
22957         -gas-locview-support from dwarf2out.  Enable
22958         -gvariable-location-views by default only with locview
22959         assembler support.  Enable -ginternal-reset-location-views by
22960         default only if the target defines the corresponding hook.
22961         Enable -ginline-points by default if location views are
22962         enabled; force it disabled if statement frontiers are
22963         disabled.
22964         * tree-inline.c (expand_call_inline): Check option variables.
22965         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
22967 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
22969         PR tree-optimization/84321
22970         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
22971         handling.  Also check whether the anti-range contains any values
22972         that satisfy the mask; switch to a VR_RANGE if not.
22974 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
22976         PR sanitizer/84340
22977         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
22979 2018-02-13  Martin Jambor  <mjambor@suse.cz>
22981         PR c++/83990
22982         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
22983         of call statements, also set location of a load to a temporary.
22985 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
22987         * config/rl78/rl78.c (add_vector_labels): New function.
22988         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
22989         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
22990         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
22991         which checks that no arguments are passed.
22992         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
22993         * doc/extend.texi: Documentation for the new attribute.
22995 2018-02-13  Andreas Schwab  <schwab@suse.de>
22997         * config/riscv/linux.h (CPP_SPEC): Define.
22999 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
23001         PR target/84335
23002         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
23003         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
23004         OPTION_MASK_ISA_AES as first argument to def_builtin_const
23005         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
23006         instead of OPTION_MASK_ISA_PCLMUL as first argument to
23007         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
23008         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
23009         temporarily for AES and PCLMUL builtins.
23011         PR tree-optimization/84339
23012         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
23013         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
23014         Formatting fixes.
23016         PR middle-end/84309
23017         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
23018         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
23019         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
23020         inline function.
23021         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
23022         inline function.
23023         * omp-simd-clone.h: New file.
23024         * omp-simd-clone.c: Include omp-simd-clone.h.
23025         (expand_simd_clones): No longer static.
23026         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
23027         cgraph.h and omp-simd-clone.h.
23028         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
23029         (vect_recog_widen_shift_pattern): Formatting fix.
23030         (vect_pattern_recog_1): Don't check optab for calls.
23032         PR target/84336
23033         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
23034         operands[2] into a REG before using gen_lowpart on it.
23036 2018-02-12  Jeff Law  <law@redhat.com>
23038         PR target/83760
23039         * config/sh/sh.c (find_barrier): Consider a sibling call
23040         a barrier as well.
23042         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
23043         successfully back substituting a reg.
23045 2018-02-12  Richard Biener  <rguenther@suse.de>
23047         PR tree-optimization/84037
23048         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
23049         parameter, move visited init to caller.
23050         (vect_slp_analyze_operations): Separate cost from validity
23051         check, initialize visited once for all instances.
23052         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
23053         for all instances.
23054         * tree-vect-stmts.c (vect_model_simple_cost): Make early
23055         out an assert.
23056         (vect_model_promotion_demotion_cost): Likewise.
23057         (vectorizable_bswap): Guard cost modeling with !slp_node
23058         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
23059         SLP stmts.
23060         (vectorizable_call): Likewise.
23061         (vectorizable_conversion): Likewise.
23062         (vectorizable_assignment): Likewise.
23063         (vectorizable_shift): Likewise.
23064         (vectorizable_operation): Likewise.
23065         (vectorizable_store): Likewise.
23066         (vectorizable_load): Likewise.
23067         (vectorizable_condition): Likewise.
23068         (vectorizable_comparison): Likewise.
23070 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
23072         PR sanitizer/84307
23073         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
23074         (ASAN_MARK): Fix fnspec to account for return value, change pointer
23075         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
23077 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23079         PR middle-end/83665
23080         * params.def (inline-min-speedup): Increase from 8 to 15.
23081         (max-inline-insns-auto): Decrease from 40 to 30.
23082         * ipa-split.c (consider_split): Add some buffer for function to
23083         be considered inlining candidate.
23084         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
23085         default values.
23087 2018-02-12  Richard Biener  <rguenther@suse.de>
23089         PR tree-optimization/84037
23090         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
23091         matched stmts if we cannot swap the non-matched ones.
23093 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23095         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
23096         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
23097         _mm_maskz_scalef_round_ss): New intrinsics.
23098         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
23099         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
23100         __builtin_ia32_scalefss_round): Remove.
23101         (__builtin_ia32_scalefsd_mask_round,
23102         __builtin_ia32_scalefss_mask_round): New intrinsics.
23103         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
23104         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
23105         ((match_operand:VF_128 2 "<round_nimm_predicate>"
23106         "<round_constraint>")): Changed to ...
23107         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
23108         "<round_scalar_constraint>")): ... this.
23109         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
23110         %0, %1, %2<round_op3>}"): Changed to ...
23111         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
23112         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
23113         %2<round_scalar_mask_op3>}"): ... this.
23114         * config/i386/subst.md (round_scalar_nimm_predicate): New.
23116 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
23118         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
23119         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
23120         (_mm_maskz_sqrt_round_ss): New intrinsics.
23121         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
23122         (__builtin_ia32_sqrtsd_mask_round)
23123         (__builtin_ia32_sqrtss_mask_round): New builtins.
23124         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
23125         (__builtin_ia32_sqrtss_round): Remove.
23126         (__builtin_ia32_sqrtsd_mask_round)
23127         (__builtin_ia32_sqrtss_mask_round): New builtins.
23128         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
23129         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
23130         ((match_operand:VF_128 1 "vector_operand"
23131         "xBm,<round_constraint>")): Changed to ...
23132         ((match_operand:VF_128 1 "vector_operand"
23133         "xBm,<round_scalar_constraint>")): ... this.
23134         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
23135         %0, %2, %<iptr>1<round_op3>}): Changed to ...
23136         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
23137         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
23138         %<iptr>1<round_scalar_mask_op3>}): ... this.
23139         ((set_attr "prefix" "<round_prefix>")): Changed to ...
23140         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
23142 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
23144         PR target/84266
23145         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
23146         Cast vec_cmpeq result to correct type.
23147         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
23148         Cast vec_cmpgt result to correct type.
23150 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
23152         * final.c (final_scan_insn_1): Renamed from...
23153         (final_scan_insn): ... this.  New wrapper, to recover
23154         seen from the outermost call in recursive ones.
23155         * config/sparc/sparc.c (output_return): Drop seen from call.
23156         (output_sibcall): Likewise.
23157         * config/visium/visium.c (output_branch): Likewise.
23159 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
23161         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
23162         function label.
23164 2018-02-10  Alan Modra  <amodra@gmail.com>
23166         PR target/84300
23167         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
23168         Specify LR as an input.
23170 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
23172         PR sanitizer/83987
23173         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
23174         remove_member_access_dummy_vars): New functions.
23175         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
23176         lower_omp_1, execute_lower_omp): Use them.
23178         PR rtl-optimization/84308
23179         * shrink-wrap.c (spread_components): Release todo vector.
23181 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
23183         PR rtl-optimization/57193
23184         * ira-color.c (struct allocno_color_data): Add member
23185         conflict_allocno_hard_prefs.
23186         (update_conflict_allocno_hard_prefs): New.
23187         (bucket_allocno_compare_func): Add a preference based on
23188         conflict_allocno_hard_prefs.
23189         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
23190         (color_allocnos): Remove a dead code.  Initiate
23191         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
23193 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23195         PR target/84226
23196         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
23197         constraint from =wa to wa.  Avoid a subreg on the output operand,
23198         instead use a pseudo and subreg it in a move.
23199         (p9_xxbrd_<mode>): Changed to ...
23200         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
23201         (p9_xxbrd_v2df): New expander.
23202         (p9_xxbrw_<mode>): Changed to ...
23203         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
23204         (p9_xxbrw_v4sf): New expander.
23206 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23208         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
23210 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
23212         PR target/83926
23213         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
23214         multiply in 32-bit mode.
23215         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
23216         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
23217         mode.
23219 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
23221         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
23222         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
23223         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
23224         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
23226 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
23228         PR lto/84213
23229         * dwarf2out.c (is_trivial_indirect_ref): New function.
23230         (dwarf2out_late_global_decl): Do not generate a location
23231         attribute for variables that have a non-trivial DECL_VALUE_EXPR
23232         and that are not defined in the current unit.
23234 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23236         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
23237         instead of a libcall for UNORDERED.
23239 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
23241         PR target/82641
23242         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
23243         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
23245 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23247         PR target/PR84295
23248         * config/s390/s390.c (s390_set_current_function): Invoke
23249         s390_indirect_branch_settings also if fndecl didn't change.
23251 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23253         * config/rs6000/rs6000.md (blockage): Set length to zero.
23255 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
23257         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23259 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23261         PR sanitizer/84285
23262         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23263         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23264         -static-lib*san.
23266         PR debug/84252
23267         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23268         PARALLEL incoming that failed vt_get_decl_and_offset check.
23270         PR middle-end/84237
23271         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23272         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23273         TREE_READONLY bit.
23274         (get_variable_section): For decls in named .bss* sections pass true as
23275         second argument to bss_initializer_p.
23277 2018-02-09  Marek Polacek  <polacek@redhat.com>
23278             Jakub Jelinek  <jakub@redhat.com>
23280         PR c++/83659
23281         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23282         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
23283         Sync some changes from cxx_fold_indirect_ref.
23285 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23287         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23288         markers.
23289         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23290         (BLOCK_INLINE_ENTRY_LABEL): New.
23291         (dwarf2out_var_location): Disregard inline entry markers.
23292         (inline_entry_data): New struct.
23293         (inline_entry_data_hasher): New hashtable type.
23294         (inline_entry_data_hasher::hash): New.
23295         (inline_entry_data_hasher::equal): New.
23296         (inline_entry_data_table): New variable.
23297         (add_high_low_attributes): Add DW_AT_entry_pc and
23298         DW_AT_GNU_entry_view attributes if a pending entry is found
23299         in inline_entry_data_table.  Add old entry_pc attribute only
23300         if debug nonbinding markers are disabled.
23301         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23302         markers are enabled.
23303         (block_within_block_p, dwarf2out_inline_entry): New.
23304         (dwarf2out_finish): Check that no entries remained in
23305         inline_entry_data_table.
23306         * final.c (reemit_insn_block_notes): Handle inline entry notes.
23307         (final_scan_insn, notice_source_line): Likewise.
23308         (rest_of_clean_state): Skip inline entry markers.
23309         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23310         markers.
23311         * gimple.c (gimple_build_debug_inline_entry): New.
23312         * gimple.h (enum gimple_debug_subcode): Add
23313         GIMPLE_DEBUG_INLINE_ENTRY.
23314         (gimple_build_debug_inline_entry): Declare.
23315         (gimple_debug_inline_entry_p): New.
23316         (gimple_debug_nonbind_marker_p): Adjust.
23317         * insn-notes.def (INLINE_ENTRY): New.
23318         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23319         inline entry marker notes.
23320         (print_insn): Likewise.
23321         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23322         (INSN_DEBUG_MARKER_KIND): Likewise.
23323         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23324         * tree-inline.c (expand_call_inline): Build and insert
23325         debug_inline_entry stmt.
23326         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23327         inline entry blocks early, if nonbind markers are enabled.
23328         (dump_scope_block): Dump fragment info.
23329         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23330         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23331         (gimple_build_debug_inline_entry): New.
23332         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23333         Enable/disable inline entry points too.
23334         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23335         (DEBUG_INSN): Describe inline entry markers.
23337         * common.opt (gvariable-location-views): New.
23338         (gvariable-location-views=incompat5): New.
23339         * config.in: Rebuilt.
23340         * configure: Rebuilt.
23341         * configure.ac: Test assembler for view support.
23342         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23343         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23344         * dwarf2out.c (var_loc_view): New typedef.
23345         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23346         (dwarf2out_locviews_in_attribute): New.
23347         (dwarf2out_locviews_in_loclist): New.
23348         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23349         (enum dw_line_info_opcode): Add LI_adv_address.
23350         (struct dw_line_info_table): Add view.
23351         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23352         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23353         (zero_view_p): New variable.
23354         (ZERO_VIEW_P): New macro.
23355         (output_asm_line_debug_info): New.
23356         (struct var_loc_node): Add view.
23357         (add_AT_view_list, AT_loc_list): New.
23358         (add_var_loc_to_decl): Add view param.  Test it against last.
23359         (new_loc_list): Add view params.  Record them.
23360         (AT_loc_list_ptr): Handle loc and view lists.
23361         (view_list_to_loc_list_val_node): New.
23362         (print_dw_val): Handle dw_val_class_view_list.
23363         (size_of_die): Likewise.
23364         (value_format): Likewise.
23365         (loc_list_has_views): New.
23366         (gen_llsym): Set vl_symbol too.
23367         (maybe_gen_llsym, skip_loc_list_entry): New.
23368         (dwarf2out_maybe_output_loclist_view_pair): New.
23369         (output_loc_list): Output view list or entries too.
23370         (output_view_list_offset): New.
23371         (output_die): Handle dw_val_class_view_list.
23372         (output_dwarf_version): New.
23373         (output_compilation_unit_header): Use it.
23374         (output_skeleton_debug_sections): Likewise.
23375         (output_rnglists, output_line_info): Likewise.
23376         (output_pubnames, output_aranges): Update version comments.
23377         (output_one_line_info_table): Output view numbers in asm comments.
23378         (dw_loc_list): Determine current endview, pass it to new_loc_list.
23379         Call maybe_gen_llsym.
23380         (loc_list_from_tree_1): Adjust.
23381         (add_AT_location_description): Create view list attribute if
23382         needed, check it's absent otherwise.
23383         (convert_cfa_to_fb_loc_list): Adjust.
23384         (maybe_emit_file): Call output_asm_line_debug_info for test.
23385         (dwarf2out_var_location): Reset views as needed.  Precompute
23386         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
23387         attribute.  Set view.
23388         (new_line_info_table): Reset next view.
23389         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23390         (dwarf2out_source_line): Likewise.  Output view resets and labels to
23391         the assembler, or select appropriate line info opcodes.
23392         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23393         (optimize_string_length): Catch it.  Adjust.
23394         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
23395         dw_val_class_view_list, and remove it if no longer needed.
23396         (hash_loc_list): Hash view numbers.
23397         (loc_list_hasher::equal): Compare them.
23398         (optimize_location_lists): Check whether a view list symbol is
23399         needed, and whether the locview attribute is present, and
23400         whether they match.  Remove the locview attribute if no longer
23401         needed.
23402         (index_location_lists): Call skip_loc_list_entry for test.
23403         (dwarf2out_finish): Call output_asm_line_debug_info for test.
23404         Use output_dwarf_version.
23405         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23406         (struct dw_val_node): Add val_view_list.
23407         * final.c (SEEN_NEXT_VIEW): New.
23408         (set_next_view_needed): New.
23409         (clear_next_view_needed): New.
23410         (maybe_output_next_view): New.
23411         (final_start_function): Rename to...
23412         (final_start_function_1): ... this.  Take pointer to FIRST,
23413         add SEEN parameter.  Emit param bindings in the initial view.
23414         (final_start_function): Reintroduce SEEN-less interface.
23415         (final): Rename to...
23416         (final_1): ... this.  Take SEEN parameter.  Output final pending
23417         next view at the end.
23418         (final): Reintroduce seen-less interface.
23419         (final_scan_insn): Output pending next view before switching
23420         sections or ending a block.  Mark the next view as needed when
23421         outputting variable locations.  Notify debug backend of section
23422         changes, and of location view changes.
23423         (rest_of_handle_final): Adjust.
23424         * toplev.c (process_options): Autodetect value for debug variable
23425         location views option.  Warn on incompat5 without -gdwarf-5.
23426         * doc/invoke.texi (gvariable-location-views): New.
23427         (gvariable-location-views=incompat5): New.
23428         (gno-variable-location-views): New.
23430 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
23432         PR tree-optimization/84136
23433         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23434         that the result of find_edge is non-NULL.
23436 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23438         PR target/83008
23439         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
23440         storing integer register in SImode.  Fix cost of 256 and 512
23441         byte aligned SSE register store.
23443 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23445         * config/i386/i386.c (ix86_multiplication_cost): Fix
23446         multiplication cost for TARGET_AVX512DQ.
23448 2018-02-08  Marek Polacek  <polacek@redhat.com>
23450         PR tree-optimization/84238
23451         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
23452         get_range_strlen.
23454 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23456         PR tree-optimization/84265
23457         * tree-vect-stmts.c (vectorizable_store): Don't treat
23458         VMAT_CONTIGUOUS accesses as grouped.
23459         (vectorizable_load): Likewise.
23461 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23463         PR tree-optimization/81635
23464         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
23465         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
23466         (test_round_for_mask): New functions.
23467         (wide_int_cc_tests): Call test_round_for_mask.
23468         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
23469         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
23470         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
23471         range returned by get_range_info.
23473 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23475         PR ipa/81360
23476         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
23477         * symtab.c: Include builtins.h
23478         (symtab_node::output_to_lto_symbol_table_p): Move here
23479         from lto-streamer-out.c:output_symbol_p.
23480         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
23481         (output_symbol_p): Move all logic to symtab.c
23482         (produce_symtab): Update.
23484 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23486         * config/s390/s390-opts.h (enum indirect_branch): Define.
23487         * config/s390/s390-protos.h (s390_return_addr_from_memory)
23488         (s390_indirect_branch_via_thunk)
23489         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
23490         (enum s390_indirect_branch_type): Define.
23491         * config/s390/s390.c (struct s390_frame_layout, struct
23492         machine_function): Remove.
23493         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
23494         (indirect_branch_table_label_no, indirect_branch_table_name):
23495         Define variables.
23496         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
23497         (enum s390_indirect_branch_option): Define.
23498         (s390_return_addr_from_memory): New function.
23499         (s390_handle_string_attribute): New function.
23500         (s390_attribute_table): Add new attribute handler.
23501         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
23502         (s390_indirect_branch_via_thunk): New function.
23503         (s390_indirect_branch_via_inline_thunk): New function.
23504         (s390_function_ok_for_sibcall): When jumping via thunk disallow
23505         sibling call optimization for non z10 compiles.
23506         (s390_emit_call): Force indirect branch target to be a single
23507         register.  Add r1 clobber for non-z10 compiles.
23508         (s390_emit_epilogue): Emit return jump via return_use expander.
23509         (s390_reorg): Handle JUMP_INSNs as execute targets.
23510         (s390_option_override_internal): Perform validity checks for the
23511         new command line options.
23512         (s390_indirect_branch_attrvalue): New function.
23513         (s390_indirect_branch_settings): New function.
23514         (s390_set_current_function): Invoke s390_indirect_branch_settings.
23515         (s390_output_indirect_thunk_function):  New function.
23516         (s390_code_end): Implement target hook.
23517         (s390_case_values_threshold): Implement target hook.
23518         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
23519         macros.
23520         * config/s390/s390.h (struct s390_frame_layout)
23521         (struct machine_function): Move here from s390.c.
23522         (TARGET_INDIRECT_BRANCH_NOBP_RET)
23523         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
23524         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
23525         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
23526         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
23527         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
23528         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
23529         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
23530         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
23531         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
23532         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
23533         (mnemonic attribute): Add values which aren't recognized
23534         automatically.
23535         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
23536         pattern for branch conversion.  Fix mnemonic attribute.
23537         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
23538         indirect branch via thunk if requested.
23539         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
23540         ("*indirect_jump"): Disable for branch conversion using out of
23541         line thunks.
23542         ("indirect_jump_via_thunk<mode>_z10")
23543         ("indirect_jump_via_thunk<mode>")
23544         ("indirect_jump_via_inlinethunk<mode>_z10")
23545         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
23546         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
23547         ("casesi_jump_via_inlinethunk<mode>_z10")
23548         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
23549         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
23550         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
23551         ("*indirect2_jump"): Disable for branch conversion.
23552         ("casesi_jump"): Turn into expander and expand patterns for branch
23553         conversion.
23554         ("return_use"): New expander.
23555         ("*return"): Emit return via thunk and rename it to ...
23556         ("*return<mode>"): ... this one.
23557         * config/s390/s390.opt: Add new options and and enum for the
23558         option values.
23560 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23562         * lra-constraints.c (match_reload): Unconditionally use
23563         gen_lowpart_SUBREG, rather than selecting between that
23564         and equivalent gen_rtx_SUBREG code.
23566 2018-02-08  Richard Biener  <rguenther@suse.de>
23568         PR tree-optimization/84233
23569         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
23570         changed flag instead of boguously re-using phi_inserted.
23572 2018-02-08  Martin Jambor  <mjambor@suse.cz>
23574         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
23575         static local variables.
23577 2018-02-08  Richard Biener  <rguenther@suse.de>
23579         PR tree-optimization/84278
23580         * tree-vect-stmts.c (vectorizable_store): When looking for
23581         smaller vector types to perform grouped strided loads/stores
23582         make sure the mode is supported by the target.
23583         (vectorizable_load): Likewise.
23585 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23587         * config/aarch64/aarch64.c (aarch64_components_for_bb):
23588         Increase LDP/STP opportunities by adding adjacent callee-saves.
23590 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23592         PR rtl-optimization/84068
23593         PR rtl-optimization/83459
23594         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
23596 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
23598         PR tree-optimization/84224
23599         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
23600         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
23601         non-zero arguments.
23603 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
23605         PR target/84113
23606         * config/rs6000/altivec.md (*restore_world): Remove LR use.
23607         * config/rs6000/predicates.md (restore_world_operation): Adjust op
23608         count, remove one USE.
23610 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
23612         * doc/install.texi (Configuration): Document the
23613         --with-long-double-format={ibm,ieee} PowerPC configuration
23614         options.
23616         PR target/84154
23617         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
23618         Convert from define_expand to be define_insn_and_split.  Rework
23619         float/double/_Float128 conversions to QI/HI/SImode to work with
23620         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
23621         conversions to QI/HImode types did a store and then a load to
23622         truncate the value.  For conversions to VSX registers, don't split
23623         the insn, instead emit the code directly.  Use the code iterator
23624         any_fix to combine signed and unsigned conversions.
23625         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
23626         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
23627         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23628         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
23629         (fix_<mode>di2_hw): Likewise.
23630         (fixuns_<mode>di2_hw): Likewise.
23631         (fix_<mode>si2_hw): Likewise.
23632         (fixuns_<mode>si2_hw): Likewise.
23633         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
23634         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23635         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
23636         fix<uns>_trunc<SFDF:mode>si2_p8.
23637         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
23638         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23639         (fix<uns>_<mode>_mem): Likewise.
23640         (fctiw<u>z_<mode>_mem): Likewise.
23641         (fix<uns>_<mode>_mem): Likewise.
23642         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
23643         the register allocator from doing a direct move to the GPRs to do
23644         a store, and instead use the ISA 3.0 store byte/half-word from
23645         vector register instruction.  For IEEE 128-bit floating point,
23646         also optimize stores of 32-bit ints.
23647         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
23649 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
23651         * genextract.c (push_pathstr_operand): New function to support
23652         [a-zA-Z].
23653         (walk_rtx): Call push_pathstr_operand.
23654         (print_path): Support [a-zA-Z].
23656 2018-02-07  Richard Biener  <rguenther@suse.de>
23658         PR tree-optimization/84037
23659         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
23660         (cse_and_gimplify_to_preheader): Declare.
23661         (vect_get_place_in_interleaving_chain): Likewise.
23662         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23663         ivexpr_map.
23664         (_loop_vec_info::~_loop_vec_info): Delete it.
23665         (cse_and_gimplify_to_preheader): New function.
23666         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
23667         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
23668         (vectorizable_load): Likewise.  For grouped stores always base
23669         the IV on the first element.
23670         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
23671         condition before gimplifying.
23673 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
23675         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
23676         *DIV_EXPR and *MOD_EXPR.
23678 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
23680         PR target/84248
23681         * config/i386/i386.c (ix86_option_override_internal): Mask out
23682         the CF_SET bit when checking -fcf-protection.
23684 2018-02-07  Tom de Vries  <tom@codesourcery.com>
23686         PR libgomp/84217
23687         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
23688         enough.
23690 2018-02-07  Richard Biener  <rguenther@suse.de>
23692         PR tree-optimization/84204
23693         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
23694         this place.
23696         PR tree-optimization/84205
23697         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
23698         special-case isl_ast_op_zdiv_r.
23700         PR tree-optimization/84223
23701         * graphite-scop-detection.c (gather_bbs::before_dom_children):
23702         Only add conditions from within the region.
23703         (gather_bbs::after_dom_children): Adjust.
23705 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
23707         PR target/84209
23708         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
23709         * config/avr/avr.md: Only post-reload split REG-REG moves if
23710         either register is GENERAL_REG_P.
23712 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
23714         PR tree-optimization/84235
23715         * tree-ssa-scopedtables.c
23716         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
23717         if the subtraction is performed in floating point type where NaNs are
23718         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
23719         build 1.  Formatting fix.
23721 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
23723         PR target/84146
23724         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
23725         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
23726         and skip it regardless of bb boundaries.  Use CALL_P macro,
23727         don't test INSN_P (insn) together with CALL_P or JUMP_P check
23728         unnecessarily, formatting fix.
23730 2018-02-06  Michael Collison  <michael.collison@arm.com>
23732         * config/arm/thumb2.md:
23733         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
23734         (*thumb_mov_notscc): Ditto.
23736 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
23738         PR target/84154
23739         * config/rs6000/rs6000.md (su code attribute): Use "u" for
23740         unsigned_fix, not "s".
23742 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23744         * configure.ac (gcc_fn_eh_frame_ro): New function.
23745         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
23746         correct .eh_frame permissions.
23747         * configure: Regenerate.
23749 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
23751         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
23752         irrelevant options.
23754 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23756         * config/rs6000/rs6000.c (rs6000_option_override_internal):
23757         Display warning message for -mno-speculate-indirect-jumps.
23759 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
23761         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
23762         Undocumented.
23763         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
23765 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
23767         PR tree-optimization/84225
23768         * tree-eh.c (find_trapping_overflow): Only call
23769         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
23771 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23773         PR target/84145
23774         * config/i386/i386.c: Reimplement the check of possible options
23775         -mibt/-mshstk conbination. Change error messages.
23776         * doc/invoke.texi: Fix a typo: remove extra '='.
23778 2018-02-06  Marek Polacek  <polacek@redhat.com>
23780         PR tree-optimization/84228
23781         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
23783 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
23785         PR target/82641
23786         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
23787         emitted arch directives.
23788         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
23789         __ARM_FEATURE_COPROC before changing architectures.
23791 2018-02-06  Richard Biener  <rguenther@suse.de>
23793         * config/i386/i386.c (print_reg): Fix typo.
23794         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
23796 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
23798         * configure: Regenerate.
23800 2018-02-05  Martin Sebor  <msebor@redhat.com>
23802         PR tree-optimization/83369
23803         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
23804         inlining context.
23806 2018-02-05  Martin Liska  <mliska@suse.cz>
23808         * doc/invoke.texi: Cherry-pick upstream r323995.
23810 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
23812         * ira.c (ira_init_register_move_cost): Adjust comment.
23814 2018-02-05  Martin Liska  <mliska@suse.cz>
23816         PR gcov-profile/84137
23817         * doc/gcov.texi: Fix typo in documentation.
23819 2018-02-05  Martin Liska  <mliska@suse.cz>
23821         PR gcov-profile/83879
23822         * doc/gcov.texi: Document necessity of --dynamic-list-data when
23823         using dlopen functionality.
23825 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
23827         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
23828         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
23829         _mm_maskz_range_ss, _mm_mask_range_round_ss,
23830         _mm_maskz_range_round_ss): New intrinsics.
23831         (__builtin_ia32_rangesd128_round)
23832         (__builtin_ia32_rangess128_round): Remove.
23833         (__builtin_ia32_rangesd128_mask_round,
23834         __builtin_ia32_rangess128_mask_round): New builtins.
23835         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
23836         __builtin_ia32_rangess128_round): Remove.
23837         (__builtin_ia32_rangesd128_mask_round,
23838         __builtin_ia32_rangess128_mask_round): New builtins.
23839         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
23840         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
23841         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
23842         "<round_saeonly_constraint>")): Changed to ...
23843         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
23844         "<round_saeonly_scalar_constraint>")): ... this.
23845         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
23846         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
23847         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
23848         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
23849         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
23851 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
23853         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
23854         options.
23855         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
23856         Remove all values except native, 8540 and 8548.
23858 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
23860         * config/i386/i386.c (ix86_output_function_return): Pass
23861         INVALID_REGNUM, instead of -1, as invalid register number to
23862         indirect_thunk_name and output_indirect_thunk.
23864 2018-02-02  Julia Koval  <julia.koval@intel.com>
23866         * config.gcc: Add -march=icelake.
23867         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
23868         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
23869         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
23870         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
23871         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
23872         (processor_target_table): Add icelake.
23873         (ix86_option_override_internal): Handle new PTAs.
23874         (get_builtin_code_for_version): Handle icelake.
23875         (M_INTEL_COREI7_ICELAKE): New.
23876         (fold_builtin_cpu): Handle icelake.
23877         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
23878         * doc/invoke.texi: Add -march=icelake.
23880 2018-02-02  Julia Koval  <julia.koval@intel.com>
23882         * config/i386/i386.c (ix86_option_override_internal): Change flags type
23883         to wide_int_bitmask.
23884         * wide-int-bitmask.h: New.
23886 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23888         PR target/84066
23889         * config/i386/i386.md: Replace Pmode with word_mode in
23890         builtin_setjmp_setup and builtin_longjmp to support x32.
23892 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
23894         PR target/56010
23895         PR target/83743
23896         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
23897         #include "opts.h".
23898         (rs6000_supported_cpu_names): New static variable.
23899         (linux_cpu_translation_table): Likewise.
23900         (elf_platform) <cpu>: Define new static variable and use it.
23901         Translate kernel AT_PLATFORM name to canonical name if needed.
23902         Error if platform name is unknown.
23904 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
23906         PR target/84089
23907         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
23909 2018-02-01  Jeff Law  <law@redhat.com>
23911         PR target/84128
23912         * config/i386/i386.c (release_scratch_register_on_entry): Add new
23913         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
23914         the scratch if RELEASE_VIA_POP is false.
23915         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
23916         If we have to save a temporary register, decrement SIZE appropriately.
23917         Pass new arguments to release_scratch_register_on_entry.
23918         (ix86_adjust_stack_and_probe): Likewise.
23919         (ix86_emit_probe_stack_range): Pass new arguments to
23920         release_scratch_register_on_entry.
23922 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
23924         PR rtl-optimization/84157
23925         * combine.c (change_zero_ext): Use REG_P predicate in
23926         front of HARD_REGISTER_P predicate.
23928 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
23930         * config/avr/avr.c (avr_option_override): Move disabling of
23931         -fdelete-null-pointer-checks to...
23932         * common/config/avr/avr-common.c (avr_option_optimization_table):
23933         ...here.
23935 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23937         PR tree-optimization/81635
23938         * tree-data-ref.c (split_constant_offset_1): For types that
23939         wrap on overflow, try to use range info to prove that wrapping
23940         cannot occur.
23942 2018-02-01  Renlin Li  <renlin.li@arm.com>
23944         PR target/83370
23945         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
23946         TAILCALL_ADDR_REGS.
23947         (aarch64_register_move_cost): Likewise.
23948         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
23949         TAILCALL_ADDR_REGS.
23950         (REG_CLASS_NAMES): Likewise.
23951         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
23952         TAILCALL_ADDR_REGS. Remove IP registers.
23953         * config/aarch64/aarch64.md (Ucs): Update register constraint.
23955 2018-02-01  Richard Biener  <rguenther@suse.de>
23957         * domwalk.h (dom_walker::dom_walker): Add additional constructor
23958         for specifying RPO order and allow NULL for that.
23959         * domwalk.c (dom_walker::dom_walker): Likewise.
23960         (dom_walker::walk): Handle NULL RPO order.
23961         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
23962         in RPO order.
23963         (rewrite_update_dom_walker): Likewise.
23964         (mark_def_dom_walker): Likewise.
23966 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23968         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
23969         (aarch64_maybe_expand_sve_subreg_move): Declare.
23970         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
23971         * config/aarch64/predicates.md (aarch64_any_register_operand): New
23972         predicate.
23973         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
23974         that are semantically a reverse operation.
23975         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
23976         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
23977         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
23978         functions.
23979         (aarch64_can_change_mode_class): For big-endian, forbid changes
23980         between two SVE modes if they have different element sizes.
23982 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23984         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
23985         the TImode handling for big-endian targets.
23987 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23989         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
23990         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
23991         not just bytes.
23992         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
23993         Remove BSWAP handing for big-endian targets and use the form of
23994         LD1RQ appropariate for the mode.
23996 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23998         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
23999         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
24000         duplicated element.
24002 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
24004         PR tearget/83845
24005         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
24006         check for operands that need to go through aarch64_sve_reload_be.
24008 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
24010         PR tree-optimization/81661
24011         PR tree-optimization/84117
24012         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
24013         * tree-eh.c: Include gimplify.h.
24014         (find_trapping_overflow, replace_trapping_overflow,
24015         rewrite_to_non_trapping_overflow): New functions.
24016         * tree-vect-loop.c: Include tree-eh.h.
24017         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
24018         * tree-data-ref.c: Include tree-eh.h.
24019         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
24021 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
24023         PR rtl-optimization/84123
24024         * combine.c (change_zero_ext): Check if hard register satisfies
24025         can_change_dest_mode before calling gen_lowpart_SUBREG.
24027 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
24029         PR target/82444
24030         * ira.c (ira_init_register_move_cost): Remove assert.
24032 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24034         PR rtl-optimization/84071
24035         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
24036         * doc/tm.texi: Regenerate.
24038 2018-01-31  Richard Biener  <rguenther@suse.de>
24040         PR tree-optimization/84132
24041         * tree-data-ref.c (analyze_miv_subscript): Properly
24042         check whether evolution_function_is_affine_multivariate_p
24043         before calling gcd_of_steps_may_divide_p.
24045 2018-01-31  Julia Koval  <julia.koval@intel.com>
24047         PR target/83618
24048         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
24049         * config/i386/i386.md (rdpid_rex64) New.
24050         (rdpid): Make 32bit only.
24052 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
24054         PR lto/84105
24055         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
24056         an IDENTIFIER_NODE for FUNCTION_TYPE's.
24058 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24060         Revert
24061         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24063         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24065 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
24067         PR rtl-optimization/84071
24068         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
24069         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
24071 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24073         * config/arc/arc.c (arc_handle_aux_attribute): New function.
24074         (arc_attribute_table): Add 'aux' attribute.
24075         (arc_in_small_data_p): Consider aux like variables.
24076         (arc_is_aux_reg_p): New function.
24077         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
24078         (arc_get_aux_arg): New function.
24079         (prepare_move_operands): Handle aux-register access.
24080         (arc_handle_aux_attribute): New function.
24081         * doc/extend.texi (ARC Variable attributes): Add subsection.
24083 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
24085         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
24086         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
24087         (arc_attribute_table): Add 'uncached' attribute.
24088         (arc_print_operand): Print '.di' flag for uncached memory
24089         accesses.
24090         (arc_in_small_data_p): Do not consider for small data the uncached
24091         types.
24092         (arc_is_uncached_mem_p): New function.
24093         * config/arc/predicates.md (compact_store_memory_operand): Check
24094         for uncached memory accesses.
24095         (nonvol_nonimm_operand): Likewise.
24096         * doc/extend.texi (ARC Type Attribute): New subsection.
24098 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
24100         PR c/84100
24101         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
24102         falign-loops=): Add Optimization flag.
24104 2018-01-30  Jeff Law  <law@redhat.com>
24106         PR target/84064
24107         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
24108         INT_REGISTERS_SAVED.  Check it prior to calling
24109         get_scratch_register_on_entry.
24110         (ix86_adjust_stack_and_probe): Similarly.
24111         (ix86_emit_probe_stack_range): Similarly.
24112         (ix86_expand_prologue): Corresponding changes.
24114 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24116         PR target/40411
24117         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
24118         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
24120 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
24122         PR target/84112
24123         * lra-constraints.c (curr_insn_transform): Process AND in the
24124         address.
24126 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
24128         PR rtl-optimization/83986
24129         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
24130         dependence against last_pending_memory_flush in addition to
24131         pending_jump_insns.
24133 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
24135         PR tree-optimization/81611
24136         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
24137         copies.
24139 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24141         PR target/83758
24142         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
24143         a reg rtx.
24145 2018-01-30  Richard Biener  <rguenther@suse.de>
24146             Jakub Jelinek  <jakub@redhat.com>
24148         PR tree-optimization/84111
24149         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
24150         inner loops added during recursion, as they don't have up-to-date
24151         SSA form.
24153 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24155         PR ipa/81360
24156         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
24157         (can_inline_edge_by_limits_p): ... here.
24158         (can_early_inline_edge_p, check_callers,
24159         update_caller_keys, update_callee_keys, recursive_inlining,
24160         add_new_edges_to_heap, speculation_useful_p,
24161         inline_small_functions,
24162         inline_small_functions, flatten_function,
24163         inline_to_all_callers_1): Update.
24165 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
24167         * profile-count.c (profile_count::combine_with_ipa_count): Handle
24168         zeros correctly.
24170 2018-01-30  Richard Biener  <rguenther@suse.de>
24172         PR tree-optimization/83008
24173         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
24174         invariant and constant vector uses in stmts when they need
24175         more than one stmt.
24177 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24179         PR bootstrap/84017
24180         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
24181         * configure: Regenerate.
24183 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24185         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
24186         pattern.
24187         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
24188         Use gen_rtx_REG rather than gen_lowpart.
24190 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24192         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
24193         rather than 0 when creating partial subregs.
24195 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
24197         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
24198         of usage.
24200 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
24202         PR target/81550
24203         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
24204         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
24205         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
24206         flags.  This restores the settings used before the 2017-07-24.
24207         Turning off pre increment/decrement/modify allows IVOPTS to
24208         optimize DF/SF loops where the index is an int.
24210 2018-01-29  Richard Biener  <rguenther@suse.de>
24211             Kelvin Nilsen  <kelvin@gcc.gnu.org>
24213         PR bootstrap/80867
24214         * tree-vect-stmts.c (vectorizable_call): Don't call
24215         targetm.vectorize_builtin_md_vectorized_function if callee is
24216         NULL.
24218 2018-01-22  Carl Love  <cel@us.ibm.com>
24220         * doc/extend.tex: Fix typo in second arg in
24221         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
24223 2018-01-29  Richard Biener  <rguenther@suse.de>
24225         PR tree-optimization/84086
24226         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
24227         (flush_ssaname_freelist): When SSA names were released reset
24228         the SCEV hash table.
24230 2018-01-29  Richard Biener  <rguenther@suse.de>
24232         PR tree-optimization/84057
24233         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
24234         removed paths when removing edges.
24236 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
24238         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
24239         -mfunction-return=@var{choice}.
24241 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
24243         PR diagnostic/84034
24244         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
24245         Handle CR like TAB.
24246         (layout::print_source_line): Likewise.
24247         (test_get_line_width_without_trailing_whitespace): Add test cases.
24249 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
24251         PR middle-end/84040
24252         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
24253         debug insns.
24255 2018-01-26  Jim Wilson  <jimw@sifive.com>
24257         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24259         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24260         specified.
24262 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24264         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24265         and CMP + SUB-immediate -> SUBS.
24267 2018-01-26  Martin Sebor  <msebor@redhat.com>
24269         PR tree-optimization/83896
24270         * tree-ssa-strlen.c (get_string_len): Rename...
24271         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
24272         Avoid assuming length is constant.
24273         (handle_char_store): Use HOST_WIDE_INT for string length.
24275 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
24277         PR target/81763
24278         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24279         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24281 2018-01-26  Richard Biener  <rguenther@suse.de>
24283         PR rtl-optimization/84003
24284         * dse.c (record_store): Only record redundant stores when
24285         the earlier store aliases at least all accesses the later one does.
24287 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
24289         PR rtl-optimization/83985
24290         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24291         REG_CFA_RESTORE insns.
24292         (delete_unmarked_insns): Don't ignore separate shrink wrapping
24293         REG_CFA_RESTORE insns here.
24295         PR c/83989
24296         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24297         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24299 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24301         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24302         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24303         (arc_init): Likewise.
24304         (arc_override_options): Likewise.
24305         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24306         value.
24307         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24308         support.
24309         * config/arc/arc.h (TARGET_DBNZ): Define.
24310         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24311         properly set the tune attribute.
24312         (dbnz): Use TARGET_DBNZ guard.
24313         * config/arc/arc.opt (mtune): Add core3 option.
24315 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24317         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24318         recognize new pic like addresses.
24319         (arc_delegitimize_address): Clean up.
24321 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24323         * config/arc/arc-arches.def: Option mrf16 valid for all
24324         architectures.
24325         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24326         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24327         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24328         * config/arc/arc-tables.opt: Regenerate.
24329         * config/arc/arc.c (arc_conditional_register_usage): Handle
24330         reduced register file case.
24331         (arc_file_start): Set must have build attributes.
24332         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24333         mrf16 option value.
24334         * config/arc/arc.opt (mrf16): Add new option.
24335         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24336         * config/arc/genmultilib.awk: Handle new mrf16 option.
24337         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24338         * config/arc/t-multilib: Regenerate.
24339         * doc/invoke.texi (ARC Options): Document mrf16 option.
24341 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24343         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24344         * config/arc/arc.c (arc_handle_secure_attribute): New function.
24345         (arc_attribute_table): Add 'secure_call' attribute.
24346         (arc_print_operand): Print secure call operand.
24347         (arc_function_ok_for_sibcall): Don't optimize tail calls when
24348         secure.
24349         (arc_is_secure_call_p): New function.  * config/arc/arc.md
24350         (call_i): Add support for sjli instruction.
24351         (call_value_i): Likewise.
24352         * config/arc/constraints.md (Csc): New constraint.
24354 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24355             John Eric Martin  <John.Martin@emmicro-us.com>
24357         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24358         * config/arc/arc.c (_arc_jli_section): New struct.
24359         (arc_jli_section): New type.
24360         (rc_jli_sections): New static variable.
24361         (arc_handle_jli_attribute): New function.
24362         (arc_attribute_table): Add jli_always and jli_fixed attribute.
24363         (arc_file_end): New function.
24364         (TARGET_ASM_FILE_END): Define.
24365         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24366         (arc_add_jli_section): New function.
24367         (jli_call_scan): Likewise.
24368         (arc_reorg): Call jli_call_scan.
24369         (arc_output_addsi): Remove 'S' from printing asm operand.
24370         (arc_is_jli_call_p): New function.
24371         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24372         operand.
24373         (movhi_insn): Likewise.
24374         (movsi_insn): Likewise.
24375         (movsi_set_cc_insn): Likewise.
24376         (loadqi_update): Likewise.
24377         (load_zeroextendqisi_update): Likewise.
24378         (load_signextendqisi_update): Likewise.
24379         (loadhi_update): Likewise.
24380         (load_zeroextendhisi_update): Likewise.
24381         (load_signextendhisi_update): Likewise.
24382         (loadsi_update): Likewise.
24383         (loadsf_update): Likewise.
24384         (movsicc_insn): Likewise.
24385         (bset_insn): Likewise.
24386         (bxor_insn): Likewise.
24387         (bclr_insn): Likewise.
24388         (bmsk_insn): Likewise.
24389         (bicsi3_insn): Likewise.
24390         (cmpsi_cc_c_insn): Likewise.
24391         (movsi_ne): Likewise.
24392         (movsi_cond_exec): Likewise.
24393         (clrsbsi2): Likewise.
24394         (norm_f): Likewise.
24395         (normw): Likewise.
24396         (swap): Likewise.
24397         (divaw): Likewise.
24398         (flag): Likewise.
24399         (sr): Likewise.
24400         (kflag): Likewise.
24401         (ffs): Likewise.
24402         (ffs_f): Likewise.
24403         (fls): Likewise.
24404         (call_i): Remove 'S' asm letter, add jli instruction.
24405         (call_value_i): Likewise.
24406         * config/arc/arc.op (mjli-always): New option.
24407         * config/arc/constraints.md (Cji): New constraint.
24408         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24409         operand.
24410         (subsf3_fpx): Likewise.
24411         (mulsf3_fpx): Likewise.
24412         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24413         asm operand.
24414         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24415         function attrbutes.
24416         * doc/invoke.texi (ARC): Document mjli-always option.
24418 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
24420         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24421         avoid addition with 0 and use incw and decw where possible.
24423 2018-01-26  Richard Biener  <rguenther@suse.de>
24425         PR tree-optimization/81082
24426         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24427         association if it requires casting to unsigned.
24428         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24429         from fold_plusminus_mult_expr to catch important cases late when
24430         range info is available.
24432 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24434         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24435         * configure.ac (hidden_linkonce): New test.
24436         * configure: Regenerate.
24437         * config.in: Regenerate.
24439 2018-01-26  Julia Koval  <julia.koval@intel.com>
24441         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
24442         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
24443         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
24444         _mm_mask_bitshuffle_epi64_mask): Fix type.
24445         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
24446         USI_FTYPE_V4DI_V4DI_USI): Remove.
24447         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
24448         __builtin_ia32_vpshufbitqmb256_mask,
24449         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
24450         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
24451         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
24453 2018-01-26  Alan Modra  <amodra@gmail.com>
24455         PR target/84033
24456         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
24457         UNSPEC_VBPERMQ.  Sort other unspecs.
24459 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
24461         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
24463 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
24465         PR middle-end/83055
24466         * predict.c (drop_profile): Do not push/pop cfun; update also
24467         node->count.
24468         (handle_missing_profiles): Fix logic looking for zero profiles.
24470 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
24472         PR middle-end/83977
24473         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
24474         on functions with #pragma omp declare simd or functions with simd
24475         attribute.
24476         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
24477         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
24478         Remove trailing \n from warning_at calls.
24480 2018-01-25  Tom de Vries  <tom@codesourcery.com>
24482         PR target/84028
24483         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24484         for neutered workers.
24486 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
24488         PR target/68467
24489         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
24490         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
24492 2018-01-24  Jeff Law  <law@redhat.com>
24494         PR target/83994
24495         * i386.c (get_probe_interval): Move to earlier point.
24496         (ix86_compute_frame_layout): If -fstack-clash-protection and
24497         the frame is larger than the probe interval, then use pushes
24498         to save registers rather than reg->mem moves.
24499         (ix86_expand_prologue): Remove conditional for int_registers_saved
24500         assertion.
24502 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
24504         PR target/84014
24505         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
24506         min/max for never referenced object.
24508 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
24510         PR middle-end/83977
24511         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
24512         here.
24513         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
24514         attributes from DECL_ATTRIBUTES (decl) without affecting
24515         DECL_ATTRIBUTES (current_function_decl).
24516         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
24517         functions with non-NULL DECL_ABSTRACT_ORIGIN.
24519 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
24521         PR tree-optimization/83979
24522         * fold-const.c (fold_comparison): Use constant_boolean_node
24523         instead of boolean_{true,false}_node.
24525 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
24527         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
24528         with zero counts.
24530 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24532         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24533         Simplify the clause that sets the length attribute.
24534         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24535         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
24536         clause that sets the length attribute.
24537         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24539 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24541         PR target/83589
24542         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
24543         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
24544         Add strict parameter.
24545         (prevent_branch_around_nothing): Insert dummy insn between branch to
24546         label and label with no ptx insn inbetween.
24547         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
24549 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24551         PR target/81352
24552         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24553         for neutered threads in warp.
24554         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
24556 2018-01-24  Richard Biener  <rguenther@suse.de>
24558         PR tree-optimization/83176
24559         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
24560         operands.
24562 2018-01-24  Richard Biener  <rguenther@suse.de>
24564         PR tree-optimization/82819
24565         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
24566         code generating pluses that are no-ops in the target precision.
24568 2018-01-24  Richard Biener  <rguenther@suse.de>
24570         PR middle-end/84000
24571         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
24573 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24575         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
24576         to merge probabilities.
24577         * predict.c (probably_never_executed): Also mark as cold functions
24578         with global 0 profile and guessed local profile.
24579         * profile-count.c (profile_probability::combine_with_count): New
24580         member function.
24581         * profile-count.h (profile_probability::operator*,
24582         profile_probability::operator*=, profile_probability::operator/,
24583         profile_probability::operator/=): Reduce precision to adjusted
24584         and set value to guessed on contradictory divisions.
24585         (profile_probability::combine_with_freq): Remove.
24586         (profile_probability::combine_wiht_count): Declare.
24587         (profile_count::force_nonzero):: Set to adjusted.
24588         (profile_count::probability_in):: Set quality to adjusted.
24589         * tree-ssa-tail-merge.c (replace_block_by): Use
24590         combine_with_count.
24592 2018-01-23  Andrew Waterman  <andrew@sifive.com>
24593             Jim Wilson  <jimw@sifive.com>
24595         * config/riscv/riscv.c (riscv_stack_boundary): New.
24596         (riscv_option_override): Set riscv_stack_boundary.  Handle
24597         riscv_preferred_stack_boundary_arg.
24598         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
24599         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
24600         (STACK_BOUNDARY): Set to riscv_stack_boundary.
24601         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
24602         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
24603         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
24605 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
24607         PR target/83905
24608         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
24609         of struct ix86_frame.
24610         (ix86_expand_epilogue): Likewise.  Add a local variable for
24611         the reg_save_offset field in struct ix86_frame.
24613 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
24615         PR tree-optimization/82604
24616         * tree-loop-distribution.c (enum partition_kind): New enum item
24617         PKIND_PARTIAL_MEMSET.
24618         (partition_builtin_p): Support above new enum item.
24619         (generate_code_for_partition): Ditto.
24620         (compute_access_range): Differentiate cases that equality can be
24621         proven at all loops, the innermost loops or no loops.
24622         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
24623         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
24624         (finalize_partitions, distribute_loop): Don't fuse partition of
24625         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
24626         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
24627         parloop is enabled.
24629 2018-01-23  Martin Liska  <mliska@suse.cz>
24631         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
24632         order to ignore the predictor.
24633         (PRED_POLYMORPHIC_CALL): Likewise.
24634         (PRED_RECURSIVE_CALL): Likewise.
24636 2018-01-23  Martin Liska  <mliska@suse.cz>
24638         * tree-profile.c (tree_profiling): Print function header to
24639         aware reader which function we are working on.
24640         * value-prof.c (gimple_find_values_to_profile): Do not print
24641         not interesting value histograms.
24643 2018-01-23  Martin Liska  <mliska@suse.cz>
24645         * profile-count.h (enum profile_quality): Add
24646         profile_uninitialized as the first value. Do not number values
24647         as they are zero based.
24648         (profile_count::verify): Update sanity check.
24649         (profile_probability::verify): Likewise.
24651 2018-01-23  Nathan Sidwell  <nathan@acm.org>
24653         * doc/invoke.texi (ffor-scope): Deprecate.
24655 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24657         PR tree-optimization/83510
24658         * domwalk.c (set_all_edges_as_executable): New function.
24659         (dom_walker::dom_walker): Convert bool param
24660         "skip_unreachable_blocks" to enum reachability.  Move setup of
24661         edge flags to set_all_edges_as_executable and only do it when
24662         reachability is REACHABLE_BLOCKS.
24663         * domwalk.h (enum dom_walker::reachability): New enum.
24664         (dom_walker::dom_walker): Convert bool param
24665         "skip_unreachable_blocks" to enum reachability.
24666         (set_all_edges_as_executable): New decl.
24667         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
24668         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
24669         "reachability".
24670         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
24671         but converting true to REACHABLE_BLOCKS.
24672         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
24673         * tree-vrp.c
24674         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
24675         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
24676         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
24677         REACHABLE_BLOCKS.
24678         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
24679         if check_all_array_refs will be called.
24681 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24683         * tree.c (selftest::test_location_wrappers): Add more test
24684         coverage.
24686 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24688         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
24689         (selftest::test_bit_in_range): Likewise.
24691 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
24693         PR testsuite/83888
24694         * doc/sourcebuild.texi (vect_float): Say that the selector
24695         only describes the situation when -funsafe-math-optimizations is on.
24696         (vect_float_strict): Document.
24698 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
24700         PR tree-optimization/83965
24701         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
24702         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
24703         instead of checking only for a reduction.
24704         (vect_recog_widen_sum_pattern): Likewise.
24706 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24708         * predict.c (probably_never_executed): Only use precise profile info.
24709         (compute_function_frequency): Skip after inlining hack since we now
24710         have quality checking.
24712 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24714         * profile-count.h (profile_probability::very_unlikely,
24715         profile_probability::unlikely, profile_probability::even): Set
24716         precision to guessed.
24718 2018-01-23  Richard Biener  <rguenther@suse.de>
24720         PR tree-optimization/83963
24721         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
24722         Properly terminate dominator walk when crossing the exit edge not
24723         when visiting its source block.
24725 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
24727         PR c++/83918
24728         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
24729         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
24731 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
24733         PR tree-optimization/83957
24734         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
24735         semicolon after for body surrounded by braces.
24737         PR tree-optimization/83081
24738         * profile-count.h (profile_probability::split): New method.
24739         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
24740         Use profile_probability::split.
24741         (do_compare_rtx_and_jump): Fix adjustment of probabilities
24742         when splitting a single conditional jump into 2.
24744 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
24746         PR tree-optimization/69452
24747         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
24748         decl.
24750 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24752         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
24753         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
24754         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
24756 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24758         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
24759         declaration.
24760         * config/rl78/rl78.md (movdi): New define_expand.
24761         * config/rl78/rl78.c (rl78_split_movdi): New function.
24763 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
24765         PR target/83862
24766         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
24767         no longer used.
24768         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
24769         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
24770         128-bit to produce an UNSPEC move to get the double word with the
24771         signbit and then a shift directly to do signbit.
24772         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
24773         implementation with a new version that just does either a direct
24774         move or a regular move.  Move memory interface to separate insns.
24775         Move insns so they are next to the expander.
24776         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
24777         with signbit move.  Split big and little endian case.
24778         (signbit<mode>2_dm_mem_le): Likewise.
24779         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
24780         (signbit<mode>2_dm2): Likewise.
24782 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24784         * config/rl78/rl78.md (anddi3): New define_expand.
24786 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24788         * config/rl78/rl78.md (umindi3): New define_expand.
24790 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24792         * config/rl78/rl78.md (smindi3): New define_expand.
24794 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24796         * config/rl78/rl78.md (smaxdi3): New define_expand.
24798 2018-01-22  Carl Love  <cel@us.ibm.com>
24800         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
24801         LVX_V1TI): Add macro expansion.
24802         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
24803         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
24804         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
24805         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
24806         Change check to determine if the instruction is a byte reversing
24807         entry.  Fix typo in comment.
24808         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
24809         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
24810         Add def_builtin calls for new builtins.
24811         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
24812         Add define_insn expansion.
24814 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24816         * config/rl78/rl78.md (umaxdi3): New define_expand.
24818 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24820         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
24821         for non-QImode registers.
24823 2018-01-22  Richard Biener  <rguenther@suse.de>
24825         PR tree-optimization/83963
24826         * graphite-scop-detection.c (scop_detection::get_sese): Delay
24827         including the loop exit block.
24828         (scop_detection::merge_sese): Likewise.
24829         (scop_detection::add_scop): Do it here instead.
24831 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24833         * doc/sourcebuild.texi (arm_softfloat): Document.
24835 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
24837         PR gcc/77734
24838         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
24839         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
24840         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
24842 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24843             David Edelsohn  <dje.gcc@gmail.com>
24845         PR target/83946
24846         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24847         Change "crset eq" to "crset 2".
24848         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24849         (*call_indirect_aix<mode>_nospec): Likewise.
24850         (*call_value_indirect_aix<mode>_nospec): Likewise.
24851         (*call_indirect_elfv2<mode>_nospec): Likewise.
24852         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
24853         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
24854         change assembly output from . to $.
24855         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24856         (indirect_jump<mode>_nospec): Change assembly output from . to $.
24857         (*tablejump<mode>_internal1_nospec): Likewise.
24859 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
24861         PR target/80870
24862         * config/sh/sh_optimize_sett_clrt.cc:
24863         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
24865 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
24867         PR tree-optimization/83940
24868         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
24869         offset_dt to vect_constant_def rather than vect_unknown_def_type.
24870         (vect_check_load_store_mask): Add a mask_dt_out parameter and
24871         use it to pass back the definition type.
24872         (vect_check_store_rhs): Likewise rhs_dt_out.
24873         (vect_build_gather_load_calls): Add a mask_dt argument and use
24874         it instead of a call to vect_is_simple_use.
24875         (vectorizable_store): Update calls to vect_check_load_store_mask
24876         and vect_check_store_rhs.  Use the dt returned by the latter instead
24877         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
24878         instead of calls to vect_is_simple_use.  Pass the scalar rather
24879         than the vector operand to vect_is_simple_use when handling
24880         second and subsequent copies of an rhs value.
24881         (vectorizable_load): Update calls to vect_check_load_store_mask
24882         and vect_build_gather_load_calls.  Use the cached mask_dt and
24883         gs_info.offset_dt instead of calls to vect_is_simple_use.
24885 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
24887         PR middle-end/83945
24888         * tree-emutls.c: Include gimplify.h.
24889         (lower_emutls_2): New function.
24890         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
24891         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
24892         it before further processing.
24894         PR target/83930
24895         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
24896         UINTVAL (trueop1) instead of INTVAL (op1).
24898 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
24900         PR debug/81570
24901         PR debug/83728
24902         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
24903         INCOMING_FRAME_SP_OFFSET if not defined.
24904         (scan_trace): Add ENTRY argument.  If true and
24905         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
24906         emit a note to adjust the CFA offset.
24907         (create_cfi_notes): Adjust scan_trace callers.
24908         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
24909         INCOMING_FRAME_SP_OFFSET in the CIE.
24910         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
24911         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
24912         Likewise.
24913         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
24914         * doc/tm.texi: Regenerated.
24916 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24918         PR rtl-optimization/83147
24919         * lra-constraints.c (remove_inheritance_pseudos): Use
24920         lra_substitute_pseudo_within_insn.
24922 2018-01-19  Tom de Vries  <tom@codesourcery.com>
24923             Cesar Philippidis  <cesar@codesourcery.com>
24925         PR target/83920
24926         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
24928 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
24930         PR target/83790
24931         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
24932         spaces for function labels.
24934 2018-01-19  Martin Liska  <mliska@suse.cz>
24936         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
24937         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
24938         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
24939         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
24940         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
24941         (PRED_CONST_RETURN): Change from 69 to 65.
24942         (PRED_NULL_RETURN): Change from 91 to 71.
24943         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
24944         (PRED_LOOP_GUARD): Change from 66 to 73.
24946 2018-01-19  Martin Liska  <mliska@suse.cz>
24948         * predict.c (predict_insn_def): Add new assert.
24949         (struct branch_predictor): Change type to signed integer.
24950         (test_prediction_value_range): Amend test to cover
24951         PROB_UNINITIALIZED.
24952         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
24953         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
24954         (PRED_LOOP_ITERATIONS_MAX): Likewise.
24955         (PRED_LOOP_IV_COMPARE): Likewise.
24956         * predict.h (PROB_UNINITIALIZED): Define new constant.
24958 2018-01-19  Martin Liska  <mliska@suse.cz>
24960         * predict.c (dump_prediction): Add new format for
24961         analyze_brprob.py script which is enabled with -details
24962         suboption.
24963         * profile-count.h (precise_p): New function.
24965 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24967         PR tree-optimization/83922
24968         * tree-vect-loop.c (vect_verify_full_masking): Return false if
24969         there are no statements that need masking.
24970         (vect_active_double_reduction_p): New function.
24971         (vect_analyze_loop_operations): Use it when handling phis that
24972         are not in the loop header.
24974 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24976         PR tree-optimization/83914
24977         * tree-vect-loop.c (vectorizable_induction): Don't convert
24978         init_expr or apply the peeling adjustment for inductions
24979         that are nested within the vectorized loop.
24981 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24983         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
24984         instead of NEG.
24986 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24988         PR sanitizer/81715
24989         PR testsuite/83882
24990         * function.h (gimplify_parameters): Add gimple_seq * argument.
24991         * function.c: Include gimple.h and options.h.
24992         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
24993         for the added local temporaries if needed.
24994         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
24995         if there are any parameter cleanups, wrap whole body into a
24996         try/finally with the cleanups.
24998 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
25000         PR target/82964
25001         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
25002         Use GET_MODE_CLASS for scalar floating point.
25004 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25006         PR ipa/82256
25007         patch by PaX Team
25008         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
25009         Fix call of call_cgraph_insertion_hooks.
25011 2018-01-18  Martin Sebor  <msebor@redhat.com>
25013         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
25015 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
25017         PR ipa/83619
25018         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
25019         frequencies.
25021 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
25023         PR other/70268
25024         * common.opt: (-ffile-prefix-map): New option.
25025         * opts.c (common_handle_option): Defer it.
25026         * opts-global.c (handle_common_deferred_options): Handle it.
25027         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
25028         * file-prefix-map.h: New file.
25029         (remap_debug_filename, add_debug_prefix_map): ...here.
25030         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
25031         * final.c (debug_prefix_map, add_debug_prefix_map
25032         remap_debug_filename): Move to...
25033         * file-prefix-map.c: New file.
25034         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
25035         generalize, get rid of alloca(), use strrchr() instead of strchr().
25036         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
25037         Implement in terms of add_prefix_map().
25038         (remap_macro_filename, remap_debug_filename): Implement in term of
25039         remap_filename().
25040         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
25041         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
25042         * dbxout.c: Include file-prefix-map.h.
25043         * varasm.c: Likewise.
25044         * vmsdbgout.c: Likewise.
25045         * xcoffout.c: Likewise.
25046         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
25047         * doc/cppopts.texi (-fmacro-prefix-map): Document.
25048         * doc/invoke.texi (-ffile-prefix-map): Document.
25049         (-fdebug-prefix-map): Update description.
25051 2018-01-18  Martin Liska  <mliska@suse.cz>
25053         * config/i386/i386.c (indirect_thunk_name): Document that also
25054         lfence is emitted.
25055         (output_indirect_thunk): Document why both instructions
25056         (pause and lfence) are generated.
25058 2018-01-18  Richard Biener  <rguenther@suse.de>
25060         PR tree-optimization/83887
25061         * graphite-scop-detection.c
25062         (scop_detection::get_nearest_dom_with_single_entry): Remove.
25063         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
25064         (scop_detection::merge_sese): Re-implement with a flood-fill
25065         algorithm that properly finds a SESE region if it exists.
25067 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
25069         PR c/61240
25070         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
25071         pointer_diff optimizations use view_convert instead of convert.
25073 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25075         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25076         Generate different code for -mno-speculate-indirect-jumps.
25077         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25078         (*call_indirect_aix<mode>): Disable for
25079         -mno-speculate-indirect-jumps.
25080         (*call_indirect_aix<mode>_nospec): New define_insn.
25081         (*call_value_indirect_aix<mode>): Disable for
25082         -mno-speculate-indirect-jumps.
25083         (*call_value_indirect_aix<mode>_nospec): New define_insn.
25084         (*sibcall_nonlocal_sysv<mode>): Generate different code for
25085         -mno-speculate-indirect-jumps.
25086         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25088 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
25090         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
25091         long double type, set the flags for noting the default long double
25092         type, even if we don't pass or return a long double type.
25094 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
25096         PR ipa/83051
25097         * ipa-inline.c (flatten_function): Do not overwrite final inlining
25098         failure.
25100 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
25102         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
25103         support for merge[hl].
25104         (fold_mergehl_helper): New helper function.
25105         (tree-vector-builder.h): New #include for tree_vector_builder usage.
25106         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
25107         (altivec_vmrglw_direct): Add xxmrglw insn.
25109 2018-01-17  Andrew Waterman  <andrew@sifive.com>
25111         * config/riscv/riscv.c (riscv_conditional_register_usage): If
25112         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
25114 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
25116         PR lto/83121
25117         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
25118         call the lto_location_cache before reading the
25119         DECL_SOURCE_LOCATION of the types.
25121 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
25122             Richard Sandiford  <richard.sandiford@linaro.org>
25124         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
25125         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
25126         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
25127         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
25128         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
25129         Add declaration.
25130         * config/aarch64/constraints.md (aarch64_movti_operand):
25131         Limit immediates.
25132         * config/aarch64/predicates.md (Uti): Add new constraint.
25134 2018-01-17  Carl Love  <cel@us.ibm.com>
25136         * config/rs6000/vsx.md (define_expand xl_len_r,
25137         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
25138         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
25139         lxvll.
25140         (define_expand, define_insn): Move the shift left from  the
25141         define_insn to the define_expand for lxvl and stxvl instructions.
25142         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
25143         and XL_LEN_R definitions to PURE.
25145 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
25147         * config/i386/i386.c (indirect_thunk_name): Declare regno
25148         as unsigned int.  Compare regno with INVALID_REGNUM.
25149         (output_indirect_thunk): Ditto.
25150         (output_indirect_thunk_function): Ditto.
25151         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
25152         in the call to output_indirect_thunk_function.
25154 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
25156         PR middle-end/83884
25157         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
25158         rather than the size of inner_type to determine the stack slot size
25159         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
25161 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
25163         PR target/83546
25164         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
25165         to PTA_SILVERMONT.
25167 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
25169         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
25170         endian Linux systems to optionally enable multilibs for selecting
25171         the long double type if the user configured an explicit type.
25172         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
25173         have no long double multilibs if not defined.
25174         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
25175         warn if the user used -mabi={ieee,ibm}longdouble and we built
25176         multilibs for long double.
25177         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
25178         appropriate multilib option.
25179         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
25180         multilib options.
25181         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
25182         for building long double multilibs.
25183         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
25185 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
25187         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
25188         copies.
25190         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
25191         64 bits.
25192         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
25193         128 bits.
25195         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
25196         variables.
25198         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
25199         return value.
25201 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25203         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
25204         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
25206 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25208         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
25209         different rtl trees depending on TARGET_64BIT.
25210         (rs6000_gen_lvx): Likewise.
25212 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
25214         * config/visium/visium.md (nop): Tweak comment.
25215         (hazard_nop): Likewise.
25217 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25219         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
25220         -mspeculate-indirect-jumps.
25221         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
25222         for -mno-speculate-indirect-jumps.
25223         (*call_indirect_elfv2<mode>_nospec): New define_insn.
25224         (*call_value_indirect_elfv2<mode>): Disable for
25225         -mno-speculate-indirect-jumps.
25226         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
25227         (indirect_jump): Emit different RTL for
25228         -mno-speculate-indirect-jumps.
25229         (*indirect_jump<mode>): Disable for
25230         -mno-speculate-indirect-jumps.
25231         (*indirect_jump<mode>_nospec): New define_insn.
25232         (tablejump): Emit different RTL for
25233         -mno-speculate-indirect-jumps.
25234         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
25235         (tablejumpsi_nospec): New define_expand.
25236         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
25237         (tablejumpdi_nospec): New define_expand.
25238         (*tablejump<mode>_internal1): Disable for
25239         -mno-speculate-indirect-jumps.
25240         (*tablejump<mode>_internal1_nospec): New define_insn.
25241         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
25242         option.
25244 2018-01-16  Artyom Skrobov tyomitch@gmail.com
25246         * caller-save.c (insert_save): Drop unnecessary parameter.  All
25247         callers updated.
25249 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25250             Richard Biener  <rguenth@suse.de>
25252         PR libgomp/83590
25253         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
25254         return early, inline manually is_gimple_sizepos.  Make sure if we
25255         call gimplify_expr we don't end up with a gimple constant.
25256         * tree.c (variably_modified_type_p): Don't return true for
25257         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
25258         * gimplify.h (is_gimple_sizepos): Remove.
25260 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25262         PR tree-optimization/83857
25263         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25264         vectorizable_live_operation for pure SLP statements.
25265         (vectorizable_live_operation): Handle PHIs.
25267 2018-01-16  Richard Biener  <rguenther@suse.de>
25269         PR tree-optimization/83867
25270         * tree-vect-stmts.c (vect_transform_stmt): Precompute
25271         nested_in_vect_loop_p since the scalar stmt may get invalidated.
25273 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25275         PR c/83844
25276         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25277         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25278         If off is not INTEGER_CST, issue a may not be aligned warning
25279         rather than isn't aligned.  Use isn%'t rather than isn't.
25280         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25281         into MULT_EXPR.
25282         <case MULT_EXPR>: Improve the case when bottom and one of the
25283         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25284         operand, in that case check if the other operand is multiple of
25285         bottom divided by the INTEGER_CST operand.
25287 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25289         PR target/83858
25290         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25291         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25292         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25293         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25294         * config/pa/pa.c (pa_function_arg_advance): Likewise.
25295         (pa_function_arg, pa_arg_partial_bytes): Likewise.
25296         (pa_function_arg_size): New function.
25298 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25300         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25301         in a separate statement.
25303 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25305         PR tree-optimization/83847
25306         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25307         group gathers and scatters.
25309 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25311         PR rtl-optimization/86620
25312         * params.def (max-sched-ready-insns): Bump minimum value to 1.
25314         PR rtl-optimization/83213
25315         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25316         to last if both are JUMP_INSNs.
25318         PR tree-optimization/83843
25319         * gimple-ssa-store-merging.c
25320         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25321         store_immediate_info for bswap/nop orig_stores.
25323 2018-01-15  Andrew Waterman  <andrew@sifive.com>
25325         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25326         !TARGET_MUL.
25327         <UDIV>: Increase cost if !TARGET_DIV.
25329 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
25331         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25332         (define_attr "cr_logical_3op"): New.
25333         (cceq_ior_compare): Adjust.
25334         (cceq_ior_compare_complement): Adjust.
25335         (*cceq_rev_compare): Adjust.
25336         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25337         (is_cracked_insn): Adjust.
25338         (insn_must_be_first_in_group): Adjust.
25339         * config/rs6000/40x.md: Adjust.
25340         * config/rs6000/440.md: Adjust.
25341         * config/rs6000/476.md: Adjust.
25342         * config/rs6000/601.md: Adjust.
25343         * config/rs6000/603.md: Adjust.
25344         * config/rs6000/6xx.md: Adjust.
25345         * config/rs6000/7450.md: Adjust.
25346         * config/rs6000/7xx.md: Adjust.
25347         * config/rs6000/8540.md: Adjust.
25348         * config/rs6000/cell.md: Adjust.
25349         * config/rs6000/e300c2c3.md: Adjust.
25350         * config/rs6000/e500mc.md: Adjust.
25351         * config/rs6000/e500mc64.md: Adjust.
25352         * config/rs6000/e5500.md: Adjust.
25353         * config/rs6000/e6500.md: Adjust.
25354         * config/rs6000/mpc.md: Adjust.
25355         * config/rs6000/power4.md: Adjust.
25356         * config/rs6000/power5.md: Adjust.
25357         * config/rs6000/power6.md: Adjust.
25358         * config/rs6000/power7.md: Adjust.
25359         * config/rs6000/power8.md: Adjust.
25360         * config/rs6000/power9.md: Adjust.
25361         * config/rs6000/rs64.md: Adjust.
25362         * config/rs6000/titan.md: Adjust.
25364 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25366         * config/i386/predicates.md (indirect_branch_operand): Rewrite
25367         ix86_indirect_branch_register logic.
25369 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25371         * config/i386/constraints.md (Bs): Update
25372         ix86_indirect_branch_register check.  Don't check
25373         ix86_indirect_branch_register with GOT_memory_operand.
25374         (Bw): Likewise.
25375         * config/i386/predicates.md (GOT_memory_operand): Don't check
25376         ix86_indirect_branch_register here.
25377         (GOT32_symbol_operand): Likewise.
25379 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25381         * config/i386/predicates.md (constant_call_address_operand):
25382         Rewrite ix86_indirect_branch_register logic.
25383         (sibcall_insn_operand): Likewise.
25385 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25387         * config/i386/constraints.md (Bs): Replace
25388         ix86_indirect_branch_thunk_register with
25389         ix86_indirect_branch_register.
25390         (Bw): Likewise.
25391         * config/i386/i386.md (indirect_jump): Likewise.
25392         (tablejump): Likewise.
25393         (*sibcall_memory): Likewise.
25394         (*sibcall_value_memory): Likewise.
25395         Peepholes of indirect call and jump via memory: Likewise.
25396         * config/i386/i386.opt: Likewise.
25397         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25398         (GOT_memory_operand): Likewise.
25399         (call_insn_operand): Likewise.
25400         (sibcall_insn_operand): Likewise.
25401         (GOT32_symbol_operand): Likewise.
25403 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25405         PR middle-end/83837
25406         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25407         type rather than type addr's type points to.
25408         (expand_omp_atomic_mutex): Likewise.
25409         (expand_omp_atomic): Likewise.
25411 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25413         PR target/83839
25414         * config/i386/i386.c (output_indirect_thunk_function): Use
25415         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25416         for  __x86_return_thunk.
25418 2018-01-15  Richard Biener  <rguenther@suse.de>
25420         PR middle-end/83850
25421         * expmed.c (extract_bit_field_1): Fix typo.
25423 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25425         PR target/83687
25426         * config/arm/iterators.md (VF): New mode iterator.
25427         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25428         Remove integer-related logic from pattern.
25429         (neon_vabd<mode>_3): Likewise.
25431 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25433         PR middle-end/82694
25434         * common.opt (fstrict-overflow): No longer an alias.
25435         (fwrapv-pointer): New option.
25436         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25437         also for pointer types based on flag_wrapv_pointer.
25438         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25439         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
25440         opts->x_flag_wrapv got set.
25441         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
25442         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
25443         POINTER_TYPE_OVERFLOW_UNDEFINED.
25444         * match.pd: Likewise in address comparison pattern.
25445         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
25447 2018-01-15  Richard Biener  <rguenther@suse.de>
25449         PR lto/83804
25450         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
25451         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
25452         Reset type names to their identifier if their TYPE_DECL doesn't
25453         have linkage (and thus is used for ODR and devirt).
25454         (save_debug_info_for_decl): Remove.
25455         (save_debug_info_for_type): Likewise.
25456         (add_tree_to_fld_list): Adjust.
25457         * tree-pretty-print.c (dump_generic_node): Make dumping of
25458         type names more robust.
25460 2018-01-15  Richard Biener  <rguenther@suse.de>
25462         * BASE-VER: Bump to 8.0.1.
25464 2018-01-14  Martin Sebor  <msebor@redhat.com>
25466         PR other/83508
25467         * builtins.c (check_access): Avoid warning when the no-warning bit
25468         is set.
25470 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
25472         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
25473         * ira-color (allocno_hard_regs_compare): Likewise.
25475 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
25477         PR target/83013
25478         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
25479         Use .pushsection/.popsection.
25481 2018-01-14  Martin Sebor  <msebor@redhat.com>
25483         PR c++/81327
25484         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
25486 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
25488         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
25489         entry from extra_headers.
25490         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
25491         extra_headers, make the list bitwise identical to the i?86-*-* one.
25493 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25495         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
25496         -mcmodel=large with -mindirect-branch=thunk,
25497         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
25498         -mfunction-return=thunk-extern.
25499         * doc/invoke.texi: Document -mcmodel=large is incompatible with
25500         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
25501         -mfunction-return=thunk and -mfunction-return=thunk-extern.
25503 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25505         * config/i386/i386.c (print_reg): Print the name of the full
25506         integer register without '%'.
25507         (ix86_print_operand): Handle 'V'.
25508         * doc/extend.texi: Document 'V' modifier.
25510 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25512         * config/i386/constraints.md (Bs): Disallow memory operand for
25513         -mindirect-branch-register.
25514         (Bw): Likewise.
25515         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25516         (GOT_memory_operand): Likewise.
25517         (call_insn_operand): Likewise.
25518         (sibcall_insn_operand): Likewise.
25519         (GOT32_symbol_operand): Likewise.
25520         * config/i386/i386.md (indirect_jump): Call convert_memory_address
25521         for -mindirect-branch-register.
25522         (tablejump): Likewise.
25523         (*sibcall_memory): Likewise.
25524         (*sibcall_value_memory): Likewise.
25525         Disallow peepholes of indirect call and jump via memory for
25526         -mindirect-branch-register.
25527         (*call_pop): Replace m with Bw.
25528         (*call_value_pop): Likewise.
25529         (*sibcall_pop_memory): Replace m with Bs.
25530         * config/i386/i386.opt (mindirect-branch-register): New option.
25531         * doc/invoke.texi: Document -mindirect-branch-register option.
25533 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25535         * config/i386/i386-protos.h (ix86_output_function_return): New.
25536         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
25537         set function_return_type.
25538         (indirect_thunk_name): Add ret_p to indicate thunk for function
25539         return.
25540         (output_indirect_thunk_function): Pass false to
25541         indirect_thunk_name.
25542         (ix86_output_indirect_branch_via_reg): Likewise.
25543         (ix86_output_indirect_branch_via_push): Likewise.
25544         (output_indirect_thunk_function): Create alias for function
25545         return thunk if regno < 0.
25546         (ix86_output_function_return): New function.
25547         (ix86_handle_fndecl_attribute): Handle function_return.
25548         (ix86_attribute_table): Add function_return.
25549         * config/i386/i386.h (machine_function): Add
25550         function_return_type.
25551         * config/i386/i386.md (simple_return_internal): Use
25552         ix86_output_function_return.
25553         (simple_return_internal_long): Likewise.
25554         * config/i386/i386.opt (mfunction-return=): New option.
25555         (indirect_branch): Mention -mfunction-return=.
25556         * doc/extend.texi: Document function_return function attribute.
25557         * doc/invoke.texi: Document -mfunction-return= option.
25559 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25561         * config/i386/i386-opts.h (indirect_branch): New.
25562         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
25563         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
25564         with local indirect jump when converting indirect call and jump.
25565         (ix86_set_indirect_branch_type): New.
25566         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
25567         (indirectlabelno): New.
25568         (indirect_thunk_needed): Likewise.
25569         (indirect_thunk_bnd_needed): Likewise.
25570         (indirect_thunks_used): Likewise.
25571         (indirect_thunks_bnd_used): Likewise.
25572         (INDIRECT_LABEL): Likewise.
25573         (indirect_thunk_name): Likewise.
25574         (output_indirect_thunk): Likewise.
25575         (output_indirect_thunk_function): Likewise.
25576         (ix86_output_indirect_branch_via_reg): Likewise.
25577         (ix86_output_indirect_branch_via_push): Likewise.
25578         (ix86_output_indirect_branch): Likewise.
25579         (ix86_output_indirect_jmp): Likewise.
25580         (ix86_code_end): Call output_indirect_thunk_function if needed.
25581         (ix86_output_call_insn): Call ix86_output_indirect_branch if
25582         needed.
25583         (ix86_handle_fndecl_attribute): Handle indirect_branch.
25584         (ix86_attribute_table): Add indirect_branch.
25585         * config/i386/i386.h (machine_function): Add indirect_branch_type
25586         and has_local_indirect_jump.
25587         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
25588         to true.
25589         (tablejump): Likewise.
25590         (*indirect_jump): Use ix86_output_indirect_jmp.
25591         (*tablejump_1): Likewise.
25592         (simple_return_indirect_internal): Likewise.
25593         * config/i386/i386.opt (mindirect-branch=): New option.
25594         (indirect_branch): New.
25595         (keep): Likewise.
25596         (thunk): Likewise.
25597         (thunk-inline): Likewise.
25598         (thunk-extern): Likewise.
25599         * doc/extend.texi: Document indirect_branch function attribute.
25600         * doc/invoke.texi: Document -mindirect-branch= option.
25602 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
25604         PR ipa/83051
25605         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
25607 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
25609         * ipa-inline.c (want_inline_small_function_p): Return false if
25610         inlining has already failed with CIF_FINAL_ERROR.
25611         (update_caller_keys): Call want_inline_small_function_p before
25612         can_inline_edge_p.
25613         (update_callee_keys): Likewise.
25615 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25617         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
25618         New function.
25619         (rs6000_quadword_masked_address_p): Likewise.
25620         (quad_aligned_load_p): Likewise.
25621         (quad_aligned_store_p): Likewise.
25622         (const_load_sequence_p): Add comment to describe the outer-most loop.
25623         (mimic_memory_attributes_and_flags): New function.
25624         (rs6000_gen_stvx): Likewise.
25625         (replace_swapped_aligned_store): Likewise.
25626         (rs6000_gen_lvx): Likewise.
25627         (replace_swapped_aligned_load): Likewise.
25628         (replace_swapped_load_constant): Capitalize argument name in
25629         comment describing this function.
25630         (rs6000_analyze_swaps): Add a third pass to search for vector loads
25631         and stores that access quad-word aligned addresses and replace
25632         with stvx or lvx instructions when appropriate.
25633         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
25634         New function prototype.
25635         (rs6000_quadword_masked_address_p): Likewise.
25636         (rs6000_gen_lvx): Likewise.
25637         (rs6000_gen_stvx): Likewise.
25638         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
25639         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
25640         when memory address is aligned.
25641         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
25642         this split to select lvx instruction when memory address is aligned.
25643         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
25644         instruction when memory address is aligned.
25645         (*vsx_le_perm_load_v16qi): Likewise.
25646         (four unnamed splitters): Modify to select the stvx instruction
25647         when memory is aligned.
25649 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
25651         * predict.c (determine_unlikely_bbs): Handle correctly BBs
25652         which appears in the queue multiple times.
25654 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25655             Alan Hayward  <alan.hayward@arm.com>
25656             David Sherwood  <david.sherwood@arm.com>
25658         * tree-vectorizer.h (vec_lower_bound): New structure.
25659         (_loop_vec_info): Add check_nonzero and lower_bounds.
25660         (LOOP_VINFO_CHECK_NONZERO): New macro.
25661         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
25662         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
25663         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
25664         fields.  Make seg_len the distance travelled, not including the
25665         access size.
25666         (dr_direction_indicator): Declare.
25667         (dr_zero_step_indicator): Likewise.
25668         (dr_known_forward_stride_p): Likewise.
25669         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
25670         tree-ssanames.h.
25671         (runtime_alias_check_p): Allow runtime alias checks with
25672         variable strides.
25673         (operator ==): Compare access_size and align.
25674         (prune_runtime_alias_test_list): Rework for new distinction between
25675         the access_size and seg_len.
25676         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
25677         segment lengths.
25678         (get_segment_min_max): New function.
25679         (create_intersect_range_checks): Use it.
25680         (dr_step_indicator): New function.
25681         (dr_direction_indicator): Likewise.
25682         (dr_zero_step_indicator): Likewise.
25683         (dr_known_forward_stride_p): Likewise.
25684         * tree-loop-distribution.c (data_ref_segment_size): Return
25685         DR_STEP * (niters - 1).
25686         (compute_alias_check_pairs): Update call to the dr_with_seg_len
25687         constructor.
25688         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
25689         (vect_preserves_scalar_order_p): New function, split out from...
25690         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
25691         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
25692         (vect_vfa_access_size): New function.
25693         (vect_vfa_align): Likewise.
25694         (vect_compile_time_alias): Take access_size_a and access_b arguments.
25695         (dump_lower_bound): New function.
25696         (vect_check_lower_bound): Likewise.
25697         (vect_small_gap_p): Likewise.
25698         (vectorizable_with_step_bound_p): Likewise.
25699         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
25700         depencies if the vectorization factor is 1.  Convert the checks
25701         for nonzero steps into checks on the bounds of DR_STEP.  Try using
25702         a bunds check for variable steps if the minimum required step is
25703         relatively small. Update calls to the dr_with_seg_len
25704         constructor and to vect_compile_time_alias.
25705         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
25706         function.
25707         (vect_loop_versioning): Call it.
25708         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
25709         when retrying.
25710         (vect_estimate_min_profitable_iters): Account for any bounds checks.
25712 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25713             Alan Hayward  <alan.hayward@arm.com>
25714             David Sherwood  <david.sherwood@arm.com>
25716         * doc/sourcebuild.texi (vect_scatter_store): Document.
25717         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
25718         optabs.
25719         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
25720         Document.
25721         * genopinit.c (main): Add supports_vec_scatter_store and
25722         supports_vec_scatter_store_cached to target_optabs.
25723         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
25724         IFN_MASK_SCATTER_STORE.
25725         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
25726         functions.
25727         * internal-fn.h (internal_store_fn_p): Declare.
25728         (internal_fn_stored_value_index): Likewise.
25729         * internal-fn.c (scatter_store_direct): New macro.
25730         (expand_scatter_store_optab_fn): New function.
25731         (direct_scatter_store_optab_supported_p): New macro.
25732         (internal_store_fn_p): New function.
25733         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
25734         IFN_MASK_SCATTER_STORE.
25735         (internal_fn_mask_index): Likewise.
25736         (internal_fn_stored_value_index): New function.
25737         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
25738         for scatter stores.
25739         * optabs-query.h (supports_vec_scatter_store_p): Declare.
25740         * optabs-query.c (supports_vec_scatter_store_p): New function.
25741         * tree-vectorizer.h (vect_get_store_rhs): Declare.
25742         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
25743         true for scatter stores.
25744         (vect_gather_scatter_fn_p): Handle scatter stores too.
25745         (vect_check_gather_scatter): Consider using scatter stores if
25746         supports_vec_scatter_store_p.
25747         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
25748         scatter stores too.
25749         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25750         internal_fn_stored_value_index.
25751         (check_load_store_masking): Handle scatter stores too.
25752         (vect_get_store_rhs): Make public.
25753         (vectorizable_call): Use internal_store_fn_p.
25754         (vectorizable_store): Handle scatter store internal functions.
25755         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
25756         when deciding whether the end of the group has been reached.
25757         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
25758         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
25759         (mask_scatter_store<mode>): New insns.
25761 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25762             Alan Hayward  <alan.hayward@arm.com>
25763             David Sherwood  <david.sherwood@arm.com>
25765         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
25766         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
25767         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
25768         function.
25769         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
25770         Use vect_truncate_gather_scatter_offset if we can't treat the
25771         operation as a normal gather load or scatter store.
25772         (get_group_load_store_type): Take the gather_scatter_info
25773         as argument.  Try using a gather load or scatter store for
25774         single-element groups.
25775         (get_load_store_type): Update calls to get_group_load_store_type
25776         and vect_use_strided_gather_scatters_p.
25778 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25779             Alan Hayward  <alan.hayward@arm.com>
25780             David Sherwood  <david.sherwood@arm.com>
25782         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
25783         optional tree argument.
25784         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
25785         null target hooks.
25786         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
25787         but continue to use the current value as a fallback.
25788         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
25789         to compare the updates.
25790         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
25791         (get_load_store_type): Use it when handling a strided access.
25792         (vect_get_strided_load_store_ops): New function.
25793         (vect_get_data_ptr_increment): Likewise.
25794         (vectorizable_load): Handle strided gather loads.  Always pass
25795         a step to vect_create_data_ref_ptr and bump_vector_ptr.
25797 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25798             Alan Hayward  <alan.hayward@arm.com>
25799             David Sherwood  <david.sherwood@arm.com>
25801         * doc/md.texi (gather_load@var{m}): Document.
25802         (mask_gather_load@var{m}): Likewise.
25803         * genopinit.c (main): Add supports_vec_gather_load and
25804         supports_vec_gather_load_cached to target_optabs.
25805         * optabs-tree.c (init_tree_optimization_optabs): Use
25806         ggc_cleared_alloc to allocate target_optabs.
25807         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
25808         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
25809         functions.
25810         * internal-fn.h (internal_load_fn_p): Declare.
25811         (internal_gather_scatter_fn_p): Likewise.
25812         (internal_fn_mask_index): Likewise.
25813         (internal_gather_scatter_fn_supported_p): Likewise.
25814         * internal-fn.c (gather_load_direct): New macro.
25815         (expand_gather_load_optab_fn): New function.
25816         (direct_gather_load_optab_supported_p): New macro.
25817         (direct_internal_fn_optab): New function.
25818         (internal_load_fn_p): Likewise.
25819         (internal_gather_scatter_fn_p): Likewise.
25820         (internal_fn_mask_index): Likewise.
25821         (internal_gather_scatter_fn_supported_p): Likewise.
25822         * optabs-query.c (supports_at_least_one_mode_p): New function.
25823         (supports_vec_gather_load_p): Likewise.
25824         * optabs-query.h (supports_vec_gather_load_p): Declare.
25825         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
25826         and memory_type field.
25827         (NUM_PATTERNS): Bump to 15.
25828         * tree-vect-data-refs.c: Include internal-fn.h.
25829         (vect_gather_scatter_fn_p): New function.
25830         (vect_describe_gather_scatter_call): Likewise.
25831         (vect_check_gather_scatter): Try using internal functions for
25832         gather loads.  Recognize existing calls to a gather load function.
25833         (vect_analyze_data_refs): Consider using gather loads if
25834         supports_vec_gather_load_p.
25835         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
25836         (vect_get_gather_scatter_offset_type): Likewise.
25837         (vect_convert_mask_for_vectype): Likewise.
25838         (vect_add_conversion_to_patterm): Likewise.
25839         (vect_try_gather_scatter_pattern): Likewise.
25840         (vect_recog_gather_scatter_pattern): New pattern recognizer.
25841         (vect_vect_recog_func_ptrs): Add it.
25842         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25843         internal_fn_mask_index and internal_gather_scatter_fn_p.
25844         (check_load_store_masking): Take the gather_scatter_info as an
25845         argument and handle gather loads.
25846         (vect_get_gather_scatter_ops): New function.
25847         (vectorizable_call): Check internal_load_fn_p.
25848         (vectorizable_load): Likewise.  Handle gather load internal
25849         functions.
25850         (vectorizable_store): Update call to check_load_store_masking.
25851         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
25852         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
25853         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
25854         (aarch64_gather_scale_operand_d): New predicates.
25855         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
25856         (mask_gather_load<mode>): New insns.
25858 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25859             Alan Hayward  <alan.hayward@arm.com>
25860             David Sherwood  <david.sherwood@arm.com>
25862         * optabs.def (fold_left_plus_optab): New optab.
25863         * doc/md.texi (fold_left_plus_@var{m}): Document.
25864         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
25865         * internal-fn.c (fold_left_direct): Define.
25866         (expand_fold_left_optab_fn): Likewise.
25867         (direct_fold_left_optab_supported_p): Likewise.
25868         * fold-const-call.c (fold_const_fold_left): New function.
25869         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
25870         * tree-parloops.c (valid_reduction_p): New function.
25871         (gather_scalar_reductions): Use it.
25872         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
25873         (vect_finish_replace_stmt): Declare.
25874         * tree-vect-loop.c (fold_left_reduction_fn): New function.
25875         (needs_fold_left_reduction_p): New function, split out from...
25876         (vect_is_simple_reduction): ...here.  Accept reductions that
25877         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
25878         (vect_force_simple_reduction): Also store the reduction type in
25879         the assignment's STMT_VINFO_REDUC_TYPE.
25880         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
25881         (merge_with_identity): New function.
25882         (vect_expand_fold_left): Likewise.
25883         (vectorize_fold_left_reduction): Likewise.
25884         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
25885         scalar phi in place for it.  Check for target support and reject
25886         cases that would reassociate the operation.  Defer the transform
25887         phase to vectorize_fold_left_reduction.
25888         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
25889         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
25890         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
25892 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25894         * tree-if-conv.c (predicate_mem_writes): Remove redundant
25895         call to ifc_temp_var.
25897 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25898             Alan Hayward  <alan.hayward@arm.com>
25899             David Sherwood  <david.sherwood@arm.com>
25901         * target.def (legitimize_address_displacement): Take the original
25902         offset as a poly_int.
25903         * targhooks.h (default_legitimize_address_displacement): Update
25904         accordingly.
25905         * targhooks.c (default_legitimize_address_displacement): Likewise.
25906         * doc/tm.texi: Regenerate.
25907         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
25908         as an argument, moving assert of ad->disp == ad->disp_term to...
25909         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
25910         Try calling targetm.legitimize_address_displacement before expanding
25911         the address rather than afterwards, and adjust for the new interface.
25912         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
25913         Match the new hook interface.  Handle SVE addresses.
25914         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
25915         new hook interface.
25917 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25919         * Makefile.in (OBJS): Add early-remat.o.
25920         * target.def (select_early_remat_modes): New hook.
25921         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
25922         * doc/tm.texi: Regenerate.
25923         * targhooks.h (default_select_early_remat_modes): Declare.
25924         * targhooks.c (default_select_early_remat_modes): New function.
25925         * timevar.def (TV_EARLY_REMAT): New timevar.
25926         * passes.def (pass_early_remat): New pass.
25927         * tree-pass.h (make_pass_early_remat): Declare.
25928         * early-remat.c: New file.
25929         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
25930         function.
25931         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
25933 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25934             Alan Hayward  <alan.hayward@arm.com>
25935             David Sherwood  <david.sherwood@arm.com>
25937         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
25938         vfm1 with a bound_epilog parameter.
25939         (vect_do_peeling): Update calls accordingly, and move the prologue
25940         call earlier in the function.  Treat the base bound_epilog as 0 for
25941         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
25942         this base when peeling for gaps.
25943         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
25944         with fully-masked loops.
25945         (vect_estimate_min_profitable_iters): Handle the single peeled
25946         iteration in that case.
25948 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25949             Alan Hayward  <alan.hayward@arm.com>
25950             David Sherwood  <david.sherwood@arm.com>
25952         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
25953         single-element interleaving even if the size is not a power of 2.
25954         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
25955         accesses for single-element interleaving if the group size is
25956         not a power of 2.
25958 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25959             Alan Hayward  <alan.hayward@arm.com>
25960             David Sherwood  <david.sherwood@arm.com>
25962         * doc/md.texi (fold_extract_last_@var{m}): Document.
25963         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
25964         * optabs.def (fold_extract_last_optab): New optab.
25965         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
25966         * internal-fn.c (fold_extract_direct): New macro.
25967         (expand_fold_extract_optab_fn): Likewise.
25968         (direct_fold_extract_optab_supported_p): Likewise.
25969         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
25970         * tree-vect-loop.c (vect_model_reduction_cost): Handle
25971         EXTRACT_LAST_REDUCTION.
25972         (get_initial_def_for_reduction): Do not create an initial vector
25973         for EXTRACT_LAST_REDUCTION reductions.
25974         (vectorizable_reduction): Leave the scalar phi in place for
25975         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
25976         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
25977         epilogue code for EXTRACT_LAST_REDUCTION and defer the
25978         transform phase to vectorizable_condition.
25979         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
25980         split out from...
25981         (vect_finish_stmt_generation): ...here.
25982         (vect_finish_replace_stmt): New function.
25983         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
25984         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
25985         pattern.
25986         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
25988 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25989             Alan Hayward  <alan.hayward@arm.com>
25990             David Sherwood  <david.sherwood@arm.com>
25992         * doc/md.texi (extract_last_@var{m}): Document.
25993         * optabs.def (extract_last_optab): New optab.
25994         * internal-fn.def (EXTRACT_LAST): New internal function.
25995         * internal-fn.c (cond_unary_direct): New macro.
25996         (expand_cond_unary_optab_fn): Likewise.
25997         (direct_cond_unary_optab_supported_p): Likewise.
25998         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
25999         loops using EXTRACT_LAST.
26000         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
26001         (extract_last_<mode>): ...this optab.
26002         (vec_extract<mode><Vel>): Update accordingly.
26004 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26005             Alan Hayward  <alan.hayward@arm.com>
26006             David Sherwood  <david.sherwood@arm.com>
26008         * target.def (empty_mask_is_expensive): New hook.
26009         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
26010         * doc/tm.texi: Regenerate.
26011         * targhooks.h (default_empty_mask_is_expensive): Declare.
26012         * targhooks.c (default_empty_mask_is_expensive): New function.
26013         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
26014         if the target says that empty masks are expensive.
26015         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
26016         New function.
26017         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
26019 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26020             Alan Hayward  <alan.hayward@arm.com>
26021             David Sherwood  <david.sherwood@arm.com>
26023         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
26024         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
26025         (vect_use_loop_mask_for_alignment_p): New function.
26026         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
26027         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
26028         niters_skip argument.  Make sure that the first niters_skip elements
26029         of the first iteration are inactive.
26030         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
26031         Update call to vect_set_loop_masks_directly.
26032         (get_misalign_in_elems): New function, split out from...
26033         (vect_gen_prolog_loop_niters): ...here.
26034         (vect_update_init_of_dr): Take a code argument that specifies whether
26035         the adjustment should be added or subtracted.
26036         (vect_update_init_of_drs): Likewise.
26037         (vect_prepare_for_masked_peels): New function.
26038         (vect_do_peeling): Skip prologue peeling if we're using a mask
26039         instead.  Update call to vect_update_inits_of_drs.
26040         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26041         mask_skip_niters.
26042         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
26043         alignment.  Do not include the number of peeled iterations in
26044         the minimum threshold in that case.
26045         (vectorizable_induction): Adjust the start value down by
26046         LOOP_VINFO_MASK_SKIP_NITERS iterations.
26047         (vect_transform_loop): Call vect_prepare_for_masked_peels.
26048         Take the number of skipped iterations into account when calculating
26049         the loop bounds.
26050         * tree-vect-stmts.c (vect_gen_while_not): New function.
26052 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26053             Alan Hayward  <alan.hayward@arm.com>
26054             David Sherwood  <david.sherwood@arm.com>
26056         * doc/sourcebuild.texi (vect_fully_masked): Document.
26057         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
26058         default value to 0.
26059         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
26060         split out from...
26061         (vect_analyze_loop_2): ...here. Don't check the vectorization
26062         factor against the number of loop iterations if the loop is
26063         fully-masked.
26065 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26066             Alan Hayward  <alan.hayward@arm.com>
26067             David Sherwood  <david.sherwood@arm.com>
26069         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
26070         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
26071         (dump_groups): Update accordingly.
26072         (iv_use::mem_type): New member variable.
26073         (address_p): New function.
26074         (record_use): Add a mem_type argument and initialize the new
26075         mem_type field.
26076         (record_group_use): Add a mem_type argument.  Use address_p.
26077         Remove obsolete null checks of base_object.  Update call to record_use.
26078         (find_interesting_uses_op): Update call to record_group_use.
26079         (find_interesting_uses_cond): Likewise.
26080         (find_interesting_uses_address): Likewise.
26081         (get_mem_type_for_internal_fn): New function.
26082         (find_address_like_use): Likewise.
26083         (find_interesting_uses_stmt): Try find_address_like_use before
26084         calling find_interesting_uses_op.
26085         (addr_offset_valid_p): Use the iv mem_type field as the type
26086         of the addressed memory.
26087         (add_autoinc_candidates): Likewise.
26088         (get_address_cost): Likewise.
26089         (split_small_address_groups_p): Use address_p.
26090         (split_address_groups): Likewise.
26091         (add_iv_candidate_for_use): Likewise.
26092         (autoinc_possible_for_pair): Likewise.
26093         (rewrite_groups): Likewise.
26094         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
26095         (determine_group_iv_cost): Update after split of USE_ADDRESS.
26096         (get_alias_ptr_type_for_ptr_address): New function.
26097         (rewrite_use_address): Rewrite address uses in calls that were
26098         identified by find_address_like_use.
26100 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26101             Alan Hayward  <alan.hayward@arm.com>
26102             David Sherwood  <david.sherwood@arm.com>
26104         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
26105         TARGET_MEM_REFs.
26106         * gimple-expr.h (is_gimple_addressable: Likewise.
26107         * gimple-expr.c (is_gimple_address): Likewise.
26108         * internal-fn.c (expand_call_mem_ref): New function.
26109         (expand_mask_load_optab_fn): Use it.
26110         (expand_mask_store_optab_fn): Likewise.
26112 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26113             Alan Hayward  <alan.hayward@arm.com>
26114             David Sherwood  <david.sherwood@arm.com>
26116         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
26117         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
26118         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
26119         (cond_umax@var{mode}): Document.
26120         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
26121         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
26122         (cond_umin_optab, cond_umax_optab): New optabs.
26123         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
26124         (COND_IOR, COND_XOR): New internal functions.
26125         * internal-fn.h (get_conditional_internal_fn): Declare.
26126         * internal-fn.c (cond_binary_direct): New macro.
26127         (expand_cond_binary_optab_fn): Likewise.
26128         (direct_cond_binary_optab_supported_p): Likewise.
26129         (get_conditional_internal_fn): New function.
26130         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
26131         Cope with reduction statements that are vectorized as calls rather
26132         than assignments.
26133         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
26134         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
26135         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
26136         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
26137         (UNSPEC_COND_EOR): New unspecs.
26138         (optab): Add mappings for them.
26139         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
26140         (sve_int_op, sve_fp_op): New int attributes.
26142 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26143             Alan Hayward  <alan.hayward@arm.com>
26144             David Sherwood  <david.sherwood@arm.com>
26146         * optabs.def (while_ult_optab): New optab.
26147         * doc/md.texi (while_ult@var{m}@var{n}): Document.
26148         * internal-fn.def (WHILE_ULT): New internal function.
26149         * internal-fn.h (direct_internal_fn_supported_p): New override
26150         that takes two types as argument.
26151         * internal-fn.c (while_direct): New macro.
26152         (expand_while_optab_fn): New function.
26153         (convert_optab_supported_p): Likewise.
26154         (direct_while_optab_supported_p): New macro.
26155         * wide-int.h (wi::udiv_ceil): New function.
26156         * tree-vectorizer.h (rgroup_masks): New structure.
26157         (vec_loop_masks): New typedef.
26158         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
26159         and fully_masked_p.
26160         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
26161         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
26162         (vect_max_vf): New function.
26163         (slpeel_make_loop_iterate_ntimes): Delete.
26164         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
26165         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
26166         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
26167         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
26168         internal-fn.h, stor-layout.h and optabs-query.h.
26169         (vect_set_loop_mask): New function.
26170         (add_preheader_seq): Likewise.
26171         (add_header_seq): Likewise.
26172         (interleave_supported_p): Likewise.
26173         (vect_maybe_permute_loop_masks): Likewise.
26174         (vect_set_loop_masks_directly): Likewise.
26175         (vect_set_loop_condition_masked): Likewise.
26176         (vect_set_loop_condition_unmasked): New function, split out from
26177         slpeel_make_loop_iterate_ntimes.
26178         (slpeel_make_loop_iterate_ntimes): Rename to..
26179         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
26180         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
26181         (vect_do_peeling): Update call accordingly.
26182         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
26183         loops.
26184         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26185         mask_compare_type, can_fully_mask_p and fully_masked_p.
26186         (release_vec_loop_masks): New function.
26187         (_loop_vec_info): Use it to free the loop masks.
26188         (can_produce_all_loop_masks_p): New function.
26189         (vect_get_max_nscalars_per_iter): Likewise.
26190         (vect_verify_full_masking): Likewise.
26191         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
26192         retries, and free the mask rgroups before retrying.  Check loop-wide
26193         reasons for disallowing fully-masked loops.  Make the final decision
26194         about whether use a fully-masked loop or not.
26195         (vect_estimate_min_profitable_iters): Do not assume that peeling
26196         for the number of iterations will be needed for fully-masked loops.
26197         (vectorizable_reduction): Disable fully-masked loops.
26198         (vectorizable_live_operation): Likewise.
26199         (vect_halve_mask_nunits): New function.
26200         (vect_double_mask_nunits): Likewise.
26201         (vect_record_loop_mask): Likewise.
26202         (vect_get_loop_mask): Likewise.
26203         (vect_transform_loop): Handle the case in which the final loop
26204         iteration might handle a partial vector.  Call vect_set_loop_condition
26205         instead of slpeel_make_loop_iterate_ntimes.
26206         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
26207         (check_load_store_masking): New function.
26208         (prepare_load_store_mask): Likewise.
26209         (vectorizable_store): Handle fully-masked loops.
26210         (vectorizable_load): Likewise.
26211         (supportable_widening_operation): Use vect_halve_mask_nunits for
26212         booleans.
26213         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
26214         (vect_gen_while): New function.
26215         * config/aarch64/aarch64.md (umax<mode>3): New expander.
26216         (aarch64_uqdec<mode>): New insn.
26218 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26219             Alan Hayward  <alan.hayward@arm.com>
26220             David Sherwood  <david.sherwood@arm.com>
26222         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
26223         (reduc_xor_scal_optab): New optabs.
26224         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
26225         (reduc_xor_scal_@var{m}): Document.
26226         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
26227         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
26228         internal functions.
26229         * fold-const-call.c (fold_const_call): Handle them.
26230         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
26231         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
26232         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
26233         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
26234         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
26235         (UNSPEC_XORV): New unspecs.
26236         (optab): Add entries for them.
26237         (BITWISEV): New int iterator.
26238         (bit_reduc_op): New int attributes.
26240 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26241             Alan Hayward  <alan.hayward@arm.com>
26242             David Sherwood  <david.sherwood@arm.com>
26244         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
26245         * internal-fn.def (VEC_SHL_INSERT): New internal function.
26246         * optabs.def (vec_shl_insert_optab): New optab.
26247         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
26248         (duplicate_and_interleave): Likewise.
26249         * tree-vect-loop.c: Include internal-fn.h.
26250         (neutral_op_for_slp_reduction): New function, split out from
26251         get_initial_defs_for_reduction.
26252         (get_initial_def_for_reduction): Handle option 2 for variable-length
26253         vectors by loading the neutral value into a vector and then shifting
26254         the initial value into element 0.
26255         (get_initial_defs_for_reduction): Replace the code argument with
26256         the neutral value calculated by neutral_op_for_slp_reduction.
26257         Use gimple_build_vector for constant-length vectors.
26258         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26259         but the first group_size elements have a neutral value.
26260         Use duplicate_and_interleave otherwise.
26261         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26262         Update call to get_initial_defs_for_reduction.  Handle SLP
26263         reductions for variable-length vectors by creating one vector
26264         result for each scalar result, with the elements associated
26265         with other scalar results stubbed out with the neutral value.
26266         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26267         Require IFN_VEC_SHL_INSERT for double reductions on
26268         variable-length vectors, or SLP reductions that have
26269         a neutral value.  Require can_duplicate_and_interleave_p
26270         support for variable-length unchained SLP reductions if there
26271         is no neutral value, such as for MIN/MAX reductions.  Also require
26272         the number of vector elements to be a multiple of the number of
26273         SLP statements when doing variable-length unchained SLP reductions.
26274         Update call to vect_create_epilog_for_reduction.
26275         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26276         and remove initial values.
26277         (duplicate_and_interleave): Make public.
26278         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26279         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26281 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26282             Alan Hayward  <alan.hayward@arm.com>
26283             David Sherwood  <david.sherwood@arm.com>
26285         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26286         (can_duplicate_and_interleave_p): New function.
26287         (vect_get_and_check_slp_defs): Take the vector of statements
26288         rather than just the current one.  Remove excess parentheses.
26289         Restriction rejectinon of vect_constant_def and vect_external_def
26290         for variable-length vectors to boolean types, or types for which
26291         can_duplicate_and_interleave_p is false.
26292         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26293         (duplicate_and_interleave): New function.
26294         (vect_get_constant_vectors): Use gimple_build_vector for
26295         constant-length vectors and suitable variable-length constant
26296         vectors.  Use duplicate_and_interleave for other variable-length
26297         vectors.  Don't defer the update when inserting new statements.
26299 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26300             Alan Hayward  <alan.hayward@arm.com>
26301             David Sherwood  <david.sherwood@arm.com>
26303         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26304         min_profitable_iters doesn't go negative.
26306 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26307             Alan Hayward  <alan.hayward@arm.com>
26308             David Sherwood  <david.sherwood@arm.com>
26310         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26311         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26312         * optabs.def (vec_mask_load_lanes_optab): New optab.
26313         (vec_mask_store_lanes_optab): Likewise.
26314         * internal-fn.def (MASK_LOAD_LANES): New internal function.
26315         (MASK_STORE_LANES): Likewise.
26316         * internal-fn.c (mask_load_lanes_direct): New macro.
26317         (mask_store_lanes_direct): Likewise.
26318         (expand_mask_load_optab_fn): Handle masked operations.
26319         (expand_mask_load_lanes_optab_fn): New macro.
26320         (expand_mask_store_optab_fn): Handle masked operations.
26321         (expand_mask_store_lanes_optab_fn): New macro.
26322         (direct_mask_load_lanes_optab_supported_p): Likewise.
26323         (direct_mask_store_lanes_optab_supported_p): Likewise.
26324         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26325         parameter.
26326         (vect_load_lanes_supported): Likewise.
26327         * tree-vect-data-refs.c (strip_conversion): New function.
26328         (can_group_stmts_p): Likewise.
26329         (vect_analyze_data_ref_accesses): Use it instead of checking
26330         for a pair of assignments.
26331         (vect_store_lanes_supported): Take a masked_p parameter.
26332         (vect_load_lanes_supported): Likewise.
26333         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26334         vect_store_lanes_supported and vect_load_lanes_supported.
26335         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26336         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26337         parameter.  Don't allow gaps for masked accesses.
26338         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
26339         and vect_load_lanes_supported.
26340         (get_load_store_type): Take a masked_p parameter and update
26341         call to get_group_load_store_type.
26342         (vectorizable_store): Update call to get_load_store_type.
26343         Handle IFN_MASK_STORE_LANES.
26344         (vectorizable_load): Update call to get_load_store_type.
26345         Handle IFN_MASK_LOAD_LANES.
26347 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26348             Alan Hayward  <alan.hayward@arm.com>
26349             David Sherwood  <david.sherwood@arm.com>
26351         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26352         modes for SVE.
26353         * config/aarch64/aarch64-protos.h
26354         (aarch64_sve_struct_memory_operand_p): Declare.
26355         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26356         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26357         (VPRED, vpred): Handle SVE structure modes.
26358         * config/aarch64/constraints.md (Utx): New constraint.
26359         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26360         (aarch64_sve_struct_nonimmediate_operand): New predicates.
26361         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26362         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26363         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26364         structure modes.  Split into pieces after RA.
26365         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26366         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26367         New patterns.
26368         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26369         SVE structure modes.
26370         (aarch64_classify_address): Likewise.
26371         (sizetochar): Move earlier in file.
26372         (aarch64_print_operand): Handle SVE register lists.
26373         (aarch64_array_mode): New function.
26374         (aarch64_sve_struct_memory_operand_p): Likewise.
26375         (TARGET_ARRAY_MODE): Redefine.
26377 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26378             Alan Hayward  <alan.hayward@arm.com>
26379             David Sherwood  <david.sherwood@arm.com>
26381         * target.def (array_mode): New target hook.
26382         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26383         * doc/tm.texi: Regenerate.
26384         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26385         * hooks.c (hook_optmode_mode_uhwi_none): New function.
26386         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26387         targetm.array_mode.
26388         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
26389         type sizes.
26391 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26392             Alan Hayward  <alan.hayward@arm.com>
26393             David Sherwood  <david.sherwood@arm.com>
26395         * fold-const.c (fold_binary_loc): Check the argument types
26396         rather than the result type when testing for a vector operation.
26398 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26400         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26401         * doc/tm.texi: Regenerate.
26403 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26404             Alan Hayward  <alan.hayward@arm.com>
26405             David Sherwood  <david.sherwood@arm.com>
26407         * doc/invoke.texi (-msve-vector-bits=): Document new option.
26408         (sve): Document new AArch64 extension.
26409         * doc/md.texi (w): Extend the description of the AArch64
26410         constraint to include SVE vectors.
26411         (Upl, Upa): Document new AArch64 predicate constraints.
26412         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26413         enum.
26414         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26415         (msve-vector-bits=): New option.
26416         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26417         SVE when these are disabled.
26418         (sve): New extension.
26419         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26420         modes.  Adjust their number of units based on aarch64_sve_vg.
26421         (MAX_BITSIZE_MODE_ANY_MODE): Define.
26422         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26423         aarch64_addr_query_type.
26424         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26425         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26426         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26427         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26428         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26429         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26430         (aarch64_simd_imm_zero_p): Delete.
26431         (aarch64_check_zero_based_sve_index_immediate): Declare.
26432         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26433         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26434         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26435         (aarch64_sve_float_mul_immediate_p): Likewise.
26436         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26437         rather than an rtx.
26438         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26439         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
26440         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
26441         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
26442         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
26443         (aarch64_regmode_natural_size): Likewise.
26444         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
26445         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
26446         left one place.
26447         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
26448         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
26449         for VG and the SVE predicate registers.
26450         (V_ALIASES): Add a "z"-prefixed alias.
26451         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
26452         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
26453         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
26454         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
26455         (REG_CLASS_NAMES): Add entries for them.
26456         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
26457         and the predicate registers.
26458         (aarch64_sve_vg): Declare.
26459         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
26460         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
26461         (REGMODE_NATURAL_SIZE): Define.
26462         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
26463         SVE macros.
26464         * config/aarch64/aarch64.c: Include cfgrtl.h.
26465         (simd_immediate_info): Add a constructor for series vectors,
26466         and an associated step field.
26467         (aarch64_sve_vg): New variable.
26468         (aarch64_dbx_register_number): Handle VG and the predicate registers.
26469         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
26470         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
26471         (VEC_ANY_DATA, VEC_STRUCT): New constants.
26472         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
26473         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
26474         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
26475         (aarch64_get_mask_mode): New functions.
26476         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
26477         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
26478         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
26479         predicate modes and predicate registers.  Explicitly restrict
26480         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
26481         to store a vector mode if it is recognized by
26482         aarch64_classify_vector_mode.
26483         (aarch64_regmode_natural_size): New function.
26484         (aarch64_hard_regno_caller_save_mode): Return the original mode
26485         for predicates.
26486         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
26487         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
26488         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
26489         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
26490         functions.
26491         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
26492         does not overlap dest if the function is frame-related.  Handle
26493         SVE constants.
26494         (aarch64_split_add_offset): New function.
26495         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
26496         them aarch64_add_offset.
26497         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
26498         and update call to aarch64_sub_sp.
26499         (aarch64_add_cfa_expression): New function.
26500         (aarch64_expand_prologue): Pass extra temporary registers to the
26501         functions above.  Handle the case in which we need to emit new
26502         DW_CFA_expressions for registers that were originally saved
26503         relative to the stack pointer, but now have to be expressed
26504         relative to the frame pointer.
26505         (aarch64_output_mi_thunk): Pass extra temporary registers to the
26506         functions above.
26507         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
26508         IP0 and IP1 values for SVE frames.
26509         (aarch64_expand_vec_series): New function.
26510         (aarch64_expand_sve_widened_duplicate): Likewise.
26511         (aarch64_expand_sve_const_vector): Likewise.
26512         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
26513         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
26514         into the register, rather than emitting a SET directly.
26515         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
26516         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
26517         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
26518         (offset_9bit_signed_scaled_p): New functions.
26519         (aarch64_replicate_bitmask_imm): New function.
26520         (aarch64_bitmask_imm): Use it.
26521         (aarch64_cannot_force_const_mem): Reject expressions involving
26522         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
26523         (aarch64_classify_index): Handle SVE indices, by requiring
26524         a plain register index with a scale that matches the element size.
26525         (aarch64_classify_address): Handle SVE addresses.  Assert that
26526         the mode of the address is VOIDmode or an integer mode.
26527         Update call to aarch64_classify_symbol.
26528         (aarch64_classify_symbolic_expression): Update call to
26529         aarch64_classify_symbol.
26530         (aarch64_const_vec_all_in_range_p): New function.
26531         (aarch64_print_vector_float_operand): Likewise.
26532         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
26533         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
26534         and the FP immediates 1.0 and 0.5.
26535         (aarch64_print_address_internal): Handle SVE addresses.
26536         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
26537         (aarch64_regno_regclass): Handle predicate registers.
26538         (aarch64_secondary_reload): Handle big-endian reloads of SVE
26539         data modes.
26540         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
26541         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
26542         (aarch64_convert_sve_vector_bits): New function.
26543         (aarch64_override_options): Use it to handle -msve-vector-bits=.
26544         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26545         rather than an rtx.
26546         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
26547         Handle SVE vector and predicate modes.  Accept VL-based constants
26548         that need only one temporary register, and VL offsets that require
26549         no temporary registers.
26550         (aarch64_conditional_register_usage): Mark the predicate registers
26551         as fixed if SVE isn't available.
26552         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
26553         Return true for SVE vector and predicate modes.
26554         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
26555         rather than an unsigned int.  Handle SVE modes.
26556         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
26557         SVE modes.
26558         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
26559         if SVE is enabled.
26560         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26561         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26562         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26563         (aarch64_sve_float_mul_immediate_p): New functions.
26564         (aarch64_sve_valid_immediate): New function.
26565         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
26566         Explicitly reject structure modes.  Check for INDEX constants.
26567         Handle PTRUE and PFALSE constants.
26568         (aarch64_check_zero_based_sve_index_immediate): New function.
26569         (aarch64_simd_imm_zero_p): Delete.
26570         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
26571         vector modes.  Accept constants in the range of CNT[BHWD].
26572         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
26573         ask for an Advanced SIMD mode.
26574         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
26575         (aarch64_simd_vector_alignment): Handle SVE predicates.
26576         (aarch64_vectorize_preferred_vector_alignment): New function.
26577         (aarch64_simd_vector_alignment_reachable): Use it instead of
26578         the vector size.
26579         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
26580         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
26581         functions.
26582         (MAX_VECT_LEN): Delete.
26583         (expand_vec_perm_d): Add a vec_flags field.
26584         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
26585         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
26586         (aarch64_evpc_ext): Don't apply a big-endian lane correction
26587         for SVE modes.
26588         (aarch64_evpc_rev): Rename to...
26589         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
26590         (aarch64_evpc_rev_global): New function.
26591         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
26592         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
26593         MAX_VECT_LEN.
26594         (aarch64_evpc_sve_tbl): New function.
26595         (aarch64_expand_vec_perm_const_1): Update after rename of
26596         aarch64_evpc_rev.  Handle SVE permutes too, trying
26597         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
26598         than aarch64_evpc_tbl.
26599         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
26600         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
26601         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
26602         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
26603         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
26604         (aarch64_expand_sve_vcond): New functions.
26605         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
26606         of aarch64_vector_mode_p.
26607         (aarch64_dwarf_poly_indeterminate_value): New function.
26608         (aarch64_compute_pressure_classes): Likewise.
26609         (aarch64_can_change_mode_class): Likewise.
26610         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
26611         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
26612         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
26613         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
26614         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
26615         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
26616         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
26617         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
26618         constraints.
26619         (Dn, Dl, Dr): Accept const as well as const_vector.
26620         (Dz): Likewise.  Compare against CONST0_RTX.
26621         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
26622         of "vector" where appropriate.
26623         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
26624         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
26625         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
26626         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
26627         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
26628         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
26629         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
26630         (v_int_equiv): Extend to SVE modes.
26631         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
26632         mode attributes.
26633         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
26634         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
26635         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
26636         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
26637         (SVE_COND_FP_CMP): New int iterators.
26638         (perm_hilo): Handle the new unpack unspecs.
26639         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
26640         attributes.
26641         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
26642         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
26643         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
26644         (aarch64_equality_operator, aarch64_constant_vector_operand)
26645         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
26646         (aarch64_sve_nonimmediate_operand): Likewise.
26647         (aarch64_sve_general_operand): Likewise.
26648         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
26649         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
26650         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
26651         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
26652         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
26653         (aarch64_sve_float_arith_immediate): Likewise.
26654         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
26655         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
26656         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
26657         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
26658         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
26659         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
26660         (aarch64_sve_float_arith_operand): Likewise.
26661         (aarch64_sve_float_arith_with_sub_operand): Likewise.
26662         (aarch64_sve_float_mul_operand): Likewise.
26663         (aarch64_sve_vec_perm_operand): Likewise.
26664         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
26665         (aarch64_mov_operand): Accept const_poly_int and const_vector.
26666         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
26667         as well as const_vector.
26668         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
26669         in file.  Use CONST0_RTX and CONSTM1_RTX.
26670         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
26671         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
26672         Use aarch64_simd_imm_zero.
26673         * config/aarch64/aarch64-sve.md: New file.
26674         * config/aarch64/aarch64.md: Include it.
26675         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
26676         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
26677         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
26678         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
26679         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
26680         (sve): New attribute.
26681         (enabled): Disable instructions with the sve attribute unless
26682         TARGET_SVE.
26683         (movqi, movhi): Pass CONST_POLY_INT operaneds through
26684         aarch64_expand_mov_immediate.
26685         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
26686         CNT[BHSD] immediates.
26687         (movti): Split CONST_POLY_INT moves into two halves.
26688         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
26689         Split additions that need a temporary here if the destination
26690         is the stack pointer.
26691         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
26692         (*add<mode>3_poly_1): New instruction.
26693         (set_clobber_cc): New expander.
26695 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26697         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
26698         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
26699         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
26700         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
26701         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
26702         Change innermode from fixed_mode_size to machine_mode.
26703         (simplify_subreg): Update call accordingly.  Handle a constant-sized
26704         subreg of a variable-length CONST_VECTOR.
26706 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26707             Alan Hayward  <alan.hayward@arm.com>
26708             David Sherwood  <david.sherwood@arm.com>
26710         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
26711         (add_offset_to_base): New function, split out from...
26712         (create_mem_ref): ...here.  When handling a scale other than 1,
26713         check first whether the address is valid without the offset.
26714         Add it into the base if so, leaving the index and scale as-is.
26716 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
26718         PR c++/83778
26719         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
26720         fold_for_warn before checking if arg2 is INTEGER_CST.
26722 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
26724         * config/rs6000/predicates.md (load_multiple_operation): Delete.
26725         (store_multiple_operation): Delete.
26726         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
26727         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
26728         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
26729         guarded by TARGET_STRING.
26730         (rs6000_output_load_multiple): Delete.
26731         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
26732         OPTION_MASK_STRING / TARGET_STRING handling.
26733         (print_operand) <'N', 'O'>: Add comment that these are unused now.
26734         (const rs6000_opt_masks) <"string">: Change mask to 0.
26735         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
26736         (MASK_STRING): Delete.
26737         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
26738         parts.  Simplify.
26739         (load_multiple): Delete.
26740         (*ldmsi8): Delete.
26741         (*ldmsi7): Delete.
26742         (*ldmsi6): Delete.
26743         (*ldmsi5): Delete.
26744         (*ldmsi4): Delete.
26745         (*ldmsi3): Delete.
26746         (store_multiple): Delete.
26747         (*stmsi8): Delete.
26748         (*stmsi7): Delete.
26749         (*stmsi6): Delete.
26750         (*stmsi5): Delete.
26751         (*stmsi4): Delete.
26752         (*stmsi3): Delete.
26753         (movmemsi_8reg): Delete.
26754         (corresponding unnamed define_insn): Delete.
26755         (movmemsi_6reg): Delete.
26756         (corresponding unnamed define_insn): Delete.
26757         (movmemsi_4reg): Delete.
26758         (corresponding unnamed define_insn): Delete.
26759         (movmemsi_2reg): Delete.
26760         (corresponding unnamed define_insn): Delete.
26761         (movmemsi_1reg): Delete.
26762         (corresponding unnamed define_insn): Delete.
26763         * config/rs6000/rs6000.opt (mno-string): New.
26764         (mstring): Replace by deprecation warning stub.
26765         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
26767 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
26769         * regrename.c (regrename_do_replace): If replacing the same
26770         reg multiple times, try to reuse last created gen_raw_REG.
26772         PR debug/81155
26773         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
26774         main to workaround a bug in GDB.
26776 2018-01-12  Tom de Vries  <tom@codesourcery.com>
26778         PR target/83737
26779         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
26781 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
26783         PR rtl-optimization/80481
26784         * ira-color.c (get_cap_member): New function.
26785         (allocnos_conflict_by_live_ranges_p): Use it.
26786         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
26787         (setup_slot_coalesced_allocno_live_ranges): Ditto.
26789 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
26791         PR target/83628
26792         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
26793         (*saddl_se_1): Ditto.
26794         (*ssubsi_1): Ditto.
26795         (*ssubl_se_1): Ditto.
26797 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26799         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
26800         rather than wi::to_widest for DR_INITs.
26801         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
26802         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
26803         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
26804         INTEGER_CSTs.
26805         (vect_analyze_group_access_1): Note that here.
26807 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26809         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
26810         polynomial type sizes.
26812 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26814         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
26815         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
26816         (gimple_add_tmp_var): Likewise.
26818 2018-01-12  Martin Liska  <mliska@suse.cz>
26820         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
26821         (gimple_alloc_sizes): Likewise.
26822         (dump_gimple_statistics): Use PRIu64 in printf format.
26823         * gimple.h: Change uint64_t to int.
26825 2018-01-12  Martin Liska  <mliska@suse.cz>
26827         * tree-core.h: Use uint64_t instead of int.
26828         * tree.c (tree_node_counts): Likewise.
26829         (tree_node_sizes): Likewise.
26830         (dump_tree_statistics): Use PRIu64 in printf format.
26832 2018-01-12  Martin Liska  <mliska@suse.cz>
26834         * Makefile.in: As qsort_chk is implemented in vec.c, add
26835         vec.o to linkage of gencfn-macros.
26836         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
26837         passing the info to record_node_allocation_statistics.
26838         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
26839         and pass the info.
26840         * ggc-common.c (struct ggc_usage): Add operator== and use
26841         it in operator< and compare function.
26842         * mem-stats.h (struct mem_usage): Likewise.
26843         * vec.c (struct vec_usage): Remove operator< and compare
26844         function. Can be simply inherited.
26846 2018-01-12  Martin Jambor  <mjambor@suse.cz>
26848         PR target/81616
26849         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
26850         * tree-ssa-math-opts.c: Include domwalk.h.
26851         (convert_mult_to_fma_1): New function.
26852         (fma_transformation_info): New type.
26853         (fma_deferring_state): Likewise.
26854         (cancel_fma_deferring): New function.
26855         (result_of_phi): Likewise.
26856         (last_fma_candidate_feeds_initial_phi): Likewise.
26857         (convert_mult_to_fma): Added deferring logic, split actual
26858         transformation to convert_mult_to_fma_1.
26859         (math_opts_dom_walker): New type.
26860         (math_opts_dom_walker::after_dom_children): New method, body moved
26861         here from pass_optimize_widening_mul::execute, added deferring logic
26862         bits.
26863         (pass_optimize_widening_mul::execute): Moved most of code to
26864         math_opts_dom_walker::after_dom_children.
26865         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
26866         * config/i386/i386.c (ix86_option_override_internal): Added
26867         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
26869 2018-01-12  Richard Biener  <rguenther@suse.de>
26871         PR debug/83157
26872         * dwarf2out.c (gen_variable_die): Do not reset old_die for
26873         inline instance vars.
26875 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
26877         PR target/81819
26878         * config/rx/rx.c (rx_is_restricted_memory_address):
26879         Handle SUBREG case.
26881 2018-01-12  Richard Biener  <rguenther@suse.de>
26883         PR tree-optimization/80846
26884         * target.def (split_reduction): New target hook.
26885         * targhooks.c (default_split_reduction): New function.
26886         * targhooks.h (default_split_reduction): Declare.
26887         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
26888         target requests first reduce vectors by combining low and high
26889         parts.
26890         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
26891         (get_vectype_for_scalar_type_and_size): Export.
26892         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
26893         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
26894         * doc/tm.texi: Regenerate.
26895         * config/i386/i386.c (ix86_split_reduction): Implement
26896         TARGET_VECTORIZE_SPLIT_REDUCTION.
26898 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26900         PR target/83368
26901         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
26902         in PIC mode except for TARGET_VXWORKS_RTP.
26903         * config/sparc/sparc.c: Include cfgrtl.h.
26904         (TARGET_INIT_PIC_REG): Define.
26905         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
26906         (sparc_pic_register_p): New predicate.
26907         (sparc_legitimate_address_p): Use it.
26908         (sparc_legitimize_pic_address): Likewise.
26909         (sparc_delegitimize_address): Likewise.
26910         (sparc_mode_dependent_address_p): Likewise.
26911         (gen_load_pcrel_sym): Remove 4th parameter.
26912         (load_got_register): Adjust call to above.  Remove obsolete stuff.
26913         (sparc_expand_prologue): Do not call load_got_register here.
26914         (sparc_flat_expand_prologue): Likewise.
26915         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
26916         (sparc_use_pseudo_pic_reg): New function.
26917         (sparc_init_pic_reg): Likewise.
26918         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26919         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
26921 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
26923         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
26924         Add item for branch_cost.
26926 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26928         PR rtl-optimization/83565
26929         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
26930         not extend the result to a larger mode for rotate operations.
26931         (num_sign_bit_copies1): Likewise.
26933 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26935         PR target/40411
26936         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
26937         -symbolic.
26938         Use values-Xc.o for -pedantic.
26939         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
26941 2018-01-12  Martin Liska  <mliska@suse.cz>
26943         PR ipa/83054
26944         * ipa-devirt.c (final_warning_record::grow_type_warnings):
26945         New function.
26946         (possible_polymorphic_call_targets): Use it.
26947         (ipa_devirt): Likewise.
26949 2018-01-12  Martin Liska  <mliska@suse.cz>
26951         * profile-count.h (enum profile_quality): Use 0 as invalid
26952         enum value of profile_quality.
26954 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26956         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
26957         -mext-string options.
26959 2018-01-12  Richard Biener  <rguenther@suse.de>
26961         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
26962         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
26963         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
26964         Likewise.
26965         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
26967 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
26969         * configure.ac (--with-long-double-format): Add support for the
26970         configuration option to change the default long double format on
26971         PowerPC systems.
26972         * config.gcc (powerpc*-linux*-*): Likewise.
26973         * configure: Regenerate.
26974         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
26975         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
26976         used without modification.
26978 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26980         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
26981         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
26982         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
26983         MISC_BUILTIN_SPEC_BARRIER.
26984         (rs6000_init_builtins): Likewise.
26985         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
26986         enum value.
26987         (speculation_barrier): New define_insn.
26988         * doc/extend.texi: Document __builtin_speculation_barrier.
26990 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26992         PR target/83203
26993         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
26994         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
26995         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
26996         iterators.
26997         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
26998         integral modes instead of "ss" and "sd".
26999         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
27000         vectors with 32-bit and 64-bit elements.
27001         (vecdupssescalarmodesuffix): New mode attribute.
27002         (vec_dup<mode>): Use it.
27004 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
27006         PR target/83330
27007         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
27008         frame if argument is passed on stack.
27010 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
27012         PR target/82682
27013         * ree.c (combine_reaching_defs): Optimize also
27014         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
27015         reg2=any_extend(exp); reg1=reg2;, formatting fix.
27017 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27019         PR middle-end/83189
27020         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
27022 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
27024         PR middle-end/83718
27025         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
27026         after they are computed.
27028 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
27030         PR tree-optimization/83695
27031         * gimple-loop-linterchange.cc
27032         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
27033         reset cached scev information after interchange.
27034         (pass_linterchange::execute): Remove call to scev_reset_htab.
27036 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27038         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
27039         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
27040         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
27041         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
27042         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
27043         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
27044         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
27045         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
27046         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
27047         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
27048         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
27049         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
27050         (V_lane_reg): Likewise.
27051         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
27052         New define_expand.
27053         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
27054         (vfmal_lane_low<mode>_intrinsic,
27055         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
27056         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
27057         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
27058         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
27059         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
27060         vfmsl_lane_high<mode>_intrinsic): New define_insns.
27062 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27064         * config/arm/arm-cpus.in (fp16fml): New feature.
27065         (ALL_SIMD): Add fp16fml.
27066         (armv8.2-a): Add fp16fml as an option.
27067         (armv8.3-a): Likewise.
27068         (armv8.4-a): Add fp16fml as part of fp16.
27069         * config/arm/arm.h (TARGET_FP16FML): Define.
27070         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
27071         when appropriate.
27072         * config/arm/arm-modes.def (V2HF): Define.
27073         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
27074         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
27075         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
27076         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
27077         vfmsl_low, vfmsl_high): New set of builtins.
27078         * config/arm/iterators.md (PLUSMINUS): New code iterator.
27079         (vfml_op): New code attribute.
27080         (VFMLHALVES): New int iterator.
27081         (VFML, VFMLSEL): New mode attributes.
27082         (V_reg): Define mapping for V2HF.
27083         (V_hi, V_lo): New mode attributes.
27084         (VF_constraint): Likewise.
27085         (vfml_half, vfml_half_selector): New int attributes.
27086         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
27087         define_expand.
27088         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
27089         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
27090         New define_insn.
27091         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
27092         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
27093         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
27094         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
27095         documentation.
27096         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
27097         Document new effective target and option set.
27099 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27101         * config/arm/arm-cpus.in (armv8_4): New feature.
27102         (ARMv8_4a): New fgroup.
27103         (armv8.4-a): New arch.
27104         * config/arm/arm-tables.opt: Regenerate.
27105         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
27106         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
27107         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
27108         Add matching rules for -march=armv8.4-a and extensions.
27109         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
27111 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
27113         PR target/81821
27114         * config/rx/rx.md (BW): New mode attribute.
27115         (sync_lock_test_and_setsi): Add mode suffix to insn output.
27117 2018-01-11  Richard Biener  <rguenther@suse.de>
27119         PR tree-optimization/83435
27120         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
27121         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
27122         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
27124 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27125             Alan Hayward  <alan.hayward@arm.com>
27126             David Sherwood  <david.sherwood@arm.com>
27128         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
27129         field.
27130         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
27131         (aarch64_print_address_internal): Use it to check for a zero offset.
27133 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27134             Alan Hayward  <alan.hayward@arm.com>
27135             David Sherwood  <david.sherwood@arm.com>
27137         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
27138         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
27139         Return a poly_int64 rather than a HOST_WIDE_INT.
27140         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
27141         rather than a HOST_WIDE_INT.
27142         * config/aarch64/aarch64.h (aarch64_frame): Protect with
27143         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
27144         hard_fp_offset, frame_size, initial_adjust, callee_offset and
27145         final_offset from HOST_WIDE_INT to poly_int64.
27146         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
27147         to_constant when getting the number of units in an Advanced SIMD
27148         mode.
27149         (aarch64_builtin_vectorized_function): Check for a constant number
27150         of units.
27151         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
27152         GET_MODE_SIZE.
27153         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
27154         attribute instead of GET_MODE_NUNITS.
27155         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
27156         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
27157         GET_MODE_SIZE for fixed-size registers.
27158         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
27159         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
27160         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
27161         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
27162         (aarch64_print_operand, aarch64_print_address_internal)
27163         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
27164         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
27165         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
27166         Handle polynomial GET_MODE_SIZE.
27167         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
27168         wider than SImode without modification.
27169         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
27170         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
27171         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
27172         passing and returning SVE modes.
27173         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
27174         rather than GEN_INT.
27175         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
27176         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
27177         (aarch64_allocate_and_probe_stack_space): Likewise.
27178         (aarch64_layout_frame): Cope with polynomial offsets.
27179         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
27180         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
27181         polynomial offsets.
27182         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
27183         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
27184         poly_int64 rather than a HOST_WIDE_INT.
27185         (aarch64_get_separate_components, aarch64_process_components)
27186         (aarch64_expand_prologue, aarch64_expand_epilogue)
27187         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
27188         (aarch64_anchor_offset): New function, split out from...
27189         (aarch64_legitimize_address): ...here.
27190         (aarch64_builtin_vectorization_cost): Handle polynomial
27191         TYPE_VECTOR_SUBPARTS.
27192         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
27193         GET_MODE_NUNITS.
27194         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
27195         number of elements from the PARALLEL rather than the mode.
27196         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
27197         rather than GET_MODE_BITSIZE.
27198         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
27199         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
27200         (aarch64_expand_vec_perm_const_1): Handle polynomial
27201         d->perm.length () and d->perm elements.
27202         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
27203         Apply to_constant to d->perm elements.
27204         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
27205         polynomial CONST_VECTOR_NUNITS.
27206         (aarch64_move_pointer): Take amount as a poly_int64 rather
27207         than an int.
27208         (aarch64_progress_pointer): Avoid temporary variable.
27209         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
27210         the mode attribute instead of GET_MODE.
27212 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27213             Alan Hayward  <alan.hayward@arm.com>
27214             David Sherwood  <david.sherwood@arm.com>
27216         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
27217         x exists before using it.
27218         (aarch64_add_constant_internal): Rename to...
27219         (aarch64_add_offset_1): ...this.  Replace regnum with separate
27220         src and dest rtxes.  Handle the case in which they're different,
27221         including when the offset is zero.  Replace scratchreg with an rtx.
27222         Use 2 additions if there is no spare register into which we can
27223         move a 16-bit constant.
27224         (aarch64_add_constant): Delete.
27225         (aarch64_add_offset): Replace reg with separate src and dest
27226         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
27227         Use aarch64_add_offset_1.
27228         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
27229         an rtx rather than an int.  Take the delta as a poly_int64
27230         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
27231         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
27232         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
27233         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
27234         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
27235         and aarch64_add_sp.
27236         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
27237         aarch64_add_constant.
27239 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27241         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
27242         Use scalar_float_mode.
27244 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27246         * config/aarch64/aarch64-simd.md
27247         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
27248         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
27249         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
27250         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
27251         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
27252         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
27253         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
27254         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
27255         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
27256         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
27258 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27260         PR target/83514
27261         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27262         targ_options->x_arm_arch_string is non NULL.
27264 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
27266         * config/aarch64/aarch64.h
27267         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
27269 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
27271         PR target/82096
27272         * expmed.c (emit_store_flag_force): Swap if const op0
27273         and change VOIDmode to mode of op0.
27275 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27277         PR rtl-optimization/83761
27278         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27279         than bytes to mode_for_size.
27281 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27283         PR middle-end/83189
27284         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27285         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27286         profile.
27288 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27290         PR middle-end/83575
27291         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27292         when in layout mode.
27293         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27294         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27295         partition fixup.
27297 2018-01-10  Michael Collison  <michael.collison@arm.com>
27299         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27300         * config/aarch64/aarch64-option-extension.def: Add
27301         AARCH64_OPT_EXTENSION of 'fp16fml'.
27302         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27303         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27304         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27305         * config/aarch64/constraints.md (Ui7): New constraint.
27306         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27307         (VFMLA_SEL_W): Ditto.
27308         (f16quad): Ditto.
27309         (f16mac1): Ditto.
27310         (VFMLA16_LOW): New int iterator.
27311         (VFMLA16_HIGH): Ditto.
27312         (UNSPEC_FMLAL): New unspec.
27313         (UNSPEC_FMLSL): Ditto.
27314         (UNSPEC_FMLAL2): Ditto.
27315         (UNSPEC_FMLSL2): Ditto.
27316         (f16mac): New code attribute.
27317         * config/aarch64/aarch64-simd-builtins.def
27318         (aarch64_fmlal_lowv2sf): Ditto.
27319         (aarch64_fmlsl_lowv2sf): Ditto.
27320         (aarch64_fmlalq_lowv4sf): Ditto.
27321         (aarch64_fmlslq_lowv4sf): Ditto.
27322         (aarch64_fmlal_highv2sf): Ditto.
27323         (aarch64_fmlsl_highv2sf): Ditto.
27324         (aarch64_fmlalq_highv4sf): Ditto.
27325         (aarch64_fmlslq_highv4sf): Ditto.
27326         (aarch64_fmlal_lane_lowv2sf): Ditto.
27327         (aarch64_fmlsl_lane_lowv2sf): Ditto.
27328         (aarch64_fmlal_laneq_lowv2sf): Ditto.
27329         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27330         (aarch64_fmlalq_lane_lowv4sf): Ditto.
27331         (aarch64_fmlsl_lane_lowv4sf): Ditto.
27332         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27333         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27334         (aarch64_fmlal_lane_highv2sf): Ditto.
27335         (aarch64_fmlsl_lane_highv2sf): Ditto.
27336         (aarch64_fmlal_laneq_highv2sf): Ditto.
27337         (aarch64_fmlsl_laneq_highv2sf): Ditto.
27338         (aarch64_fmlalq_lane_highv4sf): Ditto.
27339         (aarch64_fmlsl_lane_highv4sf): Ditto.
27340         (aarch64_fmlalq_laneq_highv4sf): Ditto.
27341         (aarch64_fmlsl_laneq_highv4sf): Ditto.
27342         * config/aarch64/aarch64-simd.md:
27343         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27344         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27345         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27346         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27347         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27348         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27349         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27350         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27351         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27352         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27353         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27354         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27355         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27356         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27357         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27358         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27359         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27360         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27361         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27362         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27363         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27364         (vfmlsl_low_u32): Ditto.
27365         (vfmlalq_low_u32): Ditto.
27366         (vfmlslq_low_u32): Ditto.
27367         (vfmlal_high_u32): Ditto.
27368         (vfmlsl_high_u32): Ditto.
27369         (vfmlalq_high_u32): Ditto.
27370         (vfmlslq_high_u32): Ditto.
27371         (vfmlal_lane_low_u32): Ditto.
27372         (vfmlsl_lane_low_u32): Ditto.
27373         (vfmlal_laneq_low_u32): Ditto.
27374         (vfmlsl_laneq_low_u32): Ditto.
27375         (vfmlalq_lane_low_u32): Ditto.
27376         (vfmlslq_lane_low_u32): Ditto.
27377         (vfmlalq_laneq_low_u32): Ditto.
27378         (vfmlslq_laneq_low_u32): Ditto.
27379         (vfmlal_lane_high_u32): Ditto.
27380         (vfmlsl_lane_high_u32): Ditto.
27381         (vfmlal_laneq_high_u32): Ditto.
27382         (vfmlsl_laneq_high_u32): Ditto.
27383         (vfmlalq_lane_high_u32): Ditto.
27384         (vfmlslq_lane_high_u32): Ditto.
27385         (vfmlalq_laneq_high_u32): Ditto.
27386         (vfmlslq_laneq_high_u32): Ditto.
27387         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27388         (AARCH64_FL_FOR_ARCH8_4): New.
27389         (AARCH64_ISA_F16FML): New ISA flag.
27390         (TARGET_F16FML): New feature flag for fp16fml.
27391         (doc/invoke.texi): Document new fp16fml option.
27393 2018-01-10  Michael Collison  <michael.collison@arm.com>
27395         * config/aarch64/aarch64-builtins.c:
27396         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27397         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27398         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27399         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27400         (AARCH64_ISA_SHA3): New ISA flag.
27401         (TARGET_SHA3): New feature flag for sha3.
27402         * config/aarch64/iterators.md (sha512_op): New int attribute.
27403         (CRYPTO_SHA512): New int iterator.
27404         (UNSPEC_SHA512H): New unspec.
27405         (UNSPEC_SHA512H2): Ditto.
27406         (UNSPEC_SHA512SU0): Ditto.
27407         (UNSPEC_SHA512SU1): Ditto.
27408         * config/aarch64/aarch64-simd-builtins.def
27409         (aarch64_crypto_sha512hqv2di): New builtin.
27410         (aarch64_crypto_sha512h2qv2di): Ditto.
27411         (aarch64_crypto_sha512su0qv2di): Ditto.
27412         (aarch64_crypto_sha512su1qv2di): Ditto.
27413         (aarch64_eor3qv8hi): Ditto.
27414         (aarch64_rax1qv2di): Ditto.
27415         (aarch64_xarqv2di): Ditto.
27416         (aarch64_bcaxqv8hi): Ditto.
27417         * config/aarch64/aarch64-simd.md:
27418         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27419         (aarch64_crypto_sha512su0qv2di): Ditto.
27420         (aarch64_crypto_sha512su1qv2di): Ditto.
27421         (aarch64_eor3qv8hi): Ditto.
27422         (aarch64_rax1qv2di): Ditto.
27423         (aarch64_xarqv2di): Ditto.
27424         (aarch64_bcaxqv8hi): Ditto.
27425         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27426         (vsha512h2q_u64): Ditto.
27427         (vsha512su0q_u64): Ditto.
27428         (vsha512su1q_u64): Ditto.
27429         (veor3q_u16): Ditto.
27430         (vrax1q_u64): Ditto.
27431         (vxarq_u64): Ditto.
27432         (vbcaxq_u16): Ditto.
27433         * config/arm/types.md (crypto_sha512): New type attribute.
27434         (crypto_sha3): Ditto.
27435         (doc/invoke.texi): Document new sha3 option.
27437 2018-01-10  Michael Collison  <michael.collison@arm.com>
27439         * config/aarch64/aarch64-builtins.c:
27440         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
27441         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27442         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
27443         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
27444         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
27445         (AARCH64_ISA_SM4): New ISA flag.
27446         (TARGET_SM4): New feature flag for sm4.
27447         * config/aarch64/aarch64-simd-builtins.def
27448         (aarch64_sm3ss1qv4si): Ditto.
27449         (aarch64_sm3tt1aq4si): Ditto.
27450         (aarch64_sm3tt1bq4si): Ditto.
27451         (aarch64_sm3tt2aq4si): Ditto.
27452         (aarch64_sm3tt2bq4si): Ditto.
27453         (aarch64_sm3partw1qv4si): Ditto.
27454         (aarch64_sm3partw2qv4si): Ditto.
27455         (aarch64_sm4eqv4si): Ditto.
27456         (aarch64_sm4ekeyqv4si): Ditto.
27457         * config/aarch64/aarch64-simd.md:
27458         (aarch64_sm3ss1qv4si): Ditto.
27459         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
27460         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
27461         (aarch64_sm4eqv4si): Ditto.
27462         (aarch64_sm4ekeyqv4si): Ditto.
27463         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
27464         (sm3part_op): Ditto.
27465         (CRYPTO_SM3TT): Ditto.
27466         (CRYPTO_SM3PART): Ditto.
27467         (UNSPEC_SM3SS1): New unspec.
27468         (UNSPEC_SM3TT1A): Ditto.
27469         (UNSPEC_SM3TT1B): Ditto.
27470         (UNSPEC_SM3TT2A): Ditto.
27471         (UNSPEC_SM3TT2B): Ditto.
27472         (UNSPEC_SM3PARTW1): Ditto.
27473         (UNSPEC_SM3PARTW2): Ditto.
27474         (UNSPEC_SM4E): Ditto.
27475         (UNSPEC_SM4EKEY): Ditto.
27476         * config/aarch64/constraints.md (Ui2): New constraint.
27477         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
27478         * config/arm/types.md (crypto_sm3): New type attribute.
27479         (crypto_sm4): Ditto.
27480         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
27481         (vsm3tt1aq_u32): Ditto.
27482         (vsm3tt1bq_u32): Ditto.
27483         (vsm3tt2aq_u32): Ditto.
27484         (vsm3tt2bq_u32): Ditto.
27485         (vsm3partw1q_u32): Ditto.
27486         (vsm3partw2q_u32): Ditto.
27487         (vsm4eq_u32): Ditto.
27488         (vsm4ekeyq_u32): Ditto.
27489         (doc/invoke.texi): Document new sm4 option.
27491 2018-01-10  Michael Collison  <michael.collison@arm.com>
27493         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
27494         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
27495         (AARCH64_FL_FOR_ARCH8_4): New.
27496         (AARCH64_FL_V8_4): New flag.
27497         (doc/invoke.texi): Document new armv8.4-a option.
27499 2018-01-10  Michael Collison  <michael.collison@arm.com>
27501         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27502         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
27503         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
27504         * config/aarch64/aarch64-option-extension.def: Add
27505         AARCH64_OPT_EXTENSION of 'sha2'.
27506         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
27507         (crypto): Disable sha2 and aes if crypto disabled.
27508         (crypto): Enable aes and sha2 if enabled.
27509         (simd): Disable sha2 and aes if simd disabled.
27510         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
27511         New flags.
27512         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
27513         (TARGET_SHA2): New feature flag for sha2.
27514         (TARGET_AES): New feature flag for aes.
27515         * config/aarch64/aarch64-simd.md:
27516         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
27517         conditional on TARGET_AES.
27518         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
27519         (aarch64_crypto_sha1hsi): Make pattern conditional
27520         on TARGET_SHA2.
27521         (aarch64_crypto_sha1hv4si): Ditto.
27522         (aarch64_be_crypto_sha1hv4si): Ditto.
27523         (aarch64_crypto_sha1su1v4si): Ditto.
27524         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
27525         (aarch64_crypto_sha1su0v4si): Ditto.
27526         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
27527         (aarch64_crypto_sha256su0v4si): Ditto.
27528         (aarch64_crypto_sha256su1v4si): Ditto.
27529         (doc/invoke.texi): Document new aes and sha2 options.
27531 2018-01-10  Martin Sebor  <msebor@redhat.com>
27533         PR tree-optimization/83781
27534         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
27535         as string arrays.
27537 2018-01-11  Martin Sebor  <msebor@gmail.com>
27538             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27540         PR tree-optimization/83501
27541         PR tree-optimization/81703
27543         * tree-ssa-strlen.c (get_string_cst): Rename...
27544         (get_string_len): ...to this.  Handle global constants.
27545         (handle_char_store): Adjust.
27547 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
27548             Jim Wilson  <jimw@sifive.com>
27550         * config/riscv/riscv-protos.h (riscv_output_return): New.
27551         * config/riscv/riscv.c (struct machine_function): New naked_p field.
27552         (riscv_attribute_table, riscv_output_return),
27553         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
27554         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
27555         (riscv_compute_frame_info): Only compute frame->mask if not a naked
27556         function.
27557         (riscv_expand_prologue): Add early return for naked function.
27558         (riscv_expand_epilogue): Likewise.
27559         (riscv_function_ok_for_sibcall): Return false for naked function.
27560         (riscv_set_current_function): New.
27561         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
27562         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
27563         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
27564         * doc/extend.texi (RISC-V Function Attributes): New.
27566 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
27568         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
27569         check for 128-bit long double before checking TCmode.
27570         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
27571         128-bit long doubles before checking TFmode or TCmode.
27572         (FLOAT128_IBM_P): Likewise.
27574 2018-01-10  Martin Sebor  <msebor@redhat.com>
27576         PR tree-optimization/83671
27577         * builtins.c (c_strlen): Unconditionally return zero for the empty
27578         string.
27579         Use -Warray-bounds for warnings.
27580         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
27581         for non-constant array indices with COMPONENT_REF, arrays of
27582         arrays, and pointers to arrays.
27583         (gimple_fold_builtin_strlen): Determine and set length range for
27584         non-constant character arrays.
27586 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
27588         PR middle-end/81897
27589         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
27590         empty blocks.
27592 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
27594         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
27596 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27598         PR target/83399
27599         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
27600         VECTOR_MEM_ALTIVEC_OR_VSX_P.
27601         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
27602         indexed_or_indirect_operand predicate.
27603         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
27604         (*vsx_le_perm_load_v8hi): Likewise.
27605         (*vsx_le_perm_load_v16qi): Likewise.
27606         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
27607         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
27608         (*vsx_le_perm_store_v8hi): Likewise.
27609         (*vsx_le_perm_store_v16qi): Likewise.
27610         (eight unnamed splitters): Likewise.
27612 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27614         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
27615         * config/rs6000/emmintrin.h: Likewise.
27616         * config/rs6000/mmintrin.h: Likewise.
27617         * config/rs6000/xmmintrin.h: Likewise.
27619 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
27621         PR c++/43486
27622         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
27623         "public_flag".
27624         * tree.c (tree_nop_conversion): Return true for location wrapper
27625         nodes.
27626         (maybe_wrap_with_location): New function.
27627         (selftest::check_strip_nops): New function.
27628         (selftest::test_location_wrappers): New function.
27629         (selftest::tree_c_tests): Call it.
27630         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
27631         (maybe_wrap_with_location): New decl.
27632         (EXPR_LOCATION_WRAPPER_P): New macro.
27633         (location_wrapper_p): New inline function.
27634         (tree_strip_any_location_wrapper): New inline function.
27636 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
27638         PR target/83735
27639         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
27640         stack_realign_offset for the largest alignment of stack slot
27641         actually used.
27642         (ix86_find_max_used_stack_alignment): New function.
27643         (ix86_finalize_stack_frame_flags): Use it.  Set
27644         max_used_stack_alignment if we don't realign stack.
27645         * config/i386/i386.h (machine_function): Add
27646         max_used_stack_alignment.
27648 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
27650         * config/arm/arm.opt (-mbranch-cost): New option.
27651         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
27652         account.
27654 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
27656         PR target/83629
27657         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
27658         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
27660 2018-01-10  Richard Biener  <rguenther@suse.de>
27662         PR debug/83765
27663         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
27664         early out so it also covers the case where we have a non-NULL
27665         origin.
27667 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
27669         PR tree-optimization/83753
27670         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
27671         for non-strided grouped accesses if the number of elements is 1.
27673 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27675         PR target/81616
27676         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
27677         * i386.h (TARGET_USE_GATHER): Define.
27678         * x86-tune.def (X86_TUNE_USE_GATHER): New.
27680 2018-01-10  Martin Liska  <mliska@suse.cz>
27682         PR bootstrap/82831
27683         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
27684         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
27685         partitioning.
27686         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
27687         CLEANUP_NO_PARTITIONING is not set.
27689 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
27691         * doc/rtl.texi: Remove documentation of (const ...) wrappers
27692         for vectors, as a partial revert of r254296.
27693         * rtl.h (const_vec_p): Delete.
27694         (const_vec_duplicate_p): Don't test for vector CONSTs.
27695         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
27696         * expmed.c (make_tree): Likewise.
27698         Revert:
27699         * common.md (E, F): Use CONSTANT_P instead of checking for
27700         CONST_VECTOR.
27701         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
27702         checking for CONST_VECTOR.
27704 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27706         PR middle-end/83575
27707         * predict.c (force_edge_cold): Handle in more sane way edges
27708         with no prediction.
27710 2018-01-09  Carl Love  <cel@us.ibm.com>
27712         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
27713         V4SI, V4SF types.
27714         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
27715         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
27716         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
27717         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
27718         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
27719         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
27720         * config/rs6000/rs6000-protos.h: Add extern defition for
27721         rs6000_generate_float2_double_code.
27722         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
27723         function.
27724         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
27725         (float2_v2df): Add define_expand.
27727 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
27729         PR target/83628
27730         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
27731         op_mode in the force_to_mode call.
27733 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27735         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
27736         instead of checking each element individually.
27737         (aarch64_evpc_uzp): Likewise.
27738         (aarch64_evpc_zip): Likewise.
27739         (aarch64_evpc_ext): Likewise.
27740         (aarch64_evpc_rev): Likewise.
27741         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
27742         instead of checking each element individually.  Return true without
27743         generating rtl if
27744         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
27745         whether all selected elements come from the same input, instead of
27746         checking each element individually.  Remove calls to gen_rtx_REG,
27747         start_sequence and end_sequence and instead assert that no rtl is
27748         generated.
27750 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27752         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
27753         order of HIGH and CONST checks.
27755 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27757         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
27758         if the destination isn't an SSA_NAME.
27760 2018-01-09  Richard Biener  <rguenther@suse.de>
27762         PR tree-optimization/83668
27763         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
27764         move prologue...
27765         (canonicalize_loop_form): ... here, renamed from ...
27766         (canonicalize_loop_closed_ssa_form): ... this and amended to
27767         swap successor edges for loop exit blocks to make us use
27768         the RPO order we need for initial schedule generation.
27770 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
27772         PR tree-optimization/64811
27773         * match.pd: When optimizing comparisons with Inf, avoid
27774         introducing or losing exceptions from comparisons with NaN.
27776 2018-01-09  Martin Liska  <mliska@suse.cz>
27778         PR sanitizer/82517
27779         * asan.c (shadow_mem_size): Add gcc_assert.
27781 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
27783         Don't save registers in main().
27785         PR target/83738
27786         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
27787         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
27788         * config/avr/avr.c (avr_set_current_function): Don't error if
27789         naked, OS_task or OS_main are specified at the same time.
27790         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
27791         OS_main.
27792         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
27793         attribute.
27794         * common/config/avr/avr-common.c (avr_option_optimization_table):
27795         Switch on -mmain-is-OS_task for optimizing compilations.
27797 2018-01-09  Richard Biener  <rguenther@suse.de>
27799         PR tree-optimization/83572
27800         * graphite.c: Include cfganal.h.
27801         (graphite_transform_loops): Connect infinite loops to exit
27802         and remove fake edges at the end.
27804 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27806         * ipa-inline.c (edge_badness): Revert accidental checkin.
27808 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27810         PR ipa/80763
27811         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
27812         symbols; not inline clones.
27814 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
27816         PR target/83507
27817         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
27818         hard registers.  Formatting fixes.
27820         PR preprocessor/83722
27821         * gcc.c (try_generate_repro): Pass
27822         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
27823         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
27824         do_report_bug.
27826 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
27827             Kito Cheng  <kito.cheng@gmail.com>
27829         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
27830         (riscv_leaf_function_p): Delete.
27831         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
27833 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27835         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
27836         function.
27837         (do_ifelse): New function.
27838         (do_isel): New function.
27839         (do_sub3): New function.
27840         (do_add3): New function.
27841         (do_load_mask_compare): New function.
27842         (do_overlap_load_compare): New function.
27843         (expand_compare_loop): New function.
27844         (expand_block_compare): Call expand_compare_loop() when appropriate.
27845         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
27846         option description.
27847         (-mblock-compare-inline-loop-limit): New option.
27849 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27851         PR target/83677
27852         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
27853         Reverse order of second and third operands in first alternative.
27854         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
27855         of first and second elements in UNSPEC_VPERMR vector.
27856         (altivec_expand_vec_perm_le): Likewise.
27858 2018-01-08  Jeff Law  <law@redhat.com>
27860         PR rtl-optimizatin/81308
27861         * tree-switch-conversion.c (cfg_altered): New file scoped static.
27862         (process_switch): If group_case_labels makes a change, then set
27863         cfg_altered.
27864         (pass_convert_switch::execute): If a switch is converted, then
27865         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
27867         PR rtl-optimization/81308
27868         * recog.c (split_all_insns): Conditionally cleanup the CFG after
27869         splitting insns.
27871 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
27873         PR target/83663 - Revert r255946
27874         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
27875         generation for cases where splatting a value is not useful.
27876         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
27877         across a vec_duplicate and a paradoxical subreg forming a vector
27878         mode to a vec_concat.
27880 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27882         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
27883         -march=armv8.3-a variants.
27884         * config/arm/t-multilib: Likewise.
27885         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
27887 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27889         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
27890         to generate rtl.
27891         (cceq_ior_compare_complement): Give it a name so I can use it, and
27892         change boolean_or_operator predicate to boolean_operator so it can
27893         be used to generate a crand.
27894         (eqne): New code iterator.
27895         (bd/bd_neg): New code_attrs.
27896         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
27897         a single define_insn.
27898         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
27899         decrement (bdnzt/bdnzf/bdzt/bdzf).
27900         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
27901         with the new names of the branch decrement patterns, and added the
27902         names of the branch decrement conditional patterns.
27904 2018-01-08  Richard Biener  <rguenther@suse.de>
27906         PR tree-optimization/83563
27907         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
27908         cache.
27910 2018-01-08  Richard Biener  <rguenther@suse.de>
27912         PR middle-end/83713
27913         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
27915 2018-01-08  Richard Biener  <rguenther@suse.de>
27917         PR tree-optimization/83685
27918         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
27919         references to abnormals.
27921 2018-01-08  Richard Biener  <rguenther@suse.de>
27923         PR lto/83719
27924         * dwarf2out.c (output_indirect_strings): Handle empty
27925         skeleton_debug_str_hash.
27926         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
27928 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27930         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
27931         (emit_store_direct): Likewise.
27932         (arc_trampoline_adjust_address): Likewise.
27933         (arc_asm_trampoline_template): New function.
27934         (arc_initialize_trampoline): Use asm_trampoline_template.
27935         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
27936         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
27937         * config/arc/arc.md (flush_icache): Delete pattern.
27939 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27941         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
27942         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
27943         munaligned-access.
27945 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27947         PR target/83681
27948         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
27949         by not USED_FOR_TARGET.
27950         (make_pass_resolve_sw_modes): Likewise.
27952 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27954         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
27955         USED_FOR_TARGET.
27957 2018-01-08  Richard Biener  <rguenther@suse.de>
27959         PR middle-end/83580
27960         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
27962 2018-01-08  Richard Biener  <rguenther@suse.de>
27964         PR middle-end/83517
27965         * match.pd ((t * 2) / 2) -> t): Add missing :c.
27967 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
27969         PR middle-end/81897
27970         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
27971         basic blocks with a small number of successors.
27972         (convert_control_dep_chain_into_preds): Improve handling of
27973         forwarder blocks.
27974         (dump_predicates): Split apart into...
27975         (dump_pred_chain): ...here...
27976         (dump_pred_info): ...and here.
27977         (can_one_predicate_be_invalidated_p): Add debugging printfs.
27978         (can_chain_union_be_invalidated_p): Improve check for invalidation
27979         of paths.
27980         (uninit_uses_cannot_happen): Avoid unnecessary if
27981         convert_control_dep_chain_into_preds yielded nothing.
27983 2018-01-06  Martin Sebor  <msebor@redhat.com>
27985         PR tree-optimization/83640
27986         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
27987         subtracting negative offset from size.
27988         (builtin_access::overlap): Adjust offset bounds of the access to fall
27989         within the size of the object if possible.
27991 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
27993         PR rtl-optimization/83699
27994         * expmed.c (extract_bit_field_1): Restrict the vector usage of
27995         extract_bit_field_as_subreg to cases in which the extracted
27996         value is also a vector.
27998         * lra-constraints.c (process_alt_operands): Test for the equivalence
27999         substitutions when detecting a possible reload cycle.
28001 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
28003         PR debug/83480
28004         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
28005         by default if flag_selective_schedling{,2}.  Formatting fixes.
28007         PR rtl-optimization/83682
28008         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
28009         if it has non-VECTOR_MODE element mode.
28010         (vec_duplicate_p): Likewise.
28012         PR middle-end/83694
28013         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
28014         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
28016 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28018         PR target/83604
28019         * config/i386/i386-builtin.def
28020         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
28021         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
28022         Require also OPTION_MASK_ISA_AVX512F in addition to
28023         OPTION_MASK_ISA_GFNI.
28024         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
28025         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
28026         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
28027         to OPTION_MASK_ISA_GFNI.
28028         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
28029         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
28030         OPTION_MASK_ISA_AVX512BW.
28031         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
28032         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
28033         addition to OPTION_MASK_ISA_GFNI.
28034         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
28035         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
28036         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
28037         to OPTION_MASK_ISA_GFNI.
28038         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
28039         a requirement for all ISAs rather than any of them with a few
28040         exceptions.
28041         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
28042         processing.
28043         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
28044         bitmasks to be enabled with 3 exceptions, instead of requiring any
28045         enabled ISA with lots of exceptions.
28046         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
28047         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
28048         Change avx512bw in isa attribute to avx512f.
28049         * config/i386/sgxintrin.h: Add license boilerplate.
28050         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
28051         to __AVX512F__ and __AVX512VL to __AVX512VL__.
28052         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
28053         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
28054         defined.
28055         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
28056         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
28057         temporarily sse2 rather than sse if not enabled already.
28059         PR target/83604
28060         * config/i386/sse.md (VI248_VLBW): Rename to ...
28061         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
28062         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
28063         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
28064         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
28065         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
28066         mode iterator instead of VI248_VLBW.
28068 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
28070         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
28071         (record_modified): Skip clobbers; add debug output.
28072         (param_change_prob): Use sreal frequencies.
28074 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28076         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
28077         punt for user-aligned variables.
28079 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
28081         * tree-chrec.c (chrec_contains_symbols): Return true for
28082         POLY_INT_CST.
28084 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
28086         PR target/82439
28087         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
28088         of (x|y) == x for BICS pattern.
28090 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
28092         PR tree-optimization/83605
28093         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
28094         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
28095         can throw.
28097 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
28099         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
28100         * config/epiphany/rtems.h: New file.
28102 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28103             Uros Bizjak  <ubizjak@gmail.com>
28105         PR target/83554
28106         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
28107         QIreg_operand instead of register_operand predicate.
28108         * config/i386/i386.c (ix86_rop_should_change_byte_p,
28109         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
28110         comments instead of -fmitigate[-_]rop.
28112 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28114         PR bootstrap/81926
28115         * cgraphunit.c (symbol_table::compile): Switch to text_section
28116         before calling assembly_start debug hook.
28117         * run-rtl-passes.c (run_rtl_passes): Likewise.
28118         Include output.h.
28120 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28122         * tree-vrp.c (extract_range_from_binary_expr_1): Check
28123         range_int_cst_p rather than !symbolic_range_p before calling
28124         extract_range_from_multiplicative_op_1.
28126 2018-01-04  Jeff Law  <law@redhat.com>
28128         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
28129         redundant test in assertion.
28131 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28133         * doc/rtl.texi: Document machine_mode wrapper classes.
28135 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28137         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
28138         using tree_to_uhwi.
28140 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28142         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
28143         the VEC_PERM_EXPR fold to fail.
28145 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28147         PR debug/83585
28148         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
28149         to switched_sections.
28151 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28153         PR target/83680
28154         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
28155         test for d.testing.
28157 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
28159         PR target/83387
28160         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
28161         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
28163 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28165         PR debug/83666
28166         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
28167         is BLKmode and bitpos not zero or mode change is needed.
28169 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
28171         PR target/83675
28172         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
28173         TARGET_VIS2.
28175 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
28177         PR target/83628
28178         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
28179         instead of MULT rtx.  Update all corresponding splitters.
28180         (*saddl_se): Ditto.
28181         (*ssub<modesuffix>): Ditto.
28182         (*ssubl_se): Ditto.
28183         (*cmp_sadd_di): Update split patterns.
28184         (*cmp_sadd_si): Ditto.
28185         (*cmp_sadd_sidi): Ditto.
28186         (*cmp_ssub_di): Ditto.
28187         (*cmp_ssub_si): Ditto.
28188         (*cmp_ssub_sidi): Ditto.
28189         * config/alpha/predicates.md (const23_operand): New predicate.
28190         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
28191         Look for ASHIFT, not MULT inner operand.
28192         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
28194 2018-01-04  Martin Liska  <mliska@suse.cz>
28196         PR gcov-profile/83669
28197         * gcov.c (output_intermediate_file): Add version to intermediate
28198         gcov file.
28199         * doc/gcov.texi: Document new field 'version' in intermediate
28200         file format. Fix location of '-k' option of gcov command.
28202 2018-01-04  Martin Liska  <mliska@suse.cz>
28204         PR ipa/82352
28205         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
28207 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
28209         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
28211 2018-01-03  Martin Sebor  <msebor@redhat.com>
28213         PR tree-optimization/83655
28214         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
28215         checking calls with invalid arguments.
28217 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28219         * tree-vect-stmts.c (vect_get_store_rhs): New function.
28220         (vectorizable_mask_load_store): Delete.
28221         (vectorizable_call): Return false for masked loads and stores.
28222         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
28223         instead of gimple_assign_rhs1.
28224         (vectorizable_load): Handle IFN_MASK_LOAD.
28225         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
28227 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28229         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
28230         split out from..,
28231         (vectorizable_mask_load_store): ...here.
28232         (vectorizable_load): ...and here.
28234 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28236         * tree-vect-stmts.c (vect_build_all_ones_mask)
28237         (vect_build_zero_merge_argument): New functions, split out from...
28238         (vectorizable_load): ...here.
28240 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28242         * tree-vect-stmts.c (vect_check_store_rhs): New function,
28243         split out from...
28244         (vectorizable_mask_load_store): ...here.
28245         (vectorizable_store): ...and here.
28247 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28249         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
28250         split out from...
28251         (vectorizable_mask_load_store): ...here.
28253 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28255         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
28256         (vect_model_store_cost): Take a vec_load_store_type instead of a
28257         vect_def_type.
28258         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28259         (vect_model_store_cost): Take a vec_load_store_type instead of a
28260         vect_def_type.
28261         (vectorizable_mask_load_store): Update accordingly.
28262         (vectorizable_store): Likewise.
28263         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28265 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28267         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28268         IFN_MASK_LOAD calls here rather than...
28269         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28271 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28272             Alan Hayward  <alan.hayward@arm.com>
28273             David Sherwood  <david.sherwood@arm.com>
28275         * expmed.c (extract_bit_field_1): For vector extracts,
28276         fall back to extract_bit_field_as_subreg if vec_extract
28277         isn't available.
28279 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28280             Alan Hayward  <alan.hayward@arm.com>
28281             David Sherwood  <david.sherwood@arm.com>
28283         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28284         they are variable or constant sized.
28285         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28286         slots for constant-sized data.
28288 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28289             Alan Hayward  <alan.hayward@arm.com>
28290             David Sherwood  <david.sherwood@arm.com>
28292         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28293         handling COND_EXPRs with boolean comparisons, try to find a better
28294         basis for the mask type than the boolean itself.
28296 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28298         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28299         is calculated and how it can be overridden.
28300         * genmodes.c (max_bitsize_mode_any_mode): New variable.
28301         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28302         if defined.
28303         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28304         if nonzero.
28306 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28307             Alan Hayward  <alan.hayward@arm.com>
28308             David Sherwood  <david.sherwood@arm.com>
28310         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28311         Remove the mode argument.
28312         (aarch64_simd_valid_immediate): Remove the mode and inverse
28313         arguments.
28314         * config/aarch64/iterators.md (bitsize): New iterator.
28315         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28316         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28317         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28318         aarch64_simd_valid_immediate.
28319         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28320         (aarch64_reg_or_bic_imm): Likewise.
28321         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28322         with an insn_type enum and msl with a modifier_type enum.
28323         Replace element_width with a scalar_mode.  Change the shift
28324         to unsigned int.  Add constructors for scalar_float_mode and
28325         scalar_int_mode elements.
28326         (aarch64_vect_float_const_representable_p): Delete.
28327         (aarch64_can_const_movi_rtx_p)
28328         (aarch64_simd_scalar_immediate_valid_for_move)
28329         (aarch64_simd_make_constant): Update call to
28330         aarch64_simd_valid_immediate.
28331         (aarch64_advsimd_valid_immediate_hs): New function.
28332         (aarch64_advsimd_valid_immediate): Likewise.
28333         (aarch64_simd_valid_immediate): Remove mode and inverse
28334         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
28335         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28336         and aarch64_float_const_representable_p on the result.
28337         (aarch64_output_simd_mov_immediate): Remove mode argument.
28338         Update call to aarch64_simd_valid_immediate and use of
28339         simd_immediate_info.
28340         (aarch64_output_scalar_simd_mov_immediate): Update call
28341         accordingly.
28343 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28344             Alan Hayward  <alan.hayward@arm.com>
28345             David Sherwood  <david.sherwood@arm.com>
28347         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28348         (mode_nunits): Likewise CONST_MODE_NUNITS.
28349         * machmode.def (ADJUST_NUNITS): Document.
28350         * genmodes.c (mode_data::need_nunits_adj): New field.
28351         (blank_mode): Update accordingly.
28352         (adj_nunits): New variable.
28353         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28354         parameter.
28355         (emit_mode_size_inline): Set need_bytesize_adj for all modes
28356         listed in adj_nunits.
28357         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28358         listed in adj_nunits.  Don't emit case statements for such modes.
28359         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28360         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
28361         nothing if adj_nunits is nonnull.
28362         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28363         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28364         (emit_mode_fbit): Update use of print_maybe_const_decl.
28365         (emit_move_size): Likewise.  Treat the array as non-const
28366         if adj_nunits.
28367         (emit_mode_adjustments): Handle adj_nunits.
28369 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28371         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28372         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28373         (VECTOR_MODES): Use it.
28374         (make_vector_modes): Take the prefix as an argument.
28376 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28377             Alan Hayward  <alan.hayward@arm.com>
28378             David Sherwood  <david.sherwood@arm.com>
28380         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28381         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28382         for MODE_VECTOR_BOOL.
28383         * machmode.def (VECTOR_BOOL_MODE): Document.
28384         * genmodes.c (VECTOR_BOOL_MODE): New macro.
28385         (make_vector_bool_mode): New function.
28386         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28387         MODE_VECTOR_BOOL.
28388         * lto-streamer-in.c (lto_input_mode_table): Likewise.
28389         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28390         Likewise.
28391         * stor-layout.c (int_mode_for_mode): Likewise.
28392         * tree.c (build_vector_type_for_mode): Likewise.
28393         * varasm.c (output_constant_pool_2): Likewise.
28394         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28395         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
28396         for MODE_VECTOR_BOOL.
28397         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28398         of mode class checks.
28399         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28400         instead of a list of mode class checks.
28401         (expand_vector_scalar_condition): Likewise.
28402         (type_for_widest_vector_mode): Handle BImode as an inner mode.
28404 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28405             Alan Hayward  <alan.hayward@arm.com>
28406             David Sherwood  <david.sherwood@arm.com>
28408         * machmode.h (mode_size): Change from unsigned short to
28409         poly_uint16_pod.
28410         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28411         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28412         or if measurement_type is not polynomial.
28413         (fixed_size_mode::includes_p): Check for constant-sized modes.
28414         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28415         return a poly_uint16 rather than an unsigned short.
28416         (emit_mode_size): Change the type of mode_size from unsigned short
28417         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
28418         (emit_mode_adjustments): Cope with polynomial vector sizes.
28419         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28420         for GET_MODE_SIZE.
28421         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28422         for GET_MODE_SIZE.
28423         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28424         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28425         * caller-save.c (setup_save_areas): Likewise.
28426         (replace_reg_with_saved_mem): Likewise.
28427         * calls.c (emit_library_call_value_1): Likewise.
28428         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28429         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28430         (gen_lowpart_for_combine): Likewise.
28431         * convert.c (convert_to_integer_1): Likewise.
28432         * cse.c (equiv_constant, cse_insn): Likewise.
28433         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28434         (cselib_subst_to_values): Likewise.
28435         * dce.c (word_dce_process_block): Likewise.
28436         * df-problems.c (df_word_lr_mark_ref): Likewise.
28437         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28438         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28439         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
28440         (rtl_for_decl_location): Likewise.
28441         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
28442         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
28443         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
28444         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
28445         (expand_expr_real_1): Likewise.
28446         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
28447         (pad_below): Likewise.
28448         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28449         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
28450         * ira.c (get_subreg_tracking_sizes): Likewise.
28451         * ira-build.c (ira_create_allocno_objects): Likewise.
28452         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
28453         (ira_sort_regnos_for_alter_reg): Likewise.
28454         * ira-costs.c (record_operand_costs): Likewise.
28455         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
28456         (resolve_simple_move): Likewise.
28457         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
28458         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
28459         (lra_constraints): Likewise.
28460         (CONST_POOL_OK_P): Reject variable-sized modes.
28461         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
28462         (add_pseudo_to_slot, lra_spill): Likewise.
28463         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28464         * optabs-query.c (get_best_extraction_insn): Likewise.
28465         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28466         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
28467         (expand_mult_highpart, valid_multiword_target_p): Likewise.
28468         * recog.c (offsettable_address_addr_space_p): Likewise.
28469         * regcprop.c (maybe_mode_change): Likewise.
28470         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
28471         * regrename.c (build_def_use): Likewise.
28472         * regstat.c (dump_reg_info): Likewise.
28473         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
28474         (find_reloads, find_reloads_subreg_address): Likewise.
28475         * reload1.c (eliminate_regs_1): Likewise.
28476         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
28477         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
28478         (simplify_binary_operation_1, simplify_subreg): Likewise.
28479         * targhooks.c (default_function_arg_padding): Likewise.
28480         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
28481         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
28482         (verify_gimple_assign_ternary): Likewise.
28483         * tree-inline.c (estimate_move_cost): Likewise.
28484         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28485         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
28486         (get_address_cost_ainc): Likewise.
28487         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
28488         (vect_supportable_dr_alignment): Likewise.
28489         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28490         (vectorizable_reduction): Likewise.
28491         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
28492         (vectorizable_operation, vectorizable_load): Likewise.
28493         * tree.c (build_same_sized_truth_vector_type): Likewise.
28494         * valtrack.c (cleanup_auto_inc_dec): Likewise.
28495         * var-tracking.c (emit_note_insn_var_location): Likewise.
28496         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
28497         (ADDR_VEC_ALIGN): Likewise.
28499 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28500             Alan Hayward  <alan.hayward@arm.com>
28501             David Sherwood  <david.sherwood@arm.com>
28503         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
28504         unsigned short.
28505         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28506         or if measurement_type is polynomial.
28507         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
28508         * combine.c (make_extraction): Likewise.
28509         * dse.c (find_shift_sequence): Likewise.
28510         * dwarf2out.c (mem_loc_descriptor): Likewise.
28511         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
28512         (extract_bit_field, extract_low_bits): Likewise.
28513         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
28514         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
28515         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
28516         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
28517         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28518         * reload.c (find_reloads): Likewise.
28519         * reload1.c (alter_reg): Likewise.
28520         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
28521         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
28522         * tree-if-conv.c (predicate_mem_writes): Likewise.
28523         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
28524         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
28525         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
28526         * valtrack.c (dead_debug_insert_temp): Likewise.
28527         * varasm.c (mergeable_constant_section): Likewise.
28528         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
28530 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28531             Alan Hayward  <alan.hayward@arm.com>
28532             David Sherwood  <david.sherwood@arm.com>
28534         * expr.c (expand_assignment): Cope with polynomial mode sizes
28535         when assigning to a CONCAT.
28537 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28538             Alan Hayward  <alan.hayward@arm.com>
28539             David Sherwood  <david.sherwood@arm.com>
28541         * machmode.h (mode_precision): Change from unsigned short to
28542         poly_uint16_pod.
28543         (mode_to_precision): Return a poly_uint16 rather than an unsigned
28544         short.
28545         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
28546         or if measurement_type is not polynomial.
28547         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
28548         in which the mode is already known to be a scalar_int_mode.
28549         * genmodes.c (emit_mode_precision): Change the type of mode_precision
28550         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
28551         initializer.
28552         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28553         for GET_MODE_PRECISION.
28554         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28555         for GET_MODE_PRECISION.
28556         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
28557         as polynomial.
28558         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
28559         (expand_field_assignment, make_extraction): Likewise.
28560         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
28561         (get_last_value): Likewise.
28562         * convert.c (convert_to_integer_1): Likewise.
28563         * cse.c (cse_insn): Likewise.
28564         * expr.c (expand_expr_real_1): Likewise.
28565         * lra-constraints.c (simplify_operand_subreg): Likewise.
28566         * optabs-query.c (can_atomic_load_p): Likewise.
28567         * optabs.c (expand_atomic_load): Likewise.
28568         (expand_atomic_store): Likewise.
28569         * ree.c (combine_reaching_defs): Likewise.
28570         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
28571         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
28572         * tree.h (type_has_mode_precision_p): Likewise.
28573         * ubsan.c (instrument_si_overflow): Likewise.
28575 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28576             Alan Hayward  <alan.hayward@arm.com>
28577             David Sherwood  <david.sherwood@arm.com>
28579         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
28580         polynomial numbers of units.
28581         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
28582         (valid_vector_subparts_p): New function.
28583         (build_vector_type): Remove temporary shim and take the number
28584         of units as a poly_uint64 rather than an int.
28585         (build_opaque_vector_type): Take the number of units as a
28586         poly_uint64 rather than an int.
28587         * tree.c (build_vector_from_ctor): Handle polynomial
28588         TYPE_VECTOR_SUBPARTS.
28589         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
28590         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
28591         (build_vector_from_val): If the number of units is variable,
28592         use build_vec_duplicate_cst for constant operands and
28593         VEC_DUPLICATE_EXPR otherwise.
28594         (make_vector_type): Remove temporary is_constant ().
28595         (build_vector_type, build_opaque_vector_type): Take the number of
28596         units as a poly_uint64 rather than an int.
28597         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
28598         VECTOR_CST_NELTS.
28599         * cfgexpand.c (expand_debug_expr): Likewise.
28600         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
28601         (store_constructor, expand_expr_real_1): Likewise.
28602         (const_scalar_mask_from_tree): Likewise.
28603         * fold-const-call.c (fold_const_reduction): Likewise.
28604         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
28605         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
28606         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
28607         (fold_relational_const): Likewise.
28608         (native_interpret_vector): Likewise.  Change the size from an
28609         int to an unsigned int.
28610         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
28611         TYPE_VECTOR_SUBPARTS.
28612         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
28613         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
28614         duplicating a non-constant operand into a variable-length vector.
28615         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
28616         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
28617         * ipa-icf.c (sem_variable::equals): Likewise.
28618         * match.pd: Likewise.
28619         * omp-simd-clone.c (simd_clone_subparts): Likewise.
28620         * print-tree.c (print_node): Likewise.
28621         * stor-layout.c (layout_type): Likewise.
28622         * targhooks.c (default_builtin_vectorization_cost): Likewise.
28623         * tree-cfg.c (verify_gimple_comparison): Likewise.
28624         (verify_gimple_assign_binary): Likewise.
28625         (verify_gimple_assign_ternary): Likewise.
28626         (verify_gimple_assign_single): Likewise.
28627         * tree-pretty-print.c (dump_generic_node): Likewise.
28628         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28629         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
28630         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
28631         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
28632         (vect_shift_permute_load_chain): Likewise.
28633         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
28634         (expand_vector_condition, optimize_vector_constructor): Likewise.
28635         (lower_vec_perm, get_compute_type): Likewise.
28636         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28637         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
28638         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
28639         (vect_recog_mask_conversion_pattern): Likewise.
28640         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
28641         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
28642         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28643         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
28644         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
28645         (vectorizable_shift, vectorizable_operation, vectorizable_store)
28646         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
28647         (supportable_widening_operation): Likewise.
28648         (supportable_narrowing_operation): Likewise.
28649         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
28650         Likewise.
28651         * varasm.c (output_constant): Likewise.
28653 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28654             Alan Hayward  <alan.hayward@arm.com>
28655             David Sherwood  <david.sherwood@arm.com>
28657         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
28658         so that both the length == 3 and length != 3 cases set up their
28659         own permute vectors.  Add comments explaining why we know the
28660         number of elements is constant.
28661         (vect_permute_load_chain): Likewise.
28663 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28664             Alan Hayward  <alan.hayward@arm.com>
28665             David Sherwood  <david.sherwood@arm.com>
28667         * machmode.h (mode_nunits): Change from unsigned char to
28668         poly_uint16_pod.
28669         (ONLY_FIXED_SIZE_MODES): New macro.
28670         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
28671         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
28672         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
28673         New typedefs.
28674         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
28675         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
28676         or if measurement_type is not polynomial.
28677         * genmodes.c (ZERO_COEFFS): New macro.
28678         (emit_mode_nunits_inline): Make mode_nunits_inline return a
28679         poly_uint16.
28680         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
28681         Use ZERO_COEFFS when emitting initializers.
28682         * data-streamer.h (bp_pack_poly_value): New function.
28683         (bp_unpack_poly_value): Likewise.
28684         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28685         for GET_MODE_NUNITS.
28686         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28687         for GET_MODE_NUNITS.
28688         * tree.c (make_vector_type): Remove temporary shim and make
28689         the real function take the number of units as a poly_uint64
28690         rather than an int.
28691         (build_vector_type_for_mode): Handle polynomial nunits.
28692         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
28693         * emit-rtl.c (const_vec_series_p_1): Likewise.
28694         (gen_rtx_CONST_VECTOR): Likewise.
28695         * fold-const.c (test_vec_duplicate_folding): Likewise.
28696         * genrecog.c (validate_pattern): Likewise.
28697         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
28698         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28699         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
28700         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
28701         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
28702         * rtlanal.c (subreg_get_info): Likewise.
28703         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28704         (vect_grouped_load_supported): Likewise.
28705         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
28706         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
28707         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
28708         (simplify_const_unary_operation, simplify_binary_operation_1)
28709         (simplify_const_binary_operation, simplify_ternary_operation)
28710         (test_vector_ops_duplicate, test_vector_ops): Likewise.
28711         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
28712         instead of CONST_VECTOR_NUNITS.
28713         * varasm.c (output_constant_pool_2): Likewise.
28714         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
28715         explicit-encoded elements in the XVEC for variable-length vectors.
28717 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28719         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
28721 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28722             Alan Hayward  <alan.hayward@arm.com>
28723             David Sherwood  <david.sherwood@arm.com>
28725         * coretypes.h (fixed_size_mode): Declare.
28726         (fixed_size_mode_pod): New typedef.
28727         * builtins.h (target_builtins::x_apply_args_mode)
28728         (target_builtins::x_apply_result_mode): Change type to
28729         fixed_size_mode_pod.
28730         * builtins.c (apply_args_size, apply_result_size, result_vector)
28731         (expand_builtin_apply_args_1, expand_builtin_apply)
28732         (expand_builtin_return): Update accordingly.
28734 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28736         * cse.c (hash_rtx_cb): Hash only the encoded elements.
28737         * cselib.c (cselib_hash_rtx): Likewise.
28738         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
28739         CONST_VECTOR encoding.
28741 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28742             Jeff Law  <law@redhat.com>
28744         PR target/83641
28745         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
28746         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
28747         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
28748         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
28750         PR target/83641
28751         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
28752         explicitly probe *sp in a noreturn function if there were any callee
28753         register saves or frame pointer is needed.
28755 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28757         PR debug/83621
28758         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
28759         BLKmode for ternary, binary or unary expressions.
28761         PR debug/83645
28762         * var-tracking.c (delete_vta_debug_insn): New inline function.
28763         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
28764         insns from get_insns () to NULL instead of each bb separately.
28765         Use delete_vta_debug_insn.  No longer static.
28766         (vt_debug_insns_local, variable_tracking_main_1): Adjust
28767         delete_vta_debug_insns callers.
28768         * rtl.h (delete_vta_debug_insns): Declare.
28769         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
28770         instead of variable_tracking_main.
28772 2018-01-03  Martin Sebor  <msebor@redhat.com>
28774         PR tree-optimization/83603
28775         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
28776         arguments past the endof the argument list in functions declared
28777         without a prototype.
28778         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
28779         Avoid checking when arguments are null.
28781 2018-01-03  Martin Sebor  <msebor@redhat.com>
28783         PR c/83559
28784         * doc/extend.texi (attribute const): Fix a typo.
28785         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
28786         issuing -Wsuggest-attribute for void functions.
28788 2018-01-03  Martin Sebor  <msebor@redhat.com>
28790         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
28791         offset_int::from instead of wide_int::to_shwi.
28792         (maybe_diag_overlap): Remove assertion.
28793         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
28794         * gimple-ssa-sprintf.c (format_directive): Same.
28795         (parse_directive): Same.
28796         (sprintf_dom_walker::compute_format_length): Same.
28797         (try_substitute_return_value): Same.
28799 2018-01-03  Jeff Law  <law@redhat.com>
28801         PR middle-end/83654
28802         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
28803         non-constant residual for zero at runtime and avoid probing in
28804         that case.  Reorganize code for trailing problem to mirror handling
28805         of the residual.
28807 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28809         PR tree-optimization/83501
28810         * tree-ssa-strlen.c (get_string_cst): New.
28811         (handle_char_store): Call get_string_cst.
28813 2018-01-03  Martin Liska  <mliska@suse.cz>
28815         PR tree-optimization/83593
28816         * tree-ssa-strlen.c: Include tree-cfg.h.
28817         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
28818         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
28819         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
28820         to false.
28821         (strlen_dom_walker::before_dom_children): Call
28822         gimple_purge_dead_eh_edges. Dump tranformation with details
28823         dump flags.
28824         (strlen_dom_walker::before_dom_children): Update call by adding
28825         new argument cleanup_eh.
28826         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
28828 2018-01-03  Martin Liska  <mliska@suse.cz>
28830         PR ipa/83549
28831         * cif-code.def (VARIADIC_THUNK): New enum value.
28832         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
28833         thunks.
28835 2018-01-03  Jan Beulich  <jbeulich@suse.com>
28837         * sse.md (mov<mode>_internal): Tighten condition for when to use
28838         vmovdqu<ssescalarsize> for TI and OI modes.
28840 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28842         Update copyright years.
28844 2018-01-03  Martin Liska  <mliska@suse.cz>
28846         PR ipa/83594
28847         * ipa-visibility.c (function_and_variable_visibility): Skip
28848         functions with noipa attribure.
28850 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28852         * gcc.c (process_command): Update copyright notice dates.
28853         * gcov-dump.c (print_version): Ditto.
28854         * gcov.c (print_version): Ditto.
28855         * gcov-tool.c (print_version): Ditto.
28856         * gengtype.c (create_file): Ditto.
28857         * doc/cpp.texi: Bump @copying's copyright year.
28858         * doc/cppinternals.texi: Ditto.
28859         * doc/gcc.texi: Ditto.
28860         * doc/gccint.texi: Ditto.
28861         * doc/gcov.texi: Ditto.
28862         * doc/install.texi: Ditto.
28863         * doc/invoke.texi: Ditto.
28865 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28867         * vector-builder.h (vector_builder::m_full_nelts): Change from
28868         unsigned int to poly_uint64.
28869         (vector_builder::full_nelts): Update prototype accordingly.
28870         (vector_builder::new_vector): Likewise.
28871         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
28872         (vector_builder::operator ==): Likewise.
28873         (vector_builder::finalize): Likewise.
28874         * int-vector-builder.h (int_vector_builder::int_vector_builder):
28875         Take the number of elements as a poly_uint64 rather than an
28876         unsigned int.
28877         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
28878         from unsigned int to poly_uint64.
28879         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
28880         (vec_perm_indices::new_vector): Likewise.
28881         (vec_perm_indices::length): Likewise.
28882         (vec_perm_indices::nelts_per_input): Likewise.
28883         (vec_perm_indices::input_nelts): Likewise.
28884         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
28885         number of elements per input as a poly_uint64 rather than an
28886         unsigned int.  Use the original encoding for variable-length
28887         vectors, rather than clamping each individual element.
28888         For the second and subsequent elements in each pattern,
28889         clamp the step and base before clamping their sum.
28890         (vec_perm_indices::series_p): Handle polynomial element counts.
28891         (vec_perm_indices::all_in_range_p): Likewise.
28892         (vec_perm_indices_to_tree): Likewise.
28893         (vec_perm_indices_to_rtx): Likewise.
28894         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
28895         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
28896         (tree_vector_builder::new_binary_operation): Handle polynomial
28897         element counts.  Return false if we need to know the number
28898         of elements at compile time.
28899         * fold-const.c (fold_vec_perm): Punt if the number of elements
28900         isn't known at compile time.
28902 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28904         * vec-perm-indices.h (vec_perm_builder): Change element type
28905         from HOST_WIDE_INT to poly_int64.
28906         (vec_perm_indices::element_type): Update accordingly.
28907         (vec_perm_indices::clamp): Handle polynomial element_types.
28908         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28909         (vec_perm_indices::all_in_range_p): Likewise.
28910         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
28911         than shwi trees.
28912         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
28913         polynomial vec_perm_indices element types.
28914         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
28915         * fold-const.c (fold_vec_perm): Likewise.
28916         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
28917         * tree-vect-generic.c (lower_vec_perm): Likewise.
28918         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28919         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
28920         element type to HOST_WIDE_INT.
28922 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28923             Alan Hayward  <alan.hayward@arm.com>
28924             David Sherwood  <david.sherwood@arm.com>
28926         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
28927         rather than an int.  Use plus_constant.
28928         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
28929         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
28931 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28932             Alan Hayward  <alan.hayward@arm.com>
28933             David Sherwood  <david.sherwood@arm.com>
28935         * calls.c (emit_call_1, expand_call): Change struct_value_size from
28936         a HOST_WIDE_INT to a poly_int64.
28938 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28939             Alan Hayward  <alan.hayward@arm.com>
28940             David Sherwood  <david.sherwood@arm.com>
28942         * calls.c (load_register_parameters): Cope with polynomial
28943         mode sizes.  Require a constant size for BLKmode parameters
28944         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
28945         forces a parameter to be padded at the lsb end in order to
28946         fill a complete number of words, require the parameter size
28947         to be ordered wrt UNITS_PER_WORD.
28949 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28950             Alan Hayward  <alan.hayward@arm.com>
28951             David Sherwood  <david.sherwood@arm.com>
28953         * reload1.c (spill_stack_slot_width): Change element type
28954         from unsigned int to poly_uint64_pod.
28955         (alter_reg): Treat mode sizes as polynomial.
28957 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28958             Alan Hayward  <alan.hayward@arm.com>
28959             David Sherwood  <david.sherwood@arm.com>
28961         * reload.c (complex_word_subreg_p): New function.
28962         (reload_inner_reg_of_subreg, push_reload): Use it.
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         * lra-constraints.c (process_alt_operands): Reject matched
28969         operands whose sizes aren't ordered.
28970         (match_reload): Refer to this check here.
28972 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28973             Alan Hayward  <alan.hayward@arm.com>
28974             David Sherwood  <david.sherwood@arm.com>
28976         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
28977         that the mode size is in the set {1, 2, 4, 8, 16}.
28979 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28980             Alan Hayward  <alan.hayward@arm.com>
28981             David Sherwood  <david.sherwood@arm.com>
28983         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
28984         Use plus_constant instead of gen_rtx_PLUS.
28986 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28987             Alan Hayward  <alan.hayward@arm.com>
28988             David Sherwood  <david.sherwood@arm.com>
28990         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
28991         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
28992         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
28993         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
28994         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
28995         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
28996         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
28997         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
28998         * config/i386/i386.c (ix86_push_rounding): ...this new function.
28999         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
29000         a poly_int64.
29001         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
29002         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
29003         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
29004         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
29005         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
29006         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
29007         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
29008         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
29009         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
29010         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
29011         function.
29012         * expr.c (emit_move_resolve_push): Treat the input and result
29013         of PUSH_ROUNDING as a poly_int64.
29014         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
29015         (emit_push_insn): Likewise.
29016         * lra-eliminations.c (mark_not_eliminable): Likewise.
29017         * recog.c (push_operand): Likewise.
29018         * reload1.c (elimination_effects): Likewise.
29019         * rtlanal.c (nonzero_bits1): Likewise.
29020         * calls.c (store_one_arg): Likewise.  Require the padding to be
29021         known at compile time.
29023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29024             Alan Hayward  <alan.hayward@arm.com>
29025             David Sherwood  <david.sherwood@arm.com>
29027         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
29028         Use plus_constant instead of gen_rtx_PLUS.
29030 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29031             Alan Hayward  <alan.hayward@arm.com>
29032             David Sherwood  <david.sherwood@arm.com>
29034         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
29035         rather than an int.
29037 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29038             Alan Hayward  <alan.hayward@arm.com>
29039             David Sherwood  <david.sherwood@arm.com>
29041         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
29042         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
29043         via stack temporaries.  Treat the mode size as polynomial too.
29045 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29046             Alan Hayward  <alan.hayward@arm.com>
29047             David Sherwood  <david.sherwood@arm.com>
29049         * expr.c (expand_expr_real_2): When handling conversions involving
29050         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
29051         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
29052         as a poly_uint64 too.
29054 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29055             Alan Hayward  <alan.hayward@arm.com>
29056             David Sherwood  <david.sherwood@arm.com>
29058         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
29060 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29061             Alan Hayward  <alan.hayward@arm.com>
29062             David Sherwood  <david.sherwood@arm.com>
29064         * combine.c (can_change_dest_mode): Handle polynomial
29065         REGMODE_NATURAL_SIZE.
29066         * expmed.c (store_bit_field_1): Likewise.
29067         * expr.c (store_constructor): Likewise.
29068         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
29069         and polynomial REGMODE_NATURAL_SIZE.
29070         (gen_lowpart_common): Likewise.
29071         * reginfo.c (record_subregs_of_mode): Likewise.
29072         * rtlanal.c (read_modify_subreg_p): Likewise.
29074 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29075             Alan Hayward  <alan.hayward@arm.com>
29076             David Sherwood  <david.sherwood@arm.com>
29078         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
29079         numbers of elements.
29081 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29082             Alan Hayward  <alan.hayward@arm.com>
29083             David Sherwood  <david.sherwood@arm.com>
29085         * match.pd: Cope with polynomial numbers of vector elements.
29087 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29088             Alan Hayward  <alan.hayward@arm.com>
29089             David Sherwood  <david.sherwood@arm.com>
29091         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
29092         in a POINTER_PLUS_EXPR.
29094 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29095             Alan Hayward  <alan.hayward@arm.com>
29096             David Sherwood  <david.sherwood@arm.com>
29098         * omp-simd-clone.c (simd_clone_subparts): New function.
29099         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
29100         (ipa_simd_modify_function_body): Likewise.
29102 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29103             Alan Hayward  <alan.hayward@arm.com>
29104             David Sherwood  <david.sherwood@arm.com>
29106         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
29107         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
29108         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
29109         (expand_vector_condition, vector_element): Likewise.
29110         (subparts_gt): New function.
29111         (get_compute_type): Use subparts_gt.
29112         (count_type_subparts): Delete.
29113         (expand_vector_operations_1): Use subparts_gt instead of
29114         count_type_subparts.
29116 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29117             Alan Hayward  <alan.hayward@arm.com>
29118             David Sherwood  <david.sherwood@arm.com>
29120         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
29121         (vect_compile_time_alias): ...this new function.  Do the calculation
29122         on poly_ints rather than trees.
29123         (vect_prune_runtime_alias_test_list): Update call accordingly.
29125 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29126             Alan Hayward  <alan.hayward@arm.com>
29127             David Sherwood  <david.sherwood@arm.com>
29129         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
29130         numbers of units.
29131         (vect_schedule_slp_instance): Likewise.
29133 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29134             Alan Hayward  <alan.hayward@arm.com>
29135             David Sherwood  <david.sherwood@arm.com>
29137         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
29138         constant and extern definitions for variable-length vectors.
29139         (vect_get_constant_vectors): Note that the number of units
29140         is known to be constant.
29142 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29143             Alan Hayward  <alan.hayward@arm.com>
29144             David Sherwood  <david.sherwood@arm.com>
29146         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
29147         of units as polynomial.  Choose between WIDE and NARROW based
29148         on multiple_p.
29150 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29151             Alan Hayward  <alan.hayward@arm.com>
29152             David Sherwood  <david.sherwood@arm.com>
29154         * tree-vect-stmts.c (simd_clone_subparts): New function.
29155         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
29157 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29158             Alan Hayward  <alan.hayward@arm.com>
29159             David Sherwood  <david.sherwood@arm.com>
29161         * tree-vect-stmts.c (vectorizable_call): Treat the number of
29162         vectors as polynomial.  Use build_index_vector for
29163         IFN_GOMP_SIMD_LANE.
29165 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29166             Alan Hayward  <alan.hayward@arm.com>
29167             David Sherwood  <david.sherwood@arm.com>
29169         * tree-vect-stmts.c (get_load_store_type): Treat the number of
29170         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
29171         for variable-length vectors.
29172         (vectorizable_mask_load_store): Treat the number of units as
29173         polynomial, asserting that it is constant if the condition has
29174         already been enforced.
29175         (vectorizable_store, vectorizable_load): Likewise.
29177 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29178             Alan Hayward  <alan.hayward@arm.com>
29179             David Sherwood  <david.sherwood@arm.com>
29181         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
29182         of units as polynomial.  Punt if we can't tell at compile time
29183         which vector contains the final result.
29185 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29186             Alan Hayward  <alan.hayward@arm.com>
29187             David Sherwood  <david.sherwood@arm.com>
29189         * tree-vect-loop.c (vectorizable_induction): Treat the number
29190         of units as polynomial.  Punt on SLP inductions.  Use an integer
29191         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
29192         cast of such a series for variable-length floating-point
29193         reductions.
29195 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29196             Alan Hayward  <alan.hayward@arm.com>
29197             David Sherwood  <david.sherwood@arm.com>
29199         * tree.h (build_index_vector): Declare.
29200         * tree.c (build_index_vector): New function.
29201         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
29202         of units as polynomial, forcibly converting it to a constant if
29203         vectorizable_reduction has already enforced the condition.
29204         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
29205         to create a {1,2,3,...} vector.
29206         (vectorizable_reduction): Treat the number of units as polynomial.
29207         Choose vectype_in based on the largest scalar element size rather
29208         than the smallest number of units.  Enforce the restrictions
29209         relied on above.
29211 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29212             Alan Hayward  <alan.hayward@arm.com>
29213             David Sherwood  <david.sherwood@arm.com>
29215         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
29216         number of units as polynomial.
29218 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29219             Alan Hayward  <alan.hayward@arm.com>
29220             David Sherwood  <david.sherwood@arm.com>
29222         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
29223         * target.def (autovectorize_vector_sizes): Return the vector sizes
29224         by pointer, using vector_sizes rather than a bitmask.
29225         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
29226         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
29227         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
29228         Likewise.
29229         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
29230         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
29231         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
29232         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
29233         * omp-general.c (omp_max_vf): Likewise.
29234         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29235         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
29236         * tree-vect-loop.c (vect_analyze_loop): Likewise.
29237         * tree-vect-slp.c (vect_slp_bb): Likewise.
29238         * doc/tm.texi: Regenerate.
29239         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
29240         to a poly_uint64.
29241         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
29242         the vector size as a poly_uint64 rather than an unsigned int.
29243         (current_vector_size): Change from an unsigned int to a poly_uint64.
29244         (get_vectype_for_scalar_type): Update accordingly.
29245         * tree.h (build_truth_vector_type): Take the size and number of
29246         units as a poly_uint64 rather than an unsigned int.
29247         (build_vector_type): Add a temporary overload that takes
29248         the number of units as a poly_uint64 rather than an unsigned int.
29249         * tree.c (make_vector_type): Likewise.
29250         (build_truth_vector_type): Take the number of units as a poly_uint64
29251         rather than an unsigned int.
29253 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29254             Alan Hayward  <alan.hayward@arm.com>
29255             David Sherwood  <david.sherwood@arm.com>
29257         * target.def (get_mask_mode): Take the number of units and length
29258         as poly_uint64s rather than unsigned ints.
29259         * targhooks.h (default_get_mask_mode): Update accordingly.
29260         * targhooks.c (default_get_mask_mode): Likewise.
29261         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29262         * doc/tm.texi: Regenerate.
29264 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29265             Alan Hayward  <alan.hayward@arm.com>
29266             David Sherwood  <david.sherwood@arm.com>
29268         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29269         * omp-general.c (omp_max_vf): Likewise.
29270         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29271         (expand_omp_simd): Handle polynomial safelen.
29272         * omp-low.c (omplow_simd_context): Add a default constructor.
29273         (omplow_simd_context::max_vf): Change from int to poly_uint64.
29274         (lower_rec_simd_input_clauses): Update accordingly.
29275         (lower_rec_input_clauses): Likewise.
29277 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29278             Alan Hayward  <alan.hayward@arm.com>
29279             David Sherwood  <david.sherwood@arm.com>
29281         * tree-vectorizer.h (vect_nunits_for_cost): New function.
29282         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29283         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29284         (vect_analyze_slp_cost): Likewise.
29285         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29286         (vect_model_load_cost): Likewise.
29288 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29289             Alan Hayward  <alan.hayward@arm.com>
29290             David Sherwood  <david.sherwood@arm.com>
29292         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29293         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29294         from an unsigned int * to a poly_uint64_pod *.
29295         (calculate_unrolling_factor): New function.
29296         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
29298 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29299             Alan Hayward  <alan.hayward@arm.com>
29300             David Sherwood  <david.sherwood@arm.com>
29302         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29303         from an unsigned int to a poly_uint64.
29304         (_loop_vec_info::slp_unrolling_factor): Likewise.
29305         (_loop_vec_info::vectorization_factor): Change from an int
29306         to a poly_uint64.
29307         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29308         (vect_get_num_vectors): New function.
29309         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29310         (vect_get_num_copies): Use vect_get_num_vectors.
29311         (vect_analyze_data_ref_dependences): Change max_vf from an int *
29312         to an unsigned int *.
29313         (vect_analyze_data_refs): Change min_vf from an int * to a
29314         poly_uint64 *.
29315         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29316         than an unsigned HOST_WIDE_INT.
29317         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29318         (vect_analyze_data_ref_dependence): Change max_vf from an int *
29319         to an unsigned int *.
29320         (vect_analyze_data_ref_dependences): Likewise.
29321         (vect_compute_data_ref_alignment): Handle polynomial vf.
29322         (vect_enhance_data_refs_alignment): Likewise.
29323         (vect_prune_runtime_alias_test_list): Likewise.
29324         (vect_shift_permute_load_chain): Likewise.
29325         (vect_supportable_dr_alignment): Likewise.
29326         (dependence_distance_ge_vf): Take the vectorization factor as a
29327         poly_uint64 rather than an unsigned HOST_WIDE_INT.
29328         (vect_analyze_data_refs): Change min_vf from an int * to a
29329         poly_uint64 *.
29330         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29331         vfm1 as a poly_uint64 rather than an int.  Make the same change
29332         for the returned bound_scalar.
29333         (vect_gen_vector_loop_niters): Handle polynomial vf.
29334         (vect_do_peeling): Likewise.  Update call to
29335         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29336         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29337         be constant.
29338         * tree-vect-loop.c (vect_determine_vectorization_factor)
29339         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29340         (vect_get_known_peeling_cost): Likewise.
29341         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29342         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29343         (vect_transform_loop): Likewise.  Use the lowest possible VF when
29344         updating the upper bounds of the loop.
29345         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
29346         rather than an int.
29347         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29348         polynomial unroll factors.
29349         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29350         (vect_make_slp_decision): Likewise.
29351         (vect_supported_load_permutation_p): Likewise, and polynomial
29352         vf too.
29353         (vect_analyze_slp_cost): Handle polynomial vf.
29354         (vect_slp_analyze_node_operations): Likewise.
29355         (vect_slp_analyze_bb_1): Likewise.
29356         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29357         than an unsigned HOST_WIDE_INT.
29358         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29359         (vectorizable_load): Handle polynomial vf.
29360         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29361         a poly_uint64.
29362         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29364 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29365             Alan Hayward  <alan.hayward@arm.com>
29366             David Sherwood  <david.sherwood@arm.com>
29368         * match.pd: Handle bit operations involving three constants
29369         and try to fold one pair.
29371 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29373         * tree-vect-loop-manip.c: Include gimple-fold.h.
29374         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29375         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
29376         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29377         Add a path that uses a step of VF instead of 1, but disable it
29378         for now.
29379         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29380         and niters_no_overflow parameters.  Update calls to
29381         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29382         Create a new SSA name if the latter choses to use a ste other
29383         than zero, and return it via niters_vector_mult_vf_var.
29384         * tree-vect-loop.c (vect_transform_loop): Update calls to
29385         vect_do_peeling, vect_gen_vector_loop_niters and
29386         slpeel_make_loop_iterate_ntimes.
29387         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29388         (vect_gen_vector_loop_niters): Update declarations after above changes.
29390 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
29392         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29393         128-bit round to integer instructions.
29394         (ceil<mode>2): Likewise.
29395         (btrunc<mode>2): Likewise.
29396         (round<mode>2): Likewise.
29398 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29400         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29401         unaligned VSX load/store on P8/P9.
29402         (expand_block_clear): Allow the use of unaligned VSX
29403         load/store on P8/P9.
29405 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29407         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29408         New function.
29409         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29410         swap associated with both a load and a store.
29412 2018-01-02  Andrew Waterman  <andrew@sifive.com>
29414         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29415         * config/riscv/riscv.md (clear_cache): Use it.
29417 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
29419         * web.c: Remove out-of-date comment.
29421 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29423         * expr.c (fixup_args_size_notes): Check that any existing
29424         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29425         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29426         (emit_single_push_insn): ...here.
29428 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29430         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29431         (const_vector_encoded_nelts): New function.
29432         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29433         (const_vector_int_elt, const_vector_elt): Declare.
29434         * emit-rtl.c (const_vector_int_elt_1): New function.
29435         (const_vector_elt): Likewise.
29436         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29437         of CONST_VECTOR_ELT.
29439 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29441         * expr.c: Include rtx-vector-builder.h.
29442         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
29443         directly on the tree encoding.
29444         (const_vector_from_tree): Likewise.
29445         * optabs.c: Include rtx-vector-builder.h.
29446         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
29447         sequence of "u" values.
29448         * vec-perm-indices.c: Include rtx-vector-builder.h.
29449         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
29450         directly on the vec_perm_indices encoding.
29452 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29454         * doc/rtl.texi (const_vector): Describe new encoding scheme.
29455         * Makefile.in (OBJS): Add rtx-vector-builder.o.
29456         * rtx-vector-builder.h: New file.
29457         * rtx-vector-builder.c: Likewise.
29458         * rtl.h (rtx_def::u2): Add a const_vector field.
29459         (CONST_VECTOR_NPATTERNS): New macro.
29460         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
29461         (CONST_VECTOR_DUPLICATE_P): Likewise.
29462         (CONST_VECTOR_STEPPED_P): Likewise.
29463         (CONST_VECTOR_ENCODED_ELT): Likewise.
29464         (const_vec_duplicate_p): Check for a duplicated vector encoding.
29465         (unwrap_const_vec_duplicate): Likewise.
29466         (const_vec_series_p): Check for a non-duplicated vector encoding.
29467         Say that the function only returns true for integer vectors.
29468         * emit-rtl.c: Include rtx-vector-builder.h.
29469         (gen_const_vec_duplicate_1): Delete.
29470         (gen_const_vector): Call gen_const_vec_duplicate instead of
29471         gen_const_vec_duplicate_1.
29472         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
29473         (gen_const_vec_duplicate): Use rtx_vector_builder.
29474         (gen_const_vec_series): Likewise.
29475         (gen_rtx_CONST_VECTOR): Likewise.
29476         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
29477         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29478         Build a new vector rather than modifying a CONST_VECTOR in-place.
29479         (handle_special_swappables): Update call accordingly.
29480         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
29481         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29482         Build a new vector rather than modifying a CONST_VECTOR in-place.
29483         (handle_special_swappables): Update call accordingly.
29485 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29487         * simplify-rtx.c (simplify_const_binary_operation): Use
29488         CONST_VECTOR_ELT instead of XVECEXP.
29490 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29492         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
29493         the selector elements to be different from the data elements
29494         if the selector is a VECTOR_CST.
29495         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
29496         ssizetype for the selector.
29498 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29500         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
29501         before testing each element individually.
29502         * tree-vect-generic.c (lower_vec_perm): Likewise.
29504 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29506         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
29507         * selftest-run-tests.c (selftest::run_tests): Call it.
29508         * vector-builder.h (vector_builder::operator ==): New function.
29509         (vector_builder::operator !=): Likewise.
29510         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
29511         (vec_perm_indices::all_from_input_p): New function.
29512         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29513         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
29514         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
29515         instead of reading the VECTOR_CST directly.  Detect whether both
29516         vector inputs are the same before constructing the vec_perm_indices,
29517         and update the number of inputs argument accordingly.  Use the
29518         utility functions added above.  Only construct sel2 if we need to.
29520 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29522         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
29523         the broadcast of the low byte.
29524         (expand_mult_highpart): Use an explicit encoding for the permutes.
29525         * optabs-query.c (can_mult_highpart_p): Likewise.
29526         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
29527         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29528         (vectorizable_bswap): Likewise.
29529         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
29530         explicit encoding for the power-of-2 permutes.
29531         (vect_permute_store_chain): Likewise.
29532         (vect_grouped_load_supported): Likewise.
29533         (vect_permute_load_chain): Likewise.
29535 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29537         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
29538         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
29539         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
29540         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29541         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
29542         (vect_gen_perm_mask_any): Likewise.
29544 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29546         * int-vector-builder.h: New file.
29547         * vec-perm-indices.h: Include int-vector-builder.h.
29548         (vec_perm_indices): Redefine as an int_vector_builder.
29549         (auto_vec_perm_indices): Delete.
29550         (vec_perm_builder): Redefine as a stand-alone class.
29551         (vec_perm_indices::vec_perm_indices): New function.
29552         (vec_perm_indices::clamp): Likewise.
29553         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
29554         (vec_perm_indices::new_vector): New function.
29555         (vec_perm_indices::new_expanded_vector): Update for new
29556         vec_perm_indices class.
29557         (vec_perm_indices::rotate_inputs): New function.
29558         (vec_perm_indices::all_in_range_p): Operate directly on the
29559         encoded form, without computing elided elements.
29560         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
29561         encoding.  Update for new vec_perm_indices class.
29562         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
29563         the given vec_perm_builder.
29564         (expand_vec_perm_var): Update vec_perm_builder constructor.
29565         (expand_mult_highpart): Use vec_perm_builder instead of
29566         auto_vec_perm_indices.
29567         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
29568         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
29569         or double series encoding as appropriate.
29570         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
29571         vec_perm_indices instead of auto_vec_perm_indices.
29572         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29573         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29574         (vect_permute_store_chain): Likewise.
29575         (vect_grouped_load_supported): Likewise.
29576         (vect_permute_load_chain): Likewise.
29577         (vect_shift_permute_load_chain): Likewise.
29578         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29579         (vect_transform_slp_perm_load): Likewise.
29580         (vect_schedule_slp_instance): Likewise.
29581         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29582         (vectorizable_mask_load_store): Likewise.
29583         (vectorizable_bswap): Likewise.
29584         (vectorizable_store): Likewise.
29585         (vectorizable_load): Likewise.
29586         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
29587         vec_perm_indices instead of auto_vec_perm_indices.  Use
29588         tree_to_vec_perm_builder to read the vector from a tree.
29589         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
29590         vec_perm_builder instead of a vec_perm_indices.
29591         (have_whole_vector_shift): Use vec_perm_builder and
29592         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
29593         truncation to calc_vec_perm_mask_for_shift.
29594         (vect_create_epilog_for_reduction): Likewise.
29595         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
29596         from auto_vec_perm_indices to vec_perm_indices.
29597         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29598         instead of changing individual elements.
29599         (aarch64_vectorize_vec_perm_const): Use new_vector to install
29600         the vector in d.perm.
29601         * config/arm/arm.c (expand_vec_perm_d::perm): Change
29602         from auto_vec_perm_indices to vec_perm_indices.
29603         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29604         instead of changing individual elements.
29605         (arm_vectorize_vec_perm_const): Use new_vector to install
29606         the vector in d.perm.
29607         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
29608         Update vec_perm_builder constructor.
29609         (rs6000_expand_interleave): Likewise.
29610         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
29611         (rs6000_expand_interleave): Likewise.
29613 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29615         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
29616         to qimode could truncate the indices.
29617         * optabs.c (expand_vec_perm_var): Likewise.
29619 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29621         * Makefile.in (OBJS): Add vec-perm-indices.o.
29622         * vec-perm-indices.h: New file.
29623         * vec-perm-indices.c: Likewise.
29624         * target.h (vec_perm_indices): Replace with a forward class
29625         declaration.
29626         (auto_vec_perm_indices): Move to vec-perm-indices.h.
29627         * optabs.h: Include vec-perm-indices.h.
29628         (expand_vec_perm): Delete.
29629         (selector_fits_mode_p, expand_vec_perm_var): Declare.
29630         (expand_vec_perm_const): Declare.
29631         * target.def (vec_perm_const_ok): Replace with...
29632         (vec_perm_const): ...this new hook.
29633         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
29634         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
29635         * doc/tm.texi: Regenerate.
29636         * optabs.def (vec_perm_const): Delete.
29637         * doc/md.texi (vec_perm_const): Likewise.
29638         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
29639         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
29640         expand_vec_perm for constant permutation vectors.  Assert that
29641         the mode of variable permutation vectors is the integer equivalent
29642         of the mode that is being permuted.
29643         * optabs-query.h (selector_fits_mode_p): Declare.
29644         * optabs-query.c: Include vec-perm-indices.h.
29645         (selector_fits_mode_p): New function.
29646         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
29647         is defined, instead of checking whether the vec_perm_const_optab
29648         exists.  Use targetm.vectorize.vec_perm_const instead of
29649         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
29650         fit in the vector mode before using a variable permute.
29651         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
29652         vec_perm_indices instead of an rtx.
29653         (expand_vec_perm): Replace with...
29654         (expand_vec_perm_const): ...this new function.  Take the selector
29655         as a vec_perm_indices rather than an rtx.  Also take the mode of
29656         the selector.  Update call to shift_amt_for_vec_perm_mask.
29657         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
29658         Use vec_perm_indices::new_expanded_vector to expand the original
29659         selector into bytes.  Check whether the indices fit in the vector
29660         mode before using a variable permute.
29661         (expand_vec_perm_var): Make global.
29662         (expand_mult_highpart): Use expand_vec_perm_const.
29663         * fold-const.c: Includes vec-perm-indices.h.
29664         * tree-ssa-forwprop.c: Likewise.
29665         * tree-vect-data-refs.c: Likewise.
29666         * tree-vect-generic.c: Likewise.
29667         * tree-vect-loop.c: Likewise.
29668         * tree-vect-slp.c: Likewise.
29669         * tree-vect-stmts.c: Likewise.
29670         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
29671         Delete.
29672         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
29673         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
29674         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
29675         (aarch64_vectorize_vec_perm_const): ...this new function.
29676         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29677         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29678         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
29679         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
29680         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29681         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29682         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
29683         into...
29684         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
29685         check for NEON modes.
29686         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
29687         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
29688         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
29689         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
29690         into...
29691         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
29692         the old VEC_PERM_CONST conditions.
29693         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
29694         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
29695         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
29696         (ia64_vectorize_vec_perm_const_ok): Merge into...
29697         (ia64_vectorize_vec_perm_const): ...this new function.
29698         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
29699         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
29700         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
29701         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
29702         * config/mips/mips.c (mips_expand_vec_perm_const)
29703         (mips_vectorize_vec_perm_const_ok): Merge into...
29704         (mips_vectorize_vec_perm_const): ...this new function.
29705         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
29706         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
29707         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
29708         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
29709         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
29710         (rs6000_expand_vec_perm_const): Delete.
29711         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
29712         Delete.
29713         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29714         (altivec_expand_vec_perm_const_le): Take each operand individually.
29715         Operate on constant selectors rather than rtxes.
29716         (altivec_expand_vec_perm_const): Likewise.  Update call to
29717         altivec_expand_vec_perm_const_le.
29718         (rs6000_expand_vec_perm_const): Delete.
29719         (rs6000_vectorize_vec_perm_const_ok): Delete.
29720         (rs6000_vectorize_vec_perm_const): New function.
29721         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29722         an element count and rtx array.
29723         (rs6000_expand_extract_even): Update call accordingly.
29724         (rs6000_expand_interleave): Likewise.
29725         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
29726         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
29727         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
29728         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
29729         (rs6000_expand_vec_perm_const): Delete.
29730         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29731         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29732         (altivec_expand_vec_perm_const_le): Take each operand individually.
29733         Operate on constant selectors rather than rtxes.
29734         (altivec_expand_vec_perm_const): Likewise.  Update call to
29735         altivec_expand_vec_perm_const_le.
29736         (rs6000_expand_vec_perm_const): Delete.
29737         (rs6000_vectorize_vec_perm_const_ok): Delete.
29738         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
29739         reference to the SPE evmerge intructions.
29740         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29741         an element count and rtx array.
29742         (rs6000_expand_extract_even): Update call accordingly.
29743         (rs6000_expand_interleave): Likewise.
29744         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
29745         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
29746         new function.
29747         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29749 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29751         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
29752         vector mode and that that mode matches the mode of the data
29753         being permuted.
29754         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
29755         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
29756         directly using expand_vec_perm_1 when forcing selectors into
29757         registers.
29758         (expand_vec_perm_var): New function, split out from expand_vec_perm.
29760 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29762         * optabs-query.h (can_vec_perm_p): Delete.
29763         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
29764         * optabs-query.c (can_vec_perm_p): Split into...
29765         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
29766         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
29767         particular selector is valid.
29768         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29769         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29770         (vect_grouped_load_supported): Likewise.
29771         (vect_shift_permute_load_chain): Likewise.
29772         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29773         (vect_transform_slp_perm_load): Likewise.
29774         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29775         (vectorizable_bswap): Likewise.
29776         (vect_gen_perm_mask_checked): Likewise.
29777         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
29778         implementations of variable permutation vectors into account
29779         when deciding which selector to use.
29780         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
29781         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
29782         with a false third argument.
29783         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
29784         to test whether the constant selector is valid and can_vec_perm_var_p
29785         to test whether a variable selector is valid.
29787 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29789         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
29790         * optabs-query.c (can_vec_perm_p): Likewise.
29791         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
29792         instead of vec_perm_indices.
29793         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
29794         (vect_gen_perm_mask_checked): Likewise,
29795         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
29796         (vect_gen_perm_mask_checked): Likewise,
29798 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29800         * optabs-query.h (qimode_for_vec_perm): Declare.
29801         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
29802         (qimode_for_vec_perm): ...this new function.
29803         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
29805 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29807         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
29808         does not have a conditional at the top.
29810 2018-01-02  Richard Biener  <rguenther@suse.de>
29812         * ipa-inline.c (big_speedup_p): Fix expression.
29814 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29816         PR target/81616
29817         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
29818         for generic 4->6.
29820 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29822         PR target/81616
29823         Generic tuning.
29824         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
29825         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
29826         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
29827         cond_taken_branch_cost 3->4.
29829 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
29831         PR tree-optimization/83581
29832         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
29833         TODO_cleanup_cfg if any changes have been made.
29835         PR middle-end/83608
29836         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
29837         convert_modes if target mode has the right side, but different mode
29838         class.
29840         PR middle-end/83609
29841         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
29842         last argument when extracting from CONCAT.  If either from_real or
29843         from_imag is NULL, use expansion through memory.  If result is not
29844         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
29845         the parts directly to inner mode, if even that fails, use expansion
29846         through memory.
29848         PR middle-end/83623
29849         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
29850         check for bswap in mode rather than HImode and use that in expand_unop
29851         too.
29853 Copyright (C) 2018 Free Software Foundation, Inc.
29855 Copying and distribution of this file, with or without modification,
29856 are permitted in any medium without royalty provided the copyright
29857 notice and this notice are preserved.